mozilla
/
gecko-dev
зеркало из https://github.com/mozilla/gecko-dev.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

Merge mozilla-central to autoland. a=merge CLOSED TREE

This commit is contained in:
Margareta Eliza Balazs 2018-11-01 11:38:01 +02:00
Родитель 63f8b9a61d 39cb1e96cf
Коммит 1970272fed
222 изменённых файлов: 42134 добавлений и 1526 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

33
build/mobile/remoteautomation.py
Просмотреть файл

@ -427,7 +427,7 @@ class RemoteAutomation(Automation):
if stagedShutdown:
# Trigger an ANR report with "kill -3" (SIGQUIT)
try:
self.device.pkill(self.procName, sig=3, attempts=1)
self.device.pkill(self.procName, sig=3, attempts=1, root=True)
except ADBTimeoutError:
raise
except: # NOQA: E722
@ -435,7 +435,7 @@ class RemoteAutomation(Automation):
time.sleep(3)
# Trigger a breakpad dump with "kill -6" (SIGABRT)
try:
self.device.pkill(self.procName, sig=6, attempts=1)
self.device.pkill(self.procName, sig=6, attempts=1, root=True)
except ADBTimeoutError:
raise
except: # NOQA: E722
@ -447,15 +447,30 @@ class RemoteAutomation(Automation):
print("%s still alive after SIGABRT: waiting..." % self.procName)
time.sleep(5)
else:
return
break
retries += 1
try:
self.device.pkill(self.procName, sig=9, attempts=1)
except ADBTimeoutError:
raise
except: # NOQA: E722
print("%s still alive after SIGKILL!" % self.procName)
if self.device.process_exist(self.procName):
try:
self.device.pkill(self.procName, sig=9, attempts=1, root=True)
except ADBTimeoutError:
raise
except: # NOQA: E722
print("%s still alive after SIGKILL!" % self.procName)
if self.device.process_exist(self.procName):
self.device.stop_application(self.procName)
else:
self.device.stop_application(self.procName)
# Test harnesses use the MOZ_CRASHREPORTER environment variables to suppress
# the interactive crash reporter, but that may not always be effective;
# check for and cleanup errant crashreporters.
crashreporter = "%s.CrashReporter" % self.procName
if self.device.process_exist(crashreporter):
print("Warning: %s unexpectedly found running. Killing..." % crashreporter)
try:
self.device.pkill(crashreporter, root=True)
except ADBTimeoutError:
raise
except: # NOQA: E722
pass
if self.device.process_exist(crashreporter):
print("ERROR: %s still running!!" % crashreporter)

3
config/external/moz.build поставляемый
Просмотреть файл

@ -42,6 +42,9 @@ if CONFIG['MOZ_AV1']:
if not CONFIG['MOZ_SYSTEM_PNG']:
external_dirs += ['media/libpng']
if not CONFIG['MOZ_SYSTEM_WEBP']:
external_dirs += ['media/libwebp']
if CONFIG['CPU_ARCH'] == 'arm':
external_dirs += ['media/openmax_dl/dl']

8
config/system-headers.mozbuild
Просмотреть файл

@ -1255,6 +1255,14 @@ if CONFIG['MOZ_SYSTEM_PNG']:
'png.h',
]
if CONFIG['MOZ_SYSTEM_WEBP']:
system_headers += [
'webp/decode.h',
'webp/demux.h',
'webp/mux_types.h',
'webp/types.h',
]
if CONFIG['MOZ_SYSTEM_ZLIB']:
system_headers += [
'zlib.h',

6
dom/bindings/BindingUtils.cpp
Просмотреть файл

@ -1070,6 +1070,9 @@ CreateInterfaceObjects(JSContext* cx, JS::Handle<JSObject*> global,
isChrome ? chromeOnlyProperties : nullptr,
unscopableNames, toStringTag, isGlobal);
if (!proto) {
if (name && !strcmp(name, "Document")) {
MOZ_CRASH("Bug 1405521/1488480: CreateInterfacePrototypeObject failed for Document.prototype");
}
return;
}
@ -1088,6 +1091,9 @@ CreateInterfaceObjects(JSContext* cx, JS::Handle<JSObject*> global,
isChrome,
defineOnGlobal);
if (!interface) {
if (name && !strcmp(name, "Document")) {
MOZ_CRASH("Bug 1405521/1488480: CreateInterfaceObject failed for Document");
}
if (protoCache) {
// If we fail we need to make sure to clear the value of protoCache we
// set above.

6
dom/bindings/Codegen.py
Просмотреть файл

@ -3113,6 +3113,9 @@ class CGCreateInterfaceObjectsMethod(CGAbstractMethod):
${name}, aDefineOnGlobal,
${unscopableNames},
${isGlobal});
if (protoCache && !*protoCache) {
$*{maybeCrash}
}
""",
protoClass=protoClass,
parentProto=parentProto,
@ -3127,7 +3130,8 @@ class CGCreateInterfaceObjectsMethod(CGAbstractMethod):
chromeProperties=chromeProperties,
name='"' + self.descriptor.interface.identifier.name + '"' if needInterfaceObject else "nullptr",
unscopableNames="unscopableNames" if self.haveUnscopables else "nullptr",
isGlobal=toStringBool(isGlobal))
isGlobal=toStringBool(isGlobal),
maybeCrash=maybecrash("dom::CreateInterfaceObjects failed for Document"))
# If we fail after here, we must clear interface and prototype caches
# using this code: intermediate failure must not expose the interface in

20
dom/media/test/mochitest.ini
Просмотреть файл

@ -709,19 +709,19 @@ skip-if = true # bug 475110 - disabled since we don't play Wave files standalone
[test_autoplay_contentEditable.html]
skip-if = android_version == '17' || android_version == '22' # android(bug 1232305, bug 1232318, bug 1372457)
[test_autoplay_policy.html]
skip-if = android_version == '23' # bug 1424903
skip-if = android_version >= '23' # bug 1424903
[test_autoplay_policy_activation.html]
skip-if = android_version == '23' # bug 1424903
skip-if = android_version >= '23' # bug 1424903
[test_autoplay_policy_eventdown_activation.html]
skip-if = android_version == '23' # bug 1424903
skip-if = android_version >= '23' # bug 1424903
[test_autoplay_policy_key_blacklist.html]
skip-if = android_version == '23' || (verify && debug && (os == 'win')) # bug 1424903
skip-if = android_version >= '23' || (verify && debug && (os == 'win')) # bug 1424903
[test_autoplay_policy_unmute_pauses.html]
skip-if = android_version == '23' # bug 1424903
skip-if = android_version >= '23' # bug 1424903
[test_autoplay_policy_play_before_loadedmetadata.html]
skip-if = android_version == '23' # bug 1424903
skip-if = android_version >= '23' # bug 1424903
[test_autoplay_policy_permission.html]
skip-if = android_version == '23' # bug 1424903
skip-if = android_version >= '23' # bug 1424903
[test_buffered.html]
skip-if = android_version == '22' # bug 1308388, android(bug 1232305)
[test_bug448534.html]
@ -757,7 +757,7 @@ skip-if = (android_version == '23' && debug) || (android_version == '25' && debu
skip-if = (android_version == '23' && debug) || (android_version == '25' && debug) # android(bug 1232305)
[test_bug1018933.html]
[test_bug1113600.html]
skip-if = android_version == '19' || android_version == '22' # bug 1198168, android(bug 1232305)
skip-if = android_version >= '19' # bug 1198168, android(bug 1232305)
tags=capturestream
[test_bug1242338.html]
skip-if = toolkit == 'android' # bug 1306916, bug 1329566, android(bug 1232305)
@ -1088,7 +1088,7 @@ skip-if = toolkit == 'android' || (verify && debug && os == 'win') # android(bug
[test_video_dimensions.html]
skip-if = toolkit == 'android' # bug 1298238, bug 1304535, android(bug 1232305)
[test_resolution_change.html]
skip-if = android_version == '19' # bug 1393866
skip-if = android_version >= '19' # bug 1393866
tags=capturestream
[test_resume.html]
skip-if = true # bug 1021673
@ -1286,7 +1286,7 @@ tags = suspend
[test_temporary_file_blob_video_plays.html]
skip-if = toolkit == 'android' # android(bug 1232305)
[test_videoPlaybackQuality_totalFrames.html]
skip-if = (os == 'win' || android_version == '19') # bug 1374189
skip-if = (os == 'win' || android_version >= '19') # bug 1374189
[test_video_gzip_encoding.html]

4
dom/workers/RuntimeService.cpp
Просмотреть файл

@ -2026,10 +2026,6 @@ RuntimeService::Cleanup()
{
AssertIsOnMainThread();
if (!mShuttingDown) {
Shutdown();
}
nsCOMPtr<nsIObserverService> obs = services::GetObserverService();
NS_WARNING_ASSERTION(obs, "Failed to get observer service?!");

16
gfx/layers/FrameMetrics.h
Просмотреть файл

@ -276,14 +276,28 @@ public:
mDoSmoothScroll = aOther.mDoSmoothScroll;
}
void ApplyRelativeScrollUpdateFrom(const FrameMetrics& aOther)
/**
* Applies the relative scroll offset update contained in aOther to the
* scroll offset contained in this. The scroll delta is clamped to the
* scrollable region.
*
* @returns The clamped scroll offset delta that was applied
*/
CSSPoint ApplyRelativeScrollUpdateFrom(const FrameMetrics& aOther)
{
MOZ_ASSERT(aOther.IsRelative());
CSSPoint origin = mScrollOffset;
CSSPoint delta = (aOther.mScrollOffset - aOther.mBaseScrollOffset);
ClampAndSetScrollOffset(mScrollOffset + delta);
mScrollGeneration = aOther.mScrollGeneration;
return mScrollOffset - origin;
}
/**
* Applies the relative scroll offset update contained in aOther to the
* smooth scroll destination offset contained in this. The scroll delta is
* clamped to the scrollable region.
*/
void ApplyRelativeSmoothScrollUpdateFrom(const FrameMetrics& aOther)
{
MOZ_ASSERT(aOther.IsRelative());

13
gfx/layers/apz/src/AsyncPanZoomAnimation.h
Просмотреть файл

@ -31,6 +31,19 @@ public:
virtual bool DoSample(FrameMetrics& aFrameMetrics,
const TimeDuration& aDelta) = 0;
/**
* Attempt to apply a translation to the animation in response to content
* providing a relative scroll offset update.
*
* @param aShiftDelta the amount to translate the animation in app units
* @returns Whether the animation was able to translate. If false, the
* animation must be canceled.
*/
virtual bool ApplyContentShift(const CSSPoint& aShiftDelta)
{
return false;
}
bool Sample(FrameMetrics& aFrameMetrics,
const TimeDuration& aDelta) {
// In some situations, particularly when handoff is involved, it's possible

18
gfx/layers/apz/src/AsyncPanZoomController.cpp
Просмотреть файл

@ -4380,6 +4380,7 @@ void AsyncPanZoomController::NotifyLayersUpdated(const ScrollMetadata& aScrollMe
// becomes incorrect for the purposes of calculating the LD transform. To
// correct this we need to update mExpectedGeckoMetrics to be the
// last thing we know was painted by Gecko.
Maybe<CSSPoint> relativeDelta;
if (gfxPrefs::APZRelativeUpdate() && aLayerMetrics.IsRelative()) {
APZC_LOG("%p relative updating scroll offset from %s by %s\n", this,
ToString(Metrics().GetScrollOffset()).c_str(),
@ -4395,7 +4396,7 @@ void AsyncPanZoomController::NotifyLayersUpdated(const ScrollMetadata& aScrollMe
userAction = true;
}
Metrics().ApplyRelativeScrollUpdateFrom(aLayerMetrics);
relativeDelta = Some(Metrics().ApplyRelativeScrollUpdateFrom(aLayerMetrics));
} else {
APZC_LOG("%p updating scroll offset from %s to %s\n", this,
ToString(Metrics().GetScrollOffset()).c_str(),
@ -4408,10 +4409,17 @@ void AsyncPanZoomController::NotifyLayersUpdated(const ScrollMetadata& aScrollMe
mCompositedScrollOffset = Metrics().GetScrollOffset();
mExpectedGeckoMetrics = aLayerMetrics;
// Cancel the animation (which might also trigger a repaint request)
// after we update the scroll offset above. Otherwise we can be left
// in a state where things are out of sync.
CancelAnimation();
// If we have applied a relative scroll update and a scroll animation is
// happening, attempt to apply a content shift and preserve the
// animation.
if (!mAnimation ||
relativeDelta.isNothing() ||
!mAnimation->ApplyContentShift(relativeDelta.value())) {
// Cancel the animation (which might also trigger a repaint request)
// after we update the scroll offset above. Otherwise we can be left
// in a state where things are out of sync.
CancelAnimation();
}
// Since the scroll offset has changed, we need to recompute the
// displayport margins and send them to layout. Otherwise there might be

7
gfx/layers/apz/src/AutoscrollAnimation.h
Просмотреть файл

@ -22,6 +22,13 @@ public:
bool DoSample(FrameMetrics& aFrameMetrics, const TimeDuration& aDelta) override;
bool ApplyContentShift(const CSSPoint& aShiftDelta) override
{
// Autoscroll works using screen space coordinates, so there's no work we
// need to do to handle a content shift
return true;
}
void Cancel(CancelAnimationFlags aFlags) override;
private:
AsyncPanZoomController& mApzc;

7
gfx/layers/apz/src/GenericScrollAnimation.cpp
Просмотреть файл

@ -109,5 +109,12 @@ GenericScrollAnimation::DoSample(FrameMetrics& aFrameMetrics, const TimeDuration
return !finished;
}
bool
GenericScrollAnimation::ApplyContentShift(const CSSPoint& aShiftDelta)
{
mAnimationPhysics->ApplyContentShift(aShiftDelta);
return true;
}
} // namespace layers
} // namespace mozilla

2
gfx/layers/apz/src/GenericScrollAnimation.h
Просмотреть файл

@ -28,6 +28,8 @@ public:
bool DoSample(FrameMetrics& aFrameMetrics, const TimeDuration& aDelta) override;
bool ApplyContentShift(const CSSPoint& aShiftDelta) override;
void UpdateDelta(TimeStamp aTime,
const nsPoint& aDelta,
const nsSize& aCurrentVelocity);

1
gfx/thebes/gfxPrefs.h
Просмотреть файл

@ -570,6 +570,7 @@ private:
DECL_GFX_PREF(Live, "image.mem.volatile.min_threshold_kb", ImageMemVolatileMinThresholdKB, int32_t, -1);
DECL_GFX_PREF(Once, "image.multithreaded_decoding.limit", ImageMTDecodingLimit, int32_t, -1);
DECL_GFX_PREF(Once, "image.multithreaded_decoding.idle_timeout", ImageMTDecodingIdleTimeout, int32_t, -1);
DECL_GFX_PREF(Live, "image.webp.enabled", ImageWebPEnabled, bool, false);
DECL_GFX_PREF(Once, "layers.acceleration.disabled", LayersAccelerationDisabledDoNotUseDirectly, bool, false);
DECL_GFX_PREF(Live, "layers.acceleration.draw-fps", LayersDrawFPS, bool, false);

13
image/DecoderFactory.cpp
Просмотреть файл

@ -5,6 +5,7 @@
#include "DecoderFactory.h"
#include "gfxPrefs.h"
#include "nsMimeTypes.h"
#include "mozilla/RefPtr.h"
@ -19,6 +20,7 @@
#include "nsBMPDecoder.h"
#include "nsICODecoder.h"
#include "nsIconDecoder.h"
#include "nsWebPDecoder.h"
namespace mozilla {
@ -67,6 +69,11 @@ DecoderFactory::GetDecoderType(const char* aMimeType)
// Icon
} else if (!strcmp(aMimeType, IMAGE_ICON_MS)) {
type = DecoderType::ICON;
// WebP
} else if (!strcmp(aMimeType, IMAGE_WEBP) &&
gfxPrefs::ImageWebPEnabled()) {
type = DecoderType::WEBP;
}
return type;
@ -102,6 +109,9 @@ DecoderFactory::GetDecoder(DecoderType aType,
case DecoderType::ICON:
decoder = new nsIconDecoder(aImage);
break;
case DecoderType::WEBP:
decoder = new nsWebPDecoder(aImage);
break;
default:
MOZ_ASSERT_UNREACHABLE("Unknown decoder type");
}
@ -182,7 +192,8 @@ DecoderFactory::CreateAnimationDecoder(DecoderType aType,
return NS_ERROR_INVALID_ARG;
}
MOZ_ASSERT(aType == DecoderType::GIF || aType == DecoderType::PNG,
MOZ_ASSERT(aType == DecoderType::GIF || aType == DecoderType::PNG ||
aType == DecoderType::WEBP,
"Calling CreateAnimationDecoder for non-animating DecoderType");
// Create an anonymous decoder. Interaction with the SurfaceCache and the

1
image/DecoderFactory.h
Просмотреть файл

@ -37,6 +37,7 @@ enum class DecoderType
BMP,
ICO,
ICON,
WEBP,
UNKNOWN
};

8
image/SourceBuffer.h
Просмотреть файл

@ -175,6 +175,14 @@ public:
return mState == READY ? mData.mIterating.mNextReadLength : 0;
}
/// If we're ready to read, returns whether or not everything available thus
/// far has been in the same contiguous buffer.
bool IsContiguous() const
{
MOZ_ASSERT(mState == READY, "Calling IsContiguous() in the wrong state");
return mState == READY ? mData.mIterating.mChunk == 0 : false;
}
/// @return a count of the chunks we've advanced through.
uint32_t ChunkCount() const { return mChunkCount; }

1
image/build/nsImageModule.cpp
Просмотреть файл

@ -82,6 +82,7 @@ static const mozilla::Module::CategoryEntry kImageCategories[] = {
{ "Gecko-Content-Viewers", IMAGE_PNG, "@mozilla.org/content/document-loader-factory;1" },
{ "Gecko-Content-Viewers", IMAGE_APNG, "@mozilla.org/content/document-loader-factory;1" },
{ "Gecko-Content-Viewers", IMAGE_X_PNG, "@mozilla.org/content/document-loader-factory;1" },
{ "Gecko-Content-Viewers", IMAGE_WEBP, "@mozilla.org/content/document-loader-factory;1" },
{ "content-sniffing-services", "@mozilla.org/image/loader;1", "@mozilla.org/image/loader;1" },
{ nullptr }
};

1
image/decoders/moz.build
Просмотреть файл

@ -28,6 +28,7 @@ UNIFIED_SOURCES += [
'nsIconDecoder.cpp',
'nsJPEGDecoder.cpp',
'nsPNGDecoder.cpp',
'nsWebPDecoder.cpp',
]
include('/ipc/chromium/chromium-config.mozbuild')

557
image/decoders/nsWebPDecoder.cpp Normal file
Просмотреть файл

@ -0,0 +1,557 @@
/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*-
*
* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "ImageLogging.h" // Must appear first
#include "nsWebPDecoder.h"
#include "RasterImage.h"
#include "SurfacePipeFactory.h"
using namespace mozilla::gfx;
namespace mozilla {
namespace image {
static LazyLogModule sWebPLog("WebPDecoder");
nsWebPDecoder::nsWebPDecoder(RasterImage* aImage)
: Decoder(aImage)
, mDecoder(nullptr)
, mBlend(BlendMethod::OVER)
, mDisposal(DisposalMethod::KEEP)
, mTimeout(FrameTimeout::Forever())
, mFormat(SurfaceFormat::B8G8R8X8)
, mLastRow(0)
, mCurrentFrame(0)
, mData(nullptr)
, mLength(0)
, mIteratorComplete(false)
, mNeedDemuxer(true)
, mGotColorProfile(false)
, mInProfile(nullptr)
, mTransform(nullptr)
{
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::nsWebPDecoder", this));
}
nsWebPDecoder::~nsWebPDecoder()
{
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::~nsWebPDecoder", this));
if (mDecoder) {
WebPIDelete(mDecoder);
WebPFreeDecBuffer(&mBuffer);
}
if (mInProfile) {
// mTransform belongs to us only if mInProfile is non-null
if (mTransform) {
qcms_transform_release(mTransform);
}
qcms_profile_release(mInProfile);
}
}
LexerResult
nsWebPDecoder::ReadData()
{
MOZ_ASSERT(mData);
MOZ_ASSERT(mLength > 0);
WebPDemuxer* demuxer = nullptr;
bool complete = mIteratorComplete;
if (mNeedDemuxer) {
WebPDemuxState state;
WebPData fragment;
fragment.bytes = mData;
fragment.size = mLength;
demuxer = WebPDemuxPartial(&fragment, &state);
if (state == WEBP_DEMUX_PARSE_ERROR) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::ReadData -- demux parse error\n", this));
WebPDemuxDelete(demuxer);
return LexerResult(TerminalState::FAILURE);
}
if (state == WEBP_DEMUX_PARSING_HEADER) {
WebPDemuxDelete(demuxer);
return LexerResult(Yield::NEED_MORE_DATA);
}
if (!demuxer) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::ReadData -- no demuxer\n", this));
return LexerResult(TerminalState::FAILURE);
}
complete = complete || state == WEBP_DEMUX_DONE;
}
LexerResult rv(TerminalState::FAILURE);
if (!HasSize()) {
rv = ReadHeader(demuxer, complete);
} else {
rv = ReadPayload(demuxer, complete);
}
WebPDemuxDelete(demuxer);
return rv;
}
LexerResult
nsWebPDecoder::DoDecode(SourceBufferIterator& aIterator, IResumable* aOnResume)
{
MOZ_ASSERT(!HasError(), "Shouldn't call DoDecode after error!");
SourceBufferIterator::State state = SourceBufferIterator::COMPLETE;
if (!mIteratorComplete) {
state = aIterator.Advance(SIZE_MAX);
// We need to remember since we can't advance a complete iterator.
mIteratorComplete = state == SourceBufferIterator::COMPLETE;
}
switch (state) {
case SourceBufferIterator::READY:
if (!aIterator.IsContiguous()) {
// We need to buffer. This should be rare, but expensive.
break;
}
if (!mData) {
// For as long as we hold onto an iterator, we know the data pointers
// to the chunks cannot change underneath us, so save the pointer to
// the first block.
MOZ_ASSERT(mLength == 0);
mData = reinterpret_cast<const uint8_t*>(aIterator.Data());
}
mLength += aIterator.Length();
return ReadData();
case SourceBufferIterator::COMPLETE:
return ReadData();
default:
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::DoDecode -- bad state\n", this));
return LexerResult(TerminalState::FAILURE);
}
// We need to buffer. If we have no data buffered, we need to get everything
// from the first chunk of the source buffer before appending the new data.
if (mBufferedData.empty()) {
MOZ_ASSERT(mData);
MOZ_ASSERT(mLength > 0);
if (!mBufferedData.append(mData, mLength)) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::DoDecode -- oom, initialize %zu\n",
this, mLength));
return LexerResult(TerminalState::FAILURE);
}
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::DoDecode -- buffered %zu bytes\n",
this, mLength));
}
// Append the incremental data from the iterator.
if (!mBufferedData.append(aIterator.Data(), aIterator.Length())) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::DoDecode -- oom, append %zu on %zu\n",
this, aIterator.Length(), mBufferedData.length()));
return LexerResult(TerminalState::FAILURE);
}
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::DoDecode -- buffered %zu -> %zu bytes\n",
this, aIterator.Length(), mBufferedData.length()));
mData = mBufferedData.begin();
mLength = mBufferedData.length();
return ReadData();
}
nsresult
nsWebPDecoder::CreateFrame(const nsIntRect& aFrameRect)
{
MOZ_ASSERT(HasSize());
MOZ_ASSERT(!mDecoder);
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::CreateFrame -- frame %u, %d x %d\n",
this, mCurrentFrame, aFrameRect.width, aFrameRect.height));
// If this is our first frame in an animation and it doesn't cover the
// full frame, then we are transparent even if there is no alpha
if (mCurrentFrame == 0 && !aFrameRect.IsEqualEdges(FullFrame())) {
MOZ_ASSERT(HasAnimation());
mFormat = SurfaceFormat::B8G8R8A8;
PostHasTransparency();
}
WebPInitDecBuffer(&mBuffer);
mBuffer.colorspace = MODE_RGBA;
mDecoder = WebPINewDecoder(&mBuffer);
if (!mDecoder) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::CreateFrame -- create decoder error\n",
this));
return NS_ERROR_FAILURE;
}
SurfacePipeFlags pipeFlags = SurfacePipeFlags();
if (ShouldBlendAnimation()) {
pipeFlags |= SurfacePipeFlags::BLEND_ANIMATION;
}
AnimationParams animParams {
aFrameRect, mTimeout, mCurrentFrame, mBlend, mDisposal
};
Maybe<SurfacePipe> pipe = SurfacePipeFactory::CreateSurfacePipe(this,
Size(), OutputSize(), aFrameRect, mFormat, Some(animParams), pipeFlags);
if (!pipe) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::CreateFrame -- no pipe\n", this));
return NS_ERROR_FAILURE;
}
mPipe = std::move(*pipe);
return NS_OK;
}
void
nsWebPDecoder::EndFrame()
{
MOZ_ASSERT(HasSize());
MOZ_ASSERT(mDecoder);
auto opacity = mFormat == SurfaceFormat::B8G8R8A8
? Opacity::SOME_TRANSPARENCY : Opacity::FULLY_OPAQUE;
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::EndFrame -- frame %u, opacity %d, "
"disposal %d, timeout %d, blend %d\n",
this, mCurrentFrame, (int)opacity, (int)mDisposal,
mTimeout.AsEncodedValueDeprecated(), (int)mBlend));
PostFrameStop(opacity);
WebPIDelete(mDecoder);
WebPFreeDecBuffer(&mBuffer);
mDecoder = nullptr;
mLastRow = 0;
++mCurrentFrame;
}
void
nsWebPDecoder::ApplyColorProfile(const char* aProfile, size_t aLength)
{
MOZ_ASSERT(!mGotColorProfile);
mGotColorProfile = true;
if (GetSurfaceFlags() & SurfaceFlags::NO_COLORSPACE_CONVERSION) {
return;
}
auto mode = gfxPlatform::GetCMSMode();
if (mode == eCMSMode_Off || (mode == eCMSMode_TaggedOnly && !aProfile)) {
return;
}
if (!aProfile || !gfxPlatform::GetCMSOutputProfile()) {
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::ApplyColorProfile -- not tagged or no output "
"profile , use sRGB transform\n", this));
mTransform = gfxPlatform::GetCMSRGBATransform();
return;
}
mInProfile = qcms_profile_from_memory(aProfile, aLength);
if (!mInProfile) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::ApplyColorProfile -- bad color profile\n",
this));
return;
}
// Calculate rendering intent.
int intent = gfxPlatform::GetRenderingIntent();
if (intent == -1) {
intent = qcms_profile_get_rendering_intent(mInProfile);
}
// Create the color management transform.
mTransform = qcms_transform_create(mInProfile,
QCMS_DATA_RGBA_8,
gfxPlatform::GetCMSOutputProfile(),
QCMS_DATA_RGBA_8,
(qcms_intent)intent);
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::ApplyColorProfile -- use tagged "
"transform\n", this));
}
LexerResult
nsWebPDecoder::ReadHeader(WebPDemuxer* aDemuxer,
bool aIsComplete)
{
MOZ_ASSERT(aDemuxer);
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::ReadHeader -- %zu bytes\n", this, mLength));
uint32_t flags = WebPDemuxGetI(aDemuxer, WEBP_FF_FORMAT_FLAGS);
if (!IsMetadataDecode() && !mGotColorProfile) {
if (flags & WebPFeatureFlags::ICCP_FLAG) {
WebPChunkIterator iter;
if (!WebPDemuxGetChunk(aDemuxer, "ICCP", 1, &iter)) {
return aIsComplete ? LexerResult(TerminalState::FAILURE)
: LexerResult(Yield::NEED_MORE_DATA);
}
ApplyColorProfile(reinterpret_cast<const char*>(iter.chunk.bytes),
iter.chunk.size);
WebPDemuxReleaseChunkIterator(&iter);
} else {
ApplyColorProfile(nullptr, 0);
}
}
if (flags & WebPFeatureFlags::ANIMATION_FLAG) {
// A metadata decode expects to get the correct first frame timeout which
// sadly is not provided by the normal WebP header parsing.
WebPIterator iter;
if (!WebPDemuxGetFrame(aDemuxer, 1, &iter)) {
return aIsComplete ? LexerResult(TerminalState::FAILURE)
: LexerResult(Yield::NEED_MORE_DATA);
}
PostIsAnimated(FrameTimeout::FromRawMilliseconds(iter.duration));
WebPDemuxReleaseIterator(&iter);
} else {
// Single frames don't need a demuxer to be created.
mNeedDemuxer = false;
}
uint32_t width = WebPDemuxGetI(aDemuxer, WEBP_FF_CANVAS_WIDTH);
uint32_t height = WebPDemuxGetI(aDemuxer, WEBP_FF_CANVAS_HEIGHT);
if (width > INT32_MAX || height > INT32_MAX) {
return LexerResult(TerminalState::FAILURE);
}
PostSize(width, height);
bool alpha = flags & WebPFeatureFlags::ALPHA_FLAG;
if (alpha) {
mFormat = SurfaceFormat::B8G8R8A8;
PostHasTransparency();
}
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::ReadHeader -- %u x %u, alpha %d, "
"animation %d, metadata decode %d, first frame decode %d\n",
this, width, height, alpha, HasAnimation(),
IsMetadataDecode(), IsFirstFrameDecode()));
if (IsMetadataDecode()) {
return LexerResult(TerminalState::SUCCESS);
}
return ReadPayload(aDemuxer, aIsComplete);
}
LexerResult
nsWebPDecoder::ReadPayload(WebPDemuxer* aDemuxer,
bool aIsComplete)
{
if (!HasAnimation()) {
auto rv = ReadSingle(mData, mLength, FullFrame());
if (rv.is<TerminalState>() &&
rv.as<TerminalState>() == TerminalState::SUCCESS) {
PostDecodeDone();
}
return rv;
}
return ReadMultiple(aDemuxer, aIsComplete);
}
LexerResult
nsWebPDecoder::ReadSingle(const uint8_t* aData, size_t aLength, const IntRect& aFrameRect)
{
MOZ_ASSERT(!IsMetadataDecode());
MOZ_ASSERT(aData);
MOZ_ASSERT(aLength > 0);
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::ReadSingle -- %zu bytes\n", this, aLength));
if (!mDecoder && NS_FAILED(CreateFrame(aFrameRect))) {
return LexerResult(TerminalState::FAILURE);
}
bool complete;
VP8StatusCode status = WebPIUpdate(mDecoder, aData, aLength);
switch (status) {
case VP8_STATUS_OK:
complete = true;
break;
case VP8_STATUS_SUSPENDED:
complete = false;
break;
default:
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::ReadSingle -- append error %d\n",
this, status));
return LexerResult(TerminalState::FAILURE);
}
int lastRow = -1;
int width = 0;
int height = 0;
int stride = 0;
uint8_t* rowStart = WebPIDecGetRGB(mDecoder, &lastRow, &width, &height, &stride);
if (!rowStart || lastRow == -1) {
return LexerResult(Yield::NEED_MORE_DATA);
}
if (width <= 0 || height <= 0 || stride <= 0) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::ReadSingle -- bad (w,h,s) = (%d, %d, %d)\n",
this, width, height, stride));
return LexerResult(TerminalState::FAILURE);
}
const bool noPremultiply =
bool(GetSurfaceFlags() & SurfaceFlags::NO_PREMULTIPLY_ALPHA);
for (int row = mLastRow; row < lastRow; row++) {
uint8_t* src = rowStart + row * stride;
if (mTransform) {
qcms_transform_data(mTransform, src, src, width);
}
WriteState result;
if (noPremultiply) {
result = mPipe.WritePixelsToRow<uint32_t>([&]() -> NextPixel<uint32_t> {
MOZ_ASSERT(mFormat == SurfaceFormat::B8G8R8A8 || src[3] == 0xFF);
const uint32_t pixel =
gfxPackedPixelNoPreMultiply(src[3], src[0], src[1], src[2]);
src += 4;
return AsVariant(pixel);
});
} else {
result = mPipe.WritePixelsToRow<uint32_t>([&]() -> NextPixel<uint32_t> {
MOZ_ASSERT(mFormat == SurfaceFormat::B8G8R8A8 || src[3] == 0xFF);
const uint32_t pixel = gfxPackedPixel(src[3], src[0], src[1], src[2]);
src += 4;
return AsVariant(pixel);
});
}
MOZ_ASSERT(result != WriteState::FAILURE);
MOZ_ASSERT_IF(result == WriteState::FINISHED, complete && row == lastRow - 1);
if (result == WriteState::FAILURE) {
MOZ_LOG(sWebPLog, LogLevel::Error,
("[this=%p] nsWebPDecoder::ReadSingle -- write pixels error\n",
this));
return LexerResult(TerminalState::FAILURE);
}
}
if (mLastRow != lastRow) {
mLastRow = lastRow;
Maybe<SurfaceInvalidRect> invalidRect = mPipe.TakeInvalidRect();
if (invalidRect) {
PostInvalidation(invalidRect->mInputSpaceRect,
Some(invalidRect->mOutputSpaceRect));
}
}
if (!complete) {
return LexerResult(Yield::NEED_MORE_DATA);
}
EndFrame();
return LexerResult(TerminalState::SUCCESS);
}
LexerResult
nsWebPDecoder::ReadMultiple(WebPDemuxer* aDemuxer, bool aIsComplete)
{
MOZ_ASSERT(!IsMetadataDecode());
MOZ_ASSERT(aDemuxer);
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::ReadMultiple\n", this));
bool complete = aIsComplete;
WebPIterator iter;
auto rv = LexerResult(Yield::NEED_MORE_DATA);
if (WebPDemuxGetFrame(aDemuxer, mCurrentFrame + 1, &iter)) {
switch (iter.blend_method) {
case WEBP_MUX_BLEND:
mBlend = BlendMethod::OVER;
break;
case WEBP_MUX_NO_BLEND:
mBlend = BlendMethod::SOURCE;
break;
default:
MOZ_ASSERT_UNREACHABLE("Unhandled blend method");
break;
}
switch (iter.dispose_method) {
case WEBP_MUX_DISPOSE_NONE:
mDisposal = DisposalMethod::KEEP;
break;
case WEBP_MUX_DISPOSE_BACKGROUND:
mDisposal = DisposalMethod::CLEAR;
break;
default:
MOZ_ASSERT_UNREACHABLE("Unhandled dispose method");
break;
}
mFormat = iter.has_alpha ? SurfaceFormat::B8G8R8A8 : SurfaceFormat::B8G8R8X8;
mTimeout = FrameTimeout::FromRawMilliseconds(iter.duration);
nsIntRect frameRect(iter.x_offset, iter.y_offset, iter.width, iter.height);
rv = ReadSingle(iter.fragment.bytes, iter.fragment.size, frameRect);
complete = complete && !WebPDemuxNextFrame(&iter);
WebPDemuxReleaseIterator(&iter);
}
if (rv.is<TerminalState>() &&
rv.as<TerminalState>() == TerminalState::SUCCESS) {
// If we extracted one frame, and it is not the last, we need to yield to
// the lexer to allow the upper layers to acknowledge the frame.
if (!complete && !IsFirstFrameDecode()) {
rv = LexerResult(Yield::OUTPUT_AVAILABLE);
} else {
uint32_t loopCount = WebPDemuxGetI(aDemuxer, WEBP_FF_LOOP_COUNT);
MOZ_LOG(sWebPLog, LogLevel::Debug,
("[this=%p] nsWebPDecoder::ReadMultiple -- loop count %u\n",
this, loopCount));
PostDecodeDone(loopCount - 1);
}
}
return rv;
}
Maybe<Telemetry::HistogramID>
nsWebPDecoder::SpeedHistogram() const
{
return Some(Telemetry::IMAGE_DECODE_SPEED_WEBP);
}
} // namespace image
} // namespace mozilla

111
image/decoders/nsWebPDecoder.h Normal file
Просмотреть файл

@ -0,0 +1,111 @@
/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*-
*
* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#ifndef mozilla_image_decoders_nsWebPDecoder_h
#define mozilla_image_decoders_nsWebPDecoder_h
#include "Decoder.h"
#include "webp/demux.h"
#include "StreamingLexer.h"
#include "SurfacePipe.h"
namespace mozilla {
namespace image {
class RasterImage;
class nsWebPDecoder final : public Decoder
{
public:
virtual ~nsWebPDecoder();
protected:
LexerResult DoDecode(SourceBufferIterator& aIterator,
IResumable* aOnResume) override;
Maybe<Telemetry::HistogramID> SpeedHistogram() const override;
private:
friend class DecoderFactory;
// Decoders should only be instantiated via DecoderFactory.
explicit nsWebPDecoder(RasterImage* aImage);
enum class State
{
WEBP_DATA,
FINISHED_WEBP_DATA
};
void ApplyColorProfile(const char* aProfile, size_t aLength);
LexerResult ReadData();
LexerResult ReadHeader(WebPDemuxer* aDemuxer, bool aIsComplete);
LexerResult ReadPayload(WebPDemuxer* aDemuxer, bool aIsComplete);
nsresult CreateFrame(const nsIntRect& aFrameRect);
void EndFrame();
LexerResult ReadSingle(const uint8_t* aData, size_t aLength,
const IntRect& aFrameRect);
LexerResult ReadMultiple(WebPDemuxer* aDemuxer, bool aIsComplete);
/// The SurfacePipe used to write to the output surface.
SurfacePipe mPipe;
/// The buffer used to accumulate data until the complete WebP header is
/// received, if and only if the iterator is discontiguous.
Vector<uint8_t> mBufferedData;
/// The libwebp output buffer descriptor pointing to the decoded data.
WebPDecBuffer mBuffer;
/// The libwebp incremental decoder descriptor, wraps mBuffer.
WebPIDecoder* mDecoder;
/// Blend method for the current frame.
BlendMethod mBlend;
/// Disposal method for the current frame.
DisposalMethod mDisposal;
/// Frame timeout for the current frame;
FrameTimeout mTimeout;
/// Surface format for the current frame.
gfx::SurfaceFormat mFormat;
/// The last row of decoded pixels written to mPipe.
int mLastRow;
/// Number of decoded frames.
uint32_t mCurrentFrame;
/// Pointer to the start of the contiguous encoded image data.
const uint8_t* mData;
/// Length of data pointed to by mData.
size_t mLength;
/// True if the iterator has reached its end.
bool mIteratorComplete;
/// True if this decoding pass requires a WebPDemuxer.
bool mNeedDemuxer;
/// True if we have setup the color profile for the image.
bool mGotColorProfile;
/// Color management profile from the ICCP chunk in the image.
qcms_profile* mInProfile;
/// Color management transform to apply to image data.
qcms_transform* mTransform;
};
} // namespace image
} // namespace mozilla
#endif // mozilla_image_decoders_nsWebPDecoder_h

29
image/test/gtest/Common.cpp
Просмотреть файл

@ -38,7 +38,12 @@ AutoInitializeImageLib::AutoInitializeImageLib()
sImageLibInitialized = true;
// Force sRGB to be consistent with reftests.
Preferences::SetBool("gfx.color_management.force_srgb", true);
nsresult rv = Preferences::SetBool("gfx.color_management.force_srgb", true);
EXPECT_TRUE(rv == NS_OK);
// Ensure WebP is enabled to run decoder tests.
rv = Preferences::SetBool("image.webp.enabled", true);
EXPECT_TRUE(rv == NS_OK);
// Ensure that ImageLib services are initialized.
nsCOMPtr<imgITools> imgTools = do_CreateInstance("@mozilla.org/image/tools;1");
@ -551,6 +556,16 @@ ImageTestCase GreenIconTestCase()
TEST_CASE_IS_TRANSPARENT);
}
ImageTestCase GreenWebPTestCase()
{
return ImageTestCase("green.webp", "image/webp", IntSize(100, 100));
}
ImageTestCase GreenWebPIccSrgbTestCase()
{
return ImageTestCase("green.icc_srgb.webp", "image/webp", IntSize(100, 100));
}
ImageTestCase GreenFirstFrameAnimatedGIFTestCase()
{
return ImageTestCase("first-frame-green.gif", "image/gif", IntSize(100, 100),
@ -563,6 +578,12 @@ ImageTestCase GreenFirstFrameAnimatedPNGTestCase()
TEST_CASE_IS_TRANSPARENT | TEST_CASE_IS_ANIMATED);
}
ImageTestCase GreenFirstFrameAnimatedWebPTestCase()
{
return ImageTestCase("first-frame-green.webp", "image/webp", IntSize(100, 100),
TEST_CASE_IS_ANIMATED);
}
ImageTestCase CorruptTestCase()
{
return ImageTestCase("corrupt.jpg", "image/jpeg", IntSize(100, 100),
@ -699,6 +720,12 @@ ImageTestCase DownscaledIconTestCase()
IntSize(20, 20), TEST_CASE_IS_TRANSPARENT);
}
ImageTestCase DownscaledWebPTestCase()
{
return ImageTestCase("downscaled.webp", "image/webp", IntSize(100, 100),
IntSize(20, 20));
}
ImageTestCase DownscaledTransparentICOWithANDMaskTestCase()
{
// This test case is an ICO with AND mask transparency. We want to ensure that

5
image/test/gtest/Common.h
Просмотреть файл

@ -433,9 +433,13 @@ ImageTestCase GreenJPGTestCase();
ImageTestCase GreenBMPTestCase();
ImageTestCase GreenICOTestCase();
ImageTestCase GreenIconTestCase();
ImageTestCase GreenWebPTestCase();
ImageTestCase GreenWebPIccSrgbTestCase();
ImageTestCase GreenFirstFrameAnimatedGIFTestCase();
ImageTestCase GreenFirstFrameAnimatedPNGTestCase();
ImageTestCase GreenFirstFrameAnimatedWebPTestCase();
ImageTestCase CorruptTestCase();
ImageTestCase CorruptBMPWithTruncatedHeader();
@ -460,6 +464,7 @@ ImageTestCase DownscaledJPGTestCase();
ImageTestCase DownscaledBMPTestCase();
ImageTestCase DownscaledICOTestCase();
ImageTestCase DownscaledIconTestCase();
ImageTestCase DownscaledWebPTestCase();
ImageTestCase DownscaledTransparentICOWithANDMaskTestCase();
ImageTestCase TruncatedSmallGIFTestCase();

1
image/test/gtest/TestDecodeToSurface.cpp
Просмотреть файл

@ -125,6 +125,7 @@ TEST_F(ImageDecodeToSurface, JPG) { RunDecodeToSurface(GreenJPGTestCase()); }
TEST_F(ImageDecodeToSurface, BMP) { RunDecodeToSurface(GreenBMPTestCase()); }
TEST_F(ImageDecodeToSurface, ICO) { RunDecodeToSurface(GreenICOTestCase()); }
TEST_F(ImageDecodeToSurface, Icon) { RunDecodeToSurface(GreenIconTestCase()); }
TEST_F(ImageDecodeToSurface, WebP) { RunDecodeToSurface(GreenWebPTestCase()); }
TEST_F(ImageDecodeToSurface, AnimatedGIF)
{

523
image/test/gtest/TestDecoders.cpp
Просмотреть файл

@ -329,6 +329,247 @@ CheckAnimationDecoderSingleChunk(const ImageTestCase& aTestCase)
});
}
static void
CheckDecoderFrameFirst(const ImageTestCase& aTestCase)
{
// Verify that we can decode this test case and retrieve the first frame using
// imgIContainer::FRAME_FIRST. This ensures that we correctly trigger a
// single-frame decode rather than an animated decode when
// imgIContainer::FRAME_FIRST is requested.
// Create an image.
RefPtr<Image> image =
ImageFactory::CreateAnonymousImage(nsDependentCString(aTestCase.mMimeType));
ASSERT_TRUE(!image->HasError());
nsCOMPtr<nsIInputStream> inputStream = LoadFile(aTestCase.mPath);
ASSERT_TRUE(inputStream);
// Figure out how much data we have.
uint64_t length;
nsresult rv = inputStream->Available(&length);
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Write the data into the image.
rv = image->OnImageDataAvailable(nullptr, nullptr, inputStream, 0,
static_cast<uint32_t>(length));
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Let the image know we've sent all the data.
rv = image->OnImageDataComplete(nullptr, nullptr, NS_OK, true);
ASSERT_TRUE(NS_SUCCEEDED(rv));
RefPtr<ProgressTracker> tracker = image->GetProgressTracker();
tracker->SyncNotifyProgress(FLAG_LOAD_COMPLETE);
// Lock the image so its surfaces don't disappear during the test.
image->LockImage();
auto unlock = mozilla::MakeScopeExit([&] {
image->UnlockImage();
});
// Use GetFrame() to force a sync decode of the image, specifying FRAME_FIRST
// to ensure that we don't get an animated decode.
RefPtr<SourceSurface> surface =
image->GetFrame(imgIContainer::FRAME_FIRST,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that the image's metadata meets our expectations.
IntSize imageSize(0, 0);
rv = image->GetWidth(&imageSize.width);
EXPECT_TRUE(NS_SUCCEEDED(rv));
rv = image->GetHeight(&imageSize.height);
EXPECT_TRUE(NS_SUCCEEDED(rv));
EXPECT_EQ(aTestCase.mSize.width, imageSize.width);
EXPECT_EQ(aTestCase.mSize.height, imageSize.height);
Progress imageProgress = tracker->GetProgress();
EXPECT_TRUE(bool(imageProgress & FLAG_HAS_TRANSPARENCY) == false);
EXPECT_TRUE(bool(imageProgress & FLAG_IS_ANIMATED) == true);
// Ensure that we decoded the static version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ false);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(bool(result.Surface()));
}
// Ensure that we didn't decode the animated version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ false);
ASSERT_EQ(MatchType::NOT_FOUND, result.Type());
}
// Use GetFrame() to force a sync decode of the image, this time specifying
// FRAME_CURRENT to ensure that we get an animated decode.
RefPtr<SourceSurface> animatedSurface =
image->GetFrame(imgIContainer::FRAME_CURRENT,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that we decoded both frames of the animated version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(NS_SUCCEEDED(result.Surface().Seek(0)));
EXPECT_TRUE(bool(result.Surface()));
RefPtr<imgFrame> partialFrame = result.Surface().GetFrame(1);
EXPECT_TRUE(bool(partialFrame));
}
// Ensure that the static version is still around.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(bool(result.Surface()));
}
}
static void
CheckDecoderFrameCurrent(const ImageTestCase& aTestCase)
{
// Verify that we can decode this test case and retrieve the entire sequence
// of frames using imgIContainer::FRAME_CURRENT. This ensures that we
// correctly trigger an animated decode rather than a single-frame decode when
// imgIContainer::FRAME_CURRENT is requested.
// Create an image.
RefPtr<Image> image =
ImageFactory::CreateAnonymousImage(nsDependentCString(aTestCase.mMimeType));
ASSERT_TRUE(!image->HasError());
nsCOMPtr<nsIInputStream> inputStream = LoadFile(aTestCase.mPath);
ASSERT_TRUE(inputStream);
// Figure out how much data we have.
uint64_t length;
nsresult rv = inputStream->Available(&length);
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Write the data into the image.
rv = image->OnImageDataAvailable(nullptr, nullptr, inputStream, 0,
static_cast<uint32_t>(length));
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Let the image know we've sent all the data.
rv = image->OnImageDataComplete(nullptr, nullptr, NS_OK, true);
ASSERT_TRUE(NS_SUCCEEDED(rv));
RefPtr<ProgressTracker> tracker = image->GetProgressTracker();
tracker->SyncNotifyProgress(FLAG_LOAD_COMPLETE);
// Lock the image so its surfaces don't disappear during the test.
image->LockImage();
// Use GetFrame() to force a sync decode of the image, specifying
// FRAME_CURRENT to ensure we get an animated decode.
RefPtr<SourceSurface> surface =
image->GetFrame(imgIContainer::FRAME_CURRENT,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that the image's metadata meets our expectations.
IntSize imageSize(0, 0);
rv = image->GetWidth(&imageSize.width);
EXPECT_TRUE(NS_SUCCEEDED(rv));
rv = image->GetHeight(&imageSize.height);
EXPECT_TRUE(NS_SUCCEEDED(rv));
EXPECT_EQ(aTestCase.mSize.width, imageSize.width);
EXPECT_EQ(aTestCase.mSize.height, imageSize.height);
Progress imageProgress = tracker->GetProgress();
EXPECT_TRUE(bool(imageProgress & FLAG_HAS_TRANSPARENCY) == false);
EXPECT_TRUE(bool(imageProgress & FLAG_IS_ANIMATED) == true);
// Ensure that we decoded both frames of the animated version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(NS_SUCCEEDED(result.Surface().Seek(0)));
EXPECT_TRUE(bool(result.Surface()));
RefPtr<imgFrame> partialFrame = result.Surface().GetFrame(1);
EXPECT_TRUE(bool(partialFrame));
}
// Ensure that we didn't decode the static version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ false);
ASSERT_EQ(MatchType::NOT_FOUND, result.Type());
}
// Use GetFrame() to force a sync decode of the image, this time specifying
// FRAME_FIRST to ensure that we get a single-frame decode.
RefPtr<SourceSurface> animatedSurface =
image->GetFrame(imgIContainer::FRAME_FIRST,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that we decoded the static version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(bool(result.Surface()));
}
// Ensure that both frames of the animated version are still around.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(NS_SUCCEEDED(result.Surface().Seek(0)));
EXPECT_TRUE(bool(result.Surface()));
RefPtr<imgFrame> partialFrame = result.Surface().GetFrame(1);
EXPECT_TRUE(bool(partialFrame));
}
}
class ImageDecoders : public ::testing::Test
{
protected:
@ -430,6 +671,26 @@ TEST_F(ImageDecoders, IconDownscaleDuringDecode)
CheckDownscaleDuringDecode(DownscaledIconTestCase());
}
TEST_F(ImageDecoders, WebPSingleChunk)
{
CheckDecoderSingleChunk(GreenWebPTestCase());
}
TEST_F(ImageDecoders, WebPMultiChunk)
{
CheckDecoderMultiChunk(GreenWebPTestCase());
}
TEST_F(ImageDecoders, WebPDownscaleDuringDecode)
{
CheckDownscaleDuringDecode(DownscaledWebPTestCase());
}
TEST_F(ImageDecoders, WebPIccSrgbMultiChunk)
{
CheckDecoderMultiChunk(GreenWebPIccSrgbTestCase());
}
TEST_F(ImageDecoders, AnimatedGIFSingleChunk)
{
CheckDecoderSingleChunk(GreenFirstFrameAnimatedGIFTestCase());
@ -460,6 +721,21 @@ TEST_F(ImageDecoders, AnimatedPNGWithBlendedFrames)
CheckAnimationDecoderSingleChunk(GreenFirstFrameAnimatedPNGTestCase());
}
TEST_F(ImageDecoders, AnimatedWebPSingleChunk)
{
CheckDecoderSingleChunk(GreenFirstFrameAnimatedWebPTestCase());
}
TEST_F(ImageDecoders, AnimatedWebPMultiChunk)
{
CheckDecoderMultiChunk(GreenFirstFrameAnimatedWebPTestCase());
}
TEST_F(ImageDecoders, AnimatedWebPWithBlendedFrames)
{
CheckAnimationDecoderSingleChunk(GreenFirstFrameAnimatedWebPTestCase());
}
TEST_F(ImageDecoders, CorruptSingleChunk)
{
CheckDecoderSingleChunk(CorruptTestCase());
@ -507,245 +783,12 @@ TEST_F(ImageDecoders, CorruptICOWithBadBppSingleChunk)
TEST_F(ImageDecoders, AnimatedGIFWithFRAME_FIRST)
{
ImageTestCase testCase = GreenFirstFrameAnimatedGIFTestCase();
// Verify that we can decode this test case and retrieve the first frame using
// imgIContainer::FRAME_FIRST. This ensures that we correctly trigger a
// single-frame decode rather than an animated decode when
// imgIContainer::FRAME_FIRST is requested.
// Create an image.
RefPtr<Image> image =
ImageFactory::CreateAnonymousImage(nsDependentCString(testCase.mMimeType));
ASSERT_TRUE(!image->HasError());
nsCOMPtr<nsIInputStream> inputStream = LoadFile(testCase.mPath);
ASSERT_TRUE(inputStream);
// Figure out how much data we have.
uint64_t length;
nsresult rv = inputStream->Available(&length);
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Write the data into the image.
rv = image->OnImageDataAvailable(nullptr, nullptr, inputStream, 0,
static_cast<uint32_t>(length));
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Let the image know we've sent all the data.
rv = image->OnImageDataComplete(nullptr, nullptr, NS_OK, true);
ASSERT_TRUE(NS_SUCCEEDED(rv));
RefPtr<ProgressTracker> tracker = image->GetProgressTracker();
tracker->SyncNotifyProgress(FLAG_LOAD_COMPLETE);
// Lock the image so its surfaces don't disappear during the test.
image->LockImage();
auto unlock = mozilla::MakeScopeExit([&] {
image->UnlockImage();
});
// Use GetFrame() to force a sync decode of the image, specifying FRAME_FIRST
// to ensure that we don't get an animated decode.
RefPtr<SourceSurface> surface =
image->GetFrame(imgIContainer::FRAME_FIRST,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that the image's metadata meets our expectations.
IntSize imageSize(0, 0);
rv = image->GetWidth(&imageSize.width);
EXPECT_TRUE(NS_SUCCEEDED(rv));
rv = image->GetHeight(&imageSize.height);
EXPECT_TRUE(NS_SUCCEEDED(rv));
EXPECT_EQ(testCase.mSize.width, imageSize.width);
EXPECT_EQ(testCase.mSize.height, imageSize.height);
Progress imageProgress = tracker->GetProgress();
EXPECT_TRUE(bool(imageProgress & FLAG_HAS_TRANSPARENCY) == false);
EXPECT_TRUE(bool(imageProgress & FLAG_IS_ANIMATED) == true);
// Ensure that we decoded the static version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ false);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(bool(result.Surface()));
}
// Ensure that we didn't decode the animated version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ false);
ASSERT_EQ(MatchType::NOT_FOUND, result.Type());
}
// Use GetFrame() to force a sync decode of the image, this time specifying
// FRAME_CURRENT to ensure that we get an animated decode.
RefPtr<SourceSurface> animatedSurface =
image->GetFrame(imgIContainer::FRAME_CURRENT,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that we decoded both frames of the animated version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(NS_SUCCEEDED(result.Surface().Seek(0)));
EXPECT_TRUE(bool(result.Surface()));
RefPtr<imgFrame> partialFrame = result.Surface().GetFrame(1);
EXPECT_TRUE(bool(partialFrame));
}
// Ensure that the static version is still around.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(bool(result.Surface()));
}
CheckDecoderFrameFirst(GreenFirstFrameAnimatedGIFTestCase());
}
TEST_F(ImageDecoders, AnimatedGIFWithFRAME_CURRENT)
{
ImageTestCase testCase = GreenFirstFrameAnimatedGIFTestCase();
// Verify that we can decode this test case and retrieve the entire sequence
// of frames using imgIContainer::FRAME_CURRENT. This ensures that we
// correctly trigger an animated decode rather than a single-frame decode when
// imgIContainer::FRAME_CURRENT is requested.
// Create an image.
RefPtr<Image> image =
ImageFactory::CreateAnonymousImage(nsDependentCString(testCase.mMimeType));
ASSERT_TRUE(!image->HasError());
nsCOMPtr<nsIInputStream> inputStream = LoadFile(testCase.mPath);
ASSERT_TRUE(inputStream);
// Figure out how much data we have.
uint64_t length;
nsresult rv = inputStream->Available(&length);
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Write the data into the image.
rv = image->OnImageDataAvailable(nullptr, nullptr, inputStream, 0,
static_cast<uint32_t>(length));
ASSERT_TRUE(NS_SUCCEEDED(rv));
// Let the image know we've sent all the data.
rv = image->OnImageDataComplete(nullptr, nullptr, NS_OK, true);
ASSERT_TRUE(NS_SUCCEEDED(rv));
RefPtr<ProgressTracker> tracker = image->GetProgressTracker();
tracker->SyncNotifyProgress(FLAG_LOAD_COMPLETE);
// Lock the image so its surfaces don't disappear during the test.
image->LockImage();
// Use GetFrame() to force a sync decode of the image, specifying
// FRAME_CURRENT to ensure we get an animated decode.
RefPtr<SourceSurface> surface =
image->GetFrame(imgIContainer::FRAME_CURRENT,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that the image's metadata meets our expectations.
IntSize imageSize(0, 0);
rv = image->GetWidth(&imageSize.width);
EXPECT_TRUE(NS_SUCCEEDED(rv));
rv = image->GetHeight(&imageSize.height);
EXPECT_TRUE(NS_SUCCEEDED(rv));
EXPECT_EQ(testCase.mSize.width, imageSize.width);
EXPECT_EQ(testCase.mSize.height, imageSize.height);
Progress imageProgress = tracker->GetProgress();
EXPECT_TRUE(bool(imageProgress & FLAG_HAS_TRANSPARENCY) == false);
EXPECT_TRUE(bool(imageProgress & FLAG_IS_ANIMATED) == true);
// Ensure that we decoded both frames of the animated version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(NS_SUCCEEDED(result.Surface().Seek(0)));
EXPECT_TRUE(bool(result.Surface()));
RefPtr<imgFrame> partialFrame = result.Surface().GetFrame(1);
EXPECT_TRUE(bool(partialFrame));
}
// Ensure that we didn't decode the static version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ false);
ASSERT_EQ(MatchType::NOT_FOUND, result.Type());
}
// Use GetFrame() to force a sync decode of the image, this time specifying
// FRAME_FIRST to ensure that we get a single-frame decode.
RefPtr<SourceSurface> animatedSurface =
image->GetFrame(imgIContainer::FRAME_FIRST,
imgIContainer::FLAG_SYNC_DECODE);
// Ensure that we decoded the static version of the image.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eStatic),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(bool(result.Surface()));
}
// Ensure that both frames of the animated version are still around.
{
LookupResult result =
SurfaceCache::Lookup(ImageKey(image.get()),
RasterSurfaceKey(imageSize,
DefaultSurfaceFlags(),
PlaybackType::eAnimated),
/* aMarkUsed = */ true);
ASSERT_EQ(MatchType::EXACT, result.Type());
EXPECT_TRUE(NS_SUCCEEDED(result.Surface().Seek(0)));
EXPECT_TRUE(bool(result.Surface()));
RefPtr<imgFrame> partialFrame = result.Surface().GetFrame(1);
EXPECT_TRUE(bool(partialFrame));
}
CheckDecoderFrameCurrent(GreenFirstFrameAnimatedGIFTestCase());
}
TEST_F(ImageDecoders, AnimatedGIFWithExtraImageSubBlocks)
@ -817,6 +860,16 @@ TEST_F(ImageDecoders, AnimatedGIFWithExtraImageSubBlocks)
EXPECT_TRUE(bool(partialFrame));
}
TEST_F(ImageDecoders, AnimatedWebPWithFRAME_FIRST)
{
CheckDecoderFrameFirst(GreenFirstFrameAnimatedWebPTestCase());
}
TEST_F(ImageDecoders, AnimatedWebPWithFRAME_CURRENT)
{
CheckDecoderFrameCurrent(GreenFirstFrameAnimatedWebPTestCase());
}
TEST_F(ImageDecoders, TruncatedSmallGIFSingleChunk)
{
CheckDecoderSingleChunk(TruncatedSmallGIFTestCase());

1
image/test/gtest/TestMetadata.cpp
Просмотреть файл

@ -153,6 +153,7 @@ TEST_F(ImageDecoderMetadata, JPG) { CheckMetadata(GreenJPGTestCase()); }
TEST_F(ImageDecoderMetadata, BMP) { CheckMetadata(GreenBMPTestCase()); }
TEST_F(ImageDecoderMetadata, ICO) { CheckMetadata(GreenICOTestCase()); }
TEST_F(ImageDecoderMetadata, Icon) { CheckMetadata(GreenIconTestCase()); }
TEST_F(ImageDecoderMetadata, WebP) { CheckMetadata(GreenWebPTestCase()); }
TEST_F(ImageDecoderMetadata, AnimatedGIF)
{

Двоичные данные
image/test/gtest/downscaled.webp Normal file
Просмотреть файл

Двоичный файл не отображается.
Режим "рядом"

После

Ширина:  |  Высота:  |  Размер: 56 B

Двоичные данные
image/test/gtest/first-frame-green.webp Normal file
Просмотреть файл

Двоичный файл не отображается.
Режим "рядом"

После

Ширина:  |  Высота:  |  Размер: 154 B

Двоичные данные
image/test/gtest/green.icc_srgb.webp Normal file
Просмотреть файл

Двоичный файл не отображается.
Режим "рядом"

После

Ширина:  |  Высота:  |  Размер: 3.0 KiB

Двоичные данные
image/test/gtest/green.webp Normal file
Просмотреть файл

Двоичный файл не отображается.
Режим "рядом"

После

Ширина:  |  Высота:  |  Размер: 42 B

4
image/test/gtest/moz.build
Просмотреть файл

@ -48,8 +48,10 @@ TEST_HARNESS_FILES.gtest += [
'downscaled.icon',
'downscaled.jpg',
'downscaled.png',
'downscaled.webp',
'first-frame-green.gif',
'first-frame-green.png',
'first-frame-green.webp',
'first-frame-padding.gif',
'green-1x1-truncated.gif',
'green-large-bmp.ico',
@ -57,10 +59,12 @@ TEST_HARNESS_FILES.gtest += [
'green-multiple-sizes.ico',
'green.bmp',
'green.gif',
'green.icc_srgb.webp',
'green.ico',
'green.icon',
'green.jpg',
'green.png',
'green.webp',
'invalid-truncated-metadata.bmp',
'no-frame-delay.gif',
'rle4.bmp',

84
js/src/frontend/moz.build Normal file
Просмотреть файл

@ -0,0 +1,84 @@
# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*-
# vim: set filetype=python:
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
FINAL_LIBRARY = 'js'
FILES_PER_UNIFIED_FILE = 6
# Includes should be relative to parent path
LOCAL_INCLUDES += [
'!..',
'..'
]
include('../js-config.mozbuild')
include('../js-cxxflags.mozbuild')
# Generate frontend/ReservedWordsGenerated.h from frontend/ReservedWords.h
GENERATED_FILES += ['ReservedWordsGenerated.h']
ReservedWordsGenerated = GENERATED_FILES['ReservedWordsGenerated.h']
ReservedWordsGenerated.script = 'GenerateReservedWords.py'
ReservedWordsGenerated.inputs += ['ReservedWords.h']
UNIFIED_SOURCES += [
'BytecodeCompiler.cpp',
'BytecodeControlStructures.cpp',
'BytecodeEmitter.cpp',
'CallOrNewEmitter.cpp',
'CForEmitter.cpp',
'DoWhileEmitter.cpp',
'ElemOpEmitter.cpp',
'EmitterScope.cpp',
'ExpressionStatementEmitter.cpp',
'FoldConstants.cpp',
'ForInEmitter.cpp',
'ForOfEmitter.cpp',
'ForOfLoopControl.cpp',
'IfEmitter.cpp',
'JumpList.cpp',
'NameFunctions.cpp',
'NameOpEmitter.cpp',
'ParseNode.cpp',
'PropOpEmitter.cpp',
'SwitchEmitter.cpp',
'TDZCheckCache.cpp',
'TokenStream.cpp',
'TryEmitter.cpp',
'WhileEmitter.cpp',
]
# Parser.cpp cannot be built in unified mode because of explicit
# template instantiations.
SOURCES += [
'Parser.cpp',
]
if CONFIG['JS_BUILD_BINAST']:
# Using SOURCES, as UNIFIED_SOURCES causes mysterious bugs on 32-bit platforms.
# These parts of BinAST are designed only to test evolutions of the
# specification.
SOURCES += ['BinTokenReaderTester.cpp']
# These parts of BinAST should eventually move to release.
SOURCES += [
'BinSource-auto.cpp',
'BinSource.cpp',
'BinSourceRuntimeSupport.cpp',
'BinToken.cpp',
'BinTokenReaderBase.cpp',
'BinTokenReaderMultipart.cpp',
]
# Instrument BinAST files for fuzzing as we have a fuzzing target for BinAST.
if CONFIG['FUZZING_INTERFACES'] and CONFIG['LIBFUZZER']:
include('/tools/fuzzing/libfuzzer-flags.mozbuild')
SOURCES['BinSource-auto.cpp'].flags += libfuzzer_flags
SOURCES['BinSource.cpp'].flags += libfuzzer_flags
SOURCES['BinToken.cpp'].flags += libfuzzer_flags
SOURCES['BinTokenReaderBase.cpp'].flags += libfuzzer_flags
SOURCES['BinTokenReaderMultipart.cpp'].flags += libfuzzer_flags

2
js/src/jit/BaselineJIT.cpp
Просмотреть файл

@ -796,7 +796,7 @@ BaselineScript::computeYieldAndAwaitNativeOffsets(JSScript* script)
return nativeCode;
};
mozilla::Span<uint32_t> pcOffsets = script->yieldAndAwaitOffsets();
mozilla::Span<const uint32_t> pcOffsets = script->yieldAndAwaitOffsets();
uint8_t** nativeOffsets = yieldEntryList();
std::transform(pcOffsets.begin(), pcOffsets.end(), nativeOffsets, computeNative);
}

58
js/src/moz.build
Просмотреть файл

@ -210,30 +210,6 @@ UNIFIED_SOURCES += [
'ds/Bitmap.cpp',
'ds/LifoAlloc.cpp',
'ds/MemoryProtectionExceptionHandler.cpp',
'frontend/BytecodeCompiler.cpp',
'frontend/BytecodeControlStructures.cpp',
'frontend/BytecodeEmitter.cpp',
'frontend/CallOrNewEmitter.cpp',
'frontend/CForEmitter.cpp',
'frontend/DoWhileEmitter.cpp',
'frontend/ElemOpEmitter.cpp',
'frontend/EmitterScope.cpp',
'frontend/ExpressionStatementEmitter.cpp',
'frontend/FoldConstants.cpp',
'frontend/ForInEmitter.cpp',
'frontend/ForOfEmitter.cpp',
'frontend/ForOfLoopControl.cpp',
'frontend/IfEmitter.cpp',
'frontend/JumpList.cpp',
'frontend/NameFunctions.cpp',
'frontend/NameOpEmitter.cpp',
'frontend/ParseNode.cpp',
'frontend/PropOpEmitter.cpp',
'frontend/SwitchEmitter.cpp',
'frontend/TDZCheckCache.cpp',
'frontend/TokenStream.cpp',
'frontend/TryEmitter.cpp',
'frontend/WhileEmitter.cpp',
'gc/Allocator.cpp',
'gc/AtomMarking.cpp',
'gc/Barrier.cpp',
@ -373,8 +349,6 @@ UNIFIED_SOURCES += [
# builtin/RegExp.cpp cannot be built in unified mode because it causes huge
# win32 test slowdowns
# frontend/Parser.cpp cannot be built in unified mode because of explicit
# template instantiations.
# jsmath.cpp cannot be built in unified mode because it needs to re-#define the
# RtlGenRandom declaration's calling convention in <ntsecapi.h> on Windows.
# jsutil.cpp cannot be built in unified mode because it is needed for
@ -391,7 +365,6 @@ UNIFIED_SOURCES += [
# files unlucky enough to be unified with it.
SOURCES += [
'builtin/RegExp.cpp',
'frontend/Parser.cpp',
'gc/StoreBuffer.cpp',
'jsmath.cpp',
'jsutil.cpp',
@ -465,15 +438,9 @@ else:
'perf/pm_stub.cpp'
]
GENERATED_FILES += ['frontend/ReservedWordsGenerated.h']
ReservedWordsGenerated = GENERATED_FILES['frontend/ReservedWordsGenerated.h']
ReservedWordsGenerated.script = 'frontend/GenerateReservedWords.py'
ReservedWordsGenerated.inputs += [
'frontend/ReservedWords.h'
]
DIRS += [
'build',
'frontend',
'jit',
]
@ -485,29 +452,6 @@ if CONFIG['ENABLE_WASM_CRANELIFT']:
'wasm/WasmCraneliftCompile.cpp',
]
if CONFIG['JS_BUILD_BINAST']:
# Using SOURCES, as UNIFIED_SOURCES causes mysterious bugs on 32-bit platforms.
# These parts of BinAST are designed only to test evolutions of the
# specification.
SOURCES += ['frontend/BinTokenReaderTester.cpp']
# These parts of BinAST should eventually move to release.
SOURCES += [
'frontend/BinSource-auto.cpp',
'frontend/BinSource.cpp',
'frontend/BinSourceRuntimeSupport.cpp',
'frontend/BinToken.cpp',
'frontend/BinTokenReaderBase.cpp',
'frontend/BinTokenReaderMultipart.cpp',
]
# Instrument BinAST files for fuzzing as we have a fuzzing target for BinAST.
if CONFIG['FUZZING_INTERFACES'] and CONFIG['LIBFUZZER']:
SOURCES['frontend/BinSource-auto.cpp'].flags += libfuzzer_flags
SOURCES['frontend/BinSource.cpp'].flags += libfuzzer_flags
SOURCES['frontend/BinToken.cpp'].flags += libfuzzer_flags
SOURCES['frontend/BinTokenReaderBase.cpp'].flags += libfuzzer_flags
SOURCES['frontend/BinTokenReaderMultipart.cpp'].flags += libfuzzer_flags
# Prepare self-hosted JS code for embedding
GENERATED_FILES += [('selfhosted.out.h', 'selfhosted.js')]
selfhosted = GENERATED_FILES[('selfhosted.out.h', 'selfhosted.js')]

521
js/src/vm/JSScript.cpp
Просмотреть файл

@ -414,13 +414,13 @@ js::XDRScript(XDRState<mode>* xdr, HandleScope scriptEnclosingScope,
nsrcnotes = script->numNotes();
nscopes = script->scopes().size();
if (script->hasConsts()) {
nconsts = script->consts().size();
}
if (script->hasObjects()) {
nobjects = script->objects().size();
}
nscopes = script->scopes().size();
if (script->hasTrynotes()) {
ntrynotes = script->trynotes().size();
}
@ -732,9 +732,11 @@ js::XDRScript(XDRState<mode>* xdr, HandleScope scriptEnclosingScope,
}
}
js::PrivateScriptData* data = script->data_;
if (nconsts) {
RootedValue val(cx);
for (GCPtrValue& elem : script->consts()) {
for (GCPtrValue& elem : data->consts()) {
if (mode == XDR_ENCODE) {
val = elem.get();
}
@ -747,7 +749,7 @@ js::XDRScript(XDRState<mode>* xdr, HandleScope scriptEnclosingScope,
{
MOZ_ASSERT(nscopes != 0);
GCPtrScope* vector = script->scopes().data();
GCPtrScope* vector = data->scopes().data();
RootedScope scope(cx);
RootedScope enclosing(cx);
ScopeKind scopeKind;
@ -844,7 +846,7 @@ js::XDRScript(XDRState<mode>* xdr, HandleScope scriptEnclosingScope,
* after the enclosing block has been XDR'd.
*/
if (nobjects) {
for (GCPtrObject& elem : script->objects()) {
for (GCPtrObject& elem : data->objects()) {
XDRClassKind classk;
if (mode == XDR_ENCODE) {
@ -939,7 +941,7 @@ js::XDRScript(XDRState<mode>* xdr, HandleScope scriptEnclosingScope,
MOZ_TRY(xdr->codeMarker(0xF83B989A));
if (ntrynotes) {
for (JSTryNote& elem : script->trynotes()) {
for (JSTryNote& elem : data->tryNotes()) {
MOZ_TRY(xdr->codeUint8(&elem.kind));
MOZ_TRY(xdr->codeUint32(&elem.stackDepth));
MOZ_TRY(xdr->codeUint32(&elem.start));
@ -948,7 +950,7 @@ js::XDRScript(XDRState<mode>* xdr, HandleScope scriptEnclosingScope,
}
if (nscopenotes) {
for (ScopeNote& elem : script->scopeNotes()) {
for (ScopeNote& elem : data->scopeNotes()) {
MOZ_TRY(xdr->codeUint32(&elem.index));
MOZ_TRY(xdr->codeUint32(&elem.start));
MOZ_TRY(xdr->codeUint32(&elem.length));
@ -957,7 +959,7 @@ js::XDRScript(XDRState<mode>* xdr, HandleScope scriptEnclosingScope,
}
if (nyieldoffsets) {
for (uint32_t& elem : script->yieldAndAwaitOffsets()) {
for (uint32_t& elem : data->yieldAndAwaitOffsets()) {
MOZ_TRY(xdr->codeUint32(&elem));
}
}
@ -3036,123 +3038,207 @@ js::FreeScriptData(JSRuntime* rt)
table.clear();
}
/*
* [SMDOC] JSScript data layout (unshared)
*
* JSScript::data and SharedScriptData::data have complex,
* manually-controlled, memory layouts.
*
* JSScript::data begins with some optional array headers. They are optional
* because they often aren't needed, i.e. the corresponding arrays often have
* zero elements. Each header has a bit in JSScript::hasArrayBits that
* indicates if it's present within |data|; from this the offset of each
* present array header can be computed. Each header has an accessor function
* in JSScript that encapsulates this offset computation.
*
* Array type Array elements Accessor
* ---------- -------------- --------
* ConstArray Consts consts()
* ObjectArray Objects objects()
* ObjectArray Regexps regexps()
* TryNoteArray Try notes trynotes()
* ScopeNoteArray Scope notes scopeNotes()
*
* Then are the elements of several arrays.
* - Most of these arrays have headers listed above (if present). For each of
* these, the array pointer and the array length is stored in the header.
* - The remaining arrays have pointers and lengths that are stored directly in
* JSScript. This is because, unlike the others, they are nearly always
* non-zero length and so the optional-header space optimization isn't
* worthwhile.
*
* Array elements Pointed to by Length
* -------------- ------------- ------
* Consts consts()->vector consts()->length
* Objects objects()->vector objects()->length
* Regexps regexps()->vector regexps()->length
* Try notes trynotes()->vector trynotes()->length
* Scope notes scopeNotes()->vector scopeNotes()->length
*
* IMPORTANT: This layout has two key properties.
* - It ensures that everything has sufficient alignment; in particular, the
* consts() elements need Value alignment.
* - It ensures there are no gaps between elements, which saves space and makes
* manual layout easy. In particular, in the second part, arrays with larger
* elements precede arrays with smaller elements.
*
* The following static assertions check JSScript::data's alignment properties.
*/
template<class T>
constexpr bool
KeepsValueAlignment() {
return alignof(JS::Value) % alignof(T) == 0 &&
sizeof(T) % sizeof(JS::Value) == 0;
}
template<class T>
constexpr bool
HasValueAlignment() {
return alignof(JS::Value) == alignof(T) &&
sizeof(T) == sizeof(JS::Value);
}
template<class T1, class T2>
constexpr bool
NoPaddingBetweenEntries() {
return alignof(T1) % alignof(T2) == 0;
}
/*
* These assertions ensure that there is no padding between the array headers,
* and also that the consts() elements (which follow immediately afterward) are
* Value-aligned. (There is an assumption that |data| itself is Value-aligned;
* we check this below).
*/
JS_STATIC_ASSERT(KeepsValueAlignment<ConstArray>());
JS_STATIC_ASSERT(KeepsValueAlignment<ObjectArray>()); /* there are two of these */
JS_STATIC_ASSERT(KeepsValueAlignment<TryNoteArray>());
JS_STATIC_ASSERT(KeepsValueAlignment<ScopeNoteArray>());
/* These assertions ensure there is no padding required between array elements. */
JS_STATIC_ASSERT(HasValueAlignment<GCPtrValue>());
JS_STATIC_ASSERT((NoPaddingBetweenEntries<GCPtrValue, GCPtrObject>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<GCPtrObject, GCPtrObject>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<GCPtrObject, JSTryNote>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<JSTryNote, uint32_t>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<uint32_t, uint32_t>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<GCPtrValue, ScopeNote>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<ScopeNote, ScopeNote>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<JSTryNote, ScopeNote>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<GCPtrObject, ScopeNote>()));
JS_STATIC_ASSERT((NoPaddingBetweenEntries<ScopeNote, uint32_t>()));
static inline size_t
ScriptDataSize(uint32_t nscopes, uint32_t nconsts, uint32_t nobjects,
uint32_t ntrynotes, uint32_t nscopenotes, uint32_t nyieldoffsets)
// Placement-new elements of an array. This should optimize away for types with
// trivial default initiation.
template <typename T>
static void
DefaultInitializeElements(void* arrayPtr, size_t length)
{
size_t size = 0;
uintptr_t elem = reinterpret_cast<uintptr_t>(arrayPtr);
MOZ_ASSERT(elem % alignof(T) == 0);
MOZ_ASSERT(nscopes != 0);
size += sizeof(ScopeArray) + nscopes * sizeof(Scope*);
if (nconsts != 0) {
size += sizeof(ConstArray) + nconsts * sizeof(Value);
for (size_t i = 0; i < length; ++i) {
new (reinterpret_cast<T*>(elem)) T;
elem += sizeof(T);
}
if (nobjects != 0) {
size += sizeof(ObjectArray) + nobjects * sizeof(NativeObject*);
}
/* static */ size_t
PrivateScriptData::AllocationSize(uint32_t nscopes, uint32_t nconsts, uint32_t nobjects,
uint32_t ntrynotes, uint32_t nscopenotes, uint32_t nyieldoffsets)
{
size_t size = sizeof(PrivateScriptData);
if (nconsts) { size += sizeof(PackedSpan); }
if (nobjects) { size += sizeof(PackedSpan); }
if (ntrynotes) { size += sizeof(PackedSpan); }
if (nscopenotes) { size += sizeof(PackedSpan); }
if (nyieldoffsets) { size += sizeof(PackedSpan); }
size += nscopes * sizeof(GCPtrScope);
if (nconsts) {
// The scope array doesn't maintain Value alignment, so compute the
// padding needed to remedy this.
size = JS_ROUNDUP(size, alignof(GCPtrValue));
size += nconsts * sizeof(GCPtrValue);
}
if (ntrynotes != 0) {
size += sizeof(TryNoteArray) + ntrynotes * sizeof(JSTryNote);
if (nobjects) {
size += nobjects * sizeof(GCPtrObject);
}
if (nscopenotes != 0) {
size += sizeof(ScopeNoteArray) + nscopenotes * sizeof(ScopeNote);
if (ntrynotes) {
size += ntrynotes * sizeof(JSTryNote);
}
if (nyieldoffsets != 0) {
size += sizeof(YieldAndAwaitOffsetArray) + nyieldoffsets * sizeof(uint32_t);
if (nscopenotes) {
size += nscopenotes * sizeof(ScopeNote);
}
if (nyieldoffsets) {
size += nyieldoffsets * sizeof(uint32_t);
}
return size;
return size;
}
// Placement-new elements of an array. This should optimize away for types with
// trivial default initiation.
template <typename T>
void
PrivateScriptData::initElements(size_t offset, size_t length)
{
uintptr_t base = reinterpret_cast<uintptr_t>(this);
DefaultInitializeElements<T>(reinterpret_cast<void*>(base + offset), length);
}
template <typename T>
void
PrivateScriptData::initSpan(size_t* cursor, uint32_t scaledSpanOffset, size_t length)
{
// PackedSpans are elided when arrays are empty
if (scaledSpanOffset == 0) {
MOZ_ASSERT(length == 0);
return;
}
// Placement-new the PackedSpan
PackedSpan* span = packedOffsetToPointer<PackedSpan>(scaledSpanOffset);
span = new (span) PackedSpan { uint32_t(*cursor), uint32_t(length) };
// Placement-new the elements
initElements<T>(*cursor, length);
// Advance cursor
(*cursor) += length * sizeof(T);
}
// Initialize PackedSpans and placement-new the trailing arrays.
PrivateScriptData::PrivateScriptData(uint32_t nscopes_, uint32_t nconsts, uint32_t nobjects,
uint32_t ntrynotes, uint32_t nscopenotes, uint32_t nyieldoffsets)
: nscopes(nscopes_)
{
// Convert cursor possition to a packed offset.
auto ToPackedOffset = [](size_t cursor) {
MOZ_ASSERT(cursor % PackedOffsets::SCALE == 0);
return cursor / PackedOffsets::SCALE;
};
// Helper to allocate a PackedSpan from the variable length data.
auto TakeSpan = [=](size_t* cursor) {
size_t packedOffset = ToPackedOffset(*cursor);
MOZ_ASSERT(packedOffset <= PackedOffsets::MAX_OFFSET);
(*cursor) += sizeof(PackedSpan);
return packedOffset;
};
// Variable-length data begins immediately after PrivateScriptData itself.
// NOTE: Alignment is computed using cursor/offset so the alignment of
// PrivateScriptData must be stricter than any trailing array type.
size_t cursor = sizeof(*this);
// Layout PackedSpan structures and initialize packedOffsets fields.
static_assert(alignof(PrivateScriptData) >= alignof(PackedSpan),
"Incompatible alignment");
if (nconsts) { packedOffsets.constsSpanOffset = TakeSpan(&cursor); }
if (nobjects) { packedOffsets.objectsSpanOffset = TakeSpan(&cursor); }
if (ntrynotes) { packedOffsets.tryNotesSpanOffset = TakeSpan(&cursor); }
if (nscopenotes) { packedOffsets.scopeNotesSpanOffset = TakeSpan(&cursor); }
if (nyieldoffsets) { packedOffsets.yieldOffsetsSpanOffset = TakeSpan(&cursor); }
// Layout and initialize the scopes array. Manually insert padding so that
// the subsequent |consts| array is aligned.
{
MOZ_ASSERT(nscopes > 0);
static_assert(alignof(PackedSpan) >= alignof(GCPtrScope),
"Incompatible alignment");
initElements<GCPtrScope>(cursor, nscopes);
packedOffsets.scopesOffset = ToPackedOffset(cursor);
cursor += nscopes * sizeof(GCPtrScope);
}
if (nconsts) {
// Pad to required alignment if we are emitting constant array.
cursor = JS_ROUNDUP(cursor, alignof(GCPtrValue));
static_assert(alignof(PrivateScriptData) >= alignof(GCPtrValue),
"Incompatible alignment");
initSpan<GCPtrValue>(&cursor, packedOffsets.constsSpanOffset, nconsts);
}
// Layout arrays, initialize PackedSpans and placement-new the elements.
static_assert(alignof(GCPtrValue) >= alignof(GCPtrObject),
"Incompatible alignment");
static_assert(alignof(GCPtrScope) >= alignof(GCPtrObject),
"Incompatible alignment");
initSpan<GCPtrObject>(&cursor, packedOffsets.objectsSpanOffset, nobjects);
static_assert(alignof(GCPtrObject) >= alignof(JSTryNote),
"Incompatible alignment");
initSpan<JSTryNote>(&cursor, packedOffsets.tryNotesSpanOffset, ntrynotes);
static_assert(alignof(JSTryNote) >= alignof(ScopeNote),
"Incompatible alignment");
initSpan<ScopeNote>(&cursor, packedOffsets.scopeNotesSpanOffset, nscopenotes);
static_assert(alignof(ScopeNote) >= alignof(uint32_t),
"Incompatible alignment");
initSpan<uint32_t>(&cursor, packedOffsets.yieldOffsetsSpanOffset, nyieldoffsets);
// Sanity check
MOZ_ASSERT(AllocationSize(nscopes_, nconsts, nobjects,
ntrynotes, nscopenotes, nyieldoffsets) == cursor);
}
/* static */ PrivateScriptData*
PrivateScriptData::new_(JSContext* cx,
uint32_t nscopes, uint32_t nconsts, uint32_t nobjects,
uint32_t ntrynotes, uint32_t nscopenotes, uint32_t nyieldoffsets,
uint32_t* dataSize)
{
// Compute size including trailing arrays
size_t size = AllocationSize(nscopes, nconsts, nobjects,
ntrynotes, nscopenotes, nyieldoffsets);
// Allocate contiguous raw buffer
void* raw = cx->pod_malloc<uint8_t>(size);
MOZ_ASSERT(uintptr_t(raw) % alignof(PrivateScriptData) == 0);
if (!raw) {
return nullptr;
}
if (dataSize) {
*dataSize = size;
}
// Constuct the PrivateScriptData. Trailing arrays are uninitialized but
// GCPtrs are put into a safe state.
return new (raw) PrivateScriptData(nscopes, nconsts, nobjects,
ntrynotes, nscopenotes, nyieldoffsets);
}
void
PrivateScriptData::traceChildren(JSTracer* trc)
{
auto scopearray = scopes();
TraceRange(trc, scopearray.size(), scopearray.data(), "scopes");
if (hasConsts()) {
auto constarray = consts();
TraceRange(trc, constarray.size(), constarray.data(), "consts");
}
if (hasObjects()) {
auto objarray = objects();
TraceRange(trc, objarray.size(), objarray.data(), "objects");
}
}
JSScript::JSScript(JS::Realm* realm, uint8_t* stubEntry, const ReadOnlyCompileOptions& options,
@ -3288,92 +3374,16 @@ JSScript::partiallyInit(JSContext* cx, HandleScript script, uint32_t nscopes,
{
cx->check(script);
size_t size = ScriptDataSize(nscopes, nconsts, nobjects, ntrynotes,
nscopenotes, nyieldoffsets);
script->data = AllocScriptData(cx, size);
if (size && !script->data) {
uint32_t dataSize;
PrivateScriptData* data = PrivateScriptData::new_(cx, nscopes, nconsts, nobjects, ntrynotes,
nscopenotes, nyieldoffsets, &dataSize);
if (!data) {
return false;
}
script->dataSize_ = size;
uint8_t* cursor = script->data;
// There must always be at least 1 scope, the body scope.
MOZ_ASSERT(nscopes != 0);
cursor += sizeof(ScopeArray);
if (nconsts != 0) {
script->setHasArray(CONSTS);
cursor += sizeof(ConstArray);
}
if (nobjects != 0) {
script->setHasArray(OBJECTS);
cursor += sizeof(ObjectArray);
}
if (ntrynotes != 0) {
script->setHasArray(TRYNOTES);
cursor += sizeof(TryNoteArray);
}
if (nscopenotes != 0) {
script->setHasArray(SCOPENOTES);
cursor += sizeof(ScopeNoteArray);
}
YieldAndAwaitOffsetArray* yieldAndAwaitOffsets = nullptr;
if (nyieldoffsets != 0) {
yieldAndAwaitOffsets = reinterpret_cast<YieldAndAwaitOffsetArray*>(cursor);
cursor += sizeof(YieldAndAwaitOffsetArray);
}
if (nconsts != 0) {
MOZ_ASSERT(reinterpret_cast<uintptr_t>(cursor) % sizeof(JS::Value) == 0);
script->constsRaw()->length = nconsts;
script->constsRaw()->vector = (GCPtrValue*)cursor;
cursor += nconsts * sizeof(script->constsRaw()->vector[0]);
}
script->scopesRaw()->length = nscopes;
script->scopesRaw()->vector = (GCPtrScope*)cursor;
cursor += nscopes * sizeof(script->scopesRaw()->vector[0]);
if (nobjects != 0) {
script->objectsRaw()->length = nobjects;
script->objectsRaw()->vector = (GCPtrObject*)cursor;
cursor += nobjects * sizeof(script->objectsRaw()->vector[0]);
}
if (ntrynotes != 0) {
script->trynotesRaw()->length = ntrynotes;
script->trynotesRaw()->vector = reinterpret_cast<JSTryNote*>(cursor);
size_t vectorSize = ntrynotes * sizeof(script->trynotesRaw()->vector[0]);
#ifdef DEBUG
memset(cursor, 0, vectorSize);
#endif
cursor += vectorSize;
}
if (nscopenotes != 0) {
script->scopeNotesRaw()->length = nscopenotes;
script->scopeNotesRaw()->vector = reinterpret_cast<ScopeNote*>(cursor);
size_t vectorSize = nscopenotes * sizeof(script->scopeNotesRaw()->vector[0]);
#ifdef DEBUG
memset(cursor, 0, vectorSize);
#endif
cursor += vectorSize;
}
if (nyieldoffsets != 0) {
yieldAndAwaitOffsets->init(reinterpret_cast<uint32_t*>(cursor), nyieldoffsets);
size_t vectorSize = nyieldoffsets * sizeof(script->yieldAndAwaitOffsetsRaw()[0]);
#ifdef DEBUG
memset(cursor, 0, vectorSize);
#endif
cursor += vectorSize;
}
MOZ_ASSERT(cursor == script->data + size);
script->data_ = data;
script->dataSize_ = dataSize;
return true;
}
@ -3401,7 +3411,9 @@ JSScript::initFunctionPrototype(JSContext* cx, Handle<JSScript*> script,
if (!functionProtoScope) {
return false;
}
script->scopesRaw()->vector[0].init(functionProtoScope);
js::PrivateScriptData* data = script->data_;
data->scopes()[0].init(functionProtoScope);
uint32_t codeLength = 1;
uint32_t srcNotesLength = 1;
@ -3539,21 +3551,23 @@ JSScript::fullyInitFromEmitter(JSContext* cx, HandleScript script, frontend::Byt
return false;
}
js::PrivateScriptData* data = script->data_;
if (bce->numberList.length() != 0) {
bce->numberList.finish(script->consts());
bce->numberList.finish(data->consts());
}
if (bce->objectList.length != 0) {
bce->objectList.finish(script->objects());
bce->objectList.finish(data->objects());
}
if (bce->scopeList.length() != 0) {
bce->scopeList.finish(script->scopes());
bce->scopeList.finish(data->scopes());
}
if (bce->tryNoteList.length() != 0) {
bce->tryNoteList.finish(script->trynotes(), prologueLength);
bce->tryNoteList.finish(data->tryNotes(), prologueLength);
}
if (bce->scopeNoteList.length() != 0) {
bce->scopeNoteList.finish(script->scopeNotes(), prologueLength);
bce->scopeNoteList.finish(data->scopeNotes(), prologueLength);
}
script->bitFields_.strict_ = bce->sc->strict();
script->bitFields_.explicitUseStrict_ = bce->sc->hasExplicitUseStrict();
script->bitFields_.bindingsAccessedDynamically_ = bce->sc->bindingsAccessedDynamically();
@ -3577,7 +3591,7 @@ JSScript::fullyInitFromEmitter(JSContext* cx, HandleScript script, frontend::Byt
// Copy yield offsets last, as the generator kind is set in
// initFromFunctionBox.
if (bce->yieldAndAwaitOffsetList.length() != 0) {
bce->yieldAndAwaitOffsetList.finish(script->yieldAndAwaitOffsets(), prologueLength);
bce->yieldAndAwaitOffsetList.finish(data->yieldAndAwaitOffsets(), prologueLength);
}
#ifdef DEBUG
@ -3658,7 +3672,7 @@ JSScript::computedSizeOfData() const
size_t
JSScript::sizeOfData(mozilla::MallocSizeOf mallocSizeOf) const
{
return mallocSizeOf(data);
return mallocSizeOf(data_);
}
size_t
@ -3700,9 +3714,9 @@ JSScript::finalize(FreeOp* fop)
destroyScriptCounts();
destroyDebugScript(fop);
if (data) {
JS_POISON(data, 0xdb, computedSizeOfData(), MemCheckKind::MakeNoAccess);
fop->free_(data);
if (data_) {
JS_POISON(data_, 0xdb, computedSizeOfData(), MemCheckKind::MakeNoAccess);
fop->free_(data_);
}
if (scriptData_) {
@ -3954,14 +3968,6 @@ js::DescribeScriptedCallerForCompilation(JSContext* cx, MutableHandleScript mayb
}
}
template <class T>
static inline T*
Rebase(JSScript* dst, JSScript* src, T* srcp)
{
size_t off = reinterpret_cast<uint8_t*>(srcp) - src->data;
return reinterpret_cast<T*>(dst->data + off);
}
static JSObject*
CloneInnerInterpretedFunction(JSContext* cx, HandleScope enclosingScope, HandleFunction srcFun)
{
@ -4024,18 +4030,17 @@ js::detail::CopyScript(JSContext* cx, HandleScript src, HandleScript dst,
/* Some embeddings are not careful to use ExposeObjectToActiveJS as needed. */
MOZ_ASSERT(!src->sourceObject()->isMarkedGray());
uint32_t nconsts = src->hasConsts() ? src->consts().size() : 0;
uint32_t nobjects = src->hasObjects() ? src->objects().size() : 0;
uint32_t nscopes = src->scopes().size();
uint32_t ntrynotes = src->hasTrynotes() ? src->trynotes().size() : 0;
uint32_t nscopenotes = src->hasScopeNotes() ? src->scopeNotes().size() : 0;
uint32_t nyieldoffsets = src->hasYieldAndAwaitOffsets() ? src->yieldAndAwaitOffsets().size() : 0;
#ifdef DEBUG
uint32_t nconsts = src->hasConsts() ? src->consts().size() : 0;
#endif
uint32_t nobjects = src->hasObjects() ? src->objects().size() : 0;
/* Script data */
size_t size = src->dataSize();
UniquePtr<uint8_t, JS::FreePolicy> data(AllocScriptData(cx, size));
if (size && !data) {
if (!data) {
return false;
}
@ -4101,13 +4106,9 @@ js::detail::CopyScript(JSContext* cx, HandleScript src, HandleScript dst,
}
}
/* This assignment must occur before all the Rebase calls. */
dst->data = data.release();
dst->data_ = reinterpret_cast<js::PrivateScriptData*>(data.release());
dst->dataSize_ = size;
MOZ_ASSERT(bool(dst->data) == bool(src->data));
if (dst->data) {
memcpy(dst->data, src->data, size);
}
memcpy(dst->data_, src->data_, size);
if (cx->zone() != src->zoneFromAnyThread()) {
for (size_t i = 0; i < src->scriptData()->natoms(); i++) {
@ -4134,7 +4135,6 @@ js::detail::CopyScript(JSContext* cx, HandleScript src, HandleScript dst,
dst->bitFields_.hasMappedArgsObj_ = src->hasMappedArgsObj();
dst->bitFields_.functionHasThisBinding_ = src->functionHasThisBinding();
dst->bitFields_.functionHasExtraBodyVarScope_ = src->functionHasExtraBodyVarScope();
dst->cloneHasArray(src);
dst->bitFields_.strict_ = src->strict();
dst->bitFields_.explicitUseStrict_ = src->explicitUseStrict();
dst->bitFields_.hasNonSyntacticScope_ = scopes[0]->hasOnChain(ScopeKind::NonSyntactic);
@ -4152,36 +4152,26 @@ js::detail::CopyScript(JSContext* cx, HandleScript src, HandleScript dst,
dst->bitFields_.hasRest_ = src->bitFields_.hasRest_;
dst->bitFields_.hideScriptFromDebugger_ = src->bitFields_.hideScriptFromDebugger_;
if (nconsts != 0) {
GCPtrValue* vector = Rebase<GCPtrValue>(dst, src, src->constsRaw()->vector);
dst->constsRaw()->vector = vector;
for (unsigned i = 0; i < nconsts; ++i) {
MOZ_ASSERT_IF(vector[i].isGCThing(), vector[i].toString()->isAtom());
}
}
if (nobjects != 0) {
GCPtrObject* vector = Rebase<GCPtrObject>(dst, src, src->objectsRaw()->vector);
dst->objectsRaw()->vector = vector;
for (unsigned i = 0; i < nobjects; ++i) {
vector[i].init(&objects[i]->as<NativeObject>());
}
}
{
GCPtrScope* vector = Rebase<GCPtrScope>(dst, src, src->scopesRaw()->vector);
dst->scopesRaw()->vector = vector;
auto array = dst->data_->scopes();
for (uint32_t i = 0; i < nscopes; ++i) {
vector[i].init(scopes[i]);
array[i].init(scopes[i]);
}
}
if (ntrynotes != 0) {
dst->trynotesRaw()->vector = Rebase<JSTryNote>(dst, src, src->trynotesRaw()->vector);
#ifdef DEBUG
if (nconsts) {
auto array = dst->data_->consts();
for (unsigned i = 0; i < nconsts; ++i) {
// We don't support GCThings here and thus don't need to call |init|.
MOZ_ASSERT(!array[i].isGCThing());
}
}
if (nscopenotes != 0) {
dst->scopeNotesRaw()->vector = Rebase<ScopeNote>(dst, src, src->scopeNotesRaw()->vector);
}
if (nyieldoffsets != 0) {
dst->yieldAndAwaitOffsetsRaw().vector_ =
Rebase<uint32_t>(dst, src, src->yieldAndAwaitOffsetsRaw().vector_);
#endif
if (nobjects) {
auto array = dst->data_->objects();
for (unsigned i = 0; i < nobjects; ++i) {
array[i].init(objects[i]);
}
}
return true;
@ -4526,25 +4516,14 @@ JSScript::traceChildren(JSTracer* trc)
GCMarker::fromTracer(trc)->shouldCheckCompartments(),
zone()->isCollecting());
if (data_) {
data_->traceChildren(trc);
}
if (scriptData()) {
scriptData()->traceChildren(trc);
}
if (data) {
auto array = scopes();
TraceRange(trc, array.size(), array.data(), "scopes");
}
if (hasConsts()) {
auto array = consts();
TraceRange(trc, array.size(), array.data(), "consts");
}
if (hasObjects()) {
auto array = objects();
TraceRange(trc, array.size(), array.data(), "objects");
}
MOZ_ASSERT_IF(sourceObject(), MaybeForwarded(sourceObject())->compartment() == compartment());
TraceNullableEdge(trc, &sourceObject_, "sourceObject");

337
js/src/vm/JSScript.h
Просмотреть файл

@ -147,53 +147,6 @@ struct ScopeNote {
uint32_t parent; // Index of parent block scope in notes, or NoScopeNote.
};
struct ConstArray {
js::GCPtrValue* vector; // array of indexed constant values
uint32_t length;
};
struct ObjectArray {
js::GCPtrObject* vector; // Array of indexed objects.
uint32_t length; // Count of indexed objects.
};
struct ScopeArray {
js::GCPtrScope* vector; // Array of indexed scopes.
uint32_t length; // Count of indexed scopes.
};
struct TryNoteArray {
JSTryNote* vector; // Array of indexed try notes.
uint32_t length; // Count of indexed try notes.
};
struct ScopeNoteArray {
ScopeNote* vector; // Array of indexed ScopeNote records.
uint32_t length; // Count of indexed try notes.
};
class YieldAndAwaitOffsetArray {
friend bool
detail::CopyScript(JSContext* cx, HandleScript src, HandleScript dst,
MutableHandle<GCVector<Scope*>> scopes);
uint32_t* vector_; // Array of bytecode offsets.
uint32_t length_; // Count of bytecode offsets.
public:
void init(uint32_t* vector, uint32_t length) {
vector_ = vector;
length_ = length;
}
uint32_t& operator[](uint32_t index) {
MOZ_ASSERT(index < length_);
return vector_[index];
}
uint32_t length() const {
return length_;
}
};
class ScriptCounts
{
public:
@ -1274,6 +1227,182 @@ template<XDRMode mode>
XDRResult
XDRScriptConst(XDRState<mode>* xdr, MutableHandleValue vp);
// [SMDOC] - JSScript data layout (unshared)
//
// PrivateScriptData stores variable-length data associated with a script.
// Abstractly a PrivateScriptData consists of all these arrays:
//
// * A non-empty array of GCPtrScope in scopes()
// * A possibly-empty array of GCPtrValue in consts()
// * A possibly-empty array of JSObject* in objects()
// * A possibly-empty array of JSTryNote in tryNotes()
// * A possibly-empty array of ScopeNote in scopeNotes()
// * A possibly-empty array of uint32_t in yieldAndAwaitOffsets()
//
// Accessing any of these arrays just requires calling the appropriate public
// Span-computing function.
//
// Under the hood, PrivateScriptData is a small class followed by a memory
// layout that compactly encodes all these arrays, in this manner (only
// explicit padding, "--" separators for readability only):
//
// <PrivateScriptData itself>
// --
// (OPTIONAL) PackedSpan for consts()
// (OPTIONAL) PackedSpan for objects()
// (OPTIONAL) PackedSpan for tryNotes()
// (OPTIONAL) PackedSpan for scopeNotes()
// (OPTIONAL) PackedSpan for yieldAndAwaitOffsets()
// --
// (REQUIRED) All the GCPtrScopes that constitute scopes()
// --
// (OPTIONAL) If there are consts, padding needed for space so far to be
// GCPtrValue-aligned
// (OPTIONAL) All the GCPtrValues that constitute consts()
// --
// (OPTIONAL) All the GCPtrObjects that constitute objects()
// --
// (OPTIONAL) All the JSTryNotes that constitute tryNotes()
// --
// (OPTIONAL) All the ScopeNotes that constitute scopeNotes()
// --
// (OPTIONAL) All the uint32_t's that constitute yieldAndAwaitOffsets()
//
// The contents of PrivateScriptData indicate which optional items are present.
// PrivateScriptData::packedOffsets contains bit-fields, one per array.
// Multiply each packed offset by sizeof(uint32_t) to compute a *real* offset.
//
// PrivateScriptData::scopesOffset indicates where scopes() begins. The bound
// of five PackedSpans ensures we can encode this offset compactly.
// PrivateScriptData::nscopes indicates the number of GCPtrScopes in scopes().
//
// The other PackedScriptData::*Offset fields indicate where a potential
// corresponding PackedSpan resides. If the packed offset is 0, there is no
// PackedSpan, and the array is empty. Otherwise the PackedSpan's uint32_t
// offset and length fields store: 1) a *non-packed* offset (a literal count of
// bytes offset from the *start* of PrivateScriptData struct) to the
// corresponding array, and 2) the number of elements in the array,
// respectively.
//
// PrivateScriptData and PackedSpan are 64-bit-aligned, so manual alignment in
// trailing fields is only necessary before the first trailing fields with
// increased alignment -- before GCPtrValues for consts(), on 32-bit, where the
// preceding GCPtrScopes as pointers are only 32-bit-aligned.
class alignas(JS::Value) PrivateScriptData final
{
struct PackedOffsets
{
static constexpr size_t SCALE = sizeof(uint32_t);
static constexpr size_t MAX_OFFSET = 0b1111;
// (Scaled) offset to Scopes
uint32_t scopesOffset : 8;
// (Scaled) offset to Spans. These are set to 0 if they don't exist.
uint32_t constsSpanOffset : 4;
uint32_t objectsSpanOffset : 4;
uint32_t tryNotesSpanOffset : 4;
uint32_t scopeNotesSpanOffset : 4;
uint32_t yieldOffsetsSpanOffset : 4;
};
// Detect accidental size regressions.
static_assert(sizeof(PackedOffsets) == sizeof(uint32_t),
"unexpected bit-field packing");
// A span describes base offset and length of one variable length array in
// the private data.
struct alignas(uintptr_t) PackedSpan
{
uint32_t offset;
uint32_t length;
};
// Concrete Fields
PackedOffsets packedOffsets = {}; // zeroes
uint32_t nscopes;
// Translate an offset into a concrete pointer.
template <typename T>
T* offsetToPointer(size_t offset)
{
uintptr_t base = reinterpret_cast<uintptr_t>(this);
uintptr_t elem = base + offset;
return reinterpret_cast<T*>(elem);
}
// Translate a PackedOffsets member into a pointer.
template <typename T>
T* packedOffsetToPointer(size_t packedOffset)
{
return offsetToPointer<T>(packedOffset * PackedOffsets::SCALE);
}
// Translates a PackedOffsets member into a PackedSpan* and then unpacks
// that to a mozilla::Span.
template <typename T>
mozilla::Span<T> packedOffsetToSpan(size_t scaledSpanOffset)
{
PackedSpan* span = packedOffsetToPointer<PackedSpan>(scaledSpanOffset);
T* base = offsetToPointer<T>(span->offset);
return mozilla::MakeSpan(base, span->length);
}
// Helpers for creating initializing trailing data
template <typename T>
void initSpan(size_t* cursor, uint32_t scaledSpanOffset, size_t length);
template <typename T>
void initElements(size_t offset, size_t length);
// Size to allocate
static size_t AllocationSize(uint32_t nscopes, uint32_t nconsts, uint32_t nobjects,
uint32_t ntrynotes, uint32_t nscopenotes, uint32_t nyieldoffsets);
// Initialize header and PackedSpans
PrivateScriptData(uint32_t nscopes_, uint32_t nconsts, uint32_t nobjects,
uint32_t ntrynotes, uint32_t nscopenotes, uint32_t nyieldoffsets);
public:
// Accessors for typed array spans.
mozilla::Span<GCPtrScope> scopes() {
GCPtrScope* base = packedOffsetToPointer<GCPtrScope>(packedOffsets.scopesOffset);
return mozilla::MakeSpan(base, nscopes);
}
mozilla::Span<GCPtrValue> consts() {
return packedOffsetToSpan<GCPtrValue>(packedOffsets.constsSpanOffset);
}
mozilla::Span<GCPtrObject> objects() {
return packedOffsetToSpan<GCPtrObject>(packedOffsets.objectsSpanOffset);
}
mozilla::Span<JSTryNote> tryNotes() {
return packedOffsetToSpan<JSTryNote>(packedOffsets.tryNotesSpanOffset);
}
mozilla::Span<ScopeNote> scopeNotes() {
return packedOffsetToSpan<ScopeNote>(packedOffsets.scopeNotesSpanOffset);
}
mozilla::Span<uint32_t> yieldAndAwaitOffsets() {
return packedOffsetToSpan<uint32_t>(packedOffsets.yieldOffsetsSpanOffset);
}
// Fast tests for if array exists
bool hasConsts() const { return packedOffsets.constsSpanOffset != 0; }
bool hasObjects() const { return packedOffsets.objectsSpanOffset != 0; }
bool hasTryNotes() const { return packedOffsets.tryNotesSpanOffset != 0; }
bool hasScopeNotes() const { return packedOffsets.scopeNotesSpanOffset != 0; }
bool hasYieldOffsets() const { return packedOffsets.yieldOffsetsSpanOffset != 0; }
// Allocate a new PrivateScriptData. Headers and GCPtrs are initialized.
// The size of allocation is returned as an out parameter.
static PrivateScriptData* new_(JSContext* cx,
uint32_t nscopes, uint32_t nconsts, uint32_t nobjects,
uint32_t ntrynotes, uint32_t nscopenotes, uint32_t nyieldoffsets,
uint32_t* dataSize);
void traceChildren(JSTracer* trc);
};
/*
* Common data that can be shared between many scripts in a single runtime.
*/
@ -1404,13 +1533,13 @@ class JSScript : public js::gc::TenuredCell
uint8_t* jitCodeRaw_ = nullptr;
uint8_t* jitCodeSkipArgCheck_ = nullptr;
// Shareable script data
js::SharedScriptData* scriptData_ = nullptr;
public:
// Pointer to variable-length data array (see comment above Create() for
// details).
uint8_t* data = nullptr;
// Unshared variable-length data
js::PrivateScriptData* data_ = nullptr;
public:
JS::Realm* realm_ = nullptr;
private:
@ -1539,14 +1668,6 @@ class JSScript : public js::gc::TenuredCell
* custom-assign particular bit-fields in the constructor body.
*/
// The bits in this field indicate the presence/non-presence of several
// optional arrays in |data|. See the comments above Create() for details.
uint8_t hasArrayBits_ : ARRAY_KIND_BITS;
/*
* All remaining bit-fields are single-bit bools.
*/
// No need for result value of last expression statement.
bool noScriptRval_ : 1;
@ -2392,99 +2513,43 @@ class JSScript : public js::gc::TenuredCell
size_t sizeOfData(mozilla::MallocSizeOf mallocSizeOf) const;
size_t sizeOfTypeScript(mozilla::MallocSizeOf mallocSizeOf) const;
bool hasArray(ArrayKind kind) const {
return bitFields_.hasArrayBits_ & (1 << kind);
}
void setHasArray(ArrayKind kind) { bitFields_.hasArrayBits_ |= (1 << kind); }
void cloneHasArray(JSScript* script) {
bitFields_.hasArrayBits_ = script->bitFields_.hasArrayBits_;
}
bool hasConsts() const { return hasArray(CONSTS); }
bool hasObjects() const { return hasArray(OBJECTS); }
bool hasTrynotes() const { return hasArray(TRYNOTES); }
bool hasScopeNotes() const { return hasArray(SCOPENOTES); }
bool hasYieldAndAwaitOffsets() const {
return isGenerator() || isAsync();
}
#define OFF(fooOff, hasFoo, t) (fooOff() + (hasFoo() ? sizeof(t) : 0))
size_t scopesOffset() const { return 0; }
size_t constsOffset() const { return scopesOffset() + sizeof(js::ScopeArray); }
size_t objectsOffset() const { return OFF(constsOffset, hasConsts, js::ConstArray); }
size_t trynotesOffset() const { return OFF(objectsOffset, hasObjects, js::ObjectArray); }
size_t scopeNotesOffset() const { return OFF(trynotesOffset, hasTrynotes, js::TryNoteArray); }
size_t yieldAndAwaitOffsetsOffset() const {
return OFF(scopeNotesOffset, hasScopeNotes, js::ScopeNoteArray);
}
#undef OFF
size_t dataSize() const { return dataSize_; }
private:
bool hasConsts() const { return data_->hasConsts(); }
bool hasObjects() const { return data_->hasObjects(); }
bool hasTrynotes() const { return data_->hasTryNotes(); }
bool hasScopeNotes() const { return data_->hasScopeNotes(); }
bool hasYieldAndAwaitOffsets() const {
return data_->hasYieldOffsets();
}
js::ConstArray* constsRaw() const {
mozilla::Span<const js::GCPtrScope> scopes() const {
return data_->scopes();
}
mozilla::Span<const js::GCPtrValue> consts() const {
MOZ_ASSERT(hasConsts());
return reinterpret_cast<js::ConstArray*>(data + constsOffset());
return data_->consts();
}
js::ObjectArray* objectsRaw() const {
mozilla::Span<const js::GCPtrObject> objects() const {
MOZ_ASSERT(hasObjects());
return reinterpret_cast<js::ObjectArray*>(data + objectsOffset());
return data_->objects();
}
js::ScopeArray* scopesRaw() const {
return reinterpret_cast<js::ScopeArray*>(data + scopesOffset());
}
js::TryNoteArray* trynotesRaw() const {
mozilla::Span<const JSTryNote> trynotes() const {
MOZ_ASSERT(hasTrynotes());
return reinterpret_cast<js::TryNoteArray*>(data + trynotesOffset());
return data_->tryNotes();
}
js::ScopeNoteArray* scopeNotesRaw() const {
mozilla::Span<const js::ScopeNote> scopeNotes() const {
MOZ_ASSERT(hasScopeNotes());
return reinterpret_cast<js::ScopeNoteArray*>(data + scopeNotesOffset());
return data_->scopeNotes();
}
js::YieldAndAwaitOffsetArray& yieldAndAwaitOffsetsRaw() const {
mozilla::Span<const uint32_t> yieldAndAwaitOffsets() const {
MOZ_ASSERT(hasYieldAndAwaitOffsets());
return *reinterpret_cast<js::YieldAndAwaitOffsetArray*>(data +
yieldAndAwaitOffsetsOffset());
}
public:
mozilla::Span<js::GCPtrValue> consts() const {
js::ConstArray* array = constsRaw();
return mozilla::MakeSpan(array->vector, array->length);
}
mozilla::Span<js::GCPtrObject> objects() const {
js::ObjectArray* array = objectsRaw();
return mozilla::MakeSpan(array->vector, array->length);
}
mozilla::Span<js::GCPtrScope> scopes() const {
js::ScopeArray* array = scopesRaw();
return mozilla::MakeSpan(array->vector, array->length);
}
mozilla::Span<JSTryNote> trynotes() const {
js::TryNoteArray* array = trynotesRaw();
return mozilla::MakeSpan(array->vector, array->length);
}
mozilla::Span<js::ScopeNote> scopeNotes() const {
js::ScopeNoteArray* array = scopeNotesRaw();
return mozilla::MakeSpan(array->vector, array->length);
}
mozilla::Span<uint32_t> yieldAndAwaitOffsets() const {
js::YieldAndAwaitOffsetArray& array = yieldAndAwaitOffsetsRaw();
return mozilla::MakeSpan(&array[0], array.length());
return data_->yieldAndAwaitOffsets();
}
bool hasLoops();

8
layout/generic/ScrollAnimationBezierPhysics.cpp
Просмотреть файл

@ -53,6 +53,14 @@ ScrollAnimationBezierPhysics::Update(const TimeStamp& aTime,
mIsFirstIteration = false;
}
void
ScrollAnimationBezierPhysics::ApplyContentShift(const CSSPoint& aShiftDelta)
{
nsPoint shiftDelta = CSSPoint::ToAppUnits(aShiftDelta);
mStartPos += shiftDelta;
mDestination += shiftDelta;
}
TimeDuration
ScrollAnimationBezierPhysics::ComputeDuration(const TimeStamp& aTime)
{

2
layout/generic/ScrollAnimationBezierPhysics.h
Просмотреть файл

@ -35,6 +35,8 @@ public:
const nsPoint& aDestination,
const nsSize& aCurrentVelocity) override;
void ApplyContentShift(const CSSPoint& aShiftDelta) override;
// Get the velocity at a point in time in nscoords/sec.
nsSize VelocityAt(const TimeStamp& aTime) override;

8
layout/generic/ScrollAnimationMSDPhysics.cpp
Просмотреть файл

@ -48,6 +48,14 @@ ScrollAnimationMSDPhysics::Update(const TimeStamp& aTime,
mIsFirstIteration = false;
}
void
ScrollAnimationMSDPhysics::ApplyContentShift(const CSSPoint& aShiftDelta)
{
nsPoint shiftDelta = CSSPoint::ToAppUnits(aShiftDelta);
mStartPos += shiftDelta;
mDestination += shiftDelta;
}
double
ScrollAnimationMSDPhysics::ComputeSpringConstant(const TimeStamp& aTime)
{

2
layout/generic/ScrollAnimationMSDPhysics.h
Просмотреть файл

@ -25,6 +25,8 @@ public:
const nsPoint& aDestination,
const nsSize& aCurrentVelocity) override;
void ApplyContentShift(const CSSPoint& aShiftDelta) override;
// Get the velocity at a point in time in nscoords/sec.
nsSize VelocityAt(const TimeStamp& aTime) override;

2
layout/generic/ScrollAnimationPhysics.h
Просмотреть файл

@ -19,6 +19,8 @@ public:
const nsPoint& aDestination,
const nsSize& aCurrentVelocity) = 0;
virtual void ApplyContentShift(const CSSPoint& aShiftDelta) = 0;
// Get the velocity at a point in time in nscoords/sec.
virtual nsSize VelocityAt(const TimeStamp& aTime) = 0;

32
layout/reftests/canvas/image-rendering-auto.html Normal file
Просмотреть файл

@ -0,0 +1,32 @@
<!DOCTYPE HTML>
<!--
Any copyright is dedicated to the Public Domain.
http://creativecommons.org/licenses/publicdomain/
-->
<html reftest-zoom="2" class="reftest-wait">
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<title>test image-rendering auto</title>
<style>
canvas { position:absolute;left:0px;top:0px; }
</style>
<script type="text/javascript">
document.addEventListener("MozReftestInvalidate", draw);
function draw() {
var canvas = document.getElementById("canvas");
var ctx = canvas.getContext("2d");
ctx.fillStyle = "rgb(255,0,0)";
ctx.fillRect(25,25,100,100);
ctx.fillStyle = "rgb(0,255,0)";
ctx.fillRect(25,25,50,50);
document.documentElement.removeAttribute('class');
}
</script>
</head>
<body>
<canvas style="image-rendering: auto;" id="canvas" width="300" height="300"></canvas>
</body>
</html>

38
layout/reftests/canvas/image-rendering-script.html Normal file
Просмотреть файл

@ -0,0 +1,38 @@
<!DOCTYPE HTML>
<!--
Any copyright is dedicated to the Public Domain.
http://creativecommons.org/licenses/publicdomain/
-->
<html reftest-zoom="2" class="reftest-wait">
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<title>test image-rendering script change</title>
<style>
canvas { position:absolute;left:0px;top:0px; }
</style>
<script type="text/javascript">
document.addEventListener("MozReftestInvalidate", updateImageRendering);
function updateImageRendering() {
var canvas = document.getElementById("canvas");
canvas.style.imageRendering = '-moz-crisp-edges';
document.documentElement.removeAttribute('class');
}
</script>
</head>
<body>
<canvas style="image-rendering: auto;" id="canvas" width="300" height="300">
</canvas>
<script type="text/javascript">
var canvas = document.getElementById("canvas");
var ctx = canvas.getContext("2d");
ctx.fillStyle = "rgb(255,0,0)";
ctx.fillRect(25,25,100,100);
ctx.fillStyle = "rgb(0,255,0)";
ctx.fillRect(25,25,50,50);
</script>
</body>
</html>

3
layout/reftests/canvas/reftest.list
Просмотреть файл

@ -4,6 +4,9 @@ fuzzy-if(Android,0-8,0-1000) == size-1.html size-1-ref.html
== empty-transaction-1.html empty-transaction-1-ref.html
== image-rendering-test.html image-rendering-ref.html
== image-rendering-script.html image-rendering-ref.html
!= image-rendering-auto.html image-rendering-script.html
== image-shadow.html image-shadow-ref.html
asserts-if(cocoaWidget,0-2) == size-change-1.html size-change-1-ref.html

28
layout/reftests/list-item/image-rendering-css-auto.html Normal file
Просмотреть файл

@ -0,0 +1,28 @@
<!DOCTYPE HTML>
<!--
Any copyright is dedicated to the Public Domain.
http://creativecommons.org/licenses/publicdomain/
-->
<html reftest-zoom="2">
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<title>test list-style-image image-rendering css auto</title>
<style type="text/css">
div {
display: list-item;
list-style-image: url(data:image/gif;base64,R0lGODlhHAAcAMQAAAAAAP///8DP/8bU/8zZ/9Le/9jj/97o/+Tt/Ory9vD48PX96vv/5P//wP//xv//zP//0v//2P//3v///wAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEAABMALAAAAAAcABwAAAWNYCOOZGmeaKquYxSxcGMY8RoJwlujBk7vphsuByz1hr9iQzgkFiHHpkGXekQWCgOhyR0ODAiFhOoodM9oQcExKqffajbJDe+uT/Q67o7Kw/kpfmiAKkxpEjELdQoxCXUHMVFpBTEDXARmXDAOTQQMIgxbQ1QpQgMKciOhOJ8rC6epJq+MNrF9iEq5ug0hADs=);
height: 50px;
border: 1px solid black;
list-style-position: inside;
image-rendering: auto;
}
</style>
</head>
<body>
<div>
</div>
</body>
</html>

28
layout/reftests/list-item/image-rendering-css.html Normal file
Просмотреть файл

@ -0,0 +1,28 @@
<!DOCTYPE HTML>
<!--
Any copyright is dedicated to the Public Domain.
http://creativecommons.org/licenses/publicdomain/
-->
<html reftest-zoom="2">
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<title>test list-style-image image-rendering css -moz-crisp-edges</title>
<style type="text/css">
div {
display: list-item;
list-style-image: url(data:image/gif;base64,R0lGODlhHAAcAMQAAAAAAP///8DP/8bU/8zZ/9Le/9jj/97o/+Tt/Ory9vD48PX96vv/5P//wP//xv//zP//0v//2P//3v///wAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEAABMALAAAAAAcABwAAAWNYCOOZGmeaKquYxSxcGMY8RoJwlujBk7vphsuByz1hr9iQzgkFiHHpkGXekQWCgOhyR0ODAiFhOoodM9oQcExKqffajbJDe+uT/Q67o7Kw/kpfmiAKkxpEjELdQoxCXUHMVFpBTEDXARmXDAOTQQMIgxbQ1QpQgMKciOhOJ8rC6epJq+MNrF9iEq5ug0hADs=);
height: 50px;
border: 1px solid black;
list-style-position: inside;
image-rendering: -moz-crisp-edges;
}
</style>
</head>
<body>
<div>
</div>
</body>
</html>

27
layout/reftests/list-item/image-rendering-ref.html Normal file
Просмотреть файл

@ -0,0 +1,27 @@
<!DOCTYPE HTML>
<!--
Any copyright is dedicated to the Public Domain.
http://creativecommons.org/licenses/publicdomain/
-->
<html reftest-zoom="2">
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<title>reference list-style-image image-rendering</title>
<style type="text/css">
div {
height: 50px;
border: 1px solid black;
image-rendering: -moz-crisp-edges;
}
</style>
</head>
<body>
<div>
<img src=" data:image/gif;base64,R0lGODlhHAAcAMQAAAAAAP///8DP/8bU/8zZ/9Le/9jj/97o/+Tt/Ory9vD48PX96vv/5P//wP//xv//zP//0v//2P//3v///wAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEAABMALAAAAAAcABwAAAWNYCOOZGmeaKquYxSxcGMY8RoJwlujBk7vphsuByz1hr9iQzgkFiHHpkGXekQWCgOhyR0ODAiFhOoodM9oQcExKqffajbJDe+uT/Q67o7Kw/kpfmiAKkxpEjELdQoxCXUHMVFpBTEDXARmXDAOTQQMIgxbQ1QpQgMKciOhOJ8rC6epJq+MNrF9iEq5ug0hADs="
>
</div>
</body>
</html>

37
layout/reftests/list-item/image-rendering-script.html Normal file
Просмотреть файл

@ -0,0 +1,37 @@
<!DOCTYPE HTML>
<!--
Any copyright is dedicated to the Public Domain.
http://creativecommons.org/licenses/publicdomain/
-->
<html reftest-zoom="2" class="reftest-wait">
<head>
<meta http-equiv="content-type" content="text/html; charset=UTF-8">
<title>test list-style-image image-rendering script change</title>
<style type="text/css">
div {
display: list-item;
list-style-image: url(data:image/gif;base64,R0lGODlhHAAcAMQAAAAAAP///8DP/8bU/8zZ/9Le/9jj/97o/+Tt/Ory9vD48PX96vv/5P//wP//xv//zP//0v//2P//3v///wAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACH5BAEAABMALAAAAAAcABwAAAWNYCOOZGmeaKquYxSxcGMY8RoJwlujBk7vphsuByz1hr9iQzgkFiHHpkGXekQWCgOhyR0ODAiFhOoodM9oQcExKqffajbJDe+uT/Q67o7Kw/kpfmiAKkxpEjELdQoxCXUHMVFpBTEDXARmXDAOTQQMIgxbQ1QpQgMKciOhOJ8rC6epJq+MNrF9iEq5ug0hADs=);
height: 50px;
border: 1px solid black;
list-style-position: inside;
image-rendering: auto;
}
</style>
<script type="text/javascript">
document.addEventListener("MozReftestInvalidate", updateImageRendering);
function updateImageRendering() {
var div = document.getElementById("d1");
div.style.imageRendering = '-moz-crisp-edges';
document.documentElement.removeAttribute("class");
}
</script>
</head>
<body>
<div id="d1">
</div>
</body>
</html>

3
layout/reftests/list-item/reftest.list
Просмотреть файл

@ -13,3 +13,6 @@ asserts(1) == ol-reversed-1b.html ol-reversed-1-ref.html # bug 478135
== bullet-intrinsic-isize-1.html bullet-intrinsic-isize-1-ref.html
== bullet-intrinsic-isize-2.html bullet-intrinsic-isize-2-ref.html
== bullet-justify-1.html bullet-justify-1-ref.html
== image-rendering-css.html image-rendering-ref.html
== image-rendering-script.html image-rendering-ref.html
!= image-rendering-css.html image-rendering-css-auto.html

39
media/libwebp/AUTHORS Normal file
Просмотреть файл

@ -0,0 +1,39 @@
Contributors:
- Charles Munger (clm at google dot com)
- Christian Duvivier (cduvivier at google dot com)
- Djordje Pesut (djordje dot pesut at imgtec dot com)
- Hui Su (huisu at google dot com)
- James Zern (jzern at google dot com)
- Jan Engelhardt (jengelh at medozas dot de)
- Jehan (jehan at girinstud dot io)
- Johann (johann dot koenig at duck dot com)
- Jovan Zelincevic (jovan dot zelincevic at imgtec dot com)
- Jyrki Alakuijala (jyrki at google dot com)
- Lode Vandevenne (lode at google dot com)
- Lou Quillio (louquillio at google dot com)
- Mans Rullgard (mans at mansr dot com)
- Marcin Kowalczyk (qrczak at google dot com)
- Martin Olsson (mnemo at minimum dot se)
- Mikołaj Zalewski (mikolajz at google dot com)
- Mislav Bradac (mislavm at google dot com)
- Nico Weber (thakis at chromium dot org)
- Noel Chromium (noel at chromium dot org)
- Owen Rodley (orodley at google dot com)
- Parag Salasakar (img dot mips1 at gmail dot com)
- Pascal Massimino (pascal dot massimino at gmail dot com)
- Paweł Hajdan, Jr (phajdan dot jr at chromium dot org)
- Pierre Joye (pierre dot php at gmail dot com)
- Sam Clegg (sbc at chromium dot org)
- Scott Hancher (seh at google dot com)
- Scott LaVarnway (slavarnway at google dot com)
- Scott Talbot (s at chikachow dot org)
- Slobodan Prijic (slobodan dot prijic at imgtec dot com)
- Somnath Banerjee (somnath dot banerjee at gmail dot com)
- Sriraman Tallam (tmsriram at google dot com)
- Tamar Levy (tamar dot levy at intel dot com)
- Timothy Gu (timothygu99 at gmail dot com)
- Urvang Joshi (urvang at google dot com)
- Vikas Arora (vikasa at google dot com)
- Vincent Rabaud (vrabaud at google dot com)
- Vlad Tsyrklevich (vtsyrklevich at chromium dot org)
- Yang Zhang (yang dot zhang at arm dot com)

30
media/libwebp/COPYING Normal file
Просмотреть файл

@ -0,0 +1,30 @@
Copyright (c) 2010, Google Inc. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
* Neither the name of Google nor the names of its contributors may
be used to endorse or promote products derived from this software
without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

3
media/libwebp/MOZCHANGES Normal file
Просмотреть файл

@ -0,0 +1,3 @@
Changes made to pristine libwebp source by mozilla.org developers.
2018/10/04 -- Synced with libwebp-1.0.0 (bug #1294490).

199
media/libwebp/NEWS Normal file
Просмотреть файл

@ -0,0 +1,199 @@
- 4/2/2018: version 1.0.0
This is a binary compatible release.
* lossy encoder improvements to avoid chroma shifts in various circumstances
(issues #308, #340)
* big-endian fixes for decode, RGBA import and WebPPictureDistortion
Tool updates:
gifwebp, anim_diff - default duration behavior (<= 10ms) changed to match
web browsers, transcoding tools (issue #379)
img2webp, webpmux - allow options to be passed in via a file (issue #355)
- 11/24/2017: version 0.6.1
This is a binary compatible release.
* lossless performance and compression improvements + a new 'cruncher' mode
(-m 6 -q 100)
* ARM performance improvements with clang (15-20% w/ndk r15c, issue #339)
* webp-js: emscripten/webassembly based javascript decoder
* miscellaneous bug & build fixes (issue #329, #332, #343, #353, #360, #361,
#363)
Tool updates / additions:
added webpinfo - prints file format information (issue #330)
gif2webp - loop behavior modified to match Chrome M63+ (crbug.com/649264);
'-loop_compatibility' can be used for the old behavior
- 1/26/2017: version 0.6.0
* lossless performance and compression improvements
* miscellaneous performance improvements (SSE2, NEON, MSA)
* webpmux gained a -duration option allowing for frame timing modification
* new img2webp utility allowing a sequence of images to be converted to
animated webp
* API changes:
- libwebp:
WebPPictureSharpARGBToYUVA
WebPPlaneDistortion
- libwebpmux / gif2webp:
WebPAnimEncoderOptions: kmax <= 0 now disables keyframes, kmax == 1
forces all keyframes. See mux.h and the gif2webp
manpage for details.
- 12/13/2016: version 0.5.2
This is a binary compatible release.
This release covers CVE-2016-8888 and CVE-2016-9085.
* further security related hardening in the tools; fixes to
gif2webp/AnimEncoder (issues #310, #314, #316, #322), cwebp/libwebp (issue
#312)
* full libwebp (encoder & decoder) iOS framework; libwebpdecoder
WebP.framework renamed to WebPDecoder.framework (issue #307)
* CMake support for Android Studio (2.2)
* miscellaneous build related fixes (issue #306, #313)
* miscellaneous documentation improvements (issue #225)
* minor lossy encoder fixes and improvements
- 6/14/2016: version 0.5.1
This is a binary compatible release.
* miscellaneous bug fixes (issues #280, #289)
* reverted alpha plane encoding with color cache for compatibility with
libwebp 0.4.0->0.4.3 (issues #291, #298)
* lossless encoding performance improvements
* memory reduction in both lossless encoding and decoding
* force mux output to be in the extended format (VP8X) when undefined chunks
are present (issue #294)
* gradle, cmake build support
* workaround for compiler bug causing 64-bit decode failures on android
devices using clang-3.8 in the r11c NDK
* various WebPAnimEncoder improvements
- 12/17/2015: version 0.5.0
* miscellaneous bug & build fixes (issues #234, #258, #274, #275, #278)
* encoder & decoder speed-ups on x86/ARM/MIPS for lossy & lossless
- note! YUV->RGB conversion was sped-up, but the results will be slightly
different from previous releases
* various lossless encoder improvements
* gif2webp improvements, -min_size option added
* tools fully support input from stdin and output to stdout (issue #168)
* New WebPAnimEncoder API for creating animations
* New WebPAnimDecoder API for decoding animations
* other API changes:
- libwebp:
WebPPictureSmartARGBToYUVA() (-pre 4 in cwebp)
WebPConfig::exact (-exact in cwebp; -alpha_cleanup is now the default)
WebPConfig::near_lossless (-near_lossless in cwebp)
WebPFree() (free'ing webp allocated memory in other languages)
WebPConfigLosslessPreset()
WebPMemoryWriterClear()
- libwebpdemux: removed experimental fragment related fields and functions
- libwebpmux: WebPMuxSetCanvasSize()
* new libwebpextras library with some uncommon import functions:
WebPImportGray/WebPImportRGB565/WebPImportRGB4444
- 10/15/15: version 0.4.4
This is a binary compatible release.
* rescaling out-of-bounds read fix (issue #254)
* various build fixes and improvements (issues #253, #259, #262, #267, #268)
* container documentation update
* gif2webp transparency fix (issue #245)
- 3/3/15: version 0.4.3
This is a binary compatible release.
* Android / gcc / iOS / MSVS build fixes and improvements
* lossless decode fix (issue #239 -- since 0.4.0)
* documentation / vwebp updates for animation
* multi-threading fix (issue #234)
- 10/13/14: version 0.4.2
This is a binary compatible release.
* Android / gcc build fixes
* (Windows) fix reading from stdin and writing to stdout
* gif2webp: miscellaneous fixes
* fix 'alpha-leak' with lossy compression (issue #220)
* the lossless bitstream spec has been amended to reflect the current code
- 7/24/14: version 0.4.1
This is a binary compatible release.
* AArch64 (arm64) & MIPS support/optimizations
* NEON assembly additions:
- ~25% faster lossy decode / encode (-m 4)
- ~10% faster lossless decode
- ~5-10% faster lossless encode (-m 3/4)
* dwebp/vwebp can read from stdin
* cwebp/gif2webp can write to stdout
* cwebp can read webp files; useful if storing sources as webp lossless
- 12/19/13: version 0.4.0
* improved gif2webp tool
* numerous fixes, compression improvement and speed-up
* dither option added to decoder (dwebp -dither 50 ...)
* improved multi-threaded modes (-mt option)
* improved filtering strength determination
* New function: WebPMuxGetCanvasSize
* BMP and TIFF format output added to 'dwebp'
* Significant memory reduction for decoding lossy images with alpha.
* Intertwined decoding of RGB and alpha for a shorter
time-to-first-decoded-pixel.
* WebPIterator has a new member 'has_alpha' denoting whether the frame
contains transparency.
* Container spec amended with new 'blending method' for animation.
- 6/13/13: version 0.3.1
This is a binary compatible release.
* Add incremental decoding support for images containing ALPH and ICCP chunks.
* Python bindings via swig for the simple encode/decode interfaces similar to
Java.
- 3/20/13: version 0.3.0
This is a binary compatible release.
* WebPINewRGB/WebPINewYUVA accept being passed a NULL output buffer
and will perform auto-allocation.
* default filter option is now '-strong -f 60'
* encoding speed-up for lossy methods 3 to 6
* alpha encoding can be done in parallel to lossy using 'cwebp -mt ...'
* color profile, metadata (XMP/EXIF) and animation support finalized in the
container.
* various NEON assembly additions
Tool updates / additions:
* gif2webp added
* vwebp given color profile & animation support
* cwebp can preserve color profile / metadata with '-metadata'
- 10/30/12: version 0.2.1
* Various security related fixes
* cwebp.exe: fix import errors on Windows XP
* enable DLL builds for mingw targets
- 8/3/12: version 0.2.0
* Add support for ARGB -> YUVA conversion for lossless decoder
New functions: WebPINewYUVA, WebPIDecGetYUVA
* Add stats for lossless and alpha encoding
* Security related hardening: allocation and size checks
* Add PAM output support to dwebp
- 7/19/12: version 0.1.99
* This is a pre-release of 0.2.0, not an rc to allow for further
incompatible changes based on user feedback.
* Alpha channel encode/decode support.
* Lossless encoder/decoder.
* Add TIFF input support to cwebp.
Incompatible changes:
* The encode ABI has been modified to support alpha encoding.
* Deprecated function WebPINew() has been removed.
* Decode function signatures have changed to consistently use size_t over
int/uint32_t.
* decode_vp8.h is no longer installed system-wide.
* cwebp will encode the alpha channel if present.
- 9/19/11: version 0.1.3
* Advanced decoding APIs.
* On-the-fly cropping and rescaling of images.
* SSE2 instructions for decoding performance optimizations on x86 based
platforms.
* Support Multi-threaded decoding.
* 40% improvement in Decoding performance.
* Add support for RGB565, RGBA4444 & ARGB image colorspace.
* Better handling of large picture encoding.
- 3/25/11: version 0.1.2
* Incremental decoding: picture can be decoded byte-by-byte if needs be.
* lot of bug-fixes, consolidation and stabilization
- 2/23/11: initial release of version 0.1, with the new encoder
- 9/30/10: initial release version with only the lightweight decoder

23
media/libwebp/PATENTS Normal file
Просмотреть файл

@ -0,0 +1,23 @@
Additional IP Rights Grant (Patents)
------------------------------------
"These implementations" means the copyrightable works that implement the WebM
codecs distributed by Google as part of the WebM Project.
Google hereby grants to you a perpetual, worldwide, non-exclusive, no-charge,
royalty-free, irrevocable (except as stated in this section) patent license to
make, have made, use, offer to sell, sell, import, transfer, and otherwise
run, modify and propagate the contents of these implementations of WebM, where
such license applies only to those patent claims, both currently owned by
Google and acquired in the future, licensable by Google that are necessarily
infringed by these implementations of WebM. This grant does not include claims
that would be infringed only as a consequence of further modification of these
implementations. If you or your agent or exclusive licensee institute or order
or agree to the institution of patent litigation or any other patent
enforcement activity against any entity (including a cross-claim or
counterclaim in a lawsuit) alleging that any of these implementations of WebM
or any code incorporated within any of these implementations of WebM
constitute direct or contributory patent infringement, or inducement of
patent infringement, then any patent rights granted to you under this License
for these implementations of WebM shall terminate as of the date such
litigation is filed.

782
media/libwebp/README Normal file
Просмотреть файл

@ -0,0 +1,782 @@
__ __ ____ ____ ____
/ \\/ \/ _ \/ _ )/ _ \
\ / __/ _ \ __/
\__\__/\____/\_____/__/ ____ ___
/ _/ / \ \ / _ \/ _/
/ \_/ / / \ \ __/ \__
\____/____/\_____/_____/____/v1.0.0
Description:
============
WebP codec: library to encode and decode images in WebP format. This package
contains the library that can be used in other programs to add WebP support,
as well as the command line tools 'cwebp' and 'dwebp'.
See http://developers.google.com/speed/webp
The latest source tree is available at
https://chromium.googlesource.com/webm/libwebp
It is released under the same license as the WebM project.
See http://www.webmproject.org/license/software/ or the
"COPYING" file for details. An additional intellectual
property rights grant can be found in the file PATENTS.
Building:
=========
Windows build:
--------------
By running:
nmake /f Makefile.vc CFG=release-static RTLIBCFG=static OBJDIR=output
the directory output\release-static\(x64|x86)\bin will contain the tools
cwebp.exe and dwebp.exe. The directory output\release-static\(x64|x86)\lib will
contain the libwebp static library.
The target architecture (x86/x64) is detected by Makefile.vc from the Visual
Studio compiler (cl.exe) available in the system path.
Unix build using makefile.unix:
-------------------------------
On platforms with GNU tools installed (gcc and make), running
make -f makefile.unix
will build the binaries examples/cwebp and examples/dwebp, along
with the static library src/libwebp.a. No system-wide installation
is supplied, as this is a simple alternative to the full installation
system based on the autoconf tools (see below).
Please refer to makefile.unix for additional details and customizations.
Using autoconf tools:
---------------------
Prerequisites:
A compiler (e.g., gcc), make, autoconf, automake, libtool.
On a Debian-like system the following should install everything you need for a
minimal build:
$ sudo apt-get install gcc make autoconf automake libtool
When building from git sources, you will need to run autogen.sh to generate the
configure script.
./configure
make
make install
should be all you need to have the following files
/usr/local/include/webp/decode.h
/usr/local/include/webp/encode.h
/usr/local/include/webp/types.h
/usr/local/lib/libwebp.*
/usr/local/bin/cwebp
/usr/local/bin/dwebp
installed.
Note: A decode-only library, libwebpdecoder, is available using the
'--enable-libwebpdecoder' flag. The encode library is built separately and can
be installed independently using a minor modification in the corresponding
Makefile.am configure files (see comments there). See './configure --help' for
more options.
Building for MIPS Linux:
------------------------
MIPS Linux toolchain stable available releases can be found at:
https://community.imgtec.com/developers/mips/tools/codescape-mips-sdk/available-releases/
# Add toolchain to PATH
export PATH=$PATH:/path/to/toolchain/bin
# 32-bit build for mips32r5 (p5600)
HOST=mips-mti-linux-gnu
MIPS_CFLAGS="-O3 -mips32r5 -mabi=32 -mtune=p5600 -mmsa -mfp64 \
-msched-weight -mload-store-pairs -fPIE"
MIPS_LDFLAGS="-mips32r5 -mabi=32 -mmsa -mfp64 -pie"
# 64-bit build for mips64r6 (i6400)
HOST=mips-img-linux-gnu
MIPS_CFLAGS="-O3 -mips64r6 -mabi=64 -mtune=i6400 -mmsa -mfp64 \
-msched-weight -mload-store-pairs -fPIE"
MIPS_LDFLAGS="-mips64r6 -mabi=64 -mmsa -mfp64 -pie"
./configure --host=${HOST} --build=`config.guess` \
CC="${HOST}-gcc -EL" \
CFLAGS="$MIPS_CFLAGS" \
LDFLAGS="$MIPS_LDFLAGS"
make
make install
CMake:
------
With CMake, you can compile libwebp, cwebp, dwebp, gif2web, img2webp, webpinfo
and the JS bindings.
Prerequisites:
A compiler (e.g., gcc with autotools) and CMake.
On a Debian-like system the following should install everything you need for a
minimal build:
$ sudo apt-get install build-essential cmake
When building from git sources, you will need to run cmake to generate the
makefiles.
mkdir build && cd build && cmake ../
make
make install
If you also want any of the executables, you will need to enable them through
CMake, e.g.:
cmake -DWEBP_BUILD_CWEBP=ON -DWEBP_BUILD_DWEBP=ON ../
or through your favorite interface (like ccmake or cmake-qt-gui).
Finally, once installed, you can also use WebP in your CMake project by doing:
find_package(WebP)
which will define the CMake variables WebP_INCLUDE_DIRS and WebP_LIBRARIES.
Gradle:
-------
The support for Gradle is minimal: it only helps you compile libwebp, cwebp and
dwebp and webpmux_example.
Prerequisites:
A compiler (e.g., gcc with autotools) and gradle.
On a Debian-like system the following should install everything you need for a
minimal build:
$ sudo apt-get install build-essential gradle
When building from git sources, you will need to run the Gradle wrapper with the
appropriate target, e.g. :
./gradlew buildAllExecutables
SWIG bindings:
--------------
To generate language bindings from swig/libwebp.swig at least swig-1.3
(http://www.swig.org) is required.
Currently the following functions are mapped:
Decode:
WebPGetDecoderVersion
WebPGetInfo
WebPDecodeRGBA
WebPDecodeARGB
WebPDecodeBGRA
WebPDecodeBGR
WebPDecodeRGB
Encode:
WebPGetEncoderVersion
WebPEncodeRGBA
WebPEncodeBGRA
WebPEncodeRGB
WebPEncodeBGR
WebPEncodeLosslessRGBA
WebPEncodeLosslessBGRA
WebPEncodeLosslessRGB
WebPEncodeLosslessBGR
See swig/README for more detailed build instructions.
Java bindings:
To build the swig-generated JNI wrapper code at least JDK-1.5 (or equivalent)
is necessary for enum support. The output is intended to be a shared object /
DLL that can be loaded via System.loadLibrary("webp_jni").
Python bindings:
To build the swig-generated Python extension code at least Python 2.6 is
required. Python < 2.6 may build with some minor changes to libwebp.swig or the
generated code, but is untested.
Encoding tool:
==============
The examples/ directory contains tools for encoding (cwebp) and
decoding (dwebp) images.
The easiest use should look like:
cwebp input.png -q 80 -o output.webp
which will convert the input file to a WebP file using a quality factor of 80
on a 0->100 scale (0 being the lowest quality, 100 being the best. Default
value is 75).
You might want to try the -lossless flag too, which will compress the source
(in RGBA format) without any loss. The -q quality parameter will in this case
control the amount of processing time spent trying to make the output file as
small as possible.
A longer list of options is available using the -longhelp command line flag:
> cwebp -longhelp
Usage:
cwebp [-preset <...>] [options] in_file [-o out_file]
If input size (-s) for an image is not specified, it is
assumed to be a PNG, JPEG, TIFF or WebP file.
Options:
-h / -help ............. short help
-H / -longhelp ......... long help
-q <float> ............. quality factor (0:small..100:big), default=75
-alpha_q <int> ......... transparency-compression quality (0..100),
default=100
-preset <string> ....... preset setting, one of:
default, photo, picture,
drawing, icon, text
-preset must come first, as it overwrites other parameters
-z <int> ............... activates lossless preset with given
level in [0:fast, ..., 9:slowest]
-m <int> ............... compression method (0=fast, 6=slowest), default=4
-segments <int> ........ number of segments to use (1..4), default=4
-size <int> ............ target size (in bytes)
-psnr <float> .......... target PSNR (in dB. typically: 42)
-s <int> <int> ......... input size (width x height) for YUV
-sns <int> ............. spatial noise shaping (0:off, 100:max), default=50
-f <int> ............... filter strength (0=off..100), default=60
-sharpness <int> ....... filter sharpness (0:most .. 7:least sharp), default=0
-strong ................ use strong filter instead of simple (default)
-nostrong .............. use simple filter instead of strong
-sharp_yuv ............. use sharper (and slower) RGB->YUV conversion
-partition_limit <int> . limit quality to fit the 512k limit on
the first partition (0=no degradation ... 100=full)
-pass <int> ............ analysis pass number (1..10)
-crop <x> <y> <w> <h> .. crop picture with the given rectangle
-resize <w> <h> ........ resize picture (after any cropping)
-mt .................... use multi-threading if available
-low_memory ............ reduce memory usage (slower encoding)
-map <int> ............. print map of extra info
-print_psnr ............ prints averaged PSNR distortion
-print_ssim ............ prints averaged SSIM distortion
-print_lsim ............ prints local-similarity distortion
-d <file.pgm> .......... dump the compressed output (PGM file)
-alpha_method <int> .... transparency-compression method (0..1), default=1
-alpha_filter <string> . predictive filtering for alpha plane,
one of: none, fast (default) or best
-exact ................. preserve RGB values in transparent area, default=off
-blend_alpha <hex> ..... blend colors against background color
expressed as RGB values written in
hexadecimal, e.g. 0xc0e0d0 for red=0xc0
green=0xe0 and blue=0xd0
-noalpha ............... discard any transparency information
-lossless .............. encode image losslessly, default=off
-near_lossless <int> ... use near-lossless image
preprocessing (0..100=off), default=100
-hint <string> ......... specify image characteristics hint,
one of: photo, picture or graph
-metadata <string> ..... comma separated list of metadata to
copy from the input to the output if present.
Valid values: all, none (default), exif, icc, xmp
-short ................. condense printed message
-quiet ................. don't print anything
-version ............... print version number and exit
-noasm ................. disable all assembly optimizations
-v ..................... verbose, e.g. print encoding/decoding times
-progress .............. report encoding progress
Experimental Options:
-jpeg_like ............. roughly match expected JPEG size
-af .................... auto-adjust filter strength
-pre <int> ............. pre-processing filter
The main options you might want to try in order to further tune the
visual quality are:
-preset
-sns
-f
-m
Namely:
* 'preset' will set up a default encoding configuration targeting a
particular type of input. It should appear first in the list of options,
so that subsequent options can take effect on top of this preset.
Default value is 'default'.
* 'sns' will progressively turn on (when going from 0 to 100) some additional
visual optimizations (like: segmentation map re-enforcement). This option
will balance the bit allocation differently. It tries to take bits from the
"easy" parts of the picture and use them in the "difficult" ones instead.
Usually, raising the sns value (at fixed -q value) leads to larger files,
but with better quality.
Typical value is around '75'.
* 'f' option directly links to the filtering strength used by the codec's
in-loop processing. The higher the value, the smoother the
highly-compressed area will look. This is particularly useful when aiming
at very small files. Typical values are around 20-30. Note that using the
option -strong/-nostrong will change the type of filtering. Use "-f 0" to
turn filtering off.
* 'm' controls the trade-off between encoding speed and quality. Default is 4.
You can try -m 5 or -m 6 to explore more (time-consuming) encoding
possibilities. A lower value will result in faster encoding at the expense
of quality.
Decoding tool:
==============
There is a decoding sample in examples/dwebp.c which will take
a .webp file and decode it to a PNG image file (amongst other formats).
This is simply to demonstrate the use of the API. You can verify the
file test.webp decodes to exactly the same as test_ref.ppm by using:
cd examples
./dwebp test.webp -ppm -o test.ppm
diff test.ppm test_ref.ppm
The full list of options is available using -h:
> dwebp -h
Usage: dwebp in_file [options] [-o out_file]
Decodes the WebP image file to PNG format [Default]
Use following options to convert into alternate image formats:
-pam ......... save the raw RGBA samples as a color PAM
-ppm ......... save the raw RGB samples as a color PPM
-bmp ......... save as uncompressed BMP format
-tiff ........ save as uncompressed TIFF format
-pgm ......... save the raw YUV samples as a grayscale PGM
file with IMC4 layout
-yuv ......... save the raw YUV samples in flat layout
Other options are:
-version ..... print version number and exit
-nofancy ..... don't use the fancy YUV420 upscaler
-nofilter .... disable in-loop filtering
-nodither .... disable dithering
-dither <d> .. dithering strength (in 0..100)
-alpha_dither use alpha-plane dithering if needed
-mt .......... use multi-threading
-crop <x> <y> <w> <h> ... crop output with the given rectangle
-resize <w> <h> ......... scale the output (*after* any cropping)
-flip ........ flip the output vertically
-alpha ....... only save the alpha plane
-incremental . use incremental decoding (useful for tests)
-h ........... this help message
-v ........... verbose (e.g. print encoding/decoding times)
-quiet ....... quiet mode, don't print anything
-noasm ....... disable all assembly optimizations
WebP file analysis tool:
========================
'webpinfo' can be used to print out the chunk level structure and bitstream
header information of WebP files. It can also check if the files are of valid
WebP format.
Usage: webpinfo [options] in_files
Note: there could be multiple input files;
options must come before input files.
Options:
-version ........... Print version number and exit.
-quiet ............. Do not show chunk parsing information.
-diag .............. Show parsing error diagnosis.
-summary ........... Show chunk stats summary.
-bitstream_info .... Parse bitstream header.
Visualization tool:
===================
There's a little self-serve visualization tool called 'vwebp' under the
examples/ directory. It uses OpenGL to open a simple drawing window and show
a decoded WebP file. It's not yet integrated in the automake build system, but
you can try to manually compile it using the recommendations below.
Usage: vwebp in_file [options]
Decodes the WebP image file and visualize it using OpenGL
Options are:
-version ..... print version number and exit
-noicc ....... don't use the icc profile if present
-nofancy ..... don't use the fancy YUV420 upscaler
-nofilter .... disable in-loop filtering
-dither <int> dithering strength (0..100), default=50
-noalphadither disable alpha plane dithering
-mt .......... use multi-threading
-info ........ print info
-h ........... this help message
Keyboard shortcuts:
'c' ................ toggle use of color profile
'i' ................ overlay file information
'd' ................ disable blending & disposal (debug)
'q' / 'Q' / ESC .... quit
Building:
---------
Prerequisites:
1) OpenGL & OpenGL Utility Toolkit (GLUT)
Linux:
$ sudo apt-get install freeglut3-dev mesa-common-dev
Mac + XCode:
- These libraries should be available in the OpenGL / GLUT frameworks.
Windows:
http://freeglut.sourceforge.net/index.php#download
2) (Optional) qcms (Quick Color Management System)
i. Download qcms from Mozilla / Chromium:
http://hg.mozilla.org/mozilla-central/file/0e7639e3bdfb/gfx/qcms
http://src.chromium.org/viewvc/chrome/trunk/src/third_party/qcms
ii. Build and archive the source files as libqcms.a / qcms.lib
iii. Update makefile.unix / Makefile.vc
a) Define WEBP_HAVE_QCMS
b) Update include / library paths to reference the qcms directory.
Build using makefile.unix / Makefile.vc:
$ make -f makefile.unix examples/vwebp
> nmake /f Makefile.vc CFG=release-static \
../obj/x64/release-static/bin/vwebp.exe
Animation creation tool:
========================
The utility 'img2webp' can turn a sequence of input images (PNG, JPEG, ...)
into an animated WebP file. It offers fine control over duration, encoding
modes, etc.
Usage:
img2webp [file-level options] [image files...] [per-frame options...]
File-level options (only used at the start of compression):
-min_size ............ minimize size
-loop <int> .......... loop count (default: 0, = infinite loop)
-kmax <int> .......... maximum number of frame between key-frames
(0=only keyframes)
-kmin <int> .......... minimum number of frame between key-frames
(0=disable key-frames altogether)
-mixed ............... use mixed lossy/lossless automatic mode
-v ................... verbose mode
-h ................... this help
-version ............. print version number and exit
Per-frame options (only used for subsequent images input):
-d <int> ............. frame duration in ms (default: 100)
-lossless ........... use lossless mode (default)
-lossy ... ........... use lossy mode
-q <float> ........... quality
-m <int> ............. method to use
example: img2webp -loop 2 in0.png -lossy in1.jpg
-d 80 in2.tiff -o out.webp
Animated GIF conversion:
========================
Animated GIF files can be converted to WebP files with animation using the
gif2webp utility available under examples/. The files can then be viewed using
vwebp.
Usage:
gif2webp [options] gif_file -o webp_file
Options:
-h / -help ............. this help
-lossy ................. encode image using lossy compression
-mixed ................. for each frame in the image, pick lossy
or lossless compression heuristically
-q <float> ............. quality factor (0:small..100:big)
-m <int> ............... compression method (0=fast, 6=slowest)
-min_size .............. minimize output size (default:off)
lossless compression by default; can be
combined with -q, -m, -lossy or -mixed
options
-kmin <int> ............ min distance between key frames
-kmax <int> ............ max distance between key frames
-f <int> ............... filter strength (0=off..100)
-metadata <string> ..... comma separated list of metadata to
copy from the input to the output if present
Valid values: all, none, icc, xmp (default)
-loop_compatibility .... use compatibility mode for Chrome
version prior to M62 (inclusive)
-mt .................... use multi-threading if available
-version ............... print version number and exit
-v ..................... verbose
-quiet ................. don't print anything
Building:
---------
With the libgif development files installed, gif2webp can be built using
makefile.unix:
$ make -f makefile.unix examples/gif2webp
or using autoconf:
$ ./configure --enable-everything
$ make
Comparison of animated images:
==============================
Test utility anim_diff under examples/ can be used to compare two animated
images (each can be GIF or WebP).
Usage: anim_diff <image1> <image2> [options]
Options:
-dump_frames <folder> dump decoded frames in PAM format
-min_psnr <float> ... minimum per-frame PSNR
-raw_comparison ..... if this flag is not used, RGB is
premultiplied before comparison
-max_diff <int> ..... maximum allowed difference per channel
between corresponding pixels in subsequent
frames
-h .................. this help
-version ............ print version number and exit
Building:
---------
With the libgif development files and a C++ compiler installed, anim_diff can
be built using makefile.unix:
$ make -f makefile.unix examples/anim_diff
or using autoconf:
$ ./configure --enable-everything
$ make
Encoding API:
=============
The main encoding functions are available in the header src/webp/encode.h
The ready-to-use ones are:
size_t WebPEncodeRGB(const uint8_t* rgb, int width, int height, int stride,
float quality_factor, uint8_t** output);
size_t WebPEncodeBGR(const uint8_t* bgr, int width, int height, int stride,
float quality_factor, uint8_t** output);
size_t WebPEncodeRGBA(const uint8_t* rgba, int width, int height, int stride,
float quality_factor, uint8_t** output);
size_t WebPEncodeBGRA(const uint8_t* bgra, int width, int height, int stride,
float quality_factor, uint8_t** output);
They will convert raw RGB samples to a WebP data. The only control supplied
is the quality factor.
There are some variants for using the lossless format:
size_t WebPEncodeLosslessRGB(const uint8_t* rgb, int width, int height,
int stride, uint8_t** output);
size_t WebPEncodeLosslessBGR(const uint8_t* bgr, int width, int height,
int stride, uint8_t** output);
size_t WebPEncodeLosslessRGBA(const uint8_t* rgba, int width, int height,
int stride, uint8_t** output);
size_t WebPEncodeLosslessBGRA(const uint8_t* bgra, int width, int height,
int stride, uint8_t** output);
Of course in this case, no quality factor is needed since the compression
occurs without loss of the input values, at the expense of larger output sizes.
Advanced encoding API:
----------------------
A more advanced API is based on the WebPConfig and WebPPicture structures.
WebPConfig contains the encoding settings and is not tied to a particular
picture.
WebPPicture contains input data, on which some WebPConfig will be used for
compression.
The encoding flow looks like:
-------------------------------------- BEGIN PSEUDO EXAMPLE
#include <webp/encode.h>
// Setup a config, starting form a preset and tuning some additional
// parameters
WebPConfig config;
if (!WebPConfigPreset(&config, WEBP_PRESET_PHOTO, quality_factor))
return 0; // version error
}
// ... additional tuning
config.sns_strength = 90;
config.filter_sharpness = 6;
config_error = WebPValidateConfig(&config); // not mandatory, but useful
// Setup the input data
WebPPicture pic;
if (!WebPPictureInit(&pic)) {
return 0; // version error
}
pic.width = width;
pic.height = height;
// allocated picture of dimension width x height
if (!WebPPictureAllocate(&pic)) {
return 0; // memory error
}
// at this point, 'pic' has been initialized as a container,
// and can receive the Y/U/V samples.
// Alternatively, one could use ready-made import functions like
// WebPPictureImportRGB(), which will take care of memory allocation.
// In any case, past this point, one will have to call
// WebPPictureFree(&pic) to reclaim memory.
// Set up a byte-output write method. WebPMemoryWriter, for instance.
WebPMemoryWriter wrt;
WebPMemoryWriterInit(&wrt); // initialize 'wrt'
pic.writer = MyFileWriter;
pic.custom_ptr = my_opaque_structure_to_make_MyFileWriter_work;
// Compress!
int ok = WebPEncode(&config, &pic); // ok = 0 => error occurred!
WebPPictureFree(&pic); // must be called independently of the 'ok' result.
// output data should have been handled by the writer at that point.
// -> compressed data is the memory buffer described by wrt.mem / wrt.size
// deallocate the memory used by compressed data
WebPMemoryWriterClear(&wrt);
-------------------------------------- END PSEUDO EXAMPLE
Decoding API:
=============
This is mainly just one function to call:
#include "webp/decode.h"
uint8_t* WebPDecodeRGB(const uint8_t* data, size_t data_size,
int* width, int* height);
Please have a look at the file src/webp/decode.h for the details.
There are variants for decoding in BGR/RGBA/ARGB/BGRA order, along with
decoding to raw Y'CbCr samples. One can also decode the image directly into a
pre-allocated buffer.
To detect a WebP file and gather the picture's dimensions, the function:
int WebPGetInfo(const uint8_t* data, size_t data_size,
int* width, int* height);
is supplied. No decoding is involved when using it.
Incremental decoding API:
=========================
In the case when data is being progressively transmitted, pictures can still
be incrementally decoded using a slightly more complicated API. Decoder state
is stored into an instance of the WebPIDecoder object. This object can be
created with the purpose of decoding either RGB or Y'CbCr samples.
For instance:
WebPDecBuffer buffer;
WebPInitDecBuffer(&buffer);
buffer.colorspace = MODE_BGR;
...
WebPIDecoder* idec = WebPINewDecoder(&buffer);
As data is made progressively available, this incremental-decoder object
can be used to decode the picture further. There are two (mutually exclusive)
ways to pass freshly arrived data:
either by appending the fresh bytes:
WebPIAppend(idec, fresh_data, size_of_fresh_data);
or by just mentioning the new size of the transmitted data:
WebPIUpdate(idec, buffer, size_of_transmitted_buffer);
Note that 'buffer' can be modified between each call to WebPIUpdate, in
particular when the buffer is resized to accommodate larger data.
These functions will return the decoding status: either VP8_STATUS_SUSPENDED if
decoding is not finished yet or VP8_STATUS_OK when decoding is done. Any other
status is an error condition.
The 'idec' object must always be released (even upon an error condition) by
calling: WebPDelete(idec).
To retrieve partially decoded picture samples, one must use the corresponding
method: WebPIDecGetRGB or WebPIDecGetYUVA.
It will return the last displayable pixel row.
Lastly, note that decoding can also be performed into a pre-allocated pixel
buffer. This buffer must be passed when creating a WebPIDecoder, calling
WebPINewRGB() or WebPINewYUVA().
Please have a look at the src/webp/decode.h header for further details.
Advanced Decoding API:
======================
WebP decoding supports an advanced API which provides on-the-fly cropping and
rescaling, something of great usefulness on memory-constrained environments like
mobile phones. Basically, the memory usage will scale with the output's size,
not the input's, when one only needs a quick preview or a zoomed in portion of
an otherwise too-large picture. Some CPU can be saved too, incidentally.
-------------------------------------- BEGIN PSEUDO EXAMPLE
// A) Init a configuration object
WebPDecoderConfig config;
CHECK(WebPInitDecoderConfig(&config));
// B) optional: retrieve the bitstream's features.
CHECK(WebPGetFeatures(data, data_size, &config.input) == VP8_STATUS_OK);
// C) Adjust 'config' options, if needed
config.options.no_fancy_upsampling = 1;
config.options.use_scaling = 1;
config.options.scaled_width = scaledWidth();
config.options.scaled_height = scaledHeight();
// etc.
// D) Specify 'config' output options for specifying output colorspace.
// Optionally the external image decode buffer can also be specified.
config.output.colorspace = MODE_BGRA;
// Optionally, the config.output can be pointed to an external buffer as
// well for decoding the image. This externally supplied memory buffer
// should be big enough to store the decoded picture.
config.output.u.RGBA.rgba = (uint8_t*) memory_buffer;
config.output.u.RGBA.stride = scanline_stride;
config.output.u.RGBA.size = total_size_of_the_memory_buffer;
config.output.is_external_memory = 1;
// E) Decode the WebP image. There are two variants w.r.t decoding image.
// The first one (E.1) decodes the full image and the second one (E.2) is
// used to incrementally decode the image using small input buffers.
// Any one of these steps can be used to decode the WebP image.
// E.1) Decode full image.
CHECK(WebPDecode(data, data_size, &config) == VP8_STATUS_OK);
// E.2) Decode image incrementally.
WebPIDecoder* const idec = WebPIDecode(NULL, NULL, &config);
CHECK(idec != NULL);
while (bytes_remaining > 0) {
VP8StatusCode status = WebPIAppend(idec, input, bytes_read);
if (status == VP8_STATUS_OK || status == VP8_STATUS_SUSPENDED) {
bytes_remaining -= bytes_read;
} else {
break;
}
}
WebPIDelete(idec);
// F) Decoded image is now in config.output (and config.output.u.RGBA).
// It can be saved, displayed or otherwise processed.
// G) Reclaim memory allocated in config's object. It's safe to call
// this function even if the memory is external and wasn't allocated
// by WebPDecode().
WebPFreeDecBuffer(&config.output);
-------------------------------------- END PSEUDO EXAMPLE
Bugs:
=====
Please report all bugs to the issue tracker:
https://bugs.chromium.org/p/webp
Patches welcome! See this page to get started:
http://www.webmproject.org/code/contribute/submitting-patches/
Discuss:
========
Email: webp-discuss@webmproject.org
Web: http://groups.google.com/a/webmproject.org/group/webp-discuss

227
media/libwebp/README.mux Normal file
Просмотреть файл

@ -0,0 +1,227 @@
 __ __ ____ ____ ____ __ __ _ __ __
/ \\/ \/ _ \/ _ \/ _ \/ \ \/ \___/_ / _\
\ / __/ _ \ __/ / / (_/ /__
\__\__/\_____/_____/__/ \__//_/\_____/__/___/v1.0.0
Description:
============
WebPMux: set of two libraries 'Mux' and 'Demux' for creation, extraction and
manipulation of an extended format WebP file, which can have features like
color profile, metadata and animation. Reference command-line tools 'webpmux'
and 'vwebp' as well as the WebP container specification
'doc/webp-container-spec.txt' are also provided in this package.
WebP Mux tool:
==============
The examples/ directory contains a tool (webpmux) for manipulating WebP
files. The webpmux tool can be used to create an extended format WebP file and
also to extract or strip relevant data from such a file.
A list of options is available using the -help command line flag:
> webpmux -help
Usage: webpmux -get GET_OPTIONS INPUT -o OUTPUT
webpmux -set SET_OPTIONS INPUT -o OUTPUT
webpmux -duration DURATION_OPTIONS [-duration ...]
INPUT -o OUTPUT
webpmux -strip STRIP_OPTIONS INPUT -o OUTPUT
webpmux -frame FRAME_OPTIONS [-frame...] [-loop LOOP_COUNT]
[-bgcolor BACKGROUND_COLOR] -o OUTPUT
webpmux -info INPUT
webpmux [-h|-help]
webpmux -version
webpmux argument_file_name
GET_OPTIONS:
Extract relevant data:
icc get ICC profile
exif get EXIF metadata
xmp get XMP metadata
frame n get nth frame
SET_OPTIONS:
Set color profile/metadata:
icc file.icc set ICC profile
exif file.exif set EXIF metadata
xmp file.xmp set XMP metadata
where: 'file.icc' contains the ICC profile to be set,
'file.exif' contains the EXIF metadata to be set
'file.xmp' contains the XMP metadata to be set
DURATION_OPTIONS:
Set duration of selected frames:
duration set duration for each frames
duration,frame set duration of a particular frame
duration,start,end set duration of frames in the
interval [start,end])
where: 'duration' is the duration in milliseconds
'start' is the start frame index
'end' is the inclusive end frame index
The special 'end' value '0' means: last frame.
STRIP_OPTIONS:
Strip color profile/metadata:
icc strip ICC profile
exif strip EXIF metadata
xmp strip XMP metadata
FRAME_OPTIONS(i):
Create animation:
file_i +di+[xi+yi[+mi[bi]]]
where: 'file_i' is the i'th animation frame (WebP format),
'di' is the pause duration before next frame,
'xi','yi' specify the image offset for this frame,
'mi' is the dispose method for this frame (0 or 1),
'bi' is the blending method for this frame (+b or -b)
LOOP_COUNT:
Number of times to repeat the animation.
Valid range is 0 to 65535 [Default: 0 (infinite)].
BACKGROUND_COLOR:
Background color of the canvas.
A,R,G,B
where: 'A', 'R', 'G' and 'B' are integers in the range 0 to 255 specifying
the Alpha, Red, Green and Blue component values respectively
[Default: 255,255,255,255]
INPUT & OUTPUT are in WebP format.
Note: The nature of EXIF, XMP and ICC data is not checked and is assumed to be
valid.
Note: if a single file name is passed as the argument, the arguments will be
tokenized from this file. The file name must not start with the character '-'.
Visualization tool:
===================
The examples/ directory also contains a tool (vwebp) for viewing WebP files.
It decodes the image and visualizes it using OpenGL. See the libwebp README
for details on building and running this program.
Mux API:
========
The Mux API contains methods for adding data to and reading data from WebP
files. This API currently supports XMP/EXIF metadata, ICC profile and animation.
Other features may be added in subsequent releases.
Example#1 (pseudo code): Creating a WebPMux object with image data, color
profile and XMP metadata.
int copy_data = 0;
WebPMux* mux = WebPMuxNew();
// ... (Prepare image data).
WebPMuxSetImage(mux, &image, copy_data);
// ... (Prepare ICC profile data).
WebPMuxSetChunk(mux, "ICCP", &icc_profile, copy_data);
// ... (Prepare XMP metadata).
WebPMuxSetChunk(mux, "XMP ", &xmp, copy_data);
// Get data from mux in WebP RIFF format.
WebPMuxAssemble(mux, &output_data);
WebPMuxDelete(mux);
// ... (Consume output_data; e.g. write output_data.bytes to file).
WebPDataClear(&output_data);
Example#2 (pseudo code): Get image and color profile data from a WebP file.
int copy_data = 0;
// ... (Read data from file).
WebPMux* mux = WebPMuxCreate(&data, copy_data);
WebPMuxGetFrame(mux, 1, &image);
// ... (Consume image; e.g. call WebPDecode() to decode the data).
WebPMuxGetChunk(mux, "ICCP", &icc_profile);
// ... (Consume icc_profile).
WebPMuxDelete(mux);
free(data);
For a detailed Mux API reference, please refer to the header file
(src/webp/mux.h).
Demux API:
==========
The Demux API enables extraction of images and extended format data from
WebP files. This API currently supports reading of XMP/EXIF metadata, ICC
profile and animated images. Other features may be added in subsequent
releases.
Code example: Demuxing WebP data to extract all the frames, ICC profile
and EXIF/XMP metadata.
WebPDemuxer* demux = WebPDemux(&webp_data);
uint32_t width = WebPDemuxGetI(demux, WEBP_FF_CANVAS_WIDTH);
uint32_t height = WebPDemuxGetI(demux, WEBP_FF_CANVAS_HEIGHT);
// ... (Get information about the features present in the WebP file).
uint32_t flags = WebPDemuxGetI(demux, WEBP_FF_FORMAT_FLAGS);
// ... (Iterate over all frames).
WebPIterator iter;
if (WebPDemuxGetFrame(demux, 1, &iter)) {
do {
// ... (Consume 'iter'; e.g. Decode 'iter.fragment' with WebPDecode(),
// ... and get other frame properties like width, height, offsets etc.
// ... see 'struct WebPIterator' below for more info).
} while (WebPDemuxNextFrame(&iter));
WebPDemuxReleaseIterator(&iter);
}
// ... (Extract metadata).
WebPChunkIterator chunk_iter;
if (flags & ICCP_FLAG) WebPDemuxGetChunk(demux, "ICCP", 1, &chunk_iter);
// ... (Consume the ICC profile in 'chunk_iter.chunk').
WebPDemuxReleaseChunkIterator(&chunk_iter);
if (flags & EXIF_FLAG) WebPDemuxGetChunk(demux, "EXIF", 1, &chunk_iter);
// ... (Consume the EXIF metadata in 'chunk_iter.chunk').
WebPDemuxReleaseChunkIterator(&chunk_iter);
if (flags & XMP_FLAG) WebPDemuxGetChunk(demux, "XMP ", 1, &chunk_iter);
// ... (Consume the XMP metadata in 'chunk_iter.chunk').
WebPDemuxReleaseChunkIterator(&chunk_iter);
WebPDemuxDelete(demux);
For a detailed Demux API reference, please refer to the header file
(src/webp/demux.h).
AnimEncoder API:
================
The AnimEncoder API can be used to create animated WebP images.
Code example:
WebPAnimEncoderOptions enc_options;
WebPAnimEncoderOptionsInit(&enc_options);
// ... (Tune 'enc_options' as needed).
WebPAnimEncoder* enc = WebPAnimEncoderNew(width, height, &enc_options);
while(<there are more frames>) {
WebPConfig config;
WebPConfigInit(&config);
// ... (Tune 'config' as needed).
WebPAnimEncoderAdd(enc, frame, duration, &config);
}
WebPAnimEncoderAssemble(enc, webp_data);
WebPAnimEncoderDelete(enc);
// ... (Write the 'webp_data' to a file, or re-mux it further).
For a detailed AnimEncoder API reference, please refer to the header file
(src/webp/mux.h).
Bugs:
=====
Please report all bugs to the issue tracker:
https://bugs.chromium.org/p/webp
Patches welcome! See this page to get started:
http://www.webmproject.org/code/contribute/submitting-patches/
Discuss:
========
Email: webp-discuss@webmproject.org
Web: http://groups.google.com/a/webmproject.org/group/webp-discuss

28
media/libwebp/moz.build Normal file
Просмотреть файл

@ -0,0 +1,28 @@
# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*-
# vim: set filetype=python:
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
with Files('**'):
BUG_COMPONENT = ('Core', 'ImageLib')
EXPORTS.webp += [
'src/webp/decode.h',
'src/webp/demux.h',
'src/webp/mux_types.h',
'src/webp/types.h',
]
DIRS += [
'src/dec',
'src/demux',
'src/dsp',
'src/moz',
'src/utils',
]
FINAL_LIBRARY = 'gkmedias'
# We allow warnings for third-party code that can be updated from upstream.
AllowCompilerWarnings()

232
media/libwebp/src/dec/alpha_dec.c Normal file
Просмотреть файл

@ -0,0 +1,232 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Alpha-plane decompression.
//
// Author: Skal (pascal.massimino@gmail.com)
#include <stdlib.h>
#include "src/dec/alphai_dec.h"
#include "src/dec/vp8i_dec.h"
#include "src/dec/vp8li_dec.h"
#include "src/dsp/dsp.h"
#include "src/utils/quant_levels_dec_utils.h"
#include "src/utils/utils.h"
#include "src/webp/format_constants.h"
//------------------------------------------------------------------------------
// ALPHDecoder object.
// Allocates a new alpha decoder instance.
static ALPHDecoder* ALPHNew(void) {
ALPHDecoder* const dec = (ALPHDecoder*)WebPSafeCalloc(1ULL, sizeof(*dec));
return dec;
}
// Clears and deallocates an alpha decoder instance.
static void ALPHDelete(ALPHDecoder* const dec) {
if (dec != NULL) {
VP8LDelete(dec->vp8l_dec_);
dec->vp8l_dec_ = NULL;
WebPSafeFree(dec);
}
}
//------------------------------------------------------------------------------
// Decoding.
// Initialize alpha decoding by parsing the alpha header and decoding the image
// header for alpha data stored using lossless compression.
// Returns false in case of error in alpha header (data too short, invalid
// compression method or filter, error in lossless header data etc).
static int ALPHInit(ALPHDecoder* const dec, const uint8_t* data,
size_t data_size, const VP8Io* const src_io,
uint8_t* output) {
int ok = 0;
const uint8_t* const alpha_data = data + ALPHA_HEADER_LEN;
const size_t alpha_data_size = data_size - ALPHA_HEADER_LEN;
int rsrv;
VP8Io* const io = &dec->io_;
assert(data != NULL && output != NULL && src_io != NULL);
VP8FiltersInit();
dec->output_ = output;
dec->width_ = src_io->width;
dec->height_ = src_io->height;
assert(dec->width_ > 0 && dec->height_ > 0);
if (data_size <= ALPHA_HEADER_LEN) {
return 0;
}
dec->method_ = (data[0] >> 0) & 0x03;
dec->filter_ = (WEBP_FILTER_TYPE)((data[0] >> 2) & 0x03);
dec->pre_processing_ = (data[0] >> 4) & 0x03;
rsrv = (data[0] >> 6) & 0x03;
if (dec->method_ < ALPHA_NO_COMPRESSION ||
dec->method_ > ALPHA_LOSSLESS_COMPRESSION ||
dec->filter_ >= WEBP_FILTER_LAST ||
dec->pre_processing_ > ALPHA_PREPROCESSED_LEVELS ||
rsrv != 0) {
return 0;
}
// Copy the necessary parameters from src_io to io
VP8InitIo(io);
WebPInitCustomIo(NULL, io);
io->opaque = dec;
io->width = src_io->width;
io->height = src_io->height;
io->use_cropping = src_io->use_cropping;
io->crop_left = src_io->crop_left;
io->crop_right = src_io->crop_right;
io->crop_top = src_io->crop_top;
io->crop_bottom = src_io->crop_bottom;
// No need to copy the scaling parameters.
if (dec->method_ == ALPHA_NO_COMPRESSION) {
const size_t alpha_decoded_size = dec->width_ * dec->height_;
ok = (alpha_data_size >= alpha_decoded_size);
} else {
assert(dec->method_ == ALPHA_LOSSLESS_COMPRESSION);
ok = VP8LDecodeAlphaHeader(dec, alpha_data, alpha_data_size);
}
return ok;
}
// Decodes, unfilters and dequantizes *at least* 'num_rows' rows of alpha
// starting from row number 'row'. It assumes that rows up to (row - 1) have
// already been decoded.
// Returns false in case of bitstream error.
static int ALPHDecode(VP8Decoder* const dec, int row, int num_rows) {
ALPHDecoder* const alph_dec = dec->alph_dec_;
const int width = alph_dec->width_;
const int height = alph_dec->io_.crop_bottom;
if (alph_dec->method_ == ALPHA_NO_COMPRESSION) {
int y;
const uint8_t* prev_line = dec->alpha_prev_line_;
const uint8_t* deltas = dec->alpha_data_ + ALPHA_HEADER_LEN + row * width;
uint8_t* dst = dec->alpha_plane_ + row * width;
assert(deltas <= &dec->alpha_data_[dec->alpha_data_size_]);
if (alph_dec->filter_ != WEBP_FILTER_NONE) {
assert(WebPUnfilters[alph_dec->filter_] != NULL);
for (y = 0; y < num_rows; ++y) {
WebPUnfilters[alph_dec->filter_](prev_line, deltas, dst, width);
prev_line = dst;
dst += width;
deltas += width;
}
} else {
for (y = 0; y < num_rows; ++y) {
memcpy(dst, deltas, width * sizeof(*dst));
prev_line = dst;
dst += width;
deltas += width;
}
}
dec->alpha_prev_line_ = prev_line;
} else { // alph_dec->method_ == ALPHA_LOSSLESS_COMPRESSION
assert(alph_dec->vp8l_dec_ != NULL);
if (!VP8LDecodeAlphaImageStream(alph_dec, row + num_rows)) {
return 0;
}
}
if (row + num_rows >= height) {
dec->is_alpha_decoded_ = 1;
}
return 1;
}
static int AllocateAlphaPlane(VP8Decoder* const dec, const VP8Io* const io) {
const int stride = io->width;
const int height = io->crop_bottom;
const uint64_t alpha_size = (uint64_t)stride * height;
assert(dec->alpha_plane_mem_ == NULL);
dec->alpha_plane_mem_ =
(uint8_t*)WebPSafeMalloc(alpha_size, sizeof(*dec->alpha_plane_));
if (dec->alpha_plane_mem_ == NULL) {
return 0;
}
dec->alpha_plane_ = dec->alpha_plane_mem_;
dec->alpha_prev_line_ = NULL;
return 1;
}
void WebPDeallocateAlphaMemory(VP8Decoder* const dec) {
assert(dec != NULL);
WebPSafeFree(dec->alpha_plane_mem_);
dec->alpha_plane_mem_ = NULL;
dec->alpha_plane_ = NULL;
ALPHDelete(dec->alph_dec_);
dec->alph_dec_ = NULL;
}
//------------------------------------------------------------------------------
// Main entry point.
const uint8_t* VP8DecompressAlphaRows(VP8Decoder* const dec,
const VP8Io* const io,
int row, int num_rows) {
const int width = io->width;
const int height = io->crop_bottom;
assert(dec != NULL && io != NULL);
if (row < 0 || num_rows <= 0 || row + num_rows > height) {
return NULL; // sanity check.
}
if (!dec->is_alpha_decoded_) {
if (dec->alph_dec_ == NULL) { // Initialize decoder.
dec->alph_dec_ = ALPHNew();
if (dec->alph_dec_ == NULL) return NULL;
if (!AllocateAlphaPlane(dec, io)) goto Error;
if (!ALPHInit(dec->alph_dec_, dec->alpha_data_, dec->alpha_data_size_,
io, dec->alpha_plane_)) {
goto Error;
}
// if we allowed use of alpha dithering, check whether it's needed at all
if (dec->alph_dec_->pre_processing_ != ALPHA_PREPROCESSED_LEVELS) {
dec->alpha_dithering_ = 0; // disable dithering
} else {
num_rows = height - row; // decode everything in one pass
}
}
assert(dec->alph_dec_ != NULL);
assert(row + num_rows <= height);
if (!ALPHDecode(dec, row, num_rows)) goto Error;
if (dec->is_alpha_decoded_) { // finished?
ALPHDelete(dec->alph_dec_);
dec->alph_dec_ = NULL;
if (dec->alpha_dithering_ > 0) {
uint8_t* const alpha = dec->alpha_plane_ + io->crop_top * width
+ io->crop_left;
if (!WebPDequantizeLevels(alpha,
io->crop_right - io->crop_left,
io->crop_bottom - io->crop_top,
width, dec->alpha_dithering_)) {
goto Error;
}
}
}
}
// Return a pointer to the current decoded row.
return dec->alpha_plane_ + row * width;
Error:
WebPDeallocateAlphaMemory(dec);
return NULL;
}

54
media/libwebp/src/dec/alphai_dec.h Normal file
Просмотреть файл

@ -0,0 +1,54 @@
// Copyright 2013 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Alpha decoder: internal header.
//
// Author: Urvang (urvang@google.com)
#ifndef WEBP_DEC_ALPHAI_DEC_H_
#define WEBP_DEC_ALPHAI_DEC_H_
#include "src/dec/webpi_dec.h"
#include "src/utils/filters_utils.h"
#ifdef __cplusplus
extern "C" {
#endif
struct VP8LDecoder; // Defined in dec/vp8li.h.
typedef struct ALPHDecoder ALPHDecoder;
struct ALPHDecoder {
int width_;
int height_;
int method_;
WEBP_FILTER_TYPE filter_;
int pre_processing_;
struct VP8LDecoder* vp8l_dec_;
VP8Io io_;
int use_8b_decode_; // Although alpha channel requires only 1 byte per
// pixel, sometimes VP8LDecoder may need to allocate
// 4 bytes per pixel internally during decode.
uint8_t* output_;
const uint8_t* prev_line_; // last output row (or NULL)
};
//------------------------------------------------------------------------------
// internal functions. Not public.
// Deallocate memory associated to dec->alpha_plane_ decoding
void WebPDeallocateAlphaMemory(VP8Decoder* const dec);
//------------------------------------------------------------------------------
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* WEBP_DEC_ALPHAI_DEC_H_ */

311
media/libwebp/src/dec/buffer_dec.c Normal file
Просмотреть файл

@ -0,0 +1,311 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Everything about WebPDecBuffer
//
// Author: Skal (pascal.massimino@gmail.com)
#include <stdlib.h>
#include "src/dec/vp8i_dec.h"
#include "src/dec/webpi_dec.h"
#include "src/utils/utils.h"
//------------------------------------------------------------------------------
// WebPDecBuffer
// Number of bytes per pixel for the different color-spaces.
static const uint8_t kModeBpp[MODE_LAST] = {
3, 4, 3, 4, 4, 2, 2,
4, 4, 4, 2, // pre-multiplied modes
1, 1 };
// Check that webp_csp_mode is within the bounds of WEBP_CSP_MODE.
// Convert to an integer to handle both the unsigned/signed enum cases
// without the need for casting to remove type limit warnings.
static int IsValidColorspace(int webp_csp_mode) {
return (webp_csp_mode >= MODE_RGB && webp_csp_mode < MODE_LAST);
}
// strictly speaking, the very last (or first, if flipped) row
// doesn't require padding.
#define MIN_BUFFER_SIZE(WIDTH, HEIGHT, STRIDE) \
((uint64_t)(STRIDE) * ((HEIGHT) - 1) + (WIDTH))
static VP8StatusCode CheckDecBuffer(const WebPDecBuffer* const buffer) {
int ok = 1;
const WEBP_CSP_MODE mode = buffer->colorspace;
const int width = buffer->width;
const int height = buffer->height;
if (!IsValidColorspace(mode)) {
ok = 0;
} else if (!WebPIsRGBMode(mode)) { // YUV checks
const WebPYUVABuffer* const buf = &buffer->u.YUVA;
const int uv_width = (width + 1) / 2;
const int uv_height = (height + 1) / 2;
const int y_stride = abs(buf->y_stride);
const int u_stride = abs(buf->u_stride);
const int v_stride = abs(buf->v_stride);
const int a_stride = abs(buf->a_stride);
const uint64_t y_size = MIN_BUFFER_SIZE(width, height, y_stride);
const uint64_t u_size = MIN_BUFFER_SIZE(uv_width, uv_height, u_stride);
const uint64_t v_size = MIN_BUFFER_SIZE(uv_width, uv_height, v_stride);
const uint64_t a_size = MIN_BUFFER_SIZE(width, height, a_stride);
ok &= (y_size <= buf->y_size);
ok &= (u_size <= buf->u_size);
ok &= (v_size <= buf->v_size);
ok &= (y_stride >= width);
ok &= (u_stride >= uv_width);
ok &= (v_stride >= uv_width);
ok &= (buf->y != NULL);
ok &= (buf->u != NULL);
ok &= (buf->v != NULL);
if (mode == MODE_YUVA) {
ok &= (a_stride >= width);
ok &= (a_size <= buf->a_size);
ok &= (buf->a != NULL);
}
} else { // RGB checks
const WebPRGBABuffer* const buf = &buffer->u.RGBA;
const int stride = abs(buf->stride);
const uint64_t size = MIN_BUFFER_SIZE(width, height, stride);
ok &= (size <= buf->size);
ok &= (stride >= width * kModeBpp[mode]);
ok &= (buf->rgba != NULL);
}
return ok ? VP8_STATUS_OK : VP8_STATUS_INVALID_PARAM;
}
#undef MIN_BUFFER_SIZE
static VP8StatusCode AllocateBuffer(WebPDecBuffer* const buffer) {
const int w = buffer->width;
const int h = buffer->height;
const WEBP_CSP_MODE mode = buffer->colorspace;
if (w <= 0 || h <= 0 || !IsValidColorspace(mode)) {
return VP8_STATUS_INVALID_PARAM;
}
if (buffer->is_external_memory <= 0 && buffer->private_memory == NULL) {
uint8_t* output;
int uv_stride = 0, a_stride = 0;
uint64_t uv_size = 0, a_size = 0, total_size;
// We need memory and it hasn't been allocated yet.
// => initialize output buffer, now that dimensions are known.
int stride;
uint64_t size;
if ((uint64_t)w * kModeBpp[mode] >= (1ull << 32)) {
return VP8_STATUS_INVALID_PARAM;
}
stride = w * kModeBpp[mode];
size = (uint64_t)stride * h;
if (!WebPIsRGBMode(mode)) {
uv_stride = (w + 1) / 2;
uv_size = (uint64_t)uv_stride * ((h + 1) / 2);
if (mode == MODE_YUVA) {
a_stride = w;
a_size = (uint64_t)a_stride * h;
}
}
total_size = size + 2 * uv_size + a_size;
// Security/sanity checks
output = (uint8_t*)WebPSafeMalloc(total_size, sizeof(*output));
if (output == NULL) {
return VP8_STATUS_OUT_OF_MEMORY;
}
buffer->private_memory = output;
if (!WebPIsRGBMode(mode)) { // YUVA initialization
WebPYUVABuffer* const buf = &buffer->u.YUVA;
buf->y = output;
buf->y_stride = stride;
buf->y_size = (size_t)size;
buf->u = output + size;
buf->u_stride = uv_stride;
buf->u_size = (size_t)uv_size;
buf->v = output + size + uv_size;
buf->v_stride = uv_stride;
buf->v_size = (size_t)uv_size;
if (mode == MODE_YUVA) {
buf->a = output + size + 2 * uv_size;
}
buf->a_size = (size_t)a_size;
buf->a_stride = a_stride;
} else { // RGBA initialization
WebPRGBABuffer* const buf = &buffer->u.RGBA;
buf->rgba = output;
buf->stride = stride;
buf->size = (size_t)size;
}
}
return CheckDecBuffer(buffer);
}
VP8StatusCode WebPFlipBuffer(WebPDecBuffer* const buffer) {
if (buffer == NULL) {
return VP8_STATUS_INVALID_PARAM;
}
if (WebPIsRGBMode(buffer->colorspace)) {
WebPRGBABuffer* const buf = &buffer->u.RGBA;
buf->rgba += (buffer->height - 1) * buf->stride;
buf->stride = -buf->stride;
} else {
WebPYUVABuffer* const buf = &buffer->u.YUVA;
const int H = buffer->height;
buf->y += (H - 1) * buf->y_stride;
buf->y_stride = -buf->y_stride;
buf->u += ((H - 1) >> 1) * buf->u_stride;
buf->u_stride = -buf->u_stride;
buf->v += ((H - 1) >> 1) * buf->v_stride;
buf->v_stride = -buf->v_stride;
if (buf->a != NULL) {
buf->a += (H - 1) * buf->a_stride;
buf->a_stride = -buf->a_stride;
}
}
return VP8_STATUS_OK;
}
VP8StatusCode WebPAllocateDecBuffer(int width, int height,
const WebPDecoderOptions* const options,
WebPDecBuffer* const buffer) {
VP8StatusCode status;
if (buffer == NULL || width <= 0 || height <= 0) {
return VP8_STATUS_INVALID_PARAM;
}
if (options != NULL) { // First, apply options if there is any.
if (options->use_cropping) {
const int cw = options->crop_width;
const int ch = options->crop_height;
const int x = options->crop_left & ~1;
const int y = options->crop_top & ~1;
if (x < 0 || y < 0 || cw <= 0 || ch <= 0 ||
x + cw > width || y + ch > height) {
return VP8_STATUS_INVALID_PARAM; // out of frame boundary.
}
width = cw;
height = ch;
}
if (options->use_scaling) {
#if !defined(WEBP_REDUCE_SIZE)
int scaled_width = options->scaled_width;
int scaled_height = options->scaled_height;
if (!WebPRescalerGetScaledDimensions(
width, height, &scaled_width, &scaled_height)) {
return VP8_STATUS_INVALID_PARAM;
}
width = scaled_width;
height = scaled_height;
#else
return VP8_STATUS_INVALID_PARAM; // rescaling not supported
#endif
}
}
buffer->width = width;
buffer->height = height;
// Then, allocate buffer for real.
status = AllocateBuffer(buffer);
if (status != VP8_STATUS_OK) return status;
// Use the stride trick if vertical flip is needed.
if (options != NULL && options->flip) {
status = WebPFlipBuffer(buffer);
}
return status;
}
//------------------------------------------------------------------------------
// constructors / destructors
int WebPInitDecBufferInternal(WebPDecBuffer* buffer, int version) {
if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_DECODER_ABI_VERSION)) {
return 0; // version mismatch
}
if (buffer == NULL) return 0;
memset(buffer, 0, sizeof(*buffer));
return 1;
}
void WebPFreeDecBuffer(WebPDecBuffer* buffer) {
if (buffer != NULL) {
if (buffer->is_external_memory <= 0) {
WebPSafeFree(buffer->private_memory);
}
buffer->private_memory = NULL;
}
}
void WebPCopyDecBuffer(const WebPDecBuffer* const src,
WebPDecBuffer* const dst) {
if (src != NULL && dst != NULL) {
*dst = *src;
if (src->private_memory != NULL) {
dst->is_external_memory = 1; // dst buffer doesn't own the memory.
dst->private_memory = NULL;
}
}
}
// Copy and transfer ownership from src to dst (beware of parameter order!)
void WebPGrabDecBuffer(WebPDecBuffer* const src, WebPDecBuffer* const dst) {
if (src != NULL && dst != NULL) {
*dst = *src;
if (src->private_memory != NULL) {
src->is_external_memory = 1; // src relinquishes ownership
src->private_memory = NULL;
}
}
}
VP8StatusCode WebPCopyDecBufferPixels(const WebPDecBuffer* const src_buf,
WebPDecBuffer* const dst_buf) {
assert(src_buf != NULL && dst_buf != NULL);
assert(src_buf->colorspace == dst_buf->colorspace);
dst_buf->width = src_buf->width;
dst_buf->height = src_buf->height;
if (CheckDecBuffer(dst_buf) != VP8_STATUS_OK) {
return VP8_STATUS_INVALID_PARAM;
}
if (WebPIsRGBMode(src_buf->colorspace)) {
const WebPRGBABuffer* const src = &src_buf->u.RGBA;
const WebPRGBABuffer* const dst = &dst_buf->u.RGBA;
WebPCopyPlane(src->rgba, src->stride, dst->rgba, dst->stride,
src_buf->width * kModeBpp[src_buf->colorspace],
src_buf->height);
} else {
const WebPYUVABuffer* const src = &src_buf->u.YUVA;
const WebPYUVABuffer* const dst = &dst_buf->u.YUVA;
WebPCopyPlane(src->y, src->y_stride, dst->y, dst->y_stride,
src_buf->width, src_buf->height);
WebPCopyPlane(src->u, src->u_stride, dst->u, dst->u_stride,
(src_buf->width + 1) / 2, (src_buf->height + 1) / 2);
WebPCopyPlane(src->v, src->v_stride, dst->v, dst->v_stride,
(src_buf->width + 1) / 2, (src_buf->height + 1) / 2);
if (WebPIsAlphaMode(src_buf->colorspace)) {
WebPCopyPlane(src->a, src->a_stride, dst->a, dst->a_stride,
src_buf->width, src_buf->height);
}
}
return VP8_STATUS_OK;
}
int WebPAvoidSlowMemory(const WebPDecBuffer* const output,
const WebPBitstreamFeatures* const features) {
assert(output != NULL);
return (output->is_external_memory >= 2) &&
WebPIsPremultipliedMode(output->colorspace) &&
(features != NULL && features->has_alpha);
}
//------------------------------------------------------------------------------

54
media/libwebp/src/dec/common_dec.h Normal file
Просмотреть файл

@ -0,0 +1,54 @@
// Copyright 2015 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Definitions and macros common to encoding and decoding
//
// Author: Skal (pascal.massimino@gmail.com)
#ifndef WEBP_DEC_COMMON_DEC_H_
#define WEBP_DEC_COMMON_DEC_H_
// intra prediction modes
enum { B_DC_PRED = 0, // 4x4 modes
B_TM_PRED = 1,
B_VE_PRED = 2,
B_HE_PRED = 3,
B_RD_PRED = 4,
B_VR_PRED = 5,
B_LD_PRED = 6,
B_VL_PRED = 7,
B_HD_PRED = 8,
B_HU_PRED = 9,
NUM_BMODES = B_HU_PRED + 1 - B_DC_PRED, // = 10
// Luma16 or UV modes
DC_PRED = B_DC_PRED, V_PRED = B_VE_PRED,
H_PRED = B_HE_PRED, TM_PRED = B_TM_PRED,
B_PRED = NUM_BMODES, // refined I4x4 mode
NUM_PRED_MODES = 4,
// special modes
B_DC_PRED_NOTOP = 4,
B_DC_PRED_NOLEFT = 5,
B_DC_PRED_NOTOPLEFT = 6,
NUM_B_DC_MODES = 7 };
enum { MB_FEATURE_TREE_PROBS = 3,
NUM_MB_SEGMENTS = 4,
NUM_REF_LF_DELTAS = 4,
NUM_MODE_LF_DELTAS = 4, // I4x4, ZERO, *, SPLIT
MAX_NUM_PARTITIONS = 8,
// Probabilities
NUM_TYPES = 4, // 0: i16-AC, 1: i16-DC, 2:chroma-AC, 3:i4-AC
NUM_BANDS = 8,
NUM_CTX = 3,
NUM_PROBAS = 11
};
#endif // WEBP_DEC_COMMON_DEC_H_

810
media/libwebp/src/dec/frame_dec.c Normal file
Просмотреть файл

@ -0,0 +1,810 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Frame-reconstruction function. Memory allocation.
//
// Author: Skal (pascal.massimino@gmail.com)
#include <stdlib.h>
#include "src/dec/vp8i_dec.h"
#include "src/utils/utils.h"
//------------------------------------------------------------------------------
// Main reconstruction function.
static const uint16_t kScan[16] = {
0 + 0 * BPS, 4 + 0 * BPS, 8 + 0 * BPS, 12 + 0 * BPS,
0 + 4 * BPS, 4 + 4 * BPS, 8 + 4 * BPS, 12 + 4 * BPS,
0 + 8 * BPS, 4 + 8 * BPS, 8 + 8 * BPS, 12 + 8 * BPS,
0 + 12 * BPS, 4 + 12 * BPS, 8 + 12 * BPS, 12 + 12 * BPS
};
static int CheckMode(int mb_x, int mb_y, int mode) {
if (mode == B_DC_PRED) {
if (mb_x == 0) {
return (mb_y == 0) ? B_DC_PRED_NOTOPLEFT : B_DC_PRED_NOLEFT;
} else {
return (mb_y == 0) ? B_DC_PRED_NOTOP : B_DC_PRED;
}
}
return mode;
}
static void Copy32b(uint8_t* const dst, const uint8_t* const src) {
memcpy(dst, src, 4);
}
static WEBP_INLINE void DoTransform(uint32_t bits, const int16_t* const src,
uint8_t* const dst) {
switch (bits >> 30) {
case 3:
VP8Transform(src, dst, 0);
break;
case 2:
VP8TransformAC3(src, dst);
break;
case 1:
VP8TransformDC(src, dst);
break;
default:
break;
}
}
static void DoUVTransform(uint32_t bits, const int16_t* const src,
uint8_t* const dst) {
if (bits & 0xff) { // any non-zero coeff at all?
if (bits & 0xaa) { // any non-zero AC coefficient?
VP8TransformUV(src, dst); // note we don't use the AC3 variant for U/V
} else {
VP8TransformDCUV(src, dst);
}
}
}
static void ReconstructRow(const VP8Decoder* const dec,
const VP8ThreadContext* ctx) {
int j;
int mb_x;
const int mb_y = ctx->mb_y_;
const int cache_id = ctx->id_;
uint8_t* const y_dst = dec->yuv_b_ + Y_OFF;
uint8_t* const u_dst = dec->yuv_b_ + U_OFF;
uint8_t* const v_dst = dec->yuv_b_ + V_OFF;
// Initialize left-most block.
for (j = 0; j < 16; ++j) {
y_dst[j * BPS - 1] = 129;
}
for (j = 0; j < 8; ++j) {
u_dst[j * BPS - 1] = 129;
v_dst[j * BPS - 1] = 129;
}
// Init top-left sample on left column too.
if (mb_y > 0) {
y_dst[-1 - BPS] = u_dst[-1 - BPS] = v_dst[-1 - BPS] = 129;
} else {
// we only need to do this init once at block (0,0).
// Afterward, it remains valid for the whole topmost row.
memset(y_dst - BPS - 1, 127, 16 + 4 + 1);
memset(u_dst - BPS - 1, 127, 8 + 1);
memset(v_dst - BPS - 1, 127, 8 + 1);
}
// Reconstruct one row.
for (mb_x = 0; mb_x < dec->mb_w_; ++mb_x) {
const VP8MBData* const block = ctx->mb_data_ + mb_x;
// Rotate in the left samples from previously decoded block. We move four
// pixels at a time for alignment reason, and because of in-loop filter.
if (mb_x > 0) {
for (j = -1; j < 16; ++j) {
Copy32b(&y_dst[j * BPS - 4], &y_dst[j * BPS + 12]);
}
for (j = -1; j < 8; ++j) {
Copy32b(&u_dst[j * BPS - 4], &u_dst[j * BPS + 4]);
Copy32b(&v_dst[j * BPS - 4], &v_dst[j * BPS + 4]);
}
}
{
// bring top samples into the cache
VP8TopSamples* const top_yuv = dec->yuv_t_ + mb_x;
const int16_t* const coeffs = block->coeffs_;
uint32_t bits = block->non_zero_y_;
int n;
if (mb_y > 0) {
memcpy(y_dst - BPS, top_yuv[0].y, 16);
memcpy(u_dst - BPS, top_yuv[0].u, 8);
memcpy(v_dst - BPS, top_yuv[0].v, 8);
}
// predict and add residuals
if (block->is_i4x4_) { // 4x4
uint32_t* const top_right = (uint32_t*)(y_dst - BPS + 16);
if (mb_y > 0) {
if (mb_x >= dec->mb_w_ - 1) { // on rightmost border
memset(top_right, top_yuv[0].y[15], sizeof(*top_right));
} else {
memcpy(top_right, top_yuv[1].y, sizeof(*top_right));
}
}
// replicate the top-right pixels below
top_right[BPS] = top_right[2 * BPS] = top_right[3 * BPS] = top_right[0];
// predict and add residuals for all 4x4 blocks in turn.
for (n = 0; n < 16; ++n, bits <<= 2) {
uint8_t* const dst = y_dst + kScan[n];
VP8PredLuma4[block->imodes_[n]](dst);
DoTransform(bits, coeffs + n * 16, dst);
}
} else { // 16x16
const int pred_func = CheckMode(mb_x, mb_y, block->imodes_[0]);
VP8PredLuma16[pred_func](y_dst);
if (bits != 0) {
for (n = 0; n < 16; ++n, bits <<= 2) {
DoTransform(bits, coeffs + n * 16, y_dst + kScan[n]);
}
}
}
{
// Chroma
const uint32_t bits_uv = block->non_zero_uv_;
const int pred_func = CheckMode(mb_x, mb_y, block->uvmode_);
VP8PredChroma8[pred_func](u_dst);
VP8PredChroma8[pred_func](v_dst);
DoUVTransform(bits_uv >> 0, coeffs + 16 * 16, u_dst);
DoUVTransform(bits_uv >> 8, coeffs + 20 * 16, v_dst);
}
// stash away top samples for next block
if (mb_y < dec->mb_h_ - 1) {
memcpy(top_yuv[0].y, y_dst + 15 * BPS, 16);
memcpy(top_yuv[0].u, u_dst + 7 * BPS, 8);
memcpy(top_yuv[0].v, v_dst + 7 * BPS, 8);
}
}
// Transfer reconstructed samples from yuv_b_ cache to final destination.
{
const int y_offset = cache_id * 16 * dec->cache_y_stride_;
const int uv_offset = cache_id * 8 * dec->cache_uv_stride_;
uint8_t* const y_out = dec->cache_y_ + mb_x * 16 + y_offset;
uint8_t* const u_out = dec->cache_u_ + mb_x * 8 + uv_offset;
uint8_t* const v_out = dec->cache_v_ + mb_x * 8 + uv_offset;
for (j = 0; j < 16; ++j) {
memcpy(y_out + j * dec->cache_y_stride_, y_dst + j * BPS, 16);
}
for (j = 0; j < 8; ++j) {
memcpy(u_out + j * dec->cache_uv_stride_, u_dst + j * BPS, 8);
memcpy(v_out + j * dec->cache_uv_stride_, v_dst + j * BPS, 8);
}
}
}
}
//------------------------------------------------------------------------------
// Filtering
// kFilterExtraRows[] = How many extra lines are needed on the MB boundary
// for caching, given a filtering level.
// Simple filter: up to 2 luma samples are read and 1 is written.
// Complex filter: up to 4 luma samples are read and 3 are written. Same for
// U/V, so it's 8 samples total (because of the 2x upsampling).
static const uint8_t kFilterExtraRows[3] = { 0, 2, 8 };
static void DoFilter(const VP8Decoder* const dec, int mb_x, int mb_y) {
const VP8ThreadContext* const ctx = &dec->thread_ctx_;
const int cache_id = ctx->id_;
const int y_bps = dec->cache_y_stride_;
const VP8FInfo* const f_info = ctx->f_info_ + mb_x;
uint8_t* const y_dst = dec->cache_y_ + cache_id * 16 * y_bps + mb_x * 16;
const int ilevel = f_info->f_ilevel_;
const int limit = f_info->f_limit_;
if (limit == 0) {
return;
}
assert(limit >= 3);
if (dec->filter_type_ == 1) { // simple
if (mb_x > 0) {
VP8SimpleHFilter16(y_dst, y_bps, limit + 4);
}
if (f_info->f_inner_) {
VP8SimpleHFilter16i(y_dst, y_bps, limit);
}
if (mb_y > 0) {
VP8SimpleVFilter16(y_dst, y_bps, limit + 4);
}
if (f_info->f_inner_) {
VP8SimpleVFilter16i(y_dst, y_bps, limit);
}
} else { // complex
const int uv_bps = dec->cache_uv_stride_;
uint8_t* const u_dst = dec->cache_u_ + cache_id * 8 * uv_bps + mb_x * 8;
uint8_t* const v_dst = dec->cache_v_ + cache_id * 8 * uv_bps + mb_x * 8;
const int hev_thresh = f_info->hev_thresh_;
if (mb_x > 0) {
VP8HFilter16(y_dst, y_bps, limit + 4, ilevel, hev_thresh);
VP8HFilter8(u_dst, v_dst, uv_bps, limit + 4, ilevel, hev_thresh);
}
if (f_info->f_inner_) {
VP8HFilter16i(y_dst, y_bps, limit, ilevel, hev_thresh);
VP8HFilter8i(u_dst, v_dst, uv_bps, limit, ilevel, hev_thresh);
}
if (mb_y > 0) {
VP8VFilter16(y_dst, y_bps, limit + 4, ilevel, hev_thresh);
VP8VFilter8(u_dst, v_dst, uv_bps, limit + 4, ilevel, hev_thresh);
}
if (f_info->f_inner_) {
VP8VFilter16i(y_dst, y_bps, limit, ilevel, hev_thresh);
VP8VFilter8i(u_dst, v_dst, uv_bps, limit, ilevel, hev_thresh);
}
}
}
// Filter the decoded macroblock row (if needed)
static void FilterRow(const VP8Decoder* const dec) {
int mb_x;
const int mb_y = dec->thread_ctx_.mb_y_;
assert(dec->thread_ctx_.filter_row_);
for (mb_x = dec->tl_mb_x_; mb_x < dec->br_mb_x_; ++mb_x) {
DoFilter(dec, mb_x, mb_y);
}
}
//------------------------------------------------------------------------------
// Precompute the filtering strength for each segment and each i4x4/i16x16 mode.
static void PrecomputeFilterStrengths(VP8Decoder* const dec) {
if (dec->filter_type_ > 0) {
int s;
const VP8FilterHeader* const hdr = &dec->filter_hdr_;
for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
int i4x4;
// First, compute the initial level
int base_level;
if (dec->segment_hdr_.use_segment_) {
base_level = dec->segment_hdr_.filter_strength_[s];
if (!dec->segment_hdr_.absolute_delta_) {
base_level += hdr->level_;
}
} else {
base_level = hdr->level_;
}
for (i4x4 = 0; i4x4 <= 1; ++i4x4) {
VP8FInfo* const info = &dec->fstrengths_[s][i4x4];
int level = base_level;
if (hdr->use_lf_delta_) {
level += hdr->ref_lf_delta_[0];
if (i4x4) {
level += hdr->mode_lf_delta_[0];
}
}
level = (level < 0) ? 0 : (level > 63) ? 63 : level;
if (level > 0) {
int ilevel = level;
if (hdr->sharpness_ > 0) {
if (hdr->sharpness_ > 4) {
ilevel >>= 2;
} else {
ilevel >>= 1;
}
if (ilevel > 9 - hdr->sharpness_) {
ilevel = 9 - hdr->sharpness_;
}
}
if (ilevel < 1) ilevel = 1;
info->f_ilevel_ = ilevel;
info->f_limit_ = 2 * level + ilevel;
info->hev_thresh_ = (level >= 40) ? 2 : (level >= 15) ? 1 : 0;
} else {
info->f_limit_ = 0; // no filtering
}
info->f_inner_ = i4x4;
}
}
}
}
//------------------------------------------------------------------------------
// Dithering
// minimal amp that will provide a non-zero dithering effect
#define MIN_DITHER_AMP 4
#define DITHER_AMP_TAB_SIZE 12
static const uint8_t kQuantToDitherAmp[DITHER_AMP_TAB_SIZE] = {
// roughly, it's dqm->uv_mat_[1]
8, 7, 6, 4, 4, 2, 2, 2, 1, 1, 1, 1
};
void VP8InitDithering(const WebPDecoderOptions* const options,
VP8Decoder* const dec) {
assert(dec != NULL);
if (options != NULL) {
const int d = options->dithering_strength;
const int max_amp = (1 << VP8_RANDOM_DITHER_FIX) - 1;
const int f = (d < 0) ? 0 : (d > 100) ? max_amp : (d * max_amp / 100);
if (f > 0) {
int s;
int all_amp = 0;
for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
VP8QuantMatrix* const dqm = &dec->dqm_[s];
if (dqm->uv_quant_ < DITHER_AMP_TAB_SIZE) {
// TODO(skal): should we specially dither more for uv_quant_ < 0?
const int idx = (dqm->uv_quant_ < 0) ? 0 : dqm->uv_quant_;
dqm->dither_ = (f * kQuantToDitherAmp[idx]) >> 3;
}
all_amp |= dqm->dither_;
}
if (all_amp != 0) {
VP8InitRandom(&dec->dithering_rg_, 1.0f);
dec->dither_ = 1;
}
}
// potentially allow alpha dithering
dec->alpha_dithering_ = options->alpha_dithering_strength;
if (dec->alpha_dithering_ > 100) {
dec->alpha_dithering_ = 100;
} else if (dec->alpha_dithering_ < 0) {
dec->alpha_dithering_ = 0;
}
}
}
// Convert to range: [-2,2] for dither=50, [-4,4] for dither=100
static void Dither8x8(VP8Random* const rg, uint8_t* dst, int bps, int amp) {
uint8_t dither[64];
int i;
for (i = 0; i < 8 * 8; ++i) {
dither[i] = VP8RandomBits2(rg, VP8_DITHER_AMP_BITS + 1, amp);
}
VP8DitherCombine8x8(dither, dst, bps);
}
static void DitherRow(VP8Decoder* const dec) {
int mb_x;
assert(dec->dither_);
for (mb_x = dec->tl_mb_x_; mb_x < dec->br_mb_x_; ++mb_x) {
const VP8ThreadContext* const ctx = &dec->thread_ctx_;
const VP8MBData* const data = ctx->mb_data_ + mb_x;
const int cache_id = ctx->id_;
const int uv_bps = dec->cache_uv_stride_;
if (data->dither_ >= MIN_DITHER_AMP) {
uint8_t* const u_dst = dec->cache_u_ + cache_id * 8 * uv_bps + mb_x * 8;
uint8_t* const v_dst = dec->cache_v_ + cache_id * 8 * uv_bps + mb_x * 8;
Dither8x8(&dec->dithering_rg_, u_dst, uv_bps, data->dither_);
Dither8x8(&dec->dithering_rg_, v_dst, uv_bps, data->dither_);
}
}
}
//------------------------------------------------------------------------------
// This function is called after a row of macroblocks is finished decoding.
// It also takes into account the following restrictions:
// * In case of in-loop filtering, we must hold off sending some of the bottom
// pixels as they are yet unfiltered. They will be when the next macroblock
// row is decoded. Meanwhile, we must preserve them by rotating them in the
// cache area. This doesn't hold for the very bottom row of the uncropped
// picture of course.
// * we must clip the remaining pixels against the cropping area. The VP8Io
// struct must have the following fields set correctly before calling put():
#define MACROBLOCK_VPOS(mb_y) ((mb_y) * 16) // vertical position of a MB
// Finalize and transmit a complete row. Return false in case of user-abort.
static int FinishRow(void* arg1, void* arg2) {
VP8Decoder* const dec = (VP8Decoder*)arg1;
VP8Io* const io = (VP8Io*)arg2;
int ok = 1;
const VP8ThreadContext* const ctx = &dec->thread_ctx_;
const int cache_id = ctx->id_;
const int extra_y_rows = kFilterExtraRows[dec->filter_type_];
const int ysize = extra_y_rows * dec->cache_y_stride_;
const int uvsize = (extra_y_rows / 2) * dec->cache_uv_stride_;
const int y_offset = cache_id * 16 * dec->cache_y_stride_;
const int uv_offset = cache_id * 8 * dec->cache_uv_stride_;
uint8_t* const ydst = dec->cache_y_ - ysize + y_offset;
uint8_t* const udst = dec->cache_u_ - uvsize + uv_offset;
uint8_t* const vdst = dec->cache_v_ - uvsize + uv_offset;
const int mb_y = ctx->mb_y_;
const int is_first_row = (mb_y == 0);
const int is_last_row = (mb_y >= dec->br_mb_y_ - 1);
if (dec->mt_method_ == 2) {
ReconstructRow(dec, ctx);
}
if (ctx->filter_row_) {
FilterRow(dec);
}
if (dec->dither_) {
DitherRow(dec);
}
if (io->put != NULL) {
int y_start = MACROBLOCK_VPOS(mb_y);
int y_end = MACROBLOCK_VPOS(mb_y + 1);
if (!is_first_row) {
y_start -= extra_y_rows;
io->y = ydst;
io->u = udst;
io->v = vdst;
} else {
io->y = dec->cache_y_ + y_offset;
io->u = dec->cache_u_ + uv_offset;
io->v = dec->cache_v_ + uv_offset;
}
if (!is_last_row) {
y_end -= extra_y_rows;
}
if (y_end > io->crop_bottom) {
y_end = io->crop_bottom; // make sure we don't overflow on last row.
}
// If dec->alpha_data_ is not NULL, we have some alpha plane present.
io->a = NULL;
if (dec->alpha_data_ != NULL && y_start < y_end) {
io->a = VP8DecompressAlphaRows(dec, io, y_start, y_end - y_start);
if (io->a == NULL) {
return VP8SetError(dec, VP8_STATUS_BITSTREAM_ERROR,
"Could not decode alpha data.");
}
}
if (y_start < io->crop_top) {
const int delta_y = io->crop_top - y_start;
y_start = io->crop_top;
assert(!(delta_y & 1));
io->y += dec->cache_y_stride_ * delta_y;
io->u += dec->cache_uv_stride_ * (delta_y >> 1);
io->v += dec->cache_uv_stride_ * (delta_y >> 1);
if (io->a != NULL) {
io->a += io->width * delta_y;
}
}
if (y_start < y_end) {
io->y += io->crop_left;
io->u += io->crop_left >> 1;
io->v += io->crop_left >> 1;
if (io->a != NULL) {
io->a += io->crop_left;
}
io->mb_y = y_start - io->crop_top;
io->mb_w = io->crop_right - io->crop_left;
io->mb_h = y_end - y_start;
ok = io->put(io);
}
}
// rotate top samples if needed
if (cache_id + 1 == dec->num_caches_) {
if (!is_last_row) {
memcpy(dec->cache_y_ - ysize, ydst + 16 * dec->cache_y_stride_, ysize);
memcpy(dec->cache_u_ - uvsize, udst + 8 * dec->cache_uv_stride_, uvsize);
memcpy(dec->cache_v_ - uvsize, vdst + 8 * dec->cache_uv_stride_, uvsize);
}
}
return ok;
}
#undef MACROBLOCK_VPOS
//------------------------------------------------------------------------------
int VP8ProcessRow(VP8Decoder* const dec, VP8Io* const io) {
int ok = 1;
VP8ThreadContext* const ctx = &dec->thread_ctx_;
const int filter_row =
(dec->filter_type_ > 0) &&
(dec->mb_y_ >= dec->tl_mb_y_) && (dec->mb_y_ <= dec->br_mb_y_);
if (dec->mt_method_ == 0) {
// ctx->id_ and ctx->f_info_ are already set
ctx->mb_y_ = dec->mb_y_;
ctx->filter_row_ = filter_row;
ReconstructRow(dec, ctx);
ok = FinishRow(dec, io);
} else {
WebPWorker* const worker = &dec->worker_;
// Finish previous job *before* updating context
ok &= WebPGetWorkerInterface()->Sync(worker);
assert(worker->status_ == OK);
if (ok) { // spawn a new deblocking/output job
ctx->io_ = *io;
ctx->id_ = dec->cache_id_;
ctx->mb_y_ = dec->mb_y_;
ctx->filter_row_ = filter_row;
if (dec->mt_method_ == 2) { // swap macroblock data
VP8MBData* const tmp = ctx->mb_data_;
ctx->mb_data_ = dec->mb_data_;
dec->mb_data_ = tmp;
} else {
// perform reconstruction directly in main thread
ReconstructRow(dec, ctx);
}
if (filter_row) { // swap filter info
VP8FInfo* const tmp = ctx->f_info_;
ctx->f_info_ = dec->f_info_;
dec->f_info_ = tmp;
}
// (reconstruct)+filter in parallel
WebPGetWorkerInterface()->Launch(worker);
if (++dec->cache_id_ == dec->num_caches_) {
dec->cache_id_ = 0;
}
}
}
return ok;
}
//------------------------------------------------------------------------------
// Finish setting up the decoding parameter once user's setup() is called.
VP8StatusCode VP8EnterCritical(VP8Decoder* const dec, VP8Io* const io) {
// Call setup() first. This may trigger additional decoding features on 'io'.
// Note: Afterward, we must call teardown() no matter what.
if (io->setup != NULL && !io->setup(io)) {
VP8SetError(dec, VP8_STATUS_USER_ABORT, "Frame setup failed");
return dec->status_;
}
// Disable filtering per user request
if (io->bypass_filtering) {
dec->filter_type_ = 0;
}
// Define the area where we can skip in-loop filtering, in case of cropping.
//
// 'Simple' filter reads two luma samples outside of the macroblock
// and filters one. It doesn't filter the chroma samples. Hence, we can
// avoid doing the in-loop filtering before crop_top/crop_left position.
// For the 'Complex' filter, 3 samples are read and up to 3 are filtered.
// Means: there's a dependency chain that goes all the way up to the
// top-left corner of the picture (MB #0). We must filter all the previous
// macroblocks.
{
const int extra_pixels = kFilterExtraRows[dec->filter_type_];
if (dec->filter_type_ == 2) {
// For complex filter, we need to preserve the dependency chain.
dec->tl_mb_x_ = 0;
dec->tl_mb_y_ = 0;
} else {
// For simple filter, we can filter only the cropped region.
// We include 'extra_pixels' on the other side of the boundary, since
// vertical or horizontal filtering of the previous macroblock can
// modify some abutting pixels.
dec->tl_mb_x_ = (io->crop_left - extra_pixels) >> 4;
dec->tl_mb_y_ = (io->crop_top - extra_pixels) >> 4;
if (dec->tl_mb_x_ < 0) dec->tl_mb_x_ = 0;
if (dec->tl_mb_y_ < 0) dec->tl_mb_y_ = 0;
}
// We need some 'extra' pixels on the right/bottom.
dec->br_mb_y_ = (io->crop_bottom + 15 + extra_pixels) >> 4;
dec->br_mb_x_ = (io->crop_right + 15 + extra_pixels) >> 4;
if (dec->br_mb_x_ > dec->mb_w_) {
dec->br_mb_x_ = dec->mb_w_;
}
if (dec->br_mb_y_ > dec->mb_h_) {
dec->br_mb_y_ = dec->mb_h_;
}
}
PrecomputeFilterStrengths(dec);
return VP8_STATUS_OK;
}
int VP8ExitCritical(VP8Decoder* const dec, VP8Io* const io) {
int ok = 1;
if (dec->mt_method_ > 0) {
ok = WebPGetWorkerInterface()->Sync(&dec->worker_);
}
if (io->teardown != NULL) {
io->teardown(io);
}
return ok;
}
//------------------------------------------------------------------------------
// For multi-threaded decoding we need to use 3 rows of 16 pixels as delay line.
//
// Reason is: the deblocking filter cannot deblock the bottom horizontal edges
// immediately, and needs to wait for first few rows of the next macroblock to
// be decoded. Hence, deblocking is lagging behind by 4 or 8 pixels (depending
// on strength).
// With two threads, the vertical positions of the rows being decoded are:
// Decode: [ 0..15][16..31][32..47][48..63][64..79][...
// Deblock: [ 0..11][12..27][28..43][44..59][...
// If we use two threads and two caches of 16 pixels, the sequence would be:
// Decode: [ 0..15][16..31][ 0..15!!][16..31][ 0..15][...
// Deblock: [ 0..11][12..27!!][-4..11][12..27][...
// The problem occurs during row [12..15!!] that both the decoding and
// deblocking threads are writing simultaneously.
// With 3 cache lines, one get a safe write pattern:
// Decode: [ 0..15][16..31][32..47][ 0..15][16..31][32..47][0..
// Deblock: [ 0..11][12..27][28..43][-4..11][12..27][28...
// Note that multi-threaded output _without_ deblocking can make use of two
// cache lines of 16 pixels only, since there's no lagging behind. The decoding
// and output process have non-concurrent writing:
// Decode: [ 0..15][16..31][ 0..15][16..31][...
// io->put: [ 0..15][16..31][ 0..15][...
#define MT_CACHE_LINES 3
#define ST_CACHE_LINES 1 // 1 cache row only for single-threaded case
// Initialize multi/single-thread worker
static int InitThreadContext(VP8Decoder* const dec) {
dec->cache_id_ = 0;
if (dec->mt_method_ > 0) {
WebPWorker* const worker = &dec->worker_;
if (!WebPGetWorkerInterface()->Reset(worker)) {
return VP8SetError(dec, VP8_STATUS_OUT_OF_MEMORY,
"thread initialization failed.");
}
worker->data1 = dec;
worker->data2 = (void*)&dec->thread_ctx_.io_;
worker->hook = FinishRow;
dec->num_caches_ =
(dec->filter_type_ > 0) ? MT_CACHE_LINES : MT_CACHE_LINES - 1;
} else {
dec->num_caches_ = ST_CACHE_LINES;
}
return 1;
}
int VP8GetThreadMethod(const WebPDecoderOptions* const options,
const WebPHeaderStructure* const headers,
int width, int height) {
if (options == NULL || options->use_threads == 0) {
return 0;
}
(void)headers;
(void)width;
(void)height;
assert(headers == NULL || !headers->is_lossless);
#if defined(WEBP_USE_THREAD)
if (width < MIN_WIDTH_FOR_THREADS) return 0;
// TODO(skal): tune the heuristic further
#if 0
if (height < 2 * width) return 2;
#endif
return 2;
#else // !WEBP_USE_THREAD
return 0;
#endif
}
#undef MT_CACHE_LINES
#undef ST_CACHE_LINES
//------------------------------------------------------------------------------
// Memory setup
static int AllocateMemory(VP8Decoder* const dec) {
const int num_caches = dec->num_caches_;
const int mb_w = dec->mb_w_;
// Note: we use 'size_t' when there's no overflow risk, uint64_t otherwise.
const size_t intra_pred_mode_size = 4 * mb_w * sizeof(uint8_t);
const size_t top_size = sizeof(VP8TopSamples) * mb_w;
const size_t mb_info_size = (mb_w + 1) * sizeof(VP8MB);
const size_t f_info_size =
(dec->filter_type_ > 0) ?
mb_w * (dec->mt_method_ > 0 ? 2 : 1) * sizeof(VP8FInfo)
: 0;
const size_t yuv_size = YUV_SIZE * sizeof(*dec->yuv_b_);
const size_t mb_data_size =
(dec->mt_method_ == 2 ? 2 : 1) * mb_w * sizeof(*dec->mb_data_);
const size_t cache_height = (16 * num_caches
+ kFilterExtraRows[dec->filter_type_]) * 3 / 2;
const size_t cache_size = top_size * cache_height;
// alpha_size is the only one that scales as width x height.
const uint64_t alpha_size = (dec->alpha_data_ != NULL) ?
(uint64_t)dec->pic_hdr_.width_ * dec->pic_hdr_.height_ : 0ULL;
const uint64_t needed = (uint64_t)intra_pred_mode_size
+ top_size + mb_info_size + f_info_size
+ yuv_size + mb_data_size
+ cache_size + alpha_size + WEBP_ALIGN_CST;
uint8_t* mem;
if (needed != (size_t)needed) return 0; // check for overflow
if (needed > dec->mem_size_) {
WebPSafeFree(dec->mem_);
dec->mem_size_ = 0;
dec->mem_ = WebPSafeMalloc(needed, sizeof(uint8_t));
if (dec->mem_ == NULL) {
return VP8SetError(dec, VP8_STATUS_OUT_OF_MEMORY,
"no memory during frame initialization.");
}
// down-cast is ok, thanks to WebPSafeMalloc() above.
dec->mem_size_ = (size_t)needed;
}
mem = (uint8_t*)dec->mem_;
dec->intra_t_ = mem;
mem += intra_pred_mode_size;
dec->yuv_t_ = (VP8TopSamples*)mem;
mem += top_size;
dec->mb_info_ = ((VP8MB*)mem) + 1;
mem += mb_info_size;
dec->f_info_ = f_info_size ? (VP8FInfo*)mem : NULL;
mem += f_info_size;
dec->thread_ctx_.id_ = 0;
dec->thread_ctx_.f_info_ = dec->f_info_;
if (dec->mt_method_ > 0) {
// secondary cache line. The deblocking process need to make use of the
// filtering strength from previous macroblock row, while the new ones
// are being decoded in parallel. We'll just swap the pointers.
dec->thread_ctx_.f_info_ += mb_w;
}
mem = (uint8_t*)WEBP_ALIGN(mem);
assert((yuv_size & WEBP_ALIGN_CST) == 0);
dec->yuv_b_ = mem;
mem += yuv_size;
dec->mb_data_ = (VP8MBData*)mem;
dec->thread_ctx_.mb_data_ = (VP8MBData*)mem;
if (dec->mt_method_ == 2) {
dec->thread_ctx_.mb_data_ += mb_w;
}
mem += mb_data_size;
dec->cache_y_stride_ = 16 * mb_w;
dec->cache_uv_stride_ = 8 * mb_w;
{
const int extra_rows = kFilterExtraRows[dec->filter_type_];
const int extra_y = extra_rows * dec->cache_y_stride_;
const int extra_uv = (extra_rows / 2) * dec->cache_uv_stride_;
dec->cache_y_ = mem + extra_y;
dec->cache_u_ = dec->cache_y_
+ 16 * num_caches * dec->cache_y_stride_ + extra_uv;
dec->cache_v_ = dec->cache_u_
+ 8 * num_caches * dec->cache_uv_stride_ + extra_uv;
dec->cache_id_ = 0;
}
mem += cache_size;
// alpha plane
dec->alpha_plane_ = alpha_size ? mem : NULL;
mem += alpha_size;
assert(mem <= (uint8_t*)dec->mem_ + dec->mem_size_);
// note: left/top-info is initialized once for all.
memset(dec->mb_info_ - 1, 0, mb_info_size);
VP8InitScanline(dec); // initialize left too.
// initialize top
memset(dec->intra_t_, B_DC_PRED, intra_pred_mode_size);
return 1;
}
static void InitIo(VP8Decoder* const dec, VP8Io* io) {
// prepare 'io'
io->mb_y = 0;
io->y = dec->cache_y_;
io->u = dec->cache_u_;
io->v = dec->cache_v_;
io->y_stride = dec->cache_y_stride_;
io->uv_stride = dec->cache_uv_stride_;
io->a = NULL;
}
int VP8InitFrame(VP8Decoder* const dec, VP8Io* const io) {
if (!InitThreadContext(dec)) return 0; // call first. Sets dec->num_caches_.
if (!AllocateMemory(dec)) return 0;
InitIo(dec, io);
VP8DspInit(); // Init critical function pointers and look-up tables.
return 1;
}
//------------------------------------------------------------------------------

892
media/libwebp/src/dec/idec_dec.c Normal file
Просмотреть файл

@ -0,0 +1,892 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Incremental decoding
//
// Author: somnath@google.com (Somnath Banerjee)
#include <assert.h>
#include <string.h>
#include <stdlib.h>
#include "src/dec/alphai_dec.h"
#include "src/dec/webpi_dec.h"
#include "src/dec/vp8i_dec.h"
#include "src/utils/utils.h"
// In append mode, buffer allocations increase as multiples of this value.
// Needs to be a power of 2.
#define CHUNK_SIZE 4096
#define MAX_MB_SIZE 4096
//------------------------------------------------------------------------------
// Data structures for memory and states
// Decoding states. State normally flows as:
// WEBP_HEADER->VP8_HEADER->VP8_PARTS0->VP8_DATA->DONE for a lossy image, and
// WEBP_HEADER->VP8L_HEADER->VP8L_DATA->DONE for a lossless image.
// If there is any error the decoder goes into state ERROR.
typedef enum {
STATE_WEBP_HEADER, // All the data before that of the VP8/VP8L chunk.
STATE_VP8_HEADER, // The VP8 Frame header (within the VP8 chunk).
STATE_VP8_PARTS0,
STATE_VP8_DATA,
STATE_VP8L_HEADER,
STATE_VP8L_DATA,
STATE_DONE,
STATE_ERROR
} DecState;
// Operating state for the MemBuffer
typedef enum {
MEM_MODE_NONE = 0,
MEM_MODE_APPEND,
MEM_MODE_MAP
} MemBufferMode;
// storage for partition #0 and partial data (in a rolling fashion)
typedef struct {
MemBufferMode mode_; // Operation mode
size_t start_; // start location of the data to be decoded
size_t end_; // end location
size_t buf_size_; // size of the allocated buffer
uint8_t* buf_; // We don't own this buffer in case WebPIUpdate()
size_t part0_size_; // size of partition #0
const uint8_t* part0_buf_; // buffer to store partition #0
} MemBuffer;
struct WebPIDecoder {
DecState state_; // current decoding state
WebPDecParams params_; // Params to store output info
int is_lossless_; // for down-casting 'dec_'.
void* dec_; // either a VP8Decoder or a VP8LDecoder instance
VP8Io io_;
MemBuffer mem_; // input memory buffer.
WebPDecBuffer output_; // output buffer (when no external one is supplied,
// or if the external one has slow-memory)
WebPDecBuffer* final_output_; // Slow-memory output to copy to eventually.
size_t chunk_size_; // Compressed VP8/VP8L size extracted from Header.
int last_mb_y_; // last row reached for intra-mode decoding
};
// MB context to restore in case VP8DecodeMB() fails
typedef struct {
VP8MB left_;
VP8MB info_;
VP8BitReader token_br_;
} MBContext;
//------------------------------------------------------------------------------
// MemBuffer: incoming data handling
static WEBP_INLINE size_t MemDataSize(const MemBuffer* mem) {
return (mem->end_ - mem->start_);
}
// Check if we need to preserve the compressed alpha data, as it may not have
// been decoded yet.
static int NeedCompressedAlpha(const WebPIDecoder* const idec) {
if (idec->state_ == STATE_WEBP_HEADER) {
// We haven't parsed the headers yet, so we don't know whether the image is
// lossy or lossless. This also means that we haven't parsed the ALPH chunk.
return 0;
}
if (idec->is_lossless_) {
return 0; // ALPH chunk is not present for lossless images.
} else {
const VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
assert(dec != NULL); // Must be true as idec->state_ != STATE_WEBP_HEADER.
return (dec->alpha_data_ != NULL) && !dec->is_alpha_decoded_;
}
}
static void DoRemap(WebPIDecoder* const idec, ptrdiff_t offset) {
MemBuffer* const mem = &idec->mem_;
const uint8_t* const new_base = mem->buf_ + mem->start_;
// note: for VP8, setting up idec->io_ is only really needed at the beginning
// of the decoding, till partition #0 is complete.
idec->io_.data = new_base;
idec->io_.data_size = MemDataSize(mem);
if (idec->dec_ != NULL) {
if (!idec->is_lossless_) {
VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
const uint32_t last_part = dec->num_parts_minus_one_;
if (offset != 0) {
uint32_t p;
for (p = 0; p <= last_part; ++p) {
VP8RemapBitReader(dec->parts_ + p, offset);
}
// Remap partition #0 data pointer to new offset, but only in MAP
// mode (in APPEND mode, partition #0 is copied into a fixed memory).
if (mem->mode_ == MEM_MODE_MAP) {
VP8RemapBitReader(&dec->br_, offset);
}
}
{
const uint8_t* const last_start = dec->parts_[last_part].buf_;
VP8BitReaderSetBuffer(&dec->parts_[last_part], last_start,
mem->buf_ + mem->end_ - last_start);
}
if (NeedCompressedAlpha(idec)) {
ALPHDecoder* const alph_dec = dec->alph_dec_;
dec->alpha_data_ += offset;
if (alph_dec != NULL) {
if (alph_dec->method_ == ALPHA_LOSSLESS_COMPRESSION) {
VP8LDecoder* const alph_vp8l_dec = alph_dec->vp8l_dec_;
assert(alph_vp8l_dec != NULL);
assert(dec->alpha_data_size_ >= ALPHA_HEADER_LEN);
VP8LBitReaderSetBuffer(&alph_vp8l_dec->br_,
dec->alpha_data_ + ALPHA_HEADER_LEN,
dec->alpha_data_size_ - ALPHA_HEADER_LEN);
} else { // alph_dec->method_ == ALPHA_NO_COMPRESSION
// Nothing special to do in this case.
}
}
}
} else { // Resize lossless bitreader
VP8LDecoder* const dec = (VP8LDecoder*)idec->dec_;
VP8LBitReaderSetBuffer(&dec->br_, new_base, MemDataSize(mem));
}
}
}
// Appends data to the end of MemBuffer->buf_. It expands the allocated memory
// size if required and also updates VP8BitReader's if new memory is allocated.
static int AppendToMemBuffer(WebPIDecoder* const idec,
const uint8_t* const data, size_t data_size) {
VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
MemBuffer* const mem = &idec->mem_;
const int need_compressed_alpha = NeedCompressedAlpha(idec);
const uint8_t* const old_start = mem->buf_ + mem->start_;
const uint8_t* const old_base =
need_compressed_alpha ? dec->alpha_data_ : old_start;
assert(mem->mode_ == MEM_MODE_APPEND);
if (data_size > MAX_CHUNK_PAYLOAD) {
// security safeguard: trying to allocate more than what the format
// allows for a chunk should be considered a smoke smell.
return 0;
}
if (mem->end_ + data_size > mem->buf_size_) { // Need some free memory
const size_t new_mem_start = old_start - old_base;
const size_t current_size = MemDataSize(mem) + new_mem_start;
const uint64_t new_size = (uint64_t)current_size + data_size;
const uint64_t extra_size = (new_size + CHUNK_SIZE - 1) & ~(CHUNK_SIZE - 1);
uint8_t* const new_buf =
(uint8_t*)WebPSafeMalloc(extra_size, sizeof(*new_buf));
if (new_buf == NULL) return 0;
memcpy(new_buf, old_base, current_size);
WebPSafeFree(mem->buf_);
mem->buf_ = new_buf;
mem->buf_size_ = (size_t)extra_size;
mem->start_ = new_mem_start;
mem->end_ = current_size;
}
memcpy(mem->buf_ + mem->end_, data, data_size);
mem->end_ += data_size;
assert(mem->end_ <= mem->buf_size_);
DoRemap(idec, mem->buf_ + mem->start_ - old_start);
return 1;
}
static int RemapMemBuffer(WebPIDecoder* const idec,
const uint8_t* const data, size_t data_size) {
MemBuffer* const mem = &idec->mem_;
const uint8_t* const old_buf = mem->buf_;
const uint8_t* const old_start = old_buf + mem->start_;
assert(mem->mode_ == MEM_MODE_MAP);
if (data_size < mem->buf_size_) return 0; // can't remap to a shorter buffer!
mem->buf_ = (uint8_t*)data;
mem->end_ = mem->buf_size_ = data_size;
DoRemap(idec, mem->buf_ + mem->start_ - old_start);
return 1;
}
static void InitMemBuffer(MemBuffer* const mem) {
mem->mode_ = MEM_MODE_NONE;
mem->buf_ = NULL;
mem->buf_size_ = 0;
mem->part0_buf_ = NULL;
mem->part0_size_ = 0;
}
static void ClearMemBuffer(MemBuffer* const mem) {
assert(mem);
if (mem->mode_ == MEM_MODE_APPEND) {
WebPSafeFree(mem->buf_);
WebPSafeFree((void*)mem->part0_buf_);
}
}
static int CheckMemBufferMode(MemBuffer* const mem, MemBufferMode expected) {
if (mem->mode_ == MEM_MODE_NONE) {
mem->mode_ = expected; // switch to the expected mode
} else if (mem->mode_ != expected) {
return 0; // we mixed the modes => error
}
assert(mem->mode_ == expected); // mode is ok
return 1;
}
// To be called last.
static VP8StatusCode FinishDecoding(WebPIDecoder* const idec) {
const WebPDecoderOptions* const options = idec->params_.options;
WebPDecBuffer* const output = idec->params_.output;
idec->state_ = STATE_DONE;
if (options != NULL && options->flip) {
const VP8StatusCode status = WebPFlipBuffer(output);
if (status != VP8_STATUS_OK) return status;
}
if (idec->final_output_ != NULL) {
WebPCopyDecBufferPixels(output, idec->final_output_); // do the slow-copy
WebPFreeDecBuffer(&idec->output_);
*output = *idec->final_output_;
idec->final_output_ = NULL;
}
return VP8_STATUS_OK;
}
//------------------------------------------------------------------------------
// Macroblock-decoding contexts
static void SaveContext(const VP8Decoder* dec, const VP8BitReader* token_br,
MBContext* const context) {
context->left_ = dec->mb_info_[-1];
context->info_ = dec->mb_info_[dec->mb_x_];
context->token_br_ = *token_br;
}
static void RestoreContext(const MBContext* context, VP8Decoder* const dec,
VP8BitReader* const token_br) {
dec->mb_info_[-1] = context->left_;
dec->mb_info_[dec->mb_x_] = context->info_;
*token_br = context->token_br_;
}
//------------------------------------------------------------------------------
static VP8StatusCode IDecError(WebPIDecoder* const idec, VP8StatusCode error) {
if (idec->state_ == STATE_VP8_DATA) {
VP8Io* const io = &idec->io_;
if (io->teardown != NULL) {
io->teardown(io);
}
}
idec->state_ = STATE_ERROR;
return error;
}
static void ChangeState(WebPIDecoder* const idec, DecState new_state,
size_t consumed_bytes) {
MemBuffer* const mem = &idec->mem_;
idec->state_ = new_state;
mem->start_ += consumed_bytes;
assert(mem->start_ <= mem->end_);
idec->io_.data = mem->buf_ + mem->start_;
idec->io_.data_size = MemDataSize(mem);
}
// Headers
static VP8StatusCode DecodeWebPHeaders(WebPIDecoder* const idec) {
MemBuffer* const mem = &idec->mem_;
const uint8_t* data = mem->buf_ + mem->start_;
size_t curr_size = MemDataSize(mem);
VP8StatusCode status;
WebPHeaderStructure headers;
headers.data = data;
headers.data_size = curr_size;
headers.have_all_data = 0;
status = WebPParseHeaders(&headers);
if (status == VP8_STATUS_NOT_ENOUGH_DATA) {
return VP8_STATUS_SUSPENDED; // We haven't found a VP8 chunk yet.
} else if (status != VP8_STATUS_OK) {
return IDecError(idec, status);
}
idec->chunk_size_ = headers.compressed_size;
idec->is_lossless_ = headers.is_lossless;
if (!idec->is_lossless_) {
VP8Decoder* const dec = VP8New();
if (dec == NULL) {
return VP8_STATUS_OUT_OF_MEMORY;
}
idec->dec_ = dec;
dec->alpha_data_ = headers.alpha_data;
dec->alpha_data_size_ = headers.alpha_data_size;
ChangeState(idec, STATE_VP8_HEADER, headers.offset);
} else {
VP8LDecoder* const dec = VP8LNew();
if (dec == NULL) {
return VP8_STATUS_OUT_OF_MEMORY;
}
idec->dec_ = dec;
ChangeState(idec, STATE_VP8L_HEADER, headers.offset);
}
return VP8_STATUS_OK;
}
static VP8StatusCode DecodeVP8FrameHeader(WebPIDecoder* const idec) {
const uint8_t* data = idec->mem_.buf_ + idec->mem_.start_;
const size_t curr_size = MemDataSize(&idec->mem_);
int width, height;
uint32_t bits;
if (curr_size < VP8_FRAME_HEADER_SIZE) {
// Not enough data bytes to extract VP8 Frame Header.
return VP8_STATUS_SUSPENDED;
}
if (!VP8GetInfo(data, curr_size, idec->chunk_size_, &width, &height)) {
return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR);
}
bits = data[0] | (data[1] << 8) | (data[2] << 16);
idec->mem_.part0_size_ = (bits >> 5) + VP8_FRAME_HEADER_SIZE;
idec->io_.data = data;
idec->io_.data_size = curr_size;
idec->state_ = STATE_VP8_PARTS0;
return VP8_STATUS_OK;
}
// Partition #0
static VP8StatusCode CopyParts0Data(WebPIDecoder* const idec) {
VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
VP8BitReader* const br = &dec->br_;
const size_t part_size = br->buf_end_ - br->buf_;
MemBuffer* const mem = &idec->mem_;
assert(!idec->is_lossless_);
assert(mem->part0_buf_ == NULL);
// the following is a format limitation, no need for runtime check:
assert(part_size <= mem->part0_size_);
if (part_size == 0) { // can't have zero-size partition #0
return VP8_STATUS_BITSTREAM_ERROR;
}
if (mem->mode_ == MEM_MODE_APPEND) {
// We copy and grab ownership of the partition #0 data.
uint8_t* const part0_buf = (uint8_t*)WebPSafeMalloc(1ULL, part_size);
if (part0_buf == NULL) {
return VP8_STATUS_OUT_OF_MEMORY;
}
memcpy(part0_buf, br->buf_, part_size);
mem->part0_buf_ = part0_buf;
VP8BitReaderSetBuffer(br, part0_buf, part_size);
} else {
// Else: just keep pointers to the partition #0's data in dec_->br_.
}
mem->start_ += part_size;
return VP8_STATUS_OK;
}
static VP8StatusCode DecodePartition0(WebPIDecoder* const idec) {
VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
VP8Io* const io = &idec->io_;
const WebPDecParams* const params = &idec->params_;
WebPDecBuffer* const output = params->output;
// Wait till we have enough data for the whole partition #0
if (MemDataSize(&idec->mem_) < idec->mem_.part0_size_) {
return VP8_STATUS_SUSPENDED;
}
if (!VP8GetHeaders(dec, io)) {
const VP8StatusCode status = dec->status_;
if (status == VP8_STATUS_SUSPENDED ||
status == VP8_STATUS_NOT_ENOUGH_DATA) {
// treating NOT_ENOUGH_DATA as SUSPENDED state
return VP8_STATUS_SUSPENDED;
}
return IDecError(idec, status);
}
// Allocate/Verify output buffer now
dec->status_ = WebPAllocateDecBuffer(io->width, io->height, params->options,
output);
if (dec->status_ != VP8_STATUS_OK) {
return IDecError(idec, dec->status_);
}
// This change must be done before calling VP8InitFrame()
dec->mt_method_ = VP8GetThreadMethod(params->options, NULL,
io->width, io->height);
VP8InitDithering(params->options, dec);
dec->status_ = CopyParts0Data(idec);
if (dec->status_ != VP8_STATUS_OK) {
return IDecError(idec, dec->status_);
}
// Finish setting up the decoding parameters. Will call io->setup().
if (VP8EnterCritical(dec, io) != VP8_STATUS_OK) {
return IDecError(idec, dec->status_);
}
// Note: past this point, teardown() must always be called
// in case of error.
idec->state_ = STATE_VP8_DATA;
// Allocate memory and prepare everything.
if (!VP8InitFrame(dec, io)) {
return IDecError(idec, dec->status_);
}
return VP8_STATUS_OK;
}
// Remaining partitions
static VP8StatusCode DecodeRemaining(WebPIDecoder* const idec) {
VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
VP8Io* const io = &idec->io_;
assert(dec->ready_);
for (; dec->mb_y_ < dec->mb_h_; ++dec->mb_y_) {
if (idec->last_mb_y_ != dec->mb_y_) {
if (!VP8ParseIntraModeRow(&dec->br_, dec)) {
// note: normally, error shouldn't occur since we already have the whole
// partition0 available here in DecodeRemaining(). Reaching EOF while
// reading intra modes really means a BITSTREAM_ERROR.
return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR);
}
idec->last_mb_y_ = dec->mb_y_;
}
for (; dec->mb_x_ < dec->mb_w_; ++dec->mb_x_) {
VP8BitReader* const token_br =
&dec->parts_[dec->mb_y_ & dec->num_parts_minus_one_];
MBContext context;
SaveContext(dec, token_br, &context);
if (!VP8DecodeMB(dec, token_br)) {
// We shouldn't fail when MAX_MB data was available
if (dec->num_parts_minus_one_ == 0 &&
MemDataSize(&idec->mem_) > MAX_MB_SIZE) {
return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR);
}
RestoreContext(&context, dec, token_br);
return VP8_STATUS_SUSPENDED;
}
// Release buffer only if there is only one partition
if (dec->num_parts_minus_one_ == 0) {
idec->mem_.start_ = token_br->buf_ - idec->mem_.buf_;
assert(idec->mem_.start_ <= idec->mem_.end_);
}
}
VP8InitScanline(dec); // Prepare for next scanline
// Reconstruct, filter and emit the row.
if (!VP8ProcessRow(dec, io)) {
return IDecError(idec, VP8_STATUS_USER_ABORT);
}
}
// Synchronize the thread and check for errors.
if (!VP8ExitCritical(dec, io)) {
return IDecError(idec, VP8_STATUS_USER_ABORT);
}
dec->ready_ = 0;
return FinishDecoding(idec);
}
static VP8StatusCode ErrorStatusLossless(WebPIDecoder* const idec,
VP8StatusCode status) {
if (status == VP8_STATUS_SUSPENDED || status == VP8_STATUS_NOT_ENOUGH_DATA) {
return VP8_STATUS_SUSPENDED;
}
return IDecError(idec, status);
}
static VP8StatusCode DecodeVP8LHeader(WebPIDecoder* const idec) {
VP8Io* const io = &idec->io_;
VP8LDecoder* const dec = (VP8LDecoder*)idec->dec_;
const WebPDecParams* const params = &idec->params_;
WebPDecBuffer* const output = params->output;
size_t curr_size = MemDataSize(&idec->mem_);
assert(idec->is_lossless_);
// Wait until there's enough data for decoding header.
if (curr_size < (idec->chunk_size_ >> 3)) {
dec->status_ = VP8_STATUS_SUSPENDED;
return ErrorStatusLossless(idec, dec->status_);
}
if (!VP8LDecodeHeader(dec, io)) {
if (dec->status_ == VP8_STATUS_BITSTREAM_ERROR &&
curr_size < idec->chunk_size_) {
dec->status_ = VP8_STATUS_SUSPENDED;
}
return ErrorStatusLossless(idec, dec->status_);
}
// Allocate/verify output buffer now.
dec->status_ = WebPAllocateDecBuffer(io->width, io->height, params->options,
output);
if (dec->status_ != VP8_STATUS_OK) {
return IDecError(idec, dec->status_);
}
idec->state_ = STATE_VP8L_DATA;
return VP8_STATUS_OK;
}
static VP8StatusCode DecodeVP8LData(WebPIDecoder* const idec) {
VP8LDecoder* const dec = (VP8LDecoder*)idec->dec_;
const size_t curr_size = MemDataSize(&idec->mem_);
assert(idec->is_lossless_);
// Switch to incremental decoding if we don't have all the bytes available.
dec->incremental_ = (curr_size < idec->chunk_size_);
if (!VP8LDecodeImage(dec)) {
return ErrorStatusLossless(idec, dec->status_);
}
assert(dec->status_ == VP8_STATUS_OK || dec->status_ == VP8_STATUS_SUSPENDED);
return (dec->status_ == VP8_STATUS_SUSPENDED) ? dec->status_
: FinishDecoding(idec);
}
// Main decoding loop
static VP8StatusCode IDecode(WebPIDecoder* idec) {
VP8StatusCode status = VP8_STATUS_SUSPENDED;
if (idec->state_ == STATE_WEBP_HEADER) {
status = DecodeWebPHeaders(idec);
} else {
if (idec->dec_ == NULL) {
return VP8_STATUS_SUSPENDED; // can't continue if we have no decoder.
}
}
if (idec->state_ == STATE_VP8_HEADER) {
status = DecodeVP8FrameHeader(idec);
}
if (idec->state_ == STATE_VP8_PARTS0) {
status = DecodePartition0(idec);
}
if (idec->state_ == STATE_VP8_DATA) {
status = DecodeRemaining(idec);
}
if (idec->state_ == STATE_VP8L_HEADER) {
status = DecodeVP8LHeader(idec);
}
if (idec->state_ == STATE_VP8L_DATA) {
status = DecodeVP8LData(idec);
}
return status;
}
//------------------------------------------------------------------------------
// Internal constructor
static WebPIDecoder* NewDecoder(WebPDecBuffer* const output_buffer,
const WebPBitstreamFeatures* const features) {
WebPIDecoder* idec = (WebPIDecoder*)WebPSafeCalloc(1ULL, sizeof(*idec));
if (idec == NULL) {
return NULL;
}
idec->state_ = STATE_WEBP_HEADER;
idec->chunk_size_ = 0;
idec->last_mb_y_ = -1;
InitMemBuffer(&idec->mem_);
WebPInitDecBuffer(&idec->output_);
VP8InitIo(&idec->io_);
WebPResetDecParams(&idec->params_);
if (output_buffer == NULL || WebPAvoidSlowMemory(output_buffer, features)) {
idec->params_.output = &idec->output_;
idec->final_output_ = output_buffer;
if (output_buffer != NULL) {
idec->params_.output->colorspace = output_buffer->colorspace;
}
} else {
idec->params_.output = output_buffer;
idec->final_output_ = NULL;
}
WebPInitCustomIo(&idec->params_, &idec->io_); // Plug the I/O functions.
return idec;
}
//------------------------------------------------------------------------------
// Public functions
WebPIDecoder* WebPINewDecoder(WebPDecBuffer* output_buffer) {
return NewDecoder(output_buffer, NULL);
}
WebPIDecoder* WebPIDecode(const uint8_t* data, size_t data_size,
WebPDecoderConfig* config) {
WebPIDecoder* idec;
WebPBitstreamFeatures tmp_features;
WebPBitstreamFeatures* const features =
(config == NULL) ? &tmp_features : &config->input;
memset(&tmp_features, 0, sizeof(tmp_features));
// Parse the bitstream's features, if requested:
if (data != NULL && data_size > 0) {
if (WebPGetFeatures(data, data_size, features) != VP8_STATUS_OK) {
return NULL;
}
}
// Create an instance of the incremental decoder
idec = (config != NULL) ? NewDecoder(&config->output, features)
: NewDecoder(NULL, features);
if (idec == NULL) {
return NULL;
}
// Finish initialization
if (config != NULL) {
idec->params_.options = &config->options;
}
return idec;
}
void WebPIDelete(WebPIDecoder* idec) {
if (idec == NULL) return;
if (idec->dec_ != NULL) {
if (!idec->is_lossless_) {
if (idec->state_ == STATE_VP8_DATA) {
// Synchronize the thread, clean-up and check for errors.
VP8ExitCritical((VP8Decoder*)idec->dec_, &idec->io_);
}
VP8Delete((VP8Decoder*)idec->dec_);
} else {
VP8LDelete((VP8LDecoder*)idec->dec_);
}
}
ClearMemBuffer(&idec->mem_);
WebPFreeDecBuffer(&idec->output_);
WebPSafeFree(idec);
}
//------------------------------------------------------------------------------
// Wrapper toward WebPINewDecoder
WebPIDecoder* WebPINewRGB(WEBP_CSP_MODE csp, uint8_t* output_buffer,
size_t output_buffer_size, int output_stride) {
const int is_external_memory = (output_buffer != NULL) ? 1 : 0;
WebPIDecoder* idec;
if (csp >= MODE_YUV) return NULL;
if (is_external_memory == 0) { // Overwrite parameters to sane values.
output_buffer_size = 0;
output_stride = 0;
} else { // A buffer was passed. Validate the other params.
if (output_stride == 0 || output_buffer_size == 0) {
return NULL; // invalid parameter.
}
}
idec = WebPINewDecoder(NULL);
if (idec == NULL) return NULL;
idec->output_.colorspace = csp;
idec->output_.is_external_memory = is_external_memory;
idec->output_.u.RGBA.rgba = output_buffer;
idec->output_.u.RGBA.stride = output_stride;
idec->output_.u.RGBA.size = output_buffer_size;
return idec;
}
WebPIDecoder* WebPINewYUVA(uint8_t* luma, size_t luma_size, int luma_stride,
uint8_t* u, size_t u_size, int u_stride,
uint8_t* v, size_t v_size, int v_stride,
uint8_t* a, size_t a_size, int a_stride) {
const int is_external_memory = (luma != NULL) ? 1 : 0;
WebPIDecoder* idec;
WEBP_CSP_MODE colorspace;
if (is_external_memory == 0) { // Overwrite parameters to sane values.
luma_size = u_size = v_size = a_size = 0;
luma_stride = u_stride = v_stride = a_stride = 0;
u = v = a = NULL;
colorspace = MODE_YUVA;
} else { // A luma buffer was passed. Validate the other parameters.
if (u == NULL || v == NULL) return NULL;
if (luma_size == 0 || u_size == 0 || v_size == 0) return NULL;
if (luma_stride == 0 || u_stride == 0 || v_stride == 0) return NULL;
if (a != NULL) {
if (a_size == 0 || a_stride == 0) return NULL;
}
colorspace = (a == NULL) ? MODE_YUV : MODE_YUVA;
}
idec = WebPINewDecoder(NULL);
if (idec == NULL) return NULL;
idec->output_.colorspace = colorspace;
idec->output_.is_external_memory = is_external_memory;
idec->output_.u.YUVA.y = luma;
idec->output_.u.YUVA.y_stride = luma_stride;
idec->output_.u.YUVA.y_size = luma_size;
idec->output_.u.YUVA.u = u;
idec->output_.u.YUVA.u_stride = u_stride;
idec->output_.u.YUVA.u_size = u_size;
idec->output_.u.YUVA.v = v;
idec->output_.u.YUVA.v_stride = v_stride;
idec->output_.u.YUVA.v_size = v_size;
idec->output_.u.YUVA.a = a;
idec->output_.u.YUVA.a_stride = a_stride;
idec->output_.u.YUVA.a_size = a_size;
return idec;
}
WebPIDecoder* WebPINewYUV(uint8_t* luma, size_t luma_size, int luma_stride,
uint8_t* u, size_t u_size, int u_stride,
uint8_t* v, size_t v_size, int v_stride) {
return WebPINewYUVA(luma, luma_size, luma_stride,
u, u_size, u_stride,
v, v_size, v_stride,
NULL, 0, 0);
}
//------------------------------------------------------------------------------
static VP8StatusCode IDecCheckStatus(const WebPIDecoder* const idec) {
assert(idec);
if (idec->state_ == STATE_ERROR) {
return VP8_STATUS_BITSTREAM_ERROR;
}
if (idec->state_ == STATE_DONE) {
return VP8_STATUS_OK;
}
return VP8_STATUS_SUSPENDED;
}
VP8StatusCode WebPIAppend(WebPIDecoder* idec,
const uint8_t* data, size_t data_size) {
VP8StatusCode status;
if (idec == NULL || data == NULL) {
return VP8_STATUS_INVALID_PARAM;
}
status = IDecCheckStatus(idec);
if (status != VP8_STATUS_SUSPENDED) {
return status;
}
// Check mixed calls between RemapMemBuffer and AppendToMemBuffer.
if (!CheckMemBufferMode(&idec->mem_, MEM_MODE_APPEND)) {
return VP8_STATUS_INVALID_PARAM;
}
// Append data to memory buffer
if (!AppendToMemBuffer(idec, data, data_size)) {
return VP8_STATUS_OUT_OF_MEMORY;
}
return IDecode(idec);
}
VP8StatusCode WebPIUpdate(WebPIDecoder* idec,
const uint8_t* data, size_t data_size) {
VP8StatusCode status;
if (idec == NULL || data == NULL) {
return VP8_STATUS_INVALID_PARAM;
}
status = IDecCheckStatus(idec);
if (status != VP8_STATUS_SUSPENDED) {
return status;
}
// Check mixed calls between RemapMemBuffer and AppendToMemBuffer.
if (!CheckMemBufferMode(&idec->mem_, MEM_MODE_MAP)) {
return VP8_STATUS_INVALID_PARAM;
}
// Make the memory buffer point to the new buffer
if (!RemapMemBuffer(idec, data, data_size)) {
return VP8_STATUS_INVALID_PARAM;
}
return IDecode(idec);
}
//------------------------------------------------------------------------------
static const WebPDecBuffer* GetOutputBuffer(const WebPIDecoder* const idec) {
if (idec == NULL || idec->dec_ == NULL) {
return NULL;
}
if (idec->state_ <= STATE_VP8_PARTS0) {
return NULL;
}
if (idec->final_output_ != NULL) {
return NULL; // not yet slow-copied
}
return idec->params_.output;
}
const WebPDecBuffer* WebPIDecodedArea(const WebPIDecoder* idec,
int* left, int* top,
int* width, int* height) {
const WebPDecBuffer* const src = GetOutputBuffer(idec);
if (left != NULL) *left = 0;
if (top != NULL) *top = 0;
if (src != NULL) {
if (width != NULL) *width = src->width;
if (height != NULL) *height = idec->params_.last_y;
} else {
if (width != NULL) *width = 0;
if (height != NULL) *height = 0;
}
return src;
}
uint8_t* WebPIDecGetRGB(const WebPIDecoder* idec, int* last_y,
int* width, int* height, int* stride) {
const WebPDecBuffer* const src = GetOutputBuffer(idec);
if (src == NULL) return NULL;
if (src->colorspace >= MODE_YUV) {
return NULL;
}
if (last_y != NULL) *last_y = idec->params_.last_y;
if (width != NULL) *width = src->width;
if (height != NULL) *height = src->height;
if (stride != NULL) *stride = src->u.RGBA.stride;
return src->u.RGBA.rgba;
}
uint8_t* WebPIDecGetYUVA(const WebPIDecoder* idec, int* last_y,
uint8_t** u, uint8_t** v, uint8_t** a,
int* width, int* height,
int* stride, int* uv_stride, int* a_stride) {
const WebPDecBuffer* const src = GetOutputBuffer(idec);
if (src == NULL) return NULL;
if (src->colorspace < MODE_YUV) {
return NULL;
}
if (last_y != NULL) *last_y = idec->params_.last_y;
if (u != NULL) *u = src->u.YUVA.u;
if (v != NULL) *v = src->u.YUVA.v;
if (a != NULL) *a = src->u.YUVA.a;
if (width != NULL) *width = src->width;
if (height != NULL) *height = src->height;
if (stride != NULL) *stride = src->u.YUVA.y_stride;
if (uv_stride != NULL) *uv_stride = src->u.YUVA.u_stride;
if (a_stride != NULL) *a_stride = src->u.YUVA.a_stride;
return src->u.YUVA.y;
}
int WebPISetIOHooks(WebPIDecoder* const idec,
VP8IoPutHook put,
VP8IoSetupHook setup,
VP8IoTeardownHook teardown,
void* user_data) {
if (idec == NULL || idec->state_ > STATE_WEBP_HEADER) {
return 0;
}
idec->io_.put = put;
idec->io_.setup = setup;
idec->io_.teardown = teardown;
idec->io_.opaque = user_data;
return 1;
}

649
media/libwebp/src/dec/io_dec.c Normal file
Просмотреть файл

@ -0,0 +1,649 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// functions for sample output.
//
// Author: Skal (pascal.massimino@gmail.com)
#include <assert.h>
#include <stdlib.h>
#include "src/dec/vp8i_dec.h"
#include "src/dec/webpi_dec.h"
#include "src/dsp/dsp.h"
#include "src/dsp/yuv.h"
#include "src/utils/utils.h"
//------------------------------------------------------------------------------
// Main YUV<->RGB conversion functions
static int EmitYUV(const VP8Io* const io, WebPDecParams* const p) {
WebPDecBuffer* output = p->output;
const WebPYUVABuffer* const buf = &output->u.YUVA;
uint8_t* const y_dst = buf->y + io->mb_y * buf->y_stride;
uint8_t* const u_dst = buf->u + (io->mb_y >> 1) * buf->u_stride;
uint8_t* const v_dst = buf->v + (io->mb_y >> 1) * buf->v_stride;
const int mb_w = io->mb_w;
const int mb_h = io->mb_h;
const int uv_w = (mb_w + 1) / 2;
const int uv_h = (mb_h + 1) / 2;
int j;
for (j = 0; j < mb_h; ++j) {
memcpy(y_dst + j * buf->y_stride, io->y + j * io->y_stride, mb_w);
}
for (j = 0; j < uv_h; ++j) {
memcpy(u_dst + j * buf->u_stride, io->u + j * io->uv_stride, uv_w);
memcpy(v_dst + j * buf->v_stride, io->v + j * io->uv_stride, uv_w);
}
return io->mb_h;
}
// Point-sampling U/V sampler.
static int EmitSampledRGB(const VP8Io* const io, WebPDecParams* const p) {
WebPDecBuffer* const output = p->output;
WebPRGBABuffer* const buf = &output->u.RGBA;
uint8_t* const dst = buf->rgba + io->mb_y * buf->stride;
WebPSamplerProcessPlane(io->y, io->y_stride,
io->u, io->v, io->uv_stride,
dst, buf->stride, io->mb_w, io->mb_h,
WebPSamplers[output->colorspace]);
return io->mb_h;
}
//------------------------------------------------------------------------------
// Fancy upsampling
#ifdef FANCY_UPSAMPLING
static int EmitFancyRGB(const VP8Io* const io, WebPDecParams* const p) {
int num_lines_out = io->mb_h; // a priori guess
const WebPRGBABuffer* const buf = &p->output->u.RGBA;
uint8_t* dst = buf->rgba + io->mb_y * buf->stride;
WebPUpsampleLinePairFunc upsample = WebPUpsamplers[p->output->colorspace];
const uint8_t* cur_y = io->y;
const uint8_t* cur_u = io->u;
const uint8_t* cur_v = io->v;
const uint8_t* top_u = p->tmp_u;
const uint8_t* top_v = p->tmp_v;
int y = io->mb_y;
const int y_end = io->mb_y + io->mb_h;
const int mb_w = io->mb_w;
const int uv_w = (mb_w + 1) / 2;
if (y == 0) {
// First line is special cased. We mirror the u/v samples at boundary.
upsample(cur_y, NULL, cur_u, cur_v, cur_u, cur_v, dst, NULL, mb_w);
} else {
// We can finish the left-over line from previous call.
upsample(p->tmp_y, cur_y, top_u, top_v, cur_u, cur_v,
dst - buf->stride, dst, mb_w);
++num_lines_out;
}
// Loop over each output pairs of row.
for (; y + 2 < y_end; y += 2) {
top_u = cur_u;
top_v = cur_v;
cur_u += io->uv_stride;
cur_v += io->uv_stride;
dst += 2 * buf->stride;
cur_y += 2 * io->y_stride;
upsample(cur_y - io->y_stride, cur_y,
top_u, top_v, cur_u, cur_v,
dst - buf->stride, dst, mb_w);
}
// move to last row
cur_y += io->y_stride;
if (io->crop_top + y_end < io->crop_bottom) {
// Save the unfinished samples for next call (as we're not done yet).
memcpy(p->tmp_y, cur_y, mb_w * sizeof(*p->tmp_y));
memcpy(p->tmp_u, cur_u, uv_w * sizeof(*p->tmp_u));
memcpy(p->tmp_v, cur_v, uv_w * sizeof(*p->tmp_v));
// The fancy upsampler leaves a row unfinished behind
// (except for the very last row)
num_lines_out--;
} else {
// Process the very last row of even-sized picture
if (!(y_end & 1)) {
upsample(cur_y, NULL, cur_u, cur_v, cur_u, cur_v,
dst + buf->stride, NULL, mb_w);
}
}
return num_lines_out;
}
#endif /* FANCY_UPSAMPLING */
//------------------------------------------------------------------------------
static void FillAlphaPlane(uint8_t* dst, int w, int h, int stride) {
int j;
for (j = 0; j < h; ++j) {
memset(dst, 0xff, w * sizeof(*dst));
dst += stride;
}
}
static int EmitAlphaYUV(const VP8Io* const io, WebPDecParams* const p,
int expected_num_lines_out) {
const uint8_t* alpha = io->a;
const WebPYUVABuffer* const buf = &p->output->u.YUVA;
const int mb_w = io->mb_w;
const int mb_h = io->mb_h;
uint8_t* dst = buf->a + io->mb_y * buf->a_stride;
int j;
(void)expected_num_lines_out;
assert(expected_num_lines_out == mb_h);
if (alpha != NULL) {
for (j = 0; j < mb_h; ++j) {
memcpy(dst, alpha, mb_w * sizeof(*dst));
alpha += io->width;
dst += buf->a_stride;
}
} else if (buf->a != NULL) {
// the user requested alpha, but there is none, set it to opaque.
FillAlphaPlane(dst, mb_w, mb_h, buf->a_stride);
}
return 0;
}
static int GetAlphaSourceRow(const VP8Io* const io,
const uint8_t** alpha, int* const num_rows) {
int start_y = io->mb_y;
*num_rows = io->mb_h;
// Compensate for the 1-line delay of the fancy upscaler.
// This is similar to EmitFancyRGB().
if (io->fancy_upsampling) {
if (start_y == 0) {
// We don't process the last row yet. It'll be done during the next call.
--*num_rows;
} else {
--start_y;
// Fortunately, *alpha data is persistent, so we can go back
// one row and finish alpha blending, now that the fancy upscaler
// completed the YUV->RGB interpolation.
*alpha -= io->width;
}
if (io->crop_top + io->mb_y + io->mb_h == io->crop_bottom) {
// If it's the very last call, we process all the remaining rows!
*num_rows = io->crop_bottom - io->crop_top - start_y;
}
}
return start_y;
}
static int EmitAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
int expected_num_lines_out) {
const uint8_t* alpha = io->a;
if (alpha != NULL) {
const int mb_w = io->mb_w;
const WEBP_CSP_MODE colorspace = p->output->colorspace;
const int alpha_first =
(colorspace == MODE_ARGB || colorspace == MODE_Argb);
const WebPRGBABuffer* const buf = &p->output->u.RGBA;
int num_rows;
const int start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
uint8_t* const base_rgba = buf->rgba + start_y * buf->stride;
uint8_t* const dst = base_rgba + (alpha_first ? 0 : 3);
const int has_alpha = WebPDispatchAlpha(alpha, io->width, mb_w,
num_rows, dst, buf->stride);
(void)expected_num_lines_out;
assert(expected_num_lines_out == num_rows);
// has_alpha is true if there's non-trivial alpha to premultiply with.
if (has_alpha && WebPIsPremultipliedMode(colorspace)) {
WebPApplyAlphaMultiply(base_rgba, alpha_first,
mb_w, num_rows, buf->stride);
}
}
return 0;
}
static int EmitAlphaRGBA4444(const VP8Io* const io, WebPDecParams* const p,
int expected_num_lines_out) {
const uint8_t* alpha = io->a;
if (alpha != NULL) {
const int mb_w = io->mb_w;
const WEBP_CSP_MODE colorspace = p->output->colorspace;
const WebPRGBABuffer* const buf = &p->output->u.RGBA;
int num_rows;
const int start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
uint8_t* const base_rgba = buf->rgba + start_y * buf->stride;
#if (WEBP_SWAP_16BIT_CSP == 1)
uint8_t* alpha_dst = base_rgba;
#else
uint8_t* alpha_dst = base_rgba + 1;
#endif
uint32_t alpha_mask = 0x0f;
int i, j;
for (j = 0; j < num_rows; ++j) {
for (i = 0; i < mb_w; ++i) {
// Fill in the alpha value (converted to 4 bits).
const uint32_t alpha_value = alpha[i] >> 4;
alpha_dst[2 * i] = (alpha_dst[2 * i] & 0xf0) | alpha_value;
alpha_mask &= alpha_value;
}
alpha += io->width;
alpha_dst += buf->stride;
}
(void)expected_num_lines_out;
assert(expected_num_lines_out == num_rows);
if (alpha_mask != 0x0f && WebPIsPremultipliedMode(colorspace)) {
WebPApplyAlphaMultiply4444(base_rgba, mb_w, num_rows, buf->stride);
}
}
return 0;
}
//------------------------------------------------------------------------------
// YUV rescaling (no final RGB conversion needed)
#if !defined(WEBP_REDUCE_SIZE)
static int Rescale(const uint8_t* src, int src_stride,
int new_lines, WebPRescaler* const wrk) {
int num_lines_out = 0;
while (new_lines > 0) { // import new contributions of source rows.
const int lines_in = WebPRescalerImport(wrk, new_lines, src, src_stride);
src += lines_in * src_stride;
new_lines -= lines_in;
num_lines_out += WebPRescalerExport(wrk); // emit output row(s)
}
return num_lines_out;
}
static int EmitRescaledYUV(const VP8Io* const io, WebPDecParams* const p) {
const int mb_h = io->mb_h;
const int uv_mb_h = (mb_h + 1) >> 1;
WebPRescaler* const scaler = p->scaler_y;
int num_lines_out = 0;
if (WebPIsAlphaMode(p->output->colorspace) && io->a != NULL) {
// Before rescaling, we premultiply the luma directly into the io->y
// internal buffer. This is OK since these samples are not used for
// intra-prediction (the top samples are saved in cache_y_/u_/v_).
// But we need to cast the const away, though.
WebPMultRows((uint8_t*)io->y, io->y_stride,
io->a, io->width, io->mb_w, mb_h, 0);
}
num_lines_out = Rescale(io->y, io->y_stride, mb_h, scaler);
Rescale(io->u, io->uv_stride, uv_mb_h, p->scaler_u);
Rescale(io->v, io->uv_stride, uv_mb_h, p->scaler_v);
return num_lines_out;
}
static int EmitRescaledAlphaYUV(const VP8Io* const io, WebPDecParams* const p,
int expected_num_lines_out) {
const WebPYUVABuffer* const buf = &p->output->u.YUVA;
uint8_t* const dst_a = buf->a + p->last_y * buf->a_stride;
if (io->a != NULL) {
uint8_t* const dst_y = buf->y + p->last_y * buf->y_stride;
const int num_lines_out = Rescale(io->a, io->width, io->mb_h, p->scaler_a);
assert(expected_num_lines_out == num_lines_out);
if (num_lines_out > 0) { // unmultiply the Y
WebPMultRows(dst_y, buf->y_stride, dst_a, buf->a_stride,
p->scaler_a->dst_width, num_lines_out, 1);
}
} else if (buf->a != NULL) {
// the user requested alpha, but there is none, set it to opaque.
assert(p->last_y + expected_num_lines_out <= io->scaled_height);
FillAlphaPlane(dst_a, io->scaled_width, expected_num_lines_out,
buf->a_stride);
}
return 0;
}
static int InitYUVRescaler(const VP8Io* const io, WebPDecParams* const p) {
const int has_alpha = WebPIsAlphaMode(p->output->colorspace);
const WebPYUVABuffer* const buf = &p->output->u.YUVA;
const int out_width = io->scaled_width;
const int out_height = io->scaled_height;
const int uv_out_width = (out_width + 1) >> 1;
const int uv_out_height = (out_height + 1) >> 1;
const int uv_in_width = (io->mb_w + 1) >> 1;
const int uv_in_height = (io->mb_h + 1) >> 1;
const size_t work_size = 2 * out_width; // scratch memory for luma rescaler
const size_t uv_work_size = 2 * uv_out_width; // and for each u/v ones
size_t tmp_size, rescaler_size;
rescaler_t* work;
WebPRescaler* scalers;
const int num_rescalers = has_alpha ? 4 : 3;
tmp_size = (work_size + 2 * uv_work_size) * sizeof(*work);
if (has_alpha) {
tmp_size += work_size * sizeof(*work);
}
rescaler_size = num_rescalers * sizeof(*p->scaler_y) + WEBP_ALIGN_CST;
p->memory = WebPSafeMalloc(1ULL, tmp_size + rescaler_size);
if (p->memory == NULL) {
return 0; // memory error
}
work = (rescaler_t*)p->memory;
scalers = (WebPRescaler*)WEBP_ALIGN((const uint8_t*)work + tmp_size);
p->scaler_y = &scalers[0];
p->scaler_u = &scalers[1];
p->scaler_v = &scalers[2];
p->scaler_a = has_alpha ? &scalers[3] : NULL;
WebPRescalerInit(p->scaler_y, io->mb_w, io->mb_h,
buf->y, out_width, out_height, buf->y_stride, 1,
work);
WebPRescalerInit(p->scaler_u, uv_in_width, uv_in_height,
buf->u, uv_out_width, uv_out_height, buf->u_stride, 1,
work + work_size);
WebPRescalerInit(p->scaler_v, uv_in_width, uv_in_height,
buf->v, uv_out_width, uv_out_height, buf->v_stride, 1,
work + work_size + uv_work_size);
p->emit = EmitRescaledYUV;
if (has_alpha) {
WebPRescalerInit(p->scaler_a, io->mb_w, io->mb_h,
buf->a, out_width, out_height, buf->a_stride, 1,
work + work_size + 2 * uv_work_size);
p->emit_alpha = EmitRescaledAlphaYUV;
WebPInitAlphaProcessing();
}
return 1;
}
//------------------------------------------------------------------------------
// RGBA rescaling
static int ExportRGB(WebPDecParams* const p, int y_pos) {
const WebPYUV444Converter convert =
WebPYUV444Converters[p->output->colorspace];
const WebPRGBABuffer* const buf = &p->output->u.RGBA;
uint8_t* dst = buf->rgba + y_pos * buf->stride;
int num_lines_out = 0;
// For RGB rescaling, because of the YUV420, current scan position
// U/V can be +1/-1 line from the Y one. Hence the double test.
while (WebPRescalerHasPendingOutput(p->scaler_y) &&
WebPRescalerHasPendingOutput(p->scaler_u)) {
assert(y_pos + num_lines_out < p->output->height);
assert(p->scaler_u->y_accum == p->scaler_v->y_accum);
WebPRescalerExportRow(p->scaler_y);
WebPRescalerExportRow(p->scaler_u);
WebPRescalerExportRow(p->scaler_v);
convert(p->scaler_y->dst, p->scaler_u->dst, p->scaler_v->dst,
dst, p->scaler_y->dst_width);
dst += buf->stride;
++num_lines_out;
}
return num_lines_out;
}
static int EmitRescaledRGB(const VP8Io* const io, WebPDecParams* const p) {
const int mb_h = io->mb_h;
const int uv_mb_h = (mb_h + 1) >> 1;
int j = 0, uv_j = 0;
int num_lines_out = 0;
while (j < mb_h) {
const int y_lines_in =
WebPRescalerImport(p->scaler_y, mb_h - j,
io->y + j * io->y_stride, io->y_stride);
j += y_lines_in;
if (WebPRescaleNeededLines(p->scaler_u, uv_mb_h - uv_j)) {
const int u_lines_in =
WebPRescalerImport(p->scaler_u, uv_mb_h - uv_j,
io->u + uv_j * io->uv_stride, io->uv_stride);
const int v_lines_in =
WebPRescalerImport(p->scaler_v, uv_mb_h - uv_j,
io->v + uv_j * io->uv_stride, io->uv_stride);
(void)v_lines_in; // remove a gcc warning
assert(u_lines_in == v_lines_in);
uv_j += u_lines_in;
}
num_lines_out += ExportRGB(p, p->last_y + num_lines_out);
}
return num_lines_out;
}
static int ExportAlpha(WebPDecParams* const p, int y_pos, int max_lines_out) {
const WebPRGBABuffer* const buf = &p->output->u.RGBA;
uint8_t* const base_rgba = buf->rgba + y_pos * buf->stride;
const WEBP_CSP_MODE colorspace = p->output->colorspace;
const int alpha_first =
(colorspace == MODE_ARGB || colorspace == MODE_Argb);
uint8_t* dst = base_rgba + (alpha_first ? 0 : 3);
int num_lines_out = 0;
const int is_premult_alpha = WebPIsPremultipliedMode(colorspace);
uint32_t non_opaque = 0;
const int width = p->scaler_a->dst_width;
while (WebPRescalerHasPendingOutput(p->scaler_a) &&
num_lines_out < max_lines_out) {
assert(y_pos + num_lines_out < p->output->height);
WebPRescalerExportRow(p->scaler_a);
non_opaque |= WebPDispatchAlpha(p->scaler_a->dst, 0, width, 1, dst, 0);
dst += buf->stride;
++num_lines_out;
}
if (is_premult_alpha && non_opaque) {
WebPApplyAlphaMultiply(base_rgba, alpha_first,
width, num_lines_out, buf->stride);
}
return num_lines_out;
}
static int ExportAlphaRGBA4444(WebPDecParams* const p, int y_pos,
int max_lines_out) {
const WebPRGBABuffer* const buf = &p->output->u.RGBA;
uint8_t* const base_rgba = buf->rgba + y_pos * buf->stride;
#if (WEBP_SWAP_16BIT_CSP == 1)
uint8_t* alpha_dst = base_rgba;
#else
uint8_t* alpha_dst = base_rgba + 1;
#endif
int num_lines_out = 0;
const WEBP_CSP_MODE colorspace = p->output->colorspace;
const int width = p->scaler_a->dst_width;
const int is_premult_alpha = WebPIsPremultipliedMode(colorspace);
uint32_t alpha_mask = 0x0f;
while (WebPRescalerHasPendingOutput(p->scaler_a) &&
num_lines_out < max_lines_out) {
int i;
assert(y_pos + num_lines_out < p->output->height);
WebPRescalerExportRow(p->scaler_a);
for (i = 0; i < width; ++i) {
// Fill in the alpha value (converted to 4 bits).
const uint32_t alpha_value = p->scaler_a->dst[i] >> 4;
alpha_dst[2 * i] = (alpha_dst[2 * i] & 0xf0) | alpha_value;
alpha_mask &= alpha_value;
}
alpha_dst += buf->stride;
++num_lines_out;
}
if (is_premult_alpha && alpha_mask != 0x0f) {
WebPApplyAlphaMultiply4444(base_rgba, width, num_lines_out, buf->stride);
}
return num_lines_out;
}
static int EmitRescaledAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
int expected_num_out_lines) {
if (io->a != NULL) {
WebPRescaler* const scaler = p->scaler_a;
int lines_left = expected_num_out_lines;
const int y_end = p->last_y + lines_left;
while (lines_left > 0) {
const int row_offset = scaler->src_y - io->mb_y;
WebPRescalerImport(scaler, io->mb_h + io->mb_y - scaler->src_y,
io->a + row_offset * io->width, io->width);
lines_left -= p->emit_alpha_row(p, y_end - lines_left, lines_left);
}
}
return 0;
}
static int InitRGBRescaler(const VP8Io* const io, WebPDecParams* const p) {
const int has_alpha = WebPIsAlphaMode(p->output->colorspace);
const int out_width = io->scaled_width;
const int out_height = io->scaled_height;
const int uv_in_width = (io->mb_w + 1) >> 1;
const int uv_in_height = (io->mb_h + 1) >> 1;
const size_t work_size = 2 * out_width; // scratch memory for one rescaler
rescaler_t* work; // rescalers work area
uint8_t* tmp; // tmp storage for scaled YUV444 samples before RGB conversion
size_t tmp_size1, tmp_size2, total_size, rescaler_size;
WebPRescaler* scalers;
const int num_rescalers = has_alpha ? 4 : 3;
tmp_size1 = 3 * work_size;
tmp_size2 = 3 * out_width;
if (has_alpha) {
tmp_size1 += work_size;
tmp_size2 += out_width;
}
total_size = tmp_size1 * sizeof(*work) + tmp_size2 * sizeof(*tmp);
rescaler_size = num_rescalers * sizeof(*p->scaler_y) + WEBP_ALIGN_CST;
p->memory = WebPSafeMalloc(1ULL, total_size + rescaler_size);
if (p->memory == NULL) {
return 0; // memory error
}
work = (rescaler_t*)p->memory;
tmp = (uint8_t*)(work + tmp_size1);
scalers = (WebPRescaler*)WEBP_ALIGN((const uint8_t*)work + total_size);
p->scaler_y = &scalers[0];
p->scaler_u = &scalers[1];
p->scaler_v = &scalers[2];
p->scaler_a = has_alpha ? &scalers[3] : NULL;
WebPRescalerInit(p->scaler_y, io->mb_w, io->mb_h,
tmp + 0 * out_width, out_width, out_height, 0, 1,
work + 0 * work_size);
WebPRescalerInit(p->scaler_u, uv_in_width, uv_in_height,
tmp + 1 * out_width, out_width, out_height, 0, 1,
work + 1 * work_size);
WebPRescalerInit(p->scaler_v, uv_in_width, uv_in_height,
tmp + 2 * out_width, out_width, out_height, 0, 1,
work + 2 * work_size);
p->emit = EmitRescaledRGB;
WebPInitYUV444Converters();
if (has_alpha) {
WebPRescalerInit(p->scaler_a, io->mb_w, io->mb_h,
tmp + 3 * out_width, out_width, out_height, 0, 1,
work + 3 * work_size);
p->emit_alpha = EmitRescaledAlphaRGB;
if (p->output->colorspace == MODE_RGBA_4444 ||
p->output->colorspace == MODE_rgbA_4444) {
p->emit_alpha_row = ExportAlphaRGBA4444;
} else {
p->emit_alpha_row = ExportAlpha;
}
WebPInitAlphaProcessing();
}
return 1;
}
#endif // WEBP_REDUCE_SIZE
//------------------------------------------------------------------------------
// Default custom functions
static int CustomSetup(VP8Io* io) {
WebPDecParams* const p = (WebPDecParams*)io->opaque;
const WEBP_CSP_MODE colorspace = p->output->colorspace;
const int is_rgb = WebPIsRGBMode(colorspace);
const int is_alpha = WebPIsAlphaMode(colorspace);
p->memory = NULL;
p->emit = NULL;
p->emit_alpha = NULL;
p->emit_alpha_row = NULL;
if (!WebPIoInitFromOptions(p->options, io, is_alpha ? MODE_YUV : MODE_YUVA)) {
return 0;
}
if (is_alpha && WebPIsPremultipliedMode(colorspace)) {
WebPInitUpsamplers();
}
if (io->use_scaling) {
#if !defined(WEBP_REDUCE_SIZE)
const int ok = is_rgb ? InitRGBRescaler(io, p) : InitYUVRescaler(io, p);
if (!ok) {
return 0; // memory error
}
#else
return 0; // rescaling support not compiled
#endif
} else {
if (is_rgb) {
WebPInitSamplers();
p->emit = EmitSampledRGB; // default
if (io->fancy_upsampling) {
#ifdef FANCY_UPSAMPLING
const int uv_width = (io->mb_w + 1) >> 1;
p->memory = WebPSafeMalloc(1ULL, (size_t)(io->mb_w + 2 * uv_width));
if (p->memory == NULL) {
return 0; // memory error.
}
p->tmp_y = (uint8_t*)p->memory;
p->tmp_u = p->tmp_y + io->mb_w;
p->tmp_v = p->tmp_u + uv_width;
p->emit = EmitFancyRGB;
WebPInitUpsamplers();
#endif
}
} else {
p->emit = EmitYUV;
}
if (is_alpha) { // need transparency output
p->emit_alpha =
(colorspace == MODE_RGBA_4444 || colorspace == MODE_rgbA_4444) ?
EmitAlphaRGBA4444
: is_rgb ? EmitAlphaRGB
: EmitAlphaYUV;
if (is_rgb) {
WebPInitAlphaProcessing();
}
}
}
return 1;
}
//------------------------------------------------------------------------------
static int CustomPut(const VP8Io* io) {
WebPDecParams* const p = (WebPDecParams*)io->opaque;
const int mb_w = io->mb_w;
const int mb_h = io->mb_h;
int num_lines_out;
assert(!(io->mb_y & 1));
if (mb_w <= 0 || mb_h <= 0) {
return 0;
}
num_lines_out = p->emit(io, p);
if (p->emit_alpha != NULL) {
p->emit_alpha(io, p, num_lines_out);
}
p->last_y += num_lines_out;
return 1;
}
//------------------------------------------------------------------------------
static void CustomTeardown(const VP8Io* io) {
WebPDecParams* const p = (WebPDecParams*)io->opaque;
WebPSafeFree(p->memory);
p->memory = NULL;
}
//------------------------------------------------------------------------------
// Main entry point
void WebPInitCustomIo(WebPDecParams* const params, VP8Io* const io) {
io->put = CustomPut;
io->setup = CustomSetup;
io->teardown = CustomTeardown;
io->opaque = params;
}
//------------------------------------------------------------------------------

30
media/libwebp/src/dec/moz.build Normal file
Просмотреть файл

@ -0,0 +1,30 @@
# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*-
# vim: set filetype=python:
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
with Files('**'):
BUG_COMPONENT = ('Core', 'ImageLib')
SOURCES += [
'alpha_dec.c',
'buffer_dec.c',
'frame_dec.c',
'idec_dec.c',
'io_dec.c',
'quant_dec.c',
'tree_dec.c',
'vp8_dec.c',
'vp8l_dec.c',
'webp_dec.c',
]
LOCAL_INCLUDES += [
'/media/libwebp',
]
FINAL_LIBRARY = 'gkmedias'
# We allow warnings for third-party code that can be updated from upstream.
AllowCompilerWarnings()

110
media/libwebp/src/dec/quant_dec.c Normal file
Просмотреть файл

@ -0,0 +1,110 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Quantizer initialization
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dec/vp8i_dec.h"
static WEBP_INLINE int clip(int v, int M) {
return v < 0 ? 0 : v > M ? M : v;
}
// Paragraph 14.1
static const uint8_t kDcTable[128] = {
4, 5, 6, 7, 8, 9, 10, 10,
11, 12, 13, 14, 15, 16, 17, 17,
18, 19, 20, 20, 21, 21, 22, 22,
23, 23, 24, 25, 25, 26, 27, 28,
29, 30, 31, 32, 33, 34, 35, 36,
37, 37, 38, 39, 40, 41, 42, 43,
44, 45, 46, 46, 47, 48, 49, 50,
51, 52, 53, 54, 55, 56, 57, 58,
59, 60, 61, 62, 63, 64, 65, 66,
67, 68, 69, 70, 71, 72, 73, 74,
75, 76, 76, 77, 78, 79, 80, 81,
82, 83, 84, 85, 86, 87, 88, 89,
91, 93, 95, 96, 98, 100, 101, 102,
104, 106, 108, 110, 112, 114, 116, 118,
122, 124, 126, 128, 130, 132, 134, 136,
138, 140, 143, 145, 148, 151, 154, 157
};
static const uint16_t kAcTable[128] = {
4, 5, 6, 7, 8, 9, 10, 11,
12, 13, 14, 15, 16, 17, 18, 19,
20, 21, 22, 23, 24, 25, 26, 27,
28, 29, 30, 31, 32, 33, 34, 35,
36, 37, 38, 39, 40, 41, 42, 43,
44, 45, 46, 47, 48, 49, 50, 51,
52, 53, 54, 55, 56, 57, 58, 60,
62, 64, 66, 68, 70, 72, 74, 76,
78, 80, 82, 84, 86, 88, 90, 92,
94, 96, 98, 100, 102, 104, 106, 108,
110, 112, 114, 116, 119, 122, 125, 128,
131, 134, 137, 140, 143, 146, 149, 152,
155, 158, 161, 164, 167, 170, 173, 177,
181, 185, 189, 193, 197, 201, 205, 209,
213, 217, 221, 225, 229, 234, 239, 245,
249, 254, 259, 264, 269, 274, 279, 284
};
//------------------------------------------------------------------------------
// Paragraph 9.6
void VP8ParseQuant(VP8Decoder* const dec) {
VP8BitReader* const br = &dec->br_;
const int base_q0 = VP8GetValue(br, 7);
const int dqy1_dc = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
const int dqy2_dc = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
const int dqy2_ac = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
const int dquv_dc = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
const int dquv_ac = VP8Get(br) ? VP8GetSignedValue(br, 4) : 0;
const VP8SegmentHeader* const hdr = &dec->segment_hdr_;
int i;
for (i = 0; i < NUM_MB_SEGMENTS; ++i) {
int q;
if (hdr->use_segment_) {
q = hdr->quantizer_[i];
if (!hdr->absolute_delta_) {
q += base_q0;
}
} else {
if (i > 0) {
dec->dqm_[i] = dec->dqm_[0];
continue;
} else {
q = base_q0;
}
}
{
VP8QuantMatrix* const m = &dec->dqm_[i];
m->y1_mat_[0] = kDcTable[clip(q + dqy1_dc, 127)];
m->y1_mat_[1] = kAcTable[clip(q + 0, 127)];
m->y2_mat_[0] = kDcTable[clip(q + dqy2_dc, 127)] * 2;
// For all x in [0..284], x*155/100 is bitwise equal to (x*101581) >> 16.
// The smallest precision for that is '(x*6349) >> 12' but 16 is a good
// word size.
m->y2_mat_[1] = (kAcTable[clip(q + dqy2_ac, 127)] * 101581) >> 16;
if (m->y2_mat_[1] < 8) m->y2_mat_[1] = 8;
m->uv_mat_[0] = kDcTable[clip(q + dquv_dc, 117)];
m->uv_mat_[1] = kAcTable[clip(q + dquv_ac, 127)];
m->uv_quant_ = q + dquv_ac; // for dithering strength evaluation
}
}
}
//------------------------------------------------------------------------------

532
media/libwebp/src/dec/tree_dec.c Normal file
Просмотреть файл

@ -0,0 +1,532 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Coding trees and probas
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dec/vp8i_dec.h"
#include "src/utils/bit_reader_inl_utils.h"
#if !defined(USE_GENERIC_TREE)
#if !defined(__arm__) && !defined(_M_ARM) && !defined(__aarch64__)
// using a table is ~1-2% slower on ARM. Prefer the coded-tree approach then.
#define USE_GENERIC_TREE 1 // ALTERNATE_CODE
#else
#define USE_GENERIC_TREE 0
#endif
#endif // USE_GENERIC_TREE
#if (USE_GENERIC_TREE == 1)
static const int8_t kYModesIntra4[18] = {
-B_DC_PRED, 1,
-B_TM_PRED, 2,
-B_VE_PRED, 3,
4, 6,
-B_HE_PRED, 5,
-B_RD_PRED, -B_VR_PRED,
-B_LD_PRED, 7,
-B_VL_PRED, 8,
-B_HD_PRED, -B_HU_PRED
};
#endif
//------------------------------------------------------------------------------
// Default probabilities
// Paragraph 13.5
static const uint8_t
CoeffsProba0[NUM_TYPES][NUM_BANDS][NUM_CTX][NUM_PROBAS] = {
{ { { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 }
},
{ { 253, 136, 254, 255, 228, 219, 128, 128, 128, 128, 128 },
{ 189, 129, 242, 255, 227, 213, 255, 219, 128, 128, 128 },
{ 106, 126, 227, 252, 214, 209, 255, 255, 128, 128, 128 }
},
{ { 1, 98, 248, 255, 236, 226, 255, 255, 128, 128, 128 },
{ 181, 133, 238, 254, 221, 234, 255, 154, 128, 128, 128 },
{ 78, 134, 202, 247, 198, 180, 255, 219, 128, 128, 128 },
},
{ { 1, 185, 249, 255, 243, 255, 128, 128, 128, 128, 128 },
{ 184, 150, 247, 255, 236, 224, 128, 128, 128, 128, 128 },
{ 77, 110, 216, 255, 236, 230, 128, 128, 128, 128, 128 },
},
{ { 1, 101, 251, 255, 241, 255, 128, 128, 128, 128, 128 },
{ 170, 139, 241, 252, 236, 209, 255, 255, 128, 128, 128 },
{ 37, 116, 196, 243, 228, 255, 255, 255, 128, 128, 128 }
},
{ { 1, 204, 254, 255, 245, 255, 128, 128, 128, 128, 128 },
{ 207, 160, 250, 255, 238, 128, 128, 128, 128, 128, 128 },
{ 102, 103, 231, 255, 211, 171, 128, 128, 128, 128, 128 }
},
{ { 1, 152, 252, 255, 240, 255, 128, 128, 128, 128, 128 },
{ 177, 135, 243, 255, 234, 225, 128, 128, 128, 128, 128 },
{ 80, 129, 211, 255, 194, 224, 128, 128, 128, 128, 128 }
},
{ { 1, 1, 255, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 246, 1, 255, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 255, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 }
}
},
{ { { 198, 35, 237, 223, 193, 187, 162, 160, 145, 155, 62 },
{ 131, 45, 198, 221, 172, 176, 220, 157, 252, 221, 1 },
{ 68, 47, 146, 208, 149, 167, 221, 162, 255, 223, 128 }
},
{ { 1, 149, 241, 255, 221, 224, 255, 255, 128, 128, 128 },
{ 184, 141, 234, 253, 222, 220, 255, 199, 128, 128, 128 },
{ 81, 99, 181, 242, 176, 190, 249, 202, 255, 255, 128 }
},
{ { 1, 129, 232, 253, 214, 197, 242, 196, 255, 255, 128 },
{ 99, 121, 210, 250, 201, 198, 255, 202, 128, 128, 128 },
{ 23, 91, 163, 242, 170, 187, 247, 210, 255, 255, 128 }
},
{ { 1, 200, 246, 255, 234, 255, 128, 128, 128, 128, 128 },
{ 109, 178, 241, 255, 231, 245, 255, 255, 128, 128, 128 },
{ 44, 130, 201, 253, 205, 192, 255, 255, 128, 128, 128 }
},
{ { 1, 132, 239, 251, 219, 209, 255, 165, 128, 128, 128 },
{ 94, 136, 225, 251, 218, 190, 255, 255, 128, 128, 128 },
{ 22, 100, 174, 245, 186, 161, 255, 199, 128, 128, 128 }
},
{ { 1, 182, 249, 255, 232, 235, 128, 128, 128, 128, 128 },
{ 124, 143, 241, 255, 227, 234, 128, 128, 128, 128, 128 },
{ 35, 77, 181, 251, 193, 211, 255, 205, 128, 128, 128 }
},
{ { 1, 157, 247, 255, 236, 231, 255, 255, 128, 128, 128 },
{ 121, 141, 235, 255, 225, 227, 255, 255, 128, 128, 128 },
{ 45, 99, 188, 251, 195, 217, 255, 224, 128, 128, 128 }
},
{ { 1, 1, 251, 255, 213, 255, 128, 128, 128, 128, 128 },
{ 203, 1, 248, 255, 255, 128, 128, 128, 128, 128, 128 },
{ 137, 1, 177, 255, 224, 255, 128, 128, 128, 128, 128 }
}
},
{ { { 253, 9, 248, 251, 207, 208, 255, 192, 128, 128, 128 },
{ 175, 13, 224, 243, 193, 185, 249, 198, 255, 255, 128 },
{ 73, 17, 171, 221, 161, 179, 236, 167, 255, 234, 128 }
},
{ { 1, 95, 247, 253, 212, 183, 255, 255, 128, 128, 128 },
{ 239, 90, 244, 250, 211, 209, 255, 255, 128, 128, 128 },
{ 155, 77, 195, 248, 188, 195, 255, 255, 128, 128, 128 }
},
{ { 1, 24, 239, 251, 218, 219, 255, 205, 128, 128, 128 },
{ 201, 51, 219, 255, 196, 186, 128, 128, 128, 128, 128 },
{ 69, 46, 190, 239, 201, 218, 255, 228, 128, 128, 128 }
},
{ { 1, 191, 251, 255, 255, 128, 128, 128, 128, 128, 128 },
{ 223, 165, 249, 255, 213, 255, 128, 128, 128, 128, 128 },
{ 141, 124, 248, 255, 255, 128, 128, 128, 128, 128, 128 }
},
{ { 1, 16, 248, 255, 255, 128, 128, 128, 128, 128, 128 },
{ 190, 36, 230, 255, 236, 255, 128, 128, 128, 128, 128 },
{ 149, 1, 255, 128, 128, 128, 128, 128, 128, 128, 128 }
},
{ { 1, 226, 255, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 247, 192, 255, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 240, 128, 255, 128, 128, 128, 128, 128, 128, 128, 128 }
},
{ { 1, 134, 252, 255, 255, 128, 128, 128, 128, 128, 128 },
{ 213, 62, 250, 255, 255, 128, 128, 128, 128, 128, 128 },
{ 55, 93, 255, 128, 128, 128, 128, 128, 128, 128, 128 }
},
{ { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 }
}
},
{ { { 202, 24, 213, 235, 186, 191, 220, 160, 240, 175, 255 },
{ 126, 38, 182, 232, 169, 184, 228, 174, 255, 187, 128 },
{ 61, 46, 138, 219, 151, 178, 240, 170, 255, 216, 128 }
},
{ { 1, 112, 230, 250, 199, 191, 247, 159, 255, 255, 128 },
{ 166, 109, 228, 252, 211, 215, 255, 174, 128, 128, 128 },
{ 39, 77, 162, 232, 172, 180, 245, 178, 255, 255, 128 }
},
{ { 1, 52, 220, 246, 198, 199, 249, 220, 255, 255, 128 },
{ 124, 74, 191, 243, 183, 193, 250, 221, 255, 255, 128 },
{ 24, 71, 130, 219, 154, 170, 243, 182, 255, 255, 128 }
},
{ { 1, 182, 225, 249, 219, 240, 255, 224, 128, 128, 128 },
{ 149, 150, 226, 252, 216, 205, 255, 171, 128, 128, 128 },
{ 28, 108, 170, 242, 183, 194, 254, 223, 255, 255, 128 }
},
{ { 1, 81, 230, 252, 204, 203, 255, 192, 128, 128, 128 },
{ 123, 102, 209, 247, 188, 196, 255, 233, 128, 128, 128 },
{ 20, 95, 153, 243, 164, 173, 255, 203, 128, 128, 128 }
},
{ { 1, 222, 248, 255, 216, 213, 128, 128, 128, 128, 128 },
{ 168, 175, 246, 252, 235, 205, 255, 255, 128, 128, 128 },
{ 47, 116, 215, 255, 211, 212, 255, 255, 128, 128, 128 }
},
{ { 1, 121, 236, 253, 212, 214, 255, 255, 128, 128, 128 },
{ 141, 84, 213, 252, 201, 202, 255, 219, 128, 128, 128 },
{ 42, 80, 160, 240, 162, 185, 255, 205, 128, 128, 128 }
},
{ { 1, 1, 255, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 244, 1, 255, 128, 128, 128, 128, 128, 128, 128, 128 },
{ 238, 1, 255, 128, 128, 128, 128, 128, 128, 128, 128 }
}
}
};
// Paragraph 11.5
static const uint8_t kBModesProba[NUM_BMODES][NUM_BMODES][NUM_BMODES - 1] = {
{ { 231, 120, 48, 89, 115, 113, 120, 152, 112 },
{ 152, 179, 64, 126, 170, 118, 46, 70, 95 },
{ 175, 69, 143, 80, 85, 82, 72, 155, 103 },
{ 56, 58, 10, 171, 218, 189, 17, 13, 152 },
{ 114, 26, 17, 163, 44, 195, 21, 10, 173 },
{ 121, 24, 80, 195, 26, 62, 44, 64, 85 },
{ 144, 71, 10, 38, 171, 213, 144, 34, 26 },
{ 170, 46, 55, 19, 136, 160, 33, 206, 71 },
{ 63, 20, 8, 114, 114, 208, 12, 9, 226 },
{ 81, 40, 11, 96, 182, 84, 29, 16, 36 } },
{ { 134, 183, 89, 137, 98, 101, 106, 165, 148 },
{ 72, 187, 100, 130, 157, 111, 32, 75, 80 },
{ 66, 102, 167, 99, 74, 62, 40, 234, 128 },
{ 41, 53, 9, 178, 241, 141, 26, 8, 107 },
{ 74, 43, 26, 146, 73, 166, 49, 23, 157 },
{ 65, 38, 105, 160, 51, 52, 31, 115, 128 },
{ 104, 79, 12, 27, 217, 255, 87, 17, 7 },
{ 87, 68, 71, 44, 114, 51, 15, 186, 23 },
{ 47, 41, 14, 110, 182, 183, 21, 17, 194 },
{ 66, 45, 25, 102, 197, 189, 23, 18, 22 } },
{ { 88, 88, 147, 150, 42, 46, 45, 196, 205 },
{ 43, 97, 183, 117, 85, 38, 35, 179, 61 },
{ 39, 53, 200, 87, 26, 21, 43, 232, 171 },
{ 56, 34, 51, 104, 114, 102, 29, 93, 77 },
{ 39, 28, 85, 171, 58, 165, 90, 98, 64 },
{ 34, 22, 116, 206, 23, 34, 43, 166, 73 },
{ 107, 54, 32, 26, 51, 1, 81, 43, 31 },
{ 68, 25, 106, 22, 64, 171, 36, 225, 114 },
{ 34, 19, 21, 102, 132, 188, 16, 76, 124 },
{ 62, 18, 78, 95, 85, 57, 50, 48, 51 } },
{ { 193, 101, 35, 159, 215, 111, 89, 46, 111 },
{ 60, 148, 31, 172, 219, 228, 21, 18, 111 },
{ 112, 113, 77, 85, 179, 255, 38, 120, 114 },
{ 40, 42, 1, 196, 245, 209, 10, 25, 109 },
{ 88, 43, 29, 140, 166, 213, 37, 43, 154 },
{ 61, 63, 30, 155, 67, 45, 68, 1, 209 },
{ 100, 80, 8, 43, 154, 1, 51, 26, 71 },
{ 142, 78, 78, 16, 255, 128, 34, 197, 171 },
{ 41, 40, 5, 102, 211, 183, 4, 1, 221 },
{ 51, 50, 17, 168, 209, 192, 23, 25, 82 } },
{ { 138, 31, 36, 171, 27, 166, 38, 44, 229 },
{ 67, 87, 58, 169, 82, 115, 26, 59, 179 },
{ 63, 59, 90, 180, 59, 166, 93, 73, 154 },
{ 40, 40, 21, 116, 143, 209, 34, 39, 175 },
{ 47, 15, 16, 183, 34, 223, 49, 45, 183 },
{ 46, 17, 33, 183, 6, 98, 15, 32, 183 },
{ 57, 46, 22, 24, 128, 1, 54, 17, 37 },
{ 65, 32, 73, 115, 28, 128, 23, 128, 205 },
{ 40, 3, 9, 115, 51, 192, 18, 6, 223 },
{ 87, 37, 9, 115, 59, 77, 64, 21, 47 } },
{ { 104, 55, 44, 218, 9, 54, 53, 130, 226 },
{ 64, 90, 70, 205, 40, 41, 23, 26, 57 },
{ 54, 57, 112, 184, 5, 41, 38, 166, 213 },
{ 30, 34, 26, 133, 152, 116, 10, 32, 134 },
{ 39, 19, 53, 221, 26, 114, 32, 73, 255 },
{ 31, 9, 65, 234, 2, 15, 1, 118, 73 },
{ 75, 32, 12, 51, 192, 255, 160, 43, 51 },
{ 88, 31, 35, 67, 102, 85, 55, 186, 85 },
{ 56, 21, 23, 111, 59, 205, 45, 37, 192 },
{ 55, 38, 70, 124, 73, 102, 1, 34, 98 } },
{ { 125, 98, 42, 88, 104, 85, 117, 175, 82 },
{ 95, 84, 53, 89, 128, 100, 113, 101, 45 },
{ 75, 79, 123, 47, 51, 128, 81, 171, 1 },
{ 57, 17, 5, 71, 102, 57, 53, 41, 49 },
{ 38, 33, 13, 121, 57, 73, 26, 1, 85 },
{ 41, 10, 67, 138, 77, 110, 90, 47, 114 },
{ 115, 21, 2, 10, 102, 255, 166, 23, 6 },
{ 101, 29, 16, 10, 85, 128, 101, 196, 26 },
{ 57, 18, 10, 102, 102, 213, 34, 20, 43 },
{ 117, 20, 15, 36, 163, 128, 68, 1, 26 } },
{ { 102, 61, 71, 37, 34, 53, 31, 243, 192 },
{ 69, 60, 71, 38, 73, 119, 28, 222, 37 },
{ 68, 45, 128, 34, 1, 47, 11, 245, 171 },
{ 62, 17, 19, 70, 146, 85, 55, 62, 70 },
{ 37, 43, 37, 154, 100, 163, 85, 160, 1 },
{ 63, 9, 92, 136, 28, 64, 32, 201, 85 },
{ 75, 15, 9, 9, 64, 255, 184, 119, 16 },
{ 86, 6, 28, 5, 64, 255, 25, 248, 1 },
{ 56, 8, 17, 132, 137, 255, 55, 116, 128 },
{ 58, 15, 20, 82, 135, 57, 26, 121, 40 } },
{ { 164, 50, 31, 137, 154, 133, 25, 35, 218 },
{ 51, 103, 44, 131, 131, 123, 31, 6, 158 },
{ 86, 40, 64, 135, 148, 224, 45, 183, 128 },
{ 22, 26, 17, 131, 240, 154, 14, 1, 209 },
{ 45, 16, 21, 91, 64, 222, 7, 1, 197 },
{ 56, 21, 39, 155, 60, 138, 23, 102, 213 },
{ 83, 12, 13, 54, 192, 255, 68, 47, 28 },
{ 85, 26, 85, 85, 128, 128, 32, 146, 171 },
{ 18, 11, 7, 63, 144, 171, 4, 4, 246 },
{ 35, 27, 10, 146, 174, 171, 12, 26, 128 } },
{ { 190, 80, 35, 99, 180, 80, 126, 54, 45 },
{ 85, 126, 47, 87, 176, 51, 41, 20, 32 },
{ 101, 75, 128, 139, 118, 146, 116, 128, 85 },
{ 56, 41, 15, 176, 236, 85, 37, 9, 62 },
{ 71, 30, 17, 119, 118, 255, 17, 18, 138 },
{ 101, 38, 60, 138, 55, 70, 43, 26, 142 },
{ 146, 36, 19, 30, 171, 255, 97, 27, 20 },
{ 138, 45, 61, 62, 219, 1, 81, 188, 64 },
{ 32, 41, 20, 117, 151, 142, 20, 21, 163 },
{ 112, 19, 12, 61, 195, 128, 48, 4, 24 } }
};
void VP8ResetProba(VP8Proba* const proba) {
memset(proba->segments_, 255u, sizeof(proba->segments_));
// proba->bands_[][] is initialized later
}
static void ParseIntraMode(VP8BitReader* const br,
VP8Decoder* const dec, int mb_x) {
uint8_t* const top = dec->intra_t_ + 4 * mb_x;
uint8_t* const left = dec->intra_l_;
VP8MBData* const block = dec->mb_data_ + mb_x;
// Note: we don't save segment map (yet), as we don't expect
// to decode more than 1 keyframe.
if (dec->segment_hdr_.update_map_) {
// Hardcoded tree parsing
block->segment_ = !VP8GetBit(br, dec->proba_.segments_[0])
? VP8GetBit(br, dec->proba_.segments_[1])
: 2 + VP8GetBit(br, dec->proba_.segments_[2]);
} else {
block->segment_ = 0; // default for intra
}
if (dec->use_skip_proba_) block->skip_ = VP8GetBit(br, dec->skip_p_);
block->is_i4x4_ = !VP8GetBit(br, 145); // decide for B_PRED first
if (!block->is_i4x4_) {
// Hardcoded 16x16 intra-mode decision tree.
const int ymode =
VP8GetBit(br, 156) ? (VP8GetBit(br, 128) ? TM_PRED : H_PRED)
: (VP8GetBit(br, 163) ? V_PRED : DC_PRED);
block->imodes_[0] = ymode;
memset(top, ymode, 4 * sizeof(*top));
memset(left, ymode, 4 * sizeof(*left));
} else {
uint8_t* modes = block->imodes_;
int y;
for (y = 0; y < 4; ++y) {
int ymode = left[y];
int x;
for (x = 0; x < 4; ++x) {
const uint8_t* const prob = kBModesProba[top[x]][ymode];
#if (USE_GENERIC_TREE == 1)
// Generic tree-parsing
int i = kYModesIntra4[VP8GetBit(br, prob[0])];
while (i > 0) {
i = kYModesIntra4[2 * i + VP8GetBit(br, prob[i])];
}
ymode = -i;
#else
// Hardcoded tree parsing
ymode = !VP8GetBit(br, prob[0]) ? B_DC_PRED :
!VP8GetBit(br, prob[1]) ? B_TM_PRED :
!VP8GetBit(br, prob[2]) ? B_VE_PRED :
!VP8GetBit(br, prob[3]) ?
(!VP8GetBit(br, prob[4]) ? B_HE_PRED :
(!VP8GetBit(br, prob[5]) ? B_RD_PRED : B_VR_PRED)) :
(!VP8GetBit(br, prob[6]) ? B_LD_PRED :
(!VP8GetBit(br, prob[7]) ? B_VL_PRED :
(!VP8GetBit(br, prob[8]) ? B_HD_PRED : B_HU_PRED)));
#endif // USE_GENERIC_TREE
top[x] = ymode;
}
memcpy(modes, top, 4 * sizeof(*top));
modes += 4;
left[y] = ymode;
}
}
// Hardcoded UVMode decision tree
block->uvmode_ = !VP8GetBit(br, 142) ? DC_PRED
: !VP8GetBit(br, 114) ? V_PRED
: VP8GetBit(br, 183) ? TM_PRED : H_PRED;
}
int VP8ParseIntraModeRow(VP8BitReader* const br, VP8Decoder* const dec) {
int mb_x;
for (mb_x = 0; mb_x < dec->mb_w_; ++mb_x) {
ParseIntraMode(br, dec, mb_x);
}
return !dec->br_.eof_;
}
//------------------------------------------------------------------------------
// Paragraph 13
static const uint8_t
CoeffsUpdateProba[NUM_TYPES][NUM_BANDS][NUM_CTX][NUM_PROBAS] = {
{ { { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 176, 246, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 223, 241, 252, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 249, 253, 253, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 244, 252, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 234, 254, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 253, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 246, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 239, 253, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 255, 254, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 248, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 251, 255, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 253, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 251, 254, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 255, 254, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 254, 253, 255, 254, 255, 255, 255, 255, 255, 255 },
{ 250, 255, 254, 255, 254, 255, 255, 255, 255, 255, 255 },
{ 254, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
}
},
{ { { 217, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 225, 252, 241, 253, 255, 255, 254, 255, 255, 255, 255 },
{ 234, 250, 241, 250, 253, 255, 253, 254, 255, 255, 255 }
},
{ { 255, 254, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 223, 254, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 238, 253, 254, 254, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 248, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 249, 254, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 253, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 247, 254, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 253, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 252, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 254, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 253, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 254, 253, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 250, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
}
},
{ { { 186, 251, 250, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 234, 251, 244, 254, 255, 255, 255, 255, 255, 255, 255 },
{ 251, 251, 243, 253, 254, 255, 254, 255, 255, 255, 255 }
},
{ { 255, 253, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 236, 253, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 251, 253, 253, 254, 254, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 254, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 254, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 254, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 254, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
}
},
{ { { 248, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 250, 254, 252, 254, 255, 255, 255, 255, 255, 255, 255 },
{ 248, 254, 249, 253, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 253, 253, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 246, 253, 253, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 252, 254, 251, 254, 254, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 254, 252, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 248, 254, 253, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 253, 255, 254, 254, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 251, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 245, 251, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 253, 253, 254, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 251, 253, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 252, 253, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 254, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 252, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 249, 255, 254, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 254, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 253, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 250, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
},
{ { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 254, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }
}
}
};
// Paragraph 9.9
static const uint8_t kBands[16 + 1] = {
0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7,
0 // extra entry as sentinel
};
void VP8ParseProba(VP8BitReader* const br, VP8Decoder* const dec) {
VP8Proba* const proba = &dec->proba_;
int t, b, c, p;
for (t = 0; t < NUM_TYPES; ++t) {
for (b = 0; b < NUM_BANDS; ++b) {
for (c = 0; c < NUM_CTX; ++c) {
for (p = 0; p < NUM_PROBAS; ++p) {
const int v = VP8GetBit(br, CoeffsUpdateProba[t][b][c][p]) ?
VP8GetValue(br, 8) : CoeffsProba0[t][b][c][p];
proba->bands_[t][b].probas_[c][p] = v;
}
}
}
for (b = 0; b < 16 + 1; ++b) {
proba->bands_ptr_[t][b] = &proba->bands_[t][kBands[b]];
}
}
dec->use_skip_proba_ = VP8Get(br);
if (dec->use_skip_proba_) {
dec->skip_p_ = VP8GetValue(br, 8);
}
}

721
media/libwebp/src/dec/vp8_dec.c Normal file
Просмотреть файл

@ -0,0 +1,721 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// main entry for the decoder
//
// Author: Skal (pascal.massimino@gmail.com)
#include <stdlib.h>
#include "src/dec/alphai_dec.h"
#include "src/dec/vp8i_dec.h"
#include "src/dec/vp8li_dec.h"
#include "src/dec/webpi_dec.h"
#include "src/utils/bit_reader_inl_utils.h"
#include "src/utils/utils.h"
//------------------------------------------------------------------------------
int WebPGetDecoderVersion(void) {
return (DEC_MAJ_VERSION << 16) | (DEC_MIN_VERSION << 8) | DEC_REV_VERSION;
}
//------------------------------------------------------------------------------
// Signature and pointer-to-function for GetCoeffs() variants below.
typedef int (*GetCoeffsFunc)(VP8BitReader* const br,
const VP8BandProbas* const prob[],
int ctx, const quant_t dq, int n, int16_t* out);
static volatile GetCoeffsFunc GetCoeffs = NULL;
static void InitGetCoeffs(void);
//------------------------------------------------------------------------------
// VP8Decoder
static void SetOk(VP8Decoder* const dec) {
dec->status_ = VP8_STATUS_OK;
dec->error_msg_ = "OK";
}
int VP8InitIoInternal(VP8Io* const io, int version) {
if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_DECODER_ABI_VERSION)) {
return 0; // mismatch error
}
if (io != NULL) {
memset(io, 0, sizeof(*io));
}
return 1;
}
VP8Decoder* VP8New(void) {
VP8Decoder* const dec = (VP8Decoder*)WebPSafeCalloc(1ULL, sizeof(*dec));
if (dec != NULL) {
SetOk(dec);
WebPGetWorkerInterface()->Init(&dec->worker_);
dec->ready_ = 0;
dec->num_parts_minus_one_ = 0;
InitGetCoeffs();
}
return dec;
}
VP8StatusCode VP8Status(VP8Decoder* const dec) {
if (!dec) return VP8_STATUS_INVALID_PARAM;
return dec->status_;
}
const char* VP8StatusMessage(VP8Decoder* const dec) {
if (dec == NULL) return "no object";
if (!dec->error_msg_) return "OK";
return dec->error_msg_;
}
void VP8Delete(VP8Decoder* const dec) {
if (dec != NULL) {
VP8Clear(dec);
WebPSafeFree(dec);
}
}
int VP8SetError(VP8Decoder* const dec,
VP8StatusCode error, const char* const msg) {
// The oldest error reported takes precedence over the new one.
if (dec->status_ == VP8_STATUS_OK) {
dec->status_ = error;
dec->error_msg_ = msg;
dec->ready_ = 0;
}
return 0;
}
//------------------------------------------------------------------------------
int VP8CheckSignature(const uint8_t* const data, size_t data_size) {
return (data_size >= 3 &&
data[0] == 0x9d && data[1] == 0x01 && data[2] == 0x2a);
}
int VP8GetInfo(const uint8_t* data, size_t data_size, size_t chunk_size,
int* const width, int* const height) {
if (data == NULL || data_size < VP8_FRAME_HEADER_SIZE) {
return 0; // not enough data
}
// check signature
if (!VP8CheckSignature(data + 3, data_size - 3)) {
return 0; // Wrong signature.
} else {
const uint32_t bits = data[0] | (data[1] << 8) | (data[2] << 16);
const int key_frame = !(bits & 1);
const int w = ((data[7] << 8) | data[6]) & 0x3fff;
const int h = ((data[9] << 8) | data[8]) & 0x3fff;
if (!key_frame) { // Not a keyframe.
return 0;
}
if (((bits >> 1) & 7) > 3) {
return 0; // unknown profile
}
if (!((bits >> 4) & 1)) {
return 0; // first frame is invisible!
}
if (((bits >> 5)) >= chunk_size) { // partition_length
return 0; // inconsistent size information.
}
if (w == 0 || h == 0) {
return 0; // We don't support both width and height to be zero.
}
if (width) {
*width = w;
}
if (height) {
*height = h;
}
return 1;
}
}
//------------------------------------------------------------------------------
// Header parsing
static void ResetSegmentHeader(VP8SegmentHeader* const hdr) {
assert(hdr != NULL);
hdr->use_segment_ = 0;
hdr->update_map_ = 0;
hdr->absolute_delta_ = 1;
memset(hdr->quantizer_, 0, sizeof(hdr->quantizer_));
memset(hdr->filter_strength_, 0, sizeof(hdr->filter_strength_));
}
// Paragraph 9.3
static int ParseSegmentHeader(VP8BitReader* br,
VP8SegmentHeader* hdr, VP8Proba* proba) {
assert(br != NULL);
assert(hdr != NULL);
hdr->use_segment_ = VP8Get(br);
if (hdr->use_segment_) {
hdr->update_map_ = VP8Get(br);
if (VP8Get(br)) { // update data
int s;
hdr->absolute_delta_ = VP8Get(br);
for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
hdr->quantizer_[s] = VP8Get(br) ? VP8GetSignedValue(br, 7) : 0;
}
for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
hdr->filter_strength_[s] = VP8Get(br) ? VP8GetSignedValue(br, 6) : 0;
}
}
if (hdr->update_map_) {
int s;
for (s = 0; s < MB_FEATURE_TREE_PROBS; ++s) {
proba->segments_[s] = VP8Get(br) ? VP8GetValue(br, 8) : 255u;
}
}
} else {
hdr->update_map_ = 0;
}
return !br->eof_;
}
// Paragraph 9.5
// This function returns VP8_STATUS_SUSPENDED if we don't have all the
// necessary data in 'buf'.
// This case is not necessarily an error (for incremental decoding).
// Still, no bitreader is ever initialized to make it possible to read
// unavailable memory.
// If we don't even have the partitions' sizes, than VP8_STATUS_NOT_ENOUGH_DATA
// is returned, and this is an unrecoverable error.
// If the partitions were positioned ok, VP8_STATUS_OK is returned.
static VP8StatusCode ParsePartitions(VP8Decoder* const dec,
const uint8_t* buf, size_t size) {
VP8BitReader* const br = &dec->br_;
const uint8_t* sz = buf;
const uint8_t* buf_end = buf + size;
const uint8_t* part_start;
size_t size_left = size;
size_t last_part;
size_t p;
dec->num_parts_minus_one_ = (1 << VP8GetValue(br, 2)) - 1;
last_part = dec->num_parts_minus_one_;
if (size < 3 * last_part) {
// we can't even read the sizes with sz[]! That's a failure.
return VP8_STATUS_NOT_ENOUGH_DATA;
}
part_start = buf + last_part * 3;
size_left -= last_part * 3;
for (p = 0; p < last_part; ++p) {
size_t psize = sz[0] | (sz[1] << 8) | (sz[2] << 16);
if (psize > size_left) psize = size_left;
VP8InitBitReader(dec->parts_ + p, part_start, psize);
part_start += psize;
size_left -= psize;
sz += 3;
}
VP8InitBitReader(dec->parts_ + last_part, part_start, size_left);
return (part_start < buf_end) ? VP8_STATUS_OK :
VP8_STATUS_SUSPENDED; // Init is ok, but there's not enough data
}
// Paragraph 9.4
static int ParseFilterHeader(VP8BitReader* br, VP8Decoder* const dec) {
VP8FilterHeader* const hdr = &dec->filter_hdr_;
hdr->simple_ = VP8Get(br);
hdr->level_ = VP8GetValue(br, 6);
hdr->sharpness_ = VP8GetValue(br, 3);
hdr->use_lf_delta_ = VP8Get(br);
if (hdr->use_lf_delta_) {
if (VP8Get(br)) { // update lf-delta?
int i;
for (i = 0; i < NUM_REF_LF_DELTAS; ++i) {
if (VP8Get(br)) {
hdr->ref_lf_delta_[i] = VP8GetSignedValue(br, 6);
}
}
for (i = 0; i < NUM_MODE_LF_DELTAS; ++i) {
if (VP8Get(br)) {
hdr->mode_lf_delta_[i] = VP8GetSignedValue(br, 6);
}
}
}
}
dec->filter_type_ = (hdr->level_ == 0) ? 0 : hdr->simple_ ? 1 : 2;
return !br->eof_;
}
// Topmost call
int VP8GetHeaders(VP8Decoder* const dec, VP8Io* const io) {
const uint8_t* buf;
size_t buf_size;
VP8FrameHeader* frm_hdr;
VP8PictureHeader* pic_hdr;
VP8BitReader* br;
VP8StatusCode status;
if (dec == NULL) {
return 0;
}
SetOk(dec);
if (io == NULL) {
return VP8SetError(dec, VP8_STATUS_INVALID_PARAM,
"null VP8Io passed to VP8GetHeaders()");
}
buf = io->data;
buf_size = io->data_size;
if (buf_size < 4) {
return VP8SetError(dec, VP8_STATUS_NOT_ENOUGH_DATA,
"Truncated header.");
}
// Paragraph 9.1
{
const uint32_t bits = buf[0] | (buf[1] << 8) | (buf[2] << 16);
frm_hdr = &dec->frm_hdr_;
frm_hdr->key_frame_ = !(bits & 1);
frm_hdr->profile_ = (bits >> 1) & 7;
frm_hdr->show_ = (bits >> 4) & 1;
frm_hdr->partition_length_ = (bits >> 5);
if (frm_hdr->profile_ > 3) {
return VP8SetError(dec, VP8_STATUS_BITSTREAM_ERROR,
"Incorrect keyframe parameters.");
}
if (!frm_hdr->show_) {
return VP8SetError(dec, VP8_STATUS_UNSUPPORTED_FEATURE,
"Frame not displayable.");
}
buf += 3;
buf_size -= 3;
}
pic_hdr = &dec->pic_hdr_;
if (frm_hdr->key_frame_) {
// Paragraph 9.2
if (buf_size < 7) {
return VP8SetError(dec, VP8_STATUS_NOT_ENOUGH_DATA,
"cannot parse picture header");
}
if (!VP8CheckSignature(buf, buf_size)) {
return VP8SetError(dec, VP8_STATUS_BITSTREAM_ERROR,
"Bad code word");
}
pic_hdr->width_ = ((buf[4] << 8) | buf[3]) & 0x3fff;
pic_hdr->xscale_ = buf[4] >> 6; // ratio: 1, 5/4 5/3 or 2
pic_hdr->height_ = ((buf[6] << 8) | buf[5]) & 0x3fff;
pic_hdr->yscale_ = buf[6] >> 6;
buf += 7;
buf_size -= 7;
dec->mb_w_ = (pic_hdr->width_ + 15) >> 4;
dec->mb_h_ = (pic_hdr->height_ + 15) >> 4;
// Setup default output area (can be later modified during io->setup())
io->width = pic_hdr->width_;
io->height = pic_hdr->height_;
// IMPORTANT! use some sane dimensions in crop_* and scaled_* fields.
// So they can be used interchangeably without always testing for
// 'use_cropping'.
io->use_cropping = 0;
io->crop_top = 0;
io->crop_left = 0;
io->crop_right = io->width;
io->crop_bottom = io->height;
io->use_scaling = 0;
io->scaled_width = io->width;
io->scaled_height = io->height;
io->mb_w = io->width; // sanity check
io->mb_h = io->height; // ditto
VP8ResetProba(&dec->proba_);
ResetSegmentHeader(&dec->segment_hdr_);
}
// Check if we have all the partition #0 available, and initialize dec->br_
// to read this partition (and this partition only).
if (frm_hdr->partition_length_ > buf_size) {
return VP8SetError(dec, VP8_STATUS_NOT_ENOUGH_DATA,
"bad partition length");
}
br = &dec->br_;
VP8InitBitReader(br, buf, frm_hdr->partition_length_);
buf += frm_hdr->partition_length_;
buf_size -= frm_hdr->partition_length_;
if (frm_hdr->key_frame_) {
pic_hdr->colorspace_ = VP8Get(br);
pic_hdr->clamp_type_ = VP8Get(br);
}
if (!ParseSegmentHeader(br, &dec->segment_hdr_, &dec->proba_)) {
return VP8SetError(dec, VP8_STATUS_BITSTREAM_ERROR,
"cannot parse segment header");
}
// Filter specs
if (!ParseFilterHeader(br, dec)) {
return VP8SetError(dec, VP8_STATUS_BITSTREAM_ERROR,
"cannot parse filter header");
}
status = ParsePartitions(dec, buf, buf_size);
if (status != VP8_STATUS_OK) {
return VP8SetError(dec, status, "cannot parse partitions");
}
// quantizer change
VP8ParseQuant(dec);
// Frame buffer marking
if (!frm_hdr->key_frame_) {
return VP8SetError(dec, VP8_STATUS_UNSUPPORTED_FEATURE,
"Not a key frame.");
}
VP8Get(br); // ignore the value of update_proba_
VP8ParseProba(br, dec);
// sanitized state
dec->ready_ = 1;
return 1;
}
//------------------------------------------------------------------------------
// Residual decoding (Paragraph 13.2 / 13.3)
static const uint8_t kCat3[] = { 173, 148, 140, 0 };
static const uint8_t kCat4[] = { 176, 155, 140, 135, 0 };
static const uint8_t kCat5[] = { 180, 157, 141, 134, 130, 0 };
static const uint8_t kCat6[] =
{ 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0 };
static const uint8_t* const kCat3456[] = { kCat3, kCat4, kCat5, kCat6 };
static const uint8_t kZigzag[16] = {
0, 1, 4, 8, 5, 2, 3, 6, 9, 12, 13, 10, 7, 11, 14, 15
};
// See section 13-2: http://tools.ietf.org/html/rfc6386#section-13.2
static int GetLargeValue(VP8BitReader* const br, const uint8_t* const p) {
int v;
if (!VP8GetBit(br, p[3])) {
if (!VP8GetBit(br, p[4])) {
v = 2;
} else {
v = 3 + VP8GetBit(br, p[5]);
}
} else {
if (!VP8GetBit(br, p[6])) {
if (!VP8GetBit(br, p[7])) {
v = 5 + VP8GetBit(br, 159);
} else {
v = 7 + 2 * VP8GetBit(br, 165);
v += VP8GetBit(br, 145);
}
} else {
const uint8_t* tab;
const int bit1 = VP8GetBit(br, p[8]);
const int bit0 = VP8GetBit(br, p[9 + bit1]);
const int cat = 2 * bit1 + bit0;
v = 0;
for (tab = kCat3456[cat]; *tab; ++tab) {
v += v + VP8GetBit(br, *tab);
}
v += 3 + (8 << cat);
}
}
return v;
}
// Returns the position of the last non-zero coeff plus one
static int GetCoeffsFast(VP8BitReader* const br,
const VP8BandProbas* const prob[],
int ctx, const quant_t dq, int n, int16_t* out) {
const uint8_t* p = prob[n]->probas_[ctx];
for (; n < 16; ++n) {
if (!VP8GetBit(br, p[0])) {
return n; // previous coeff was last non-zero coeff
}
while (!VP8GetBit(br, p[1])) { // sequence of zero coeffs
p = prob[++n]->probas_[0];
if (n == 16) return 16;
}
{ // non zero coeff
const VP8ProbaArray* const p_ctx = &prob[n + 1]->probas_[0];
int v;
if (!VP8GetBit(br, p[2])) {
v = 1;
p = p_ctx[1];
} else {
v = GetLargeValue(br, p);
p = p_ctx[2];
}
out[kZigzag[n]] = VP8GetSigned(br, v) * dq[n > 0];
}
}
return 16;
}
// This version of GetCoeffs() uses VP8GetBitAlt() which is an alternate version
// of VP8GetBitAlt() targeting specific platforms.
static int GetCoeffsAlt(VP8BitReader* const br,
const VP8BandProbas* const prob[],
int ctx, const quant_t dq, int n, int16_t* out) {
const uint8_t* p = prob[n]->probas_[ctx];
for (; n < 16; ++n) {
if (!VP8GetBitAlt(br, p[0])) {
return n; // previous coeff was last non-zero coeff
}
while (!VP8GetBitAlt(br, p[1])) { // sequence of zero coeffs
p = prob[++n]->probas_[0];
if (n == 16) return 16;
}
{ // non zero coeff
const VP8ProbaArray* const p_ctx = &prob[n + 1]->probas_[0];
int v;
if (!VP8GetBitAlt(br, p[2])) {
v = 1;
p = p_ctx[1];
} else {
v = GetLargeValue(br, p);
p = p_ctx[2];
}
out[kZigzag[n]] = VP8GetSigned(br, v) * dq[n > 0];
}
}
return 16;
}
static WEBP_TSAN_IGNORE_FUNCTION void InitGetCoeffs(void) {
if (GetCoeffs == NULL) {
if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
GetCoeffs = GetCoeffsAlt;
} else {
GetCoeffs = GetCoeffsFast;
}
}
}
static WEBP_INLINE uint32_t NzCodeBits(uint32_t nz_coeffs, int nz, int dc_nz) {
nz_coeffs <<= 2;
nz_coeffs |= (nz > 3) ? 3 : (nz > 1) ? 2 : dc_nz;
return nz_coeffs;
}
static int ParseResiduals(VP8Decoder* const dec,
VP8MB* const mb, VP8BitReader* const token_br) {
const VP8BandProbas* (* const bands)[16 + 1] = dec->proba_.bands_ptr_;
const VP8BandProbas* const * ac_proba;
VP8MBData* const block = dec->mb_data_ + dec->mb_x_;
const VP8QuantMatrix* const q = &dec->dqm_[block->segment_];
int16_t* dst = block->coeffs_;
VP8MB* const left_mb = dec->mb_info_ - 1;
uint8_t tnz, lnz;
uint32_t non_zero_y = 0;
uint32_t non_zero_uv = 0;
int x, y, ch;
uint32_t out_t_nz, out_l_nz;
int first;
memset(dst, 0, 384 * sizeof(*dst));
if (!block->is_i4x4_) { // parse DC
int16_t dc[16] = { 0 };
const int ctx = mb->nz_dc_ + left_mb->nz_dc_;
const int nz = GetCoeffs(token_br, bands[1], ctx, q->y2_mat_, 0, dc);
mb->nz_dc_ = left_mb->nz_dc_ = (nz > 0);
if (nz > 1) { // more than just the DC -> perform the full transform
VP8TransformWHT(dc, dst);
} else { // only DC is non-zero -> inlined simplified transform
int i;
const int dc0 = (dc[0] + 3) >> 3;
for (i = 0; i < 16 * 16; i += 16) dst[i] = dc0;
}
first = 1;
ac_proba = bands[0];
} else {
first = 0;
ac_proba = bands[3];
}
tnz = mb->nz_ & 0x0f;
lnz = left_mb->nz_ & 0x0f;
for (y = 0; y < 4; ++y) {
int l = lnz & 1;
uint32_t nz_coeffs = 0;
for (x = 0; x < 4; ++x) {
const int ctx = l + (tnz & 1);
const int nz = GetCoeffs(token_br, ac_proba, ctx, q->y1_mat_, first, dst);
l = (nz > first);
tnz = (tnz >> 1) | (l << 7);
nz_coeffs = NzCodeBits(nz_coeffs, nz, dst[0] != 0);
dst += 16;
}
tnz >>= 4;
lnz = (lnz >> 1) | (l << 7);
non_zero_y = (non_zero_y << 8) | nz_coeffs;
}
out_t_nz = tnz;
out_l_nz = lnz >> 4;
for (ch = 0; ch < 4; ch += 2) {
uint32_t nz_coeffs = 0;
tnz = mb->nz_ >> (4 + ch);
lnz = left_mb->nz_ >> (4 + ch);
for (y = 0; y < 2; ++y) {
int l = lnz & 1;
for (x = 0; x < 2; ++x) {
const int ctx = l + (tnz & 1);
const int nz = GetCoeffs(token_br, bands[2], ctx, q->uv_mat_, 0, dst);
l = (nz > 0);
tnz = (tnz >> 1) | (l << 3);
nz_coeffs = NzCodeBits(nz_coeffs, nz, dst[0] != 0);
dst += 16;
}
tnz >>= 2;
lnz = (lnz >> 1) | (l << 5);
}
// Note: we don't really need the per-4x4 details for U/V blocks.
non_zero_uv |= nz_coeffs << (4 * ch);
out_t_nz |= (tnz << 4) << ch;
out_l_nz |= (lnz & 0xf0) << ch;
}
mb->nz_ = out_t_nz;
left_mb->nz_ = out_l_nz;
block->non_zero_y_ = non_zero_y;
block->non_zero_uv_ = non_zero_uv;
// We look at the mode-code of each block and check if some blocks have less
// than three non-zero coeffs (code < 2). This is to avoid dithering flat and
// empty blocks.
block->dither_ = (non_zero_uv & 0xaaaa) ? 0 : q->dither_;
return !(non_zero_y | non_zero_uv); // will be used for further optimization
}
//------------------------------------------------------------------------------
// Main loop
int VP8DecodeMB(VP8Decoder* const dec, VP8BitReader* const token_br) {
VP8MB* const left = dec->mb_info_ - 1;
VP8MB* const mb = dec->mb_info_ + dec->mb_x_;
VP8MBData* const block = dec->mb_data_ + dec->mb_x_;
int skip = dec->use_skip_proba_ ? block->skip_ : 0;
if (!skip) {
skip = ParseResiduals(dec, mb, token_br);
} else {
left->nz_ = mb->nz_ = 0;
if (!block->is_i4x4_) {
left->nz_dc_ = mb->nz_dc_ = 0;
}
block->non_zero_y_ = 0;
block->non_zero_uv_ = 0;
block->dither_ = 0;
}
if (dec->filter_type_ > 0) { // store filter info
VP8FInfo* const finfo = dec->f_info_ + dec->mb_x_;
*finfo = dec->fstrengths_[block->segment_][block->is_i4x4_];
finfo->f_inner_ |= !skip;
}
return !token_br->eof_;
}
void VP8InitScanline(VP8Decoder* const dec) {
VP8MB* const left = dec->mb_info_ - 1;
left->nz_ = 0;
left->nz_dc_ = 0;
memset(dec->intra_l_, B_DC_PRED, sizeof(dec->intra_l_));
dec->mb_x_ = 0;
}
static int ParseFrame(VP8Decoder* const dec, VP8Io* io) {
for (dec->mb_y_ = 0; dec->mb_y_ < dec->br_mb_y_; ++dec->mb_y_) {
// Parse bitstream for this row.
VP8BitReader* const token_br =
&dec->parts_[dec->mb_y_ & dec->num_parts_minus_one_];
if (!VP8ParseIntraModeRow(&dec->br_, dec)) {
return VP8SetError(dec, VP8_STATUS_NOT_ENOUGH_DATA,
"Premature end-of-partition0 encountered.");
}
for (; dec->mb_x_ < dec->mb_w_; ++dec->mb_x_) {
if (!VP8DecodeMB(dec, token_br)) {
return VP8SetError(dec, VP8_STATUS_NOT_ENOUGH_DATA,
"Premature end-of-file encountered.");
}
}
VP8InitScanline(dec); // Prepare for next scanline
// Reconstruct, filter and emit the row.
if (!VP8ProcessRow(dec, io)) {
return VP8SetError(dec, VP8_STATUS_USER_ABORT, "Output aborted.");
}
}
if (dec->mt_method_ > 0) {
if (!WebPGetWorkerInterface()->Sync(&dec->worker_)) return 0;
}
return 1;
}
// Main entry point
int VP8Decode(VP8Decoder* const dec, VP8Io* const io) {
int ok = 0;
if (dec == NULL) {
return 0;
}
if (io == NULL) {
return VP8SetError(dec, VP8_STATUS_INVALID_PARAM,
"NULL VP8Io parameter in VP8Decode().");
}
if (!dec->ready_) {
if (!VP8GetHeaders(dec, io)) {
return 0;
}
}
assert(dec->ready_);
// Finish setting up the decoding parameter. Will call io->setup().
ok = (VP8EnterCritical(dec, io) == VP8_STATUS_OK);
if (ok) { // good to go.
// Will allocate memory and prepare everything.
if (ok) ok = VP8InitFrame(dec, io);
// Main decoding loop
if (ok) ok = ParseFrame(dec, io);
// Exit.
ok &= VP8ExitCritical(dec, io);
}
if (!ok) {
VP8Clear(dec);
return 0;
}
dec->ready_ = 0;
return ok;
}
void VP8Clear(VP8Decoder* const dec) {
if (dec == NULL) {
return;
}
WebPGetWorkerInterface()->End(&dec->worker_);
WebPDeallocateAlphaMemory(dec);
WebPSafeFree(dec->mem_);
dec->mem_ = NULL;
dec->mem_size_ = 0;
memset(&dec->br_, 0, sizeof(dec->br_));
dec->ready_ = 0;
}
//------------------------------------------------------------------------------

185
media/libwebp/src/dec/vp8_dec.h Normal file
Просмотреть файл

@ -0,0 +1,185 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Low-level API for VP8 decoder
//
// Author: Skal (pascal.massimino@gmail.com)
#ifndef WEBP_DEC_VP8_DEC_H_
#define WEBP_DEC_VP8_DEC_H_
#include "src/webp/decode.h"
#ifdef __cplusplus
extern "C" {
#endif
//------------------------------------------------------------------------------
// Lower-level API
//
// These functions provide fine-grained control of the decoding process.
// The call flow should resemble:
//
// VP8Io io;
// VP8InitIo(&io);
// io.data = data;
// io.data_size = size;
// /* customize io's functions (setup()/put()/teardown()) if needed. */
//
// VP8Decoder* dec = VP8New();
// int ok = VP8Decode(dec, &io);
// if (!ok) printf("Error: %s\n", VP8StatusMessage(dec));
// VP8Delete(dec);
// return ok;
// Input / Output
typedef struct VP8Io VP8Io;
typedef int (*VP8IoPutHook)(const VP8Io* io);
typedef int (*VP8IoSetupHook)(VP8Io* io);
typedef void (*VP8IoTeardownHook)(const VP8Io* io);
struct VP8Io {
// set by VP8GetHeaders()
int width, height; // picture dimensions, in pixels (invariable).
// These are the original, uncropped dimensions.
// The actual area passed to put() is stored
// in mb_w / mb_h fields.
// set before calling put()
int mb_y; // position of the current rows (in pixels)
int mb_w; // number of columns in the sample
int mb_h; // number of rows in the sample
const uint8_t* y, *u, *v; // rows to copy (in yuv420 format)
int y_stride; // row stride for luma
int uv_stride; // row stride for chroma
void* opaque; // user data
// called when fresh samples are available. Currently, samples are in
// YUV420 format, and can be up to width x 24 in size (depending on the
// in-loop filtering level, e.g.). Should return false in case of error
// or abort request. The actual size of the area to update is mb_w x mb_h
// in size, taking cropping into account.
VP8IoPutHook put;
// called just before starting to decode the blocks.
// Must return false in case of setup error, true otherwise. If false is
// returned, teardown() will NOT be called. But if the setup succeeded
// and true is returned, then teardown() will always be called afterward.
VP8IoSetupHook setup;
// Called just after block decoding is finished (or when an error occurred
// during put()). Is NOT called if setup() failed.
VP8IoTeardownHook teardown;
// this is a recommendation for the user-side yuv->rgb converter. This flag
// is set when calling setup() hook and can be overwritten by it. It then
// can be taken into consideration during the put() method.
int fancy_upsampling;
// Input buffer.
size_t data_size;
const uint8_t* data;
// If true, in-loop filtering will not be performed even if present in the
// bitstream. Switching off filtering may speed up decoding at the expense
// of more visible blocking. Note that output will also be non-compliant
// with the VP8 specifications.
int bypass_filtering;
// Cropping parameters.
int use_cropping;
int crop_left, crop_right, crop_top, crop_bottom;
// Scaling parameters.
int use_scaling;
int scaled_width, scaled_height;
// If non NULL, pointer to the alpha data (if present) corresponding to the
// start of the current row (That is: it is pre-offset by mb_y and takes
// cropping into account).
const uint8_t* a;
};
// Internal, version-checked, entry point
int VP8InitIoInternal(VP8Io* const, int);
// Set the custom IO function pointers and user-data. The setter for IO hooks
// should be called before initiating incremental decoding. Returns true if
// WebPIDecoder object is successfully modified, false otherwise.
int WebPISetIOHooks(WebPIDecoder* const idec,
VP8IoPutHook put,
VP8IoSetupHook setup,
VP8IoTeardownHook teardown,
void* user_data);
// Main decoding object. This is an opaque structure.
typedef struct VP8Decoder VP8Decoder;
// Create a new decoder object.
VP8Decoder* VP8New(void);
// Must be called to make sure 'io' is initialized properly.
// Returns false in case of version mismatch. Upon such failure, no other
// decoding function should be called (VP8Decode, VP8GetHeaders, ...)
static WEBP_INLINE int VP8InitIo(VP8Io* const io) {
return VP8InitIoInternal(io, WEBP_DECODER_ABI_VERSION);
}
// Decode the VP8 frame header. Returns true if ok.
// Note: 'io->data' must be pointing to the start of the VP8 frame header.
int VP8GetHeaders(VP8Decoder* const dec, VP8Io* const io);
// Decode a picture. Will call VP8GetHeaders() if it wasn't done already.
// Returns false in case of error.
int VP8Decode(VP8Decoder* const dec, VP8Io* const io);
// Return current status of the decoder:
VP8StatusCode VP8Status(VP8Decoder* const dec);
// return readable string corresponding to the last status.
const char* VP8StatusMessage(VP8Decoder* const dec);
// Resets the decoder in its initial state, reclaiming memory.
// Not a mandatory call between calls to VP8Decode().
void VP8Clear(VP8Decoder* const dec);
// Destroy the decoder object.
void VP8Delete(VP8Decoder* const dec);
//------------------------------------------------------------------------------
// Miscellaneous VP8/VP8L bitstream probing functions.
// Returns true if the next 3 bytes in data contain the VP8 signature.
WEBP_EXTERN int VP8CheckSignature(const uint8_t* const data, size_t data_size);
// Validates the VP8 data-header and retrieves basic header information viz
// width and height. Returns 0 in case of formatting error. *width/*height
// can be passed NULL.
WEBP_EXTERN int VP8GetInfo(
const uint8_t* data,
size_t data_size, // data available so far
size_t chunk_size, // total data size expected in the chunk
int* const width, int* const height);
// Returns true if the next byte(s) in data is a VP8L signature.
WEBP_EXTERN int VP8LCheckSignature(const uint8_t* const data, size_t size);
// Validates the VP8L data-header and retrieves basic header information viz
// width, height and alpha. Returns 0 in case of formatting error.
// width/height/has_alpha can be passed NULL.
WEBP_EXTERN int VP8LGetInfo(
const uint8_t* data, size_t data_size, // data available so far
int* const width, int* const height, int* const has_alpha);
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* WEBP_DEC_VP8_DEC_H_ */

319
media/libwebp/src/dec/vp8i_dec.h Normal file
Просмотреть файл

@ -0,0 +1,319 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// VP8 decoder: internal header.
//
// Author: Skal (pascal.massimino@gmail.com)
#ifndef WEBP_DEC_VP8I_DEC_H_
#define WEBP_DEC_VP8I_DEC_H_
#include <string.h> // for memcpy()
#include "src/dec/common_dec.h"
#include "src/dec/vp8li_dec.h"
#include "src/utils/bit_reader_utils.h"
#include "src/utils/random_utils.h"
#include "src/utils/thread_utils.h"
#include "src/dsp/dsp.h"
#ifdef __cplusplus
extern "C" {
#endif
//------------------------------------------------------------------------------
// Various defines and enums
// version numbers
#define DEC_MAJ_VERSION 1
#define DEC_MIN_VERSION 0
#define DEC_REV_VERSION 0
// YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).
// Constraints are: We need to store one 16x16 block of luma samples (y),
// and two 8x8 chroma blocks (u/v). These are better be 16-bytes aligned,
// in order to be SIMD-friendly. We also need to store the top, left and
// top-left samples (from previously decoded blocks), along with four
// extra top-right samples for luma (intra4x4 prediction only).
// One possible layout is, using 32 * (17 + 9) bytes:
//
// .+------ <- only 1 pixel high
// .|yyyyt.
// .|yyyyt.
// .|yyyyt.
// .|yyyy..
// .+--.+-- <- only 1 pixel high
// .|uu.|vv
// .|uu.|vv
//
// Every character is a 4x4 block, with legend:
// '.' = unused
// 'y' = y-samples 'u' = u-samples 'v' = u-samples
// '|' = left sample, '-' = top sample, '+' = top-left sample
// 't' = extra top-right sample for 4x4 modes
#define YUV_SIZE (BPS * 17 + BPS * 9)
#define Y_OFF (BPS * 1 + 8)
#define U_OFF (Y_OFF + BPS * 16 + BPS)
#define V_OFF (U_OFF + 16)
// minimal width under which lossy multi-threading is always disabled
#define MIN_WIDTH_FOR_THREADS 512
//------------------------------------------------------------------------------
// Headers
typedef struct {
uint8_t key_frame_;
uint8_t profile_;
uint8_t show_;
uint32_t partition_length_;
} VP8FrameHeader;
typedef struct {
uint16_t width_;
uint16_t height_;
uint8_t xscale_;
uint8_t yscale_;
uint8_t colorspace_; // 0 = YCbCr
uint8_t clamp_type_;
} VP8PictureHeader;
// segment features
typedef struct {
int use_segment_;
int update_map_; // whether to update the segment map or not
int absolute_delta_; // absolute or delta values for quantizer and filter
int8_t quantizer_[NUM_MB_SEGMENTS]; // quantization changes
int8_t filter_strength_[NUM_MB_SEGMENTS]; // filter strength for segments
} VP8SegmentHeader;
// probas associated to one of the contexts
typedef uint8_t VP8ProbaArray[NUM_PROBAS];
typedef struct { // all the probas associated to one band
VP8ProbaArray probas_[NUM_CTX];
} VP8BandProbas;
// Struct collecting all frame-persistent probabilities.
typedef struct {
uint8_t segments_[MB_FEATURE_TREE_PROBS];
// Type: 0:Intra16-AC 1:Intra16-DC 2:Chroma 3:Intra4
VP8BandProbas bands_[NUM_TYPES][NUM_BANDS];
const VP8BandProbas* bands_ptr_[NUM_TYPES][16 + 1];
} VP8Proba;
// Filter parameters
typedef struct {
int simple_; // 0=complex, 1=simple
int level_; // [0..63]
int sharpness_; // [0..7]
int use_lf_delta_;
int ref_lf_delta_[NUM_REF_LF_DELTAS];
int mode_lf_delta_[NUM_MODE_LF_DELTAS];
} VP8FilterHeader;
//------------------------------------------------------------------------------
// Informations about the macroblocks.
typedef struct { // filter specs
uint8_t f_limit_; // filter limit in [3..189], or 0 if no filtering
uint8_t f_ilevel_; // inner limit in [1..63]
uint8_t f_inner_; // do inner filtering?
uint8_t hev_thresh_; // high edge variance threshold in [0..2]
} VP8FInfo;
typedef struct { // Top/Left Contexts used for syntax-parsing
uint8_t nz_; // non-zero AC/DC coeffs (4bit for luma + 4bit for chroma)
uint8_t nz_dc_; // non-zero DC coeff (1bit)
} VP8MB;
// Dequantization matrices
typedef int quant_t[2]; // [DC / AC]. Can be 'uint16_t[2]' too (~slower).
typedef struct {
quant_t y1_mat_, y2_mat_, uv_mat_;
int uv_quant_; // U/V quantizer value
int dither_; // dithering amplitude (0 = off, max=255)
} VP8QuantMatrix;
// Data needed to reconstruct a macroblock
typedef struct {
int16_t coeffs_[384]; // 384 coeffs = (16+4+4) * 4*4
uint8_t is_i4x4_; // true if intra4x4
uint8_t imodes_[16]; // one 16x16 mode (#0) or sixteen 4x4 modes
uint8_t uvmode_; // chroma prediction mode
// bit-wise info about the content of each sub-4x4 blocks (in decoding order).
// Each of the 4x4 blocks for y/u/v is associated with a 2b code according to:
// code=0 -> no coefficient
// code=1 -> only DC
// code=2 -> first three coefficients are non-zero
// code=3 -> more than three coefficients are non-zero
// This allows to call specialized transform functions.
uint32_t non_zero_y_;
uint32_t non_zero_uv_;
uint8_t dither_; // local dithering strength (deduced from non_zero_*)
uint8_t skip_;
uint8_t segment_;
} VP8MBData;
// Persistent information needed by the parallel processing
typedef struct {
int id_; // cache row to process (in [0..2])
int mb_y_; // macroblock position of the row
int filter_row_; // true if row-filtering is needed
VP8FInfo* f_info_; // filter strengths (swapped with dec->f_info_)
VP8MBData* mb_data_; // reconstruction data (swapped with dec->mb_data_)
VP8Io io_; // copy of the VP8Io to pass to put()
} VP8ThreadContext;
// Saved top samples, per macroblock. Fits into a cache-line.
typedef struct {
uint8_t y[16], u[8], v[8];
} VP8TopSamples;
//------------------------------------------------------------------------------
// VP8Decoder: the main opaque structure handed over to user
struct VP8Decoder {
VP8StatusCode status_;
int ready_; // true if ready to decode a picture with VP8Decode()
const char* error_msg_; // set when status_ is not OK.
// Main data source
VP8BitReader br_;
// headers
VP8FrameHeader frm_hdr_;
VP8PictureHeader pic_hdr_;
VP8FilterHeader filter_hdr_;
VP8SegmentHeader segment_hdr_;
// Worker
WebPWorker worker_;
int mt_method_; // multi-thread method: 0=off, 1=[parse+recon][filter]
// 2=[parse][recon+filter]
int cache_id_; // current cache row
int num_caches_; // number of cached rows of 16 pixels (1, 2 or 3)
VP8ThreadContext thread_ctx_; // Thread context
// dimension, in macroblock units.
int mb_w_, mb_h_;
// Macroblock to process/filter, depending on cropping and filter_type.
int tl_mb_x_, tl_mb_y_; // top-left MB that must be in-loop filtered
int br_mb_x_, br_mb_y_; // last bottom-right MB that must be decoded
// number of partitions minus one.
uint32_t num_parts_minus_one_;
// per-partition boolean decoders.
VP8BitReader parts_[MAX_NUM_PARTITIONS];
// Dithering strength, deduced from decoding options
int dither_; // whether to use dithering or not
VP8Random dithering_rg_; // random generator for dithering
// dequantization (one set of DC/AC dequant factor per segment)
VP8QuantMatrix dqm_[NUM_MB_SEGMENTS];
// probabilities
VP8Proba proba_;
int use_skip_proba_;
uint8_t skip_p_;
// Boundary data cache and persistent buffers.
uint8_t* intra_t_; // top intra modes values: 4 * mb_w_
uint8_t intra_l_[4]; // left intra modes values
VP8TopSamples* yuv_t_; // top y/u/v samples
VP8MB* mb_info_; // contextual macroblock info (mb_w_ + 1)
VP8FInfo* f_info_; // filter strength info
uint8_t* yuv_b_; // main block for Y/U/V (size = YUV_SIZE)
uint8_t* cache_y_; // macroblock row for storing unfiltered samples
uint8_t* cache_u_;
uint8_t* cache_v_;
int cache_y_stride_;
int cache_uv_stride_;
// main memory chunk for the above data. Persistent.
void* mem_;
size_t mem_size_;
// Per macroblock non-persistent infos.
int mb_x_, mb_y_; // current position, in macroblock units
VP8MBData* mb_data_; // parsed reconstruction data
// Filtering side-info
int filter_type_; // 0=off, 1=simple, 2=complex
VP8FInfo fstrengths_[NUM_MB_SEGMENTS][2]; // precalculated per-segment/type
// Alpha
struct ALPHDecoder* alph_dec_; // alpha-plane decoder object
const uint8_t* alpha_data_; // compressed alpha data (if present)
size_t alpha_data_size_;
int is_alpha_decoded_; // true if alpha_data_ is decoded in alpha_plane_
uint8_t* alpha_plane_mem_; // memory allocated for alpha_plane_
uint8_t* alpha_plane_; // output. Persistent, contains the whole data.
const uint8_t* alpha_prev_line_; // last decoded alpha row (or NULL)
int alpha_dithering_; // derived from decoding options (0=off, 100=full)
};
//------------------------------------------------------------------------------
// internal functions. Not public.
// in vp8.c
int VP8SetError(VP8Decoder* const dec,
VP8StatusCode error, const char* const msg);
// in tree.c
void VP8ResetProba(VP8Proba* const proba);
void VP8ParseProba(VP8BitReader* const br, VP8Decoder* const dec);
// parses one row of intra mode data in partition 0, returns !eof
int VP8ParseIntraModeRow(VP8BitReader* const br, VP8Decoder* const dec);
// in quant.c
void VP8ParseQuant(VP8Decoder* const dec);
// in frame.c
int VP8InitFrame(VP8Decoder* const dec, VP8Io* const io);
// Call io->setup() and finish setting up scan parameters.
// After this call returns, one must always call VP8ExitCritical() with the
// same parameters. Both functions should be used in pair. Returns VP8_STATUS_OK
// if ok, otherwise sets and returns the error status on *dec.
VP8StatusCode VP8EnterCritical(VP8Decoder* const dec, VP8Io* const io);
// Must always be called in pair with VP8EnterCritical().
// Returns false in case of error.
int VP8ExitCritical(VP8Decoder* const dec, VP8Io* const io);
// Return the multi-threading method to use (0=off), depending
// on options and bitstream size. Only for lossy decoding.
int VP8GetThreadMethod(const WebPDecoderOptions* const options,
const WebPHeaderStructure* const headers,
int width, int height);
// Initialize dithering post-process if needed.
void VP8InitDithering(const WebPDecoderOptions* const options,
VP8Decoder* const dec);
// Process the last decoded row (filtering + output).
int VP8ProcessRow(VP8Decoder* const dec, VP8Io* const io);
// To be called at the start of a new scanline, to initialize predictors.
void VP8InitScanline(VP8Decoder* const dec);
// Decode one macroblock. Returns false if there is not enough data.
int VP8DecodeMB(VP8Decoder* const dec, VP8BitReader* const token_br);
// in alpha.c
const uint8_t* VP8DecompressAlphaRows(VP8Decoder* const dec,
const VP8Io* const io,
int row, int num_rows);
//------------------------------------------------------------------------------
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* WEBP_DEC_VP8I_DEC_H_ */

1689
media/libwebp/src/dec/vp8l_dec.c Normal file
Просмотреть файл

Разница между файлами не показана из-за своего большого размера Загрузить разницу

135
media/libwebp/src/dec/vp8li_dec.h Normal file
Просмотреть файл

@ -0,0 +1,135 @@
// Copyright 2012 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Lossless decoder: internal header.
//
// Author: Skal (pascal.massimino@gmail.com)
// Vikas Arora(vikaas.arora@gmail.com)
#ifndef WEBP_DEC_VP8LI_DEC_H_
#define WEBP_DEC_VP8LI_DEC_H_
#include <string.h> // for memcpy()
#include "src/dec/webpi_dec.h"
#include "src/utils/bit_reader_utils.h"
#include "src/utils/color_cache_utils.h"
#include "src/utils/huffman_utils.h"
#ifdef __cplusplus
extern "C" {
#endif
typedef enum {
READ_DATA = 0,
READ_HDR = 1,
READ_DIM = 2
} VP8LDecodeState;
typedef struct VP8LTransform VP8LTransform;
struct VP8LTransform {
VP8LImageTransformType type_; // transform type.
int bits_; // subsampling bits defining transform window.
int xsize_; // transform window X index.
int ysize_; // transform window Y index.
uint32_t *data_; // transform data.
};
typedef struct {
int color_cache_size_;
VP8LColorCache color_cache_;
VP8LColorCache saved_color_cache_; // for incremental
int huffman_mask_;
int huffman_subsample_bits_;
int huffman_xsize_;
uint32_t *huffman_image_;
int num_htree_groups_;
HTreeGroup *htree_groups_;
HuffmanCode *huffman_tables_;
} VP8LMetadata;
typedef struct VP8LDecoder VP8LDecoder;
struct VP8LDecoder {
VP8StatusCode status_;
VP8LDecodeState state_;
VP8Io *io_;
const WebPDecBuffer *output_; // shortcut to io->opaque->output
uint32_t *pixels_; // Internal data: either uint8_t* for alpha
// or uint32_t* for BGRA.
uint32_t *argb_cache_; // Scratch buffer for temporary BGRA storage.
VP8LBitReader br_;
int incremental_; // if true, incremental decoding is expected
VP8LBitReader saved_br_; // note: could be local variables too
int saved_last_pixel_;
int width_;
int height_;
int last_row_; // last input row decoded so far.
int last_pixel_; // last pixel decoded so far. However, it may
// not be transformed, scaled and
// color-converted yet.
int last_out_row_; // last row output so far.
VP8LMetadata hdr_;
int next_transform_;
VP8LTransform transforms_[NUM_TRANSFORMS];
// or'd bitset storing the transforms types.
uint32_t transforms_seen_;
uint8_t *rescaler_memory; // Working memory for rescaling work.
WebPRescaler *rescaler; // Common rescaler for all channels.
};
//------------------------------------------------------------------------------
// internal functions. Not public.
struct ALPHDecoder; // Defined in dec/alphai.h.
// in vp8l.c
// Decodes image header for alpha data stored using lossless compression.
// Returns false in case of error.
int VP8LDecodeAlphaHeader(struct ALPHDecoder* const alph_dec,
const uint8_t* const data, size_t data_size);
// Decodes *at least* 'last_row' rows of alpha. If some of the initial rows are
// already decoded in previous call(s), it will resume decoding from where it
// was paused.
// Returns false in case of bitstream error.
int VP8LDecodeAlphaImageStream(struct ALPHDecoder* const alph_dec,
int last_row);
// Allocates and initialize a new lossless decoder instance.
VP8LDecoder* VP8LNew(void);
// Decodes the image header. Returns false in case of error.
int VP8LDecodeHeader(VP8LDecoder* const dec, VP8Io* const io);
// Decodes an image. It's required to decode the lossless header before calling
// this function. Returns false in case of error, with updated dec->status_.
int VP8LDecodeImage(VP8LDecoder* const dec);
// Resets the decoder in its initial state, reclaiming memory.
// Preserves the dec->status_ value.
void VP8LClear(VP8LDecoder* const dec);
// Clears and deallocate a lossless decoder instance.
void VP8LDelete(VP8LDecoder* const dec);
//------------------------------------------------------------------------------
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* WEBP_DEC_VP8LI_DEC_H_ */

845
media/libwebp/src/dec/webp_dec.c Normal file
Просмотреть файл

@ -0,0 +1,845 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Main decoding functions for WEBP images.
//
// Author: Skal (pascal.massimino@gmail.com)
#include <stdlib.h>
#include "src/dec/vp8i_dec.h"
#include "src/dec/vp8li_dec.h"
#include "src/dec/webpi_dec.h"
#include "src/utils/utils.h"
#include "src/webp/mux_types.h" // ALPHA_FLAG
//------------------------------------------------------------------------------
// RIFF layout is:
// Offset tag
// 0...3 "RIFF" 4-byte tag
// 4...7 size of image data (including metadata) starting at offset 8
// 8...11 "WEBP" our form-type signature
// The RIFF container (12 bytes) is followed by appropriate chunks:
// 12..15 "VP8 ": 4-bytes tags, signaling the use of VP8 video format
// 16..19 size of the raw VP8 image data, starting at offset 20
// 20.... the VP8 bytes
// Or,
// 12..15 "VP8L": 4-bytes tags, signaling the use of VP8L lossless format
// 16..19 size of the raw VP8L image data, starting at offset 20
// 20.... the VP8L bytes
// Or,
// 12..15 "VP8X": 4-bytes tags, describing the extended-VP8 chunk.
// 16..19 size of the VP8X chunk starting at offset 20.
// 20..23 VP8X flags bit-map corresponding to the chunk-types present.
// 24..26 Width of the Canvas Image.
// 27..29 Height of the Canvas Image.
// There can be extra chunks after the "VP8X" chunk (ICCP, ANMF, VP8, VP8L,
// XMP, EXIF ...)
// All sizes are in little-endian order.
// Note: chunk data size must be padded to multiple of 2 when written.
// Validates the RIFF container (if detected) and skips over it.
// If a RIFF container is detected, returns:
// VP8_STATUS_BITSTREAM_ERROR for invalid header,
// VP8_STATUS_NOT_ENOUGH_DATA for truncated data if have_all_data is true,
// and VP8_STATUS_OK otherwise.
// In case there are not enough bytes (partial RIFF container), return 0 for
// *riff_size. Else return the RIFF size extracted from the header.
static VP8StatusCode ParseRIFF(const uint8_t** const data,
size_t* const data_size, int have_all_data,
size_t* const riff_size) {
assert(data != NULL);
assert(data_size != NULL);
assert(riff_size != NULL);
*riff_size = 0; // Default: no RIFF present.
if (*data_size >= RIFF_HEADER_SIZE && !memcmp(*data, "RIFF", TAG_SIZE)) {
if (memcmp(*data + 8, "WEBP", TAG_SIZE)) {
return VP8_STATUS_BITSTREAM_ERROR; // Wrong image file signature.
} else {
const uint32_t size = GetLE32(*data + TAG_SIZE);
// Check that we have at least one chunk (i.e "WEBP" + "VP8?nnnn").
if (size < TAG_SIZE + CHUNK_HEADER_SIZE) {
return VP8_STATUS_BITSTREAM_ERROR;
}
if (size > MAX_CHUNK_PAYLOAD) {
return VP8_STATUS_BITSTREAM_ERROR;
}
if (have_all_data && (size > *data_size - CHUNK_HEADER_SIZE)) {
return VP8_STATUS_NOT_ENOUGH_DATA; // Truncated bitstream.
}
// We have a RIFF container. Skip it.
*riff_size = size;
*data += RIFF_HEADER_SIZE;
*data_size -= RIFF_HEADER_SIZE;
}
}
return VP8_STATUS_OK;
}
// Validates the VP8X header and skips over it.
// Returns VP8_STATUS_BITSTREAM_ERROR for invalid VP8X header,
// VP8_STATUS_NOT_ENOUGH_DATA in case of insufficient data, and
// VP8_STATUS_OK otherwise.
// If a VP8X chunk is found, found_vp8x is set to true and *width_ptr,
// *height_ptr and *flags_ptr are set to the corresponding values extracted
// from the VP8X chunk.
static VP8StatusCode ParseVP8X(const uint8_t** const data,
size_t* const data_size,
int* const found_vp8x,
int* const width_ptr, int* const height_ptr,
uint32_t* const flags_ptr) {
const uint32_t vp8x_size = CHUNK_HEADER_SIZE + VP8X_CHUNK_SIZE;
assert(data != NULL);
assert(data_size != NULL);
assert(found_vp8x != NULL);
*found_vp8x = 0;
if (*data_size < CHUNK_HEADER_SIZE) {
return VP8_STATUS_NOT_ENOUGH_DATA; // Insufficient data.
}
if (!memcmp(*data, "VP8X", TAG_SIZE)) {
int width, height;
uint32_t flags;
const uint32_t chunk_size = GetLE32(*data + TAG_SIZE);
if (chunk_size != VP8X_CHUNK_SIZE) {
return VP8_STATUS_BITSTREAM_ERROR; // Wrong chunk size.
}
// Verify if enough data is available to validate the VP8X chunk.
if (*data_size < vp8x_size) {
return VP8_STATUS_NOT_ENOUGH_DATA; // Insufficient data.
}
flags = GetLE32(*data + 8);
width = 1 + GetLE24(*data + 12);
height = 1 + GetLE24(*data + 15);
if (width * (uint64_t)height >= MAX_IMAGE_AREA) {
return VP8_STATUS_BITSTREAM_ERROR; // image is too large
}
if (flags_ptr != NULL) *flags_ptr = flags;
if (width_ptr != NULL) *width_ptr = width;
if (height_ptr != NULL) *height_ptr = height;
// Skip over VP8X header bytes.
*data += vp8x_size;
*data_size -= vp8x_size;
*found_vp8x = 1;
}
return VP8_STATUS_OK;
}
// Skips to the next VP8/VP8L chunk header in the data given the size of the
// RIFF chunk 'riff_size'.
// Returns VP8_STATUS_BITSTREAM_ERROR if any invalid chunk size is encountered,
// VP8_STATUS_NOT_ENOUGH_DATA in case of insufficient data, and
// VP8_STATUS_OK otherwise.
// If an alpha chunk is found, *alpha_data and *alpha_size are set
// appropriately.
static VP8StatusCode ParseOptionalChunks(const uint8_t** const data,
size_t* const data_size,
size_t const riff_size,
const uint8_t** const alpha_data,
size_t* const alpha_size) {
const uint8_t* buf;
size_t buf_size;
uint32_t total_size = TAG_SIZE + // "WEBP".
CHUNK_HEADER_SIZE + // "VP8Xnnnn".
VP8X_CHUNK_SIZE; // data.
assert(data != NULL);
assert(data_size != NULL);
buf = *data;
buf_size = *data_size;
assert(alpha_data != NULL);
assert(alpha_size != NULL);
*alpha_data = NULL;
*alpha_size = 0;
while (1) {
uint32_t chunk_size;
uint32_t disk_chunk_size; // chunk_size with padding
*data = buf;
*data_size = buf_size;
if (buf_size < CHUNK_HEADER_SIZE) { // Insufficient data.
return VP8_STATUS_NOT_ENOUGH_DATA;
}
chunk_size = GetLE32(buf + TAG_SIZE);
if (chunk_size > MAX_CHUNK_PAYLOAD) {
return VP8_STATUS_BITSTREAM_ERROR; // Not a valid chunk size.
}
// For odd-sized chunk-payload, there's one byte padding at the end.
disk_chunk_size = (CHUNK_HEADER_SIZE + chunk_size + 1) & ~1;
total_size += disk_chunk_size;
// Check that total bytes skipped so far does not exceed riff_size.
if (riff_size > 0 && (total_size > riff_size)) {
return VP8_STATUS_BITSTREAM_ERROR; // Not a valid chunk size.
}
// Start of a (possibly incomplete) VP8/VP8L chunk implies that we have
// parsed all the optional chunks.
// Note: This check must occur before the check 'buf_size < disk_chunk_size'
// below to allow incomplete VP8/VP8L chunks.
if (!memcmp(buf, "VP8 ", TAG_SIZE) ||
!memcmp(buf, "VP8L", TAG_SIZE)) {
return VP8_STATUS_OK;
}
if (buf_size < disk_chunk_size) { // Insufficient data.
return VP8_STATUS_NOT_ENOUGH_DATA;
}
if (!memcmp(buf, "ALPH", TAG_SIZE)) { // A valid ALPH header.
*alpha_data = buf + CHUNK_HEADER_SIZE;
*alpha_size = chunk_size;
}
// We have a full and valid chunk; skip it.
buf += disk_chunk_size;
buf_size -= disk_chunk_size;
}
}
// Validates the VP8/VP8L Header ("VP8 nnnn" or "VP8L nnnn") and skips over it.
// Returns VP8_STATUS_BITSTREAM_ERROR for invalid (chunk larger than
// riff_size) VP8/VP8L header,
// VP8_STATUS_NOT_ENOUGH_DATA in case of insufficient data, and
// VP8_STATUS_OK otherwise.
// If a VP8/VP8L chunk is found, *chunk_size is set to the total number of bytes
// extracted from the VP8/VP8L chunk header.
// The flag '*is_lossless' is set to 1 in case of VP8L chunk / raw VP8L data.
static VP8StatusCode ParseVP8Header(const uint8_t** const data_ptr,
size_t* const data_size, int have_all_data,
size_t riff_size, size_t* const chunk_size,
int* const is_lossless) {
const uint8_t* const data = *data_ptr;
const int is_vp8 = !memcmp(data, "VP8 ", TAG_SIZE);
const int is_vp8l = !memcmp(data, "VP8L", TAG_SIZE);
const uint32_t minimal_size =
TAG_SIZE + CHUNK_HEADER_SIZE; // "WEBP" + "VP8 nnnn" OR
// "WEBP" + "VP8Lnnnn"
assert(data != NULL);
assert(data_size != NULL);
assert(chunk_size != NULL);
assert(is_lossless != NULL);
if (*data_size < CHUNK_HEADER_SIZE) {
return VP8_STATUS_NOT_ENOUGH_DATA; // Insufficient data.
}
if (is_vp8 || is_vp8l) {
// Bitstream contains VP8/VP8L header.
const uint32_t size = GetLE32(data + TAG_SIZE);
if ((riff_size >= minimal_size) && (size > riff_size - minimal_size)) {
return VP8_STATUS_BITSTREAM_ERROR; // Inconsistent size information.
}
if (have_all_data && (size > *data_size - CHUNK_HEADER_SIZE)) {
return VP8_STATUS_NOT_ENOUGH_DATA; // Truncated bitstream.
}
// Skip over CHUNK_HEADER_SIZE bytes from VP8/VP8L Header.
*chunk_size = size;
*data_ptr += CHUNK_HEADER_SIZE;
*data_size -= CHUNK_HEADER_SIZE;
*is_lossless = is_vp8l;
} else {
// Raw VP8/VP8L bitstream (no header).
*is_lossless = VP8LCheckSignature(data, *data_size);
*chunk_size = *data_size;
}
return VP8_STATUS_OK;
}
//------------------------------------------------------------------------------
// Fetch '*width', '*height', '*has_alpha' and fill out 'headers' based on
// 'data'. All the output parameters may be NULL. If 'headers' is NULL only the
// minimal amount will be read to fetch the remaining parameters.
// If 'headers' is non-NULL this function will attempt to locate both alpha
// data (with or without a VP8X chunk) and the bitstream chunk (VP8/VP8L).
// Note: The following chunk sequences (before the raw VP8/VP8L data) are
// considered valid by this function:
// RIFF + VP8(L)
// RIFF + VP8X + (optional chunks) + VP8(L)
// ALPH + VP8 <-- Not a valid WebP format: only allowed for internal purpose.
// VP8(L) <-- Not a valid WebP format: only allowed for internal purpose.
static VP8StatusCode ParseHeadersInternal(const uint8_t* data,
size_t data_size,
int* const width,
int* const height,
int* const has_alpha,
int* const has_animation,
int* const format,
WebPHeaderStructure* const headers) {
int canvas_width = 0;
int canvas_height = 0;
int image_width = 0;
int image_height = 0;
int found_riff = 0;
int found_vp8x = 0;
int animation_present = 0;
const int have_all_data = (headers != NULL) ? headers->have_all_data : 0;
VP8StatusCode status;
WebPHeaderStructure hdrs;
if (data == NULL || data_size < RIFF_HEADER_SIZE) {
return VP8_STATUS_NOT_ENOUGH_DATA;
}
memset(&hdrs, 0, sizeof(hdrs));
hdrs.data = data;
hdrs.data_size = data_size;
// Skip over RIFF header.
status = ParseRIFF(&data, &data_size, have_all_data, &hdrs.riff_size);
if (status != VP8_STATUS_OK) {
return status; // Wrong RIFF header / insufficient data.
}
found_riff = (hdrs.riff_size > 0);
// Skip over VP8X.
{
uint32_t flags = 0;
status = ParseVP8X(&data, &data_size, &found_vp8x,
&canvas_width, &canvas_height, &flags);
if (status != VP8_STATUS_OK) {
return status; // Wrong VP8X / insufficient data.
}
animation_present = !!(flags & ANIMATION_FLAG);
if (!found_riff && found_vp8x) {
// Note: This restriction may be removed in the future, if it becomes
// necessary to send VP8X chunk to the decoder.
return VP8_STATUS_BITSTREAM_ERROR;
}
if (has_alpha != NULL) *has_alpha = !!(flags & ALPHA_FLAG);
if (has_animation != NULL) *has_animation = animation_present;
if (format != NULL) *format = 0; // default = undefined
image_width = canvas_width;
image_height = canvas_height;
if (found_vp8x && animation_present && headers == NULL) {
status = VP8_STATUS_OK;
goto ReturnWidthHeight; // Just return features from VP8X header.
}
}
if (data_size < TAG_SIZE) {
status = VP8_STATUS_NOT_ENOUGH_DATA;
goto ReturnWidthHeight;
}
// Skip over optional chunks if data started with "RIFF + VP8X" or "ALPH".
if ((found_riff && found_vp8x) ||
(!found_riff && !found_vp8x && !memcmp(data, "ALPH", TAG_SIZE))) {
status = ParseOptionalChunks(&data, &data_size, hdrs.riff_size,
&hdrs.alpha_data, &hdrs.alpha_data_size);
if (status != VP8_STATUS_OK) {
goto ReturnWidthHeight; // Invalid chunk size / insufficient data.
}
}
// Skip over VP8/VP8L header.
status = ParseVP8Header(&data, &data_size, have_all_data, hdrs.riff_size,
&hdrs.compressed_size, &hdrs.is_lossless);
if (status != VP8_STATUS_OK) {
goto ReturnWidthHeight; // Wrong VP8/VP8L chunk-header / insufficient data.
}
if (hdrs.compressed_size > MAX_CHUNK_PAYLOAD) {
return VP8_STATUS_BITSTREAM_ERROR;
}
if (format != NULL && !animation_present) {
*format = hdrs.is_lossless ? 2 : 1;
}
if (!hdrs.is_lossless) {
if (data_size < VP8_FRAME_HEADER_SIZE) {
status = VP8_STATUS_NOT_ENOUGH_DATA;
goto ReturnWidthHeight;
}
// Validates raw VP8 data.
if (!VP8GetInfo(data, data_size, (uint32_t)hdrs.compressed_size,
&image_width, &image_height)) {
return VP8_STATUS_BITSTREAM_ERROR;
}
} else {
if (data_size < VP8L_FRAME_HEADER_SIZE) {
status = VP8_STATUS_NOT_ENOUGH_DATA;
goto ReturnWidthHeight;
}
// Validates raw VP8L data.
if (!VP8LGetInfo(data, data_size, &image_width, &image_height, has_alpha)) {
return VP8_STATUS_BITSTREAM_ERROR;
}
}
// Validates image size coherency.
if (found_vp8x) {
if (canvas_width != image_width || canvas_height != image_height) {
return VP8_STATUS_BITSTREAM_ERROR;
}
}
if (headers != NULL) {
*headers = hdrs;
headers->offset = data - headers->data;
assert((uint64_t)(data - headers->data) < MAX_CHUNK_PAYLOAD);
assert(headers->offset == headers->data_size - data_size);
}
ReturnWidthHeight:
if (status == VP8_STATUS_OK ||
(status == VP8_STATUS_NOT_ENOUGH_DATA && found_vp8x && headers == NULL)) {
if (has_alpha != NULL) {
// If the data did not contain a VP8X/VP8L chunk the only definitive way
// to set this is by looking for alpha data (from an ALPH chunk).
*has_alpha |= (hdrs.alpha_data != NULL);
}
if (width != NULL) *width = image_width;
if (height != NULL) *height = image_height;
return VP8_STATUS_OK;
} else {
return status;
}
}
VP8StatusCode WebPParseHeaders(WebPHeaderStructure* const headers) {
// status is marked volatile as a workaround for a clang-3.8 (aarch64) bug
volatile VP8StatusCode status;
int has_animation = 0;
assert(headers != NULL);
// fill out headers, ignore width/height/has_alpha.
status = ParseHeadersInternal(headers->data, headers->data_size,
NULL, NULL, NULL, &has_animation,
NULL, headers);
if (status == VP8_STATUS_OK || status == VP8_STATUS_NOT_ENOUGH_DATA) {
// The WebPDemux API + libwebp can be used to decode individual
// uncomposited frames or the WebPAnimDecoder can be used to fully
// reconstruct them (see webp/demux.h).
if (has_animation) {
status = VP8_STATUS_UNSUPPORTED_FEATURE;
}
}
return status;
}
//------------------------------------------------------------------------------
// WebPDecParams
void WebPResetDecParams(WebPDecParams* const params) {
if (params != NULL) {
memset(params, 0, sizeof(*params));
}
}
//------------------------------------------------------------------------------
// "Into" decoding variants
// Main flow
static VP8StatusCode DecodeInto(const uint8_t* const data, size_t data_size,
WebPDecParams* const params) {
VP8StatusCode status;
VP8Io io;
WebPHeaderStructure headers;
headers.data = data;
headers.data_size = data_size;
headers.have_all_data = 1;
status = WebPParseHeaders(&headers); // Process Pre-VP8 chunks.
if (status != VP8_STATUS_OK) {
return status;
}
assert(params != NULL);
VP8InitIo(&io);
io.data = headers.data + headers.offset;
io.data_size = headers.data_size - headers.offset;
WebPInitCustomIo(params, &io); // Plug the I/O functions.
if (!headers.is_lossless) {
VP8Decoder* const dec = VP8New();
if (dec == NULL) {
return VP8_STATUS_OUT_OF_MEMORY;
}
dec->alpha_data_ = headers.alpha_data;
dec->alpha_data_size_ = headers.alpha_data_size;
// Decode bitstream header, update io->width/io->height.
if (!VP8GetHeaders(dec, &io)) {
status = dec->status_; // An error occurred. Grab error status.
} else {
// Allocate/check output buffers.
status = WebPAllocateDecBuffer(io.width, io.height, params->options,
params->output);
if (status == VP8_STATUS_OK) { // Decode
// This change must be done before calling VP8Decode()
dec->mt_method_ = VP8GetThreadMethod(params->options, &headers,
io.width, io.height);
VP8InitDithering(params->options, dec);
if (!VP8Decode(dec, &io)) {
status = dec->status_;
}
}
}
VP8Delete(dec);
} else {
VP8LDecoder* const dec = VP8LNew();
if (dec == NULL) {
return VP8_STATUS_OUT_OF_MEMORY;
}
if (!VP8LDecodeHeader(dec, &io)) {
status = dec->status_; // An error occurred. Grab error status.
} else {
// Allocate/check output buffers.
status = WebPAllocateDecBuffer(io.width, io.height, params->options,
params->output);
if (status == VP8_STATUS_OK) { // Decode
if (!VP8LDecodeImage(dec)) {
status = dec->status_;
}
}
}
VP8LDelete(dec);
}
if (status != VP8_STATUS_OK) {
WebPFreeDecBuffer(params->output);
} else {
if (params->options != NULL && params->options->flip) {
// This restores the original stride values if options->flip was used
// during the call to WebPAllocateDecBuffer above.
status = WebPFlipBuffer(params->output);
}
}
return status;
}
// Helpers
static uint8_t* DecodeIntoRGBABuffer(WEBP_CSP_MODE colorspace,
const uint8_t* const data,
size_t data_size,
uint8_t* const rgba,
int stride, size_t size) {
WebPDecParams params;
WebPDecBuffer buf;
if (rgba == NULL) {
return NULL;
}
WebPInitDecBuffer(&buf);
WebPResetDecParams(&params);
params.output = &buf;
buf.colorspace = colorspace;
buf.u.RGBA.rgba = rgba;
buf.u.RGBA.stride = stride;
buf.u.RGBA.size = size;
buf.is_external_memory = 1;
if (DecodeInto(data, data_size, &params) != VP8_STATUS_OK) {
return NULL;
}
return rgba;
}
uint8_t* WebPDecodeRGBInto(const uint8_t* data, size_t data_size,
uint8_t* output, size_t size, int stride) {
return DecodeIntoRGBABuffer(MODE_RGB, data, data_size, output, stride, size);
}
uint8_t* WebPDecodeRGBAInto(const uint8_t* data, size_t data_size,
uint8_t* output, size_t size, int stride) {
return DecodeIntoRGBABuffer(MODE_RGBA, data, data_size, output, stride, size);
}
uint8_t* WebPDecodeARGBInto(const uint8_t* data, size_t data_size,
uint8_t* output, size_t size, int stride) {
return DecodeIntoRGBABuffer(MODE_ARGB, data, data_size, output, stride, size);
}
uint8_t* WebPDecodeBGRInto(const uint8_t* data, size_t data_size,
uint8_t* output, size_t size, int stride) {
return DecodeIntoRGBABuffer(MODE_BGR, data, data_size, output, stride, size);
}
uint8_t* WebPDecodeBGRAInto(const uint8_t* data, size_t data_size,
uint8_t* output, size_t size, int stride) {
return DecodeIntoRGBABuffer(MODE_BGRA, data, data_size, output, stride, size);
}
uint8_t* WebPDecodeYUVInto(const uint8_t* data, size_t data_size,
uint8_t* luma, size_t luma_size, int luma_stride,
uint8_t* u, size_t u_size, int u_stride,
uint8_t* v, size_t v_size, int v_stride) {
WebPDecParams params;
WebPDecBuffer output;
if (luma == NULL) return NULL;
WebPInitDecBuffer(&output);
WebPResetDecParams(&params);
params.output = &output;
output.colorspace = MODE_YUV;
output.u.YUVA.y = luma;
output.u.YUVA.y_stride = luma_stride;
output.u.YUVA.y_size = luma_size;
output.u.YUVA.u = u;
output.u.YUVA.u_stride = u_stride;
output.u.YUVA.u_size = u_size;
output.u.YUVA.v = v;
output.u.YUVA.v_stride = v_stride;
output.u.YUVA.v_size = v_size;
output.is_external_memory = 1;
if (DecodeInto(data, data_size, &params) != VP8_STATUS_OK) {
return NULL;
}
return luma;
}
//------------------------------------------------------------------------------
static uint8_t* Decode(WEBP_CSP_MODE mode, const uint8_t* const data,
size_t data_size, int* const width, int* const height,
WebPDecBuffer* const keep_info) {
WebPDecParams params;
WebPDecBuffer output;
WebPInitDecBuffer(&output);
WebPResetDecParams(&params);
params.output = &output;
output.colorspace = mode;
// Retrieve (and report back) the required dimensions from bitstream.
if (!WebPGetInfo(data, data_size, &output.width, &output.height)) {
return NULL;
}
if (width != NULL) *width = output.width;
if (height != NULL) *height = output.height;
// Decode
if (DecodeInto(data, data_size, &params) != VP8_STATUS_OK) {
return NULL;
}
if (keep_info != NULL) { // keep track of the side-info
WebPCopyDecBuffer(&output, keep_info);
}
// return decoded samples (don't clear 'output'!)
return WebPIsRGBMode(mode) ? output.u.RGBA.rgba : output.u.YUVA.y;
}
uint8_t* WebPDecodeRGB(const uint8_t* data, size_t data_size,
int* width, int* height) {
return Decode(MODE_RGB, data, data_size, width, height, NULL);
}
uint8_t* WebPDecodeRGBA(const uint8_t* data, size_t data_size,
int* width, int* height) {
return Decode(MODE_RGBA, data, data_size, width, height, NULL);
}
uint8_t* WebPDecodeARGB(const uint8_t* data, size_t data_size,
int* width, int* height) {
return Decode(MODE_ARGB, data, data_size, width, height, NULL);
}
uint8_t* WebPDecodeBGR(const uint8_t* data, size_t data_size,
int* width, int* height) {
return Decode(MODE_BGR, data, data_size, width, height, NULL);
}
uint8_t* WebPDecodeBGRA(const uint8_t* data, size_t data_size,
int* width, int* height) {
return Decode(MODE_BGRA, data, data_size, width, height, NULL);
}
uint8_t* WebPDecodeYUV(const uint8_t* data, size_t data_size,
int* width, int* height, uint8_t** u, uint8_t** v,
int* stride, int* uv_stride) {
WebPDecBuffer output; // only to preserve the side-infos
uint8_t* const out = Decode(MODE_YUV, data, data_size,
width, height, &output);
if (out != NULL) {
const WebPYUVABuffer* const buf = &output.u.YUVA;
*u = buf->u;
*v = buf->v;
*stride = buf->y_stride;
*uv_stride = buf->u_stride;
assert(buf->u_stride == buf->v_stride);
}
return out;
}
static void DefaultFeatures(WebPBitstreamFeatures* const features) {
assert(features != NULL);
memset(features, 0, sizeof(*features));
}
static VP8StatusCode GetFeatures(const uint8_t* const data, size_t data_size,
WebPBitstreamFeatures* const features) {
if (features == NULL || data == NULL) {
return VP8_STATUS_INVALID_PARAM;
}
DefaultFeatures(features);
// Only parse enough of the data to retrieve the features.
return ParseHeadersInternal(data, data_size,
&features->width, &features->height,
&features->has_alpha, &features->has_animation,
&features->format, NULL);
}
//------------------------------------------------------------------------------
// WebPGetInfo()
int WebPGetInfo(const uint8_t* data, size_t data_size,
int* width, int* height) {
WebPBitstreamFeatures features;
if (GetFeatures(data, data_size, &features) != VP8_STATUS_OK) {
return 0;
}
if (width != NULL) {
*width = features.width;
}
if (height != NULL) {
*height = features.height;
}
return 1;
}
//------------------------------------------------------------------------------
// Advance decoding API
int WebPInitDecoderConfigInternal(WebPDecoderConfig* config,
int version) {
if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_DECODER_ABI_VERSION)) {
return 0; // version mismatch
}
if (config == NULL) {
return 0;
}
memset(config, 0, sizeof(*config));
DefaultFeatures(&config->input);
WebPInitDecBuffer(&config->output);
return 1;
}
VP8StatusCode WebPGetFeaturesInternal(const uint8_t* data, size_t data_size,
WebPBitstreamFeatures* features,
int version) {
if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_DECODER_ABI_VERSION)) {
return VP8_STATUS_INVALID_PARAM; // version mismatch
}
if (features == NULL) {
return VP8_STATUS_INVALID_PARAM;
}
return GetFeatures(data, data_size, features);
}
VP8StatusCode WebPDecode(const uint8_t* data, size_t data_size,
WebPDecoderConfig* config) {
WebPDecParams params;
VP8StatusCode status;
if (config == NULL) {
return VP8_STATUS_INVALID_PARAM;
}
status = GetFeatures(data, data_size, &config->input);
if (status != VP8_STATUS_OK) {
if (status == VP8_STATUS_NOT_ENOUGH_DATA) {
return VP8_STATUS_BITSTREAM_ERROR; // Not-enough-data treated as error.
}
return status;
}
WebPResetDecParams(&params);
params.options = &config->options;
params.output = &config->output;
if (WebPAvoidSlowMemory(params.output, &config->input)) {
// decoding to slow memory: use a temporary in-mem buffer to decode into.
WebPDecBuffer in_mem_buffer;
WebPInitDecBuffer(&in_mem_buffer);
in_mem_buffer.colorspace = config->output.colorspace;
in_mem_buffer.width = config->input.width;
in_mem_buffer.height = config->input.height;
params.output = &in_mem_buffer;
status = DecodeInto(data, data_size, &params);
if (status == VP8_STATUS_OK) { // do the slow-copy
status = WebPCopyDecBufferPixels(&in_mem_buffer, &config->output);
}
WebPFreeDecBuffer(&in_mem_buffer);
} else {
status = DecodeInto(data, data_size, &params);
}
return status;
}
//------------------------------------------------------------------------------
// Cropping and rescaling.
int WebPIoInitFromOptions(const WebPDecoderOptions* const options,
VP8Io* const io, WEBP_CSP_MODE src_colorspace) {
const int W = io->width;
const int H = io->height;
int x = 0, y = 0, w = W, h = H;
// Cropping
io->use_cropping = (options != NULL) && (options->use_cropping > 0);
if (io->use_cropping) {
w = options->crop_width;
h = options->crop_height;
x = options->crop_left;
y = options->crop_top;
if (!WebPIsRGBMode(src_colorspace)) { // only snap for YUV420
x &= ~1;
y &= ~1;
}
if (x < 0 || y < 0 || w <= 0 || h <= 0 || x + w > W || y + h > H) {
return 0; // out of frame boundary error
}
}
io->crop_left = x;
io->crop_top = y;
io->crop_right = x + w;
io->crop_bottom = y + h;
io->mb_w = w;
io->mb_h = h;
// Scaling
io->use_scaling = (options != NULL) && (options->use_scaling > 0);
if (io->use_scaling) {
int scaled_width = options->scaled_width;
int scaled_height = options->scaled_height;
if (!WebPRescalerGetScaledDimensions(w, h, &scaled_width, &scaled_height)) {
return 0;
}
io->scaled_width = scaled_width;
io->scaled_height = scaled_height;
}
// Filter
io->bypass_filtering = (options != NULL) && options->bypass_filtering;
// Fancy upsampler
#ifdef FANCY_UPSAMPLING
io->fancy_upsampling = (options == NULL) || (!options->no_fancy_upsampling);
#endif
if (io->use_scaling) {
// disable filter (only for large downscaling ratio).
io->bypass_filtering = (io->scaled_width < W * 3 / 4) &&
(io->scaled_height < H * 3 / 4);
io->fancy_upsampling = 0;
}
return 1;
}
//------------------------------------------------------------------------------

133
media/libwebp/src/dec/webpi_dec.h Normal file
Просмотреть файл

@ -0,0 +1,133 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Internal header: WebP decoding parameters and custom IO on buffer
//
// Author: somnath@google.com (Somnath Banerjee)
#ifndef WEBP_DEC_WEBPI_DEC_H_
#define WEBP_DEC_WEBPI_DEC_H_
#ifdef __cplusplus
extern "C" {
#endif
#include "src/utils/rescaler_utils.h"
#include "src/dec/vp8_dec.h"
//------------------------------------------------------------------------------
// WebPDecParams: Decoding output parameters. Transient internal object.
typedef struct WebPDecParams WebPDecParams;
typedef int (*OutputFunc)(const VP8Io* const io, WebPDecParams* const p);
typedef int (*OutputAlphaFunc)(const VP8Io* const io, WebPDecParams* const p,
int expected_num_out_lines);
typedef int (*OutputRowFunc)(WebPDecParams* const p, int y_pos,
int max_out_lines);
struct WebPDecParams {
WebPDecBuffer* output; // output buffer.
uint8_t* tmp_y, *tmp_u, *tmp_v; // cache for the fancy upsampler
// or used for tmp rescaling
int last_y; // coordinate of the line that was last output
const WebPDecoderOptions* options; // if not NULL, use alt decoding features
WebPRescaler* scaler_y, *scaler_u, *scaler_v, *scaler_a; // rescalers
void* memory; // overall scratch memory for the output work.
OutputFunc emit; // output RGB or YUV samples
OutputAlphaFunc emit_alpha; // output alpha channel
OutputRowFunc emit_alpha_row; // output one line of rescaled alpha values
};
// Should be called first, before any use of the WebPDecParams object.
void WebPResetDecParams(WebPDecParams* const params);
//------------------------------------------------------------------------------
// Header parsing helpers
// Structure storing a description of the RIFF headers.
typedef struct {
const uint8_t* data; // input buffer
size_t data_size; // input buffer size
int have_all_data; // true if all data is known to be available
size_t offset; // offset to main data chunk (VP8 or VP8L)
const uint8_t* alpha_data; // points to alpha chunk (if present)
size_t alpha_data_size; // alpha chunk size
size_t compressed_size; // VP8/VP8L compressed data size
size_t riff_size; // size of the riff payload (or 0 if absent)
int is_lossless; // true if a VP8L chunk is present
} WebPHeaderStructure;
// Skips over all valid chunks prior to the first VP8/VP8L frame header.
// Returns: VP8_STATUS_OK, VP8_STATUS_BITSTREAM_ERROR (invalid header/chunk),
// VP8_STATUS_NOT_ENOUGH_DATA (partial input) or VP8_STATUS_UNSUPPORTED_FEATURE
// in the case of non-decodable features (animation for instance).
// In 'headers', compressed_size, offset, alpha_data, alpha_size, and lossless
// fields are updated appropriately upon success.
VP8StatusCode WebPParseHeaders(WebPHeaderStructure* const headers);
//------------------------------------------------------------------------------
// Misc utils
// Initializes VP8Io with custom setup, io and teardown functions. The default
// hooks will use the supplied 'params' as io->opaque handle.
void WebPInitCustomIo(WebPDecParams* const params, VP8Io* const io);
// Setup crop_xxx fields, mb_w and mb_h in io. 'src_colorspace' refers
// to the *compressed* format, not the output one.
int WebPIoInitFromOptions(const WebPDecoderOptions* const options,
VP8Io* const io, WEBP_CSP_MODE src_colorspace);
//------------------------------------------------------------------------------
// Internal functions regarding WebPDecBuffer memory (in buffer.c).
// Don't really need to be externally visible for now.
// Prepare 'buffer' with the requested initial dimensions width/height.
// If no external storage is supplied, initializes buffer by allocating output
// memory and setting up the stride information. Validate the parameters. Return
// an error code in case of problem (no memory, or invalid stride / size /
// dimension / etc.). If *options is not NULL, also verify that the options'
// parameters are valid and apply them to the width/height dimensions of the
// output buffer. This takes cropping / scaling / rotation into account.
// Also incorporates the options->flip flag to flip the buffer parameters if
// needed.
VP8StatusCode WebPAllocateDecBuffer(int width, int height,
const WebPDecoderOptions* const options,
WebPDecBuffer* const buffer);
// Flip buffer vertically by negating the various strides.
VP8StatusCode WebPFlipBuffer(WebPDecBuffer* const buffer);
// Copy 'src' into 'dst' buffer, making sure 'dst' is not marked as owner of the
// memory (still held by 'src'). No pixels are copied.
void WebPCopyDecBuffer(const WebPDecBuffer* const src,
WebPDecBuffer* const dst);
// Copy and transfer ownership from src to dst (beware of parameter order!)
void WebPGrabDecBuffer(WebPDecBuffer* const src, WebPDecBuffer* const dst);
// Copy pixels from 'src' into a *preallocated* 'dst' buffer. Returns
// VP8_STATUS_INVALID_PARAM if the 'dst' is not set up correctly for the copy.
VP8StatusCode WebPCopyDecBufferPixels(const WebPDecBuffer* const src,
WebPDecBuffer* const dst);
// Returns true if decoding will be slow with the current configuration
// and bitstream features.
int WebPAvoidSlowMemory(const WebPDecBuffer* const output,
const WebPBitstreamFeatures* const features);
//------------------------------------------------------------------------------
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* WEBP_DEC_WEBPI_DEC_H_ */

967
media/libwebp/src/demux/demux.c Normal file
Просмотреть файл

@ -0,0 +1,967 @@
// Copyright 2012 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// WebP container demux.
//
#ifdef HAVE_CONFIG_H
#include "src/webp/config.h"
#endif
#include <assert.h>
#include <stdlib.h>
#include <string.h>
#include "src/utils/utils.h"
#include "src/webp/decode.h" // WebPGetFeatures
#include "src/webp/demux.h"
#include "src/webp/format_constants.h"
#define DMUX_MAJ_VERSION 1
#define DMUX_MIN_VERSION 0
#define DMUX_REV_VERSION 0
typedef struct {
size_t start_; // start location of the data
size_t end_; // end location
size_t riff_end_; // riff chunk end location, can be > end_.
size_t buf_size_; // size of the buffer
const uint8_t* buf_;
} MemBuffer;
typedef struct {
size_t offset_;
size_t size_;
} ChunkData;
typedef struct Frame {
int x_offset_, y_offset_;
int width_, height_;
int has_alpha_;
int duration_;
WebPMuxAnimDispose dispose_method_;
WebPMuxAnimBlend blend_method_;
int frame_num_;
int complete_; // img_components_ contains a full image.
ChunkData img_components_[2]; // 0=VP8{,L} 1=ALPH
struct Frame* next_;
} Frame;
typedef struct Chunk {
ChunkData data_;
struct Chunk* next_;
} Chunk;
struct WebPDemuxer {
MemBuffer mem_;
WebPDemuxState state_;
int is_ext_format_;
uint32_t feature_flags_;
int canvas_width_, canvas_height_;
int loop_count_;
uint32_t bgcolor_;
int num_frames_;
Frame* frames_;
Frame** frames_tail_;
Chunk* chunks_; // non-image chunks
Chunk** chunks_tail_;
};
typedef enum {
PARSE_OK,
PARSE_NEED_MORE_DATA,
PARSE_ERROR
} ParseStatus;
typedef struct ChunkParser {
uint8_t id[4];
ParseStatus (*parse)(WebPDemuxer* const dmux);
int (*valid)(const WebPDemuxer* const dmux);
} ChunkParser;
static ParseStatus ParseSingleImage(WebPDemuxer* const dmux);
static ParseStatus ParseVP8X(WebPDemuxer* const dmux);
static int IsValidSimpleFormat(const WebPDemuxer* const dmux);
static int IsValidExtendedFormat(const WebPDemuxer* const dmux);
static const ChunkParser kMasterChunks[] = {
{ { 'V', 'P', '8', ' ' }, ParseSingleImage, IsValidSimpleFormat },
{ { 'V', 'P', '8', 'L' }, ParseSingleImage, IsValidSimpleFormat },
{ { 'V', 'P', '8', 'X' }, ParseVP8X, IsValidExtendedFormat },
{ { '0', '0', '0', '0' }, NULL, NULL },
};
//------------------------------------------------------------------------------
int WebPGetDemuxVersion(void) {
return (DMUX_MAJ_VERSION << 16) | (DMUX_MIN_VERSION << 8) | DMUX_REV_VERSION;
}
// -----------------------------------------------------------------------------
// MemBuffer
static int RemapMemBuffer(MemBuffer* const mem,
const uint8_t* data, size_t size) {
if (size < mem->buf_size_) return 0; // can't remap to a shorter buffer!
mem->buf_ = data;
mem->end_ = mem->buf_size_ = size;
return 1;
}
static int InitMemBuffer(MemBuffer* const mem,
const uint8_t* data, size_t size) {
memset(mem, 0, sizeof(*mem));
return RemapMemBuffer(mem, data, size);
}
// Return the remaining data size available in 'mem'.
static WEBP_INLINE size_t MemDataSize(const MemBuffer* const mem) {
return (mem->end_ - mem->start_);
}
// Return true if 'size' exceeds the end of the RIFF chunk.
static WEBP_INLINE int SizeIsInvalid(const MemBuffer* const mem, size_t size) {
return (size > mem->riff_end_ - mem->start_);
}
static WEBP_INLINE void Skip(MemBuffer* const mem, size_t size) {
mem->start_ += size;
}
static WEBP_INLINE void Rewind(MemBuffer* const mem, size_t size) {
mem->start_ -= size;
}
static WEBP_INLINE const uint8_t* GetBuffer(MemBuffer* const mem) {
return mem->buf_ + mem->start_;
}
// Read from 'mem' and skip the read bytes.
static WEBP_INLINE uint8_t ReadByte(MemBuffer* const mem) {
const uint8_t byte = mem->buf_[mem->start_];
Skip(mem, 1);
return byte;
}
static WEBP_INLINE int ReadLE16s(MemBuffer* const mem) {
const uint8_t* const data = mem->buf_ + mem->start_;
const int val = GetLE16(data);
Skip(mem, 2);
return val;
}
static WEBP_INLINE int ReadLE24s(MemBuffer* const mem) {
const uint8_t* const data = mem->buf_ + mem->start_;
const int val = GetLE24(data);
Skip(mem, 3);
return val;
}
static WEBP_INLINE uint32_t ReadLE32(MemBuffer* const mem) {
const uint8_t* const data = mem->buf_ + mem->start_;
const uint32_t val = GetLE32(data);
Skip(mem, 4);
return val;
}
// -----------------------------------------------------------------------------
// Secondary chunk parsing
static void AddChunk(WebPDemuxer* const dmux, Chunk* const chunk) {
*dmux->chunks_tail_ = chunk;
chunk->next_ = NULL;
dmux->chunks_tail_ = &chunk->next_;
}
// Add a frame to the end of the list, ensuring the last frame is complete.
// Returns true on success, false otherwise.
static int AddFrame(WebPDemuxer* const dmux, Frame* const frame) {
const Frame* const last_frame = *dmux->frames_tail_;
if (last_frame != NULL && !last_frame->complete_) return 0;
*dmux->frames_tail_ = frame;
frame->next_ = NULL;
dmux->frames_tail_ = &frame->next_;
return 1;
}
static void SetFrameInfo(size_t start_offset, size_t size,
int frame_num, int complete,
const WebPBitstreamFeatures* const features,
Frame* const frame) {
frame->img_components_[0].offset_ = start_offset;
frame->img_components_[0].size_ = size;
frame->width_ = features->width;
frame->height_ = features->height;
frame->has_alpha_ |= features->has_alpha;
frame->frame_num_ = frame_num;
frame->complete_ = complete;
}
// Store image bearing chunks to 'frame'. 'min_size' is an optional size
// requirement, it may be zero.
static ParseStatus StoreFrame(int frame_num, uint32_t min_size,
MemBuffer* const mem, Frame* const frame) {
int alpha_chunks = 0;
int image_chunks = 0;
int done = (MemDataSize(mem) < CHUNK_HEADER_SIZE ||
MemDataSize(mem) < min_size);
ParseStatus status = PARSE_OK;
if (done) return PARSE_NEED_MORE_DATA;
do {
const size_t chunk_start_offset = mem->start_;
const uint32_t fourcc = ReadLE32(mem);
const uint32_t payload_size = ReadLE32(mem);
const uint32_t payload_size_padded = payload_size + (payload_size & 1);
const size_t payload_available = (payload_size_padded > MemDataSize(mem))
? MemDataSize(mem) : payload_size_padded;
const size_t chunk_size = CHUNK_HEADER_SIZE + payload_available;
if (payload_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
if (SizeIsInvalid(mem, payload_size_padded)) return PARSE_ERROR;
if (payload_size_padded > MemDataSize(mem)) status = PARSE_NEED_MORE_DATA;
switch (fourcc) {
case MKFOURCC('A', 'L', 'P', 'H'):
if (alpha_chunks == 0) {
++alpha_chunks;
frame->img_components_[1].offset_ = chunk_start_offset;
frame->img_components_[1].size_ = chunk_size;
frame->has_alpha_ = 1;
frame->frame_num_ = frame_num;
Skip(mem, payload_available);
} else {
goto Done;
}
break;
case MKFOURCC('V', 'P', '8', 'L'):
if (alpha_chunks > 0) return PARSE_ERROR; // VP8L has its own alpha
// fall through
case MKFOURCC('V', 'P', '8', ' '):
if (image_chunks == 0) {
// Extract the bitstream features, tolerating failures when the data
// is incomplete.
WebPBitstreamFeatures features;
const VP8StatusCode vp8_status =
WebPGetFeatures(mem->buf_ + chunk_start_offset, chunk_size,
&features);
if (status == PARSE_NEED_MORE_DATA &&
vp8_status == VP8_STATUS_NOT_ENOUGH_DATA) {
return PARSE_NEED_MORE_DATA;
} else if (vp8_status != VP8_STATUS_OK) {
// We have enough data, and yet WebPGetFeatures() failed.
return PARSE_ERROR;
}
++image_chunks;
SetFrameInfo(chunk_start_offset, chunk_size, frame_num,
status == PARSE_OK, &features, frame);
Skip(mem, payload_available);
} else {
goto Done;
}
break;
Done:
default:
// Restore fourcc/size when moving up one level in parsing.
Rewind(mem, CHUNK_HEADER_SIZE);
done = 1;
break;
}
if (mem->start_ == mem->riff_end_) {
done = 1;
} else if (MemDataSize(mem) < CHUNK_HEADER_SIZE) {
status = PARSE_NEED_MORE_DATA;
}
} while (!done && status == PARSE_OK);
return status;
}
// Creates a new Frame if 'actual_size' is within bounds and 'mem' contains
// enough data ('min_size') to parse the payload.
// Returns PARSE_OK on success with *frame pointing to the new Frame.
// Returns PARSE_NEED_MORE_DATA with insufficient data, PARSE_ERROR otherwise.
static ParseStatus NewFrame(const MemBuffer* const mem,
uint32_t min_size, uint32_t actual_size,
Frame** frame) {
if (SizeIsInvalid(mem, min_size)) return PARSE_ERROR;
if (actual_size < min_size) return PARSE_ERROR;
if (MemDataSize(mem) < min_size) return PARSE_NEED_MORE_DATA;
*frame = (Frame*)WebPSafeCalloc(1ULL, sizeof(**frame));
return (*frame == NULL) ? PARSE_ERROR : PARSE_OK;
}
// Parse a 'ANMF' chunk and any image bearing chunks that immediately follow.
// 'frame_chunk_size' is the previously validated, padded chunk size.
static ParseStatus ParseAnimationFrame(
WebPDemuxer* const dmux, uint32_t frame_chunk_size) {
const int is_animation = !!(dmux->feature_flags_ & ANIMATION_FLAG);
const uint32_t anmf_payload_size = frame_chunk_size - ANMF_CHUNK_SIZE;
int added_frame = 0;
int bits;
MemBuffer* const mem = &dmux->mem_;
Frame* frame;
ParseStatus status =
NewFrame(mem, ANMF_CHUNK_SIZE, frame_chunk_size, &frame);
if (status != PARSE_OK) return status;
frame->x_offset_ = 2 * ReadLE24s(mem);
frame->y_offset_ = 2 * ReadLE24s(mem);
frame->width_ = 1 + ReadLE24s(mem);
frame->height_ = 1 + ReadLE24s(mem);
frame->duration_ = ReadLE24s(mem);
bits = ReadByte(mem);
frame->dispose_method_ =
(bits & 1) ? WEBP_MUX_DISPOSE_BACKGROUND : WEBP_MUX_DISPOSE_NONE;
frame->blend_method_ = (bits & 2) ? WEBP_MUX_NO_BLEND : WEBP_MUX_BLEND;
if (frame->width_ * (uint64_t)frame->height_ >= MAX_IMAGE_AREA) {
WebPSafeFree(frame);
return PARSE_ERROR;
}
// Store a frame only if the animation flag is set there is some data for
// this frame is available.
status = StoreFrame(dmux->num_frames_ + 1, anmf_payload_size, mem, frame);
if (status != PARSE_ERROR && is_animation && frame->frame_num_ > 0) {
added_frame = AddFrame(dmux, frame);
if (added_frame) {
++dmux->num_frames_;
} else {
status = PARSE_ERROR;
}
}
if (!added_frame) WebPSafeFree(frame);
return status;
}
// General chunk storage, starting with the header at 'start_offset', allowing
// the user to request the payload via a fourcc string. 'size' includes the
// header and the unpadded payload size.
// Returns true on success, false otherwise.
static int StoreChunk(WebPDemuxer* const dmux,
size_t start_offset, uint32_t size) {
Chunk* const chunk = (Chunk*)WebPSafeCalloc(1ULL, sizeof(*chunk));
if (chunk == NULL) return 0;
chunk->data_.offset_ = start_offset;
chunk->data_.size_ = size;
AddChunk(dmux, chunk);
return 1;
}
// -----------------------------------------------------------------------------
// Primary chunk parsing
static ParseStatus ReadHeader(MemBuffer* const mem) {
const size_t min_size = RIFF_HEADER_SIZE + CHUNK_HEADER_SIZE;
uint32_t riff_size;
// Basic file level validation.
if (MemDataSize(mem) < min_size) return PARSE_NEED_MORE_DATA;
if (memcmp(GetBuffer(mem), "RIFF", CHUNK_SIZE_BYTES) ||
memcmp(GetBuffer(mem) + CHUNK_HEADER_SIZE, "WEBP", CHUNK_SIZE_BYTES)) {
return PARSE_ERROR;
}
riff_size = GetLE32(GetBuffer(mem) + TAG_SIZE);
if (riff_size < CHUNK_HEADER_SIZE) return PARSE_ERROR;
if (riff_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
// There's no point in reading past the end of the RIFF chunk
mem->riff_end_ = riff_size + CHUNK_HEADER_SIZE;
if (mem->buf_size_ > mem->riff_end_) {
mem->buf_size_ = mem->end_ = mem->riff_end_;
}
Skip(mem, RIFF_HEADER_SIZE);
return PARSE_OK;
}
static ParseStatus ParseSingleImage(WebPDemuxer* const dmux) {
const size_t min_size = CHUNK_HEADER_SIZE;
MemBuffer* const mem = &dmux->mem_;
Frame* frame;
ParseStatus status;
int image_added = 0;
if (dmux->frames_ != NULL) return PARSE_ERROR;
if (SizeIsInvalid(mem, min_size)) return PARSE_ERROR;
if (MemDataSize(mem) < min_size) return PARSE_NEED_MORE_DATA;
frame = (Frame*)WebPSafeCalloc(1ULL, sizeof(*frame));
if (frame == NULL) return PARSE_ERROR;
// For the single image case we allow parsing of a partial frame, so no
// minimum size is imposed here.
status = StoreFrame(1, 0, &dmux->mem_, frame);
if (status != PARSE_ERROR) {
const int has_alpha = !!(dmux->feature_flags_ & ALPHA_FLAG);
// Clear any alpha when the alpha flag is missing.
if (!has_alpha && frame->img_components_[1].size_ > 0) {
frame->img_components_[1].offset_ = 0;
frame->img_components_[1].size_ = 0;
frame->has_alpha_ = 0;
}
// Use the frame width/height as the canvas values for non-vp8x files.
// Also, set ALPHA_FLAG if this is a lossless image with alpha.
if (!dmux->is_ext_format_ && frame->width_ > 0 && frame->height_ > 0) {
dmux->state_ = WEBP_DEMUX_PARSED_HEADER;
dmux->canvas_width_ = frame->width_;
dmux->canvas_height_ = frame->height_;
dmux->feature_flags_ |= frame->has_alpha_ ? ALPHA_FLAG : 0;
}
if (!AddFrame(dmux, frame)) {
status = PARSE_ERROR; // last frame was left incomplete
} else {
image_added = 1;
dmux->num_frames_ = 1;
}
}
if (!image_added) WebPSafeFree(frame);
return status;
}
static ParseStatus ParseVP8XChunks(WebPDemuxer* const dmux) {
const int is_animation = !!(dmux->feature_flags_ & ANIMATION_FLAG);
MemBuffer* const mem = &dmux->mem_;
int anim_chunks = 0;
ParseStatus status = PARSE_OK;
do {
int store_chunk = 1;
const size_t chunk_start_offset = mem->start_;
const uint32_t fourcc = ReadLE32(mem);
const uint32_t chunk_size = ReadLE32(mem);
const uint32_t chunk_size_padded = chunk_size + (chunk_size & 1);
if (chunk_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
if (SizeIsInvalid(mem, chunk_size_padded)) return PARSE_ERROR;
switch (fourcc) {
case MKFOURCC('V', 'P', '8', 'X'): {
return PARSE_ERROR;
}
case MKFOURCC('A', 'L', 'P', 'H'):
case MKFOURCC('V', 'P', '8', ' '):
case MKFOURCC('V', 'P', '8', 'L'): {
// check that this isn't an animation (all frames should be in an ANMF).
if (anim_chunks > 0 || is_animation) return PARSE_ERROR;
Rewind(mem, CHUNK_HEADER_SIZE);
status = ParseSingleImage(dmux);
break;
}
case MKFOURCC('A', 'N', 'I', 'M'): {
if (chunk_size_padded < ANIM_CHUNK_SIZE) return PARSE_ERROR;
if (MemDataSize(mem) < chunk_size_padded) {
status = PARSE_NEED_MORE_DATA;
} else if (anim_chunks == 0) {
++anim_chunks;
dmux->bgcolor_ = ReadLE32(mem);
dmux->loop_count_ = ReadLE16s(mem);
Skip(mem, chunk_size_padded - ANIM_CHUNK_SIZE);
} else {
store_chunk = 0;
goto Skip;
}
break;
}
case MKFOURCC('A', 'N', 'M', 'F'): {
if (anim_chunks == 0) return PARSE_ERROR; // 'ANIM' precedes frames.
status = ParseAnimationFrame(dmux, chunk_size_padded);
break;
}
case MKFOURCC('I', 'C', 'C', 'P'): {
store_chunk = !!(dmux->feature_flags_ & ICCP_FLAG);
goto Skip;
}
case MKFOURCC('E', 'X', 'I', 'F'): {
store_chunk = !!(dmux->feature_flags_ & EXIF_FLAG);
goto Skip;
}
case MKFOURCC('X', 'M', 'P', ' '): {
store_chunk = !!(dmux->feature_flags_ & XMP_FLAG);
goto Skip;
}
Skip:
default: {
if (chunk_size_padded <= MemDataSize(mem)) {
if (store_chunk) {
// Store only the chunk header and unpadded size as only the payload
// will be returned to the user.
if (!StoreChunk(dmux, chunk_start_offset,
CHUNK_HEADER_SIZE + chunk_size)) {
return PARSE_ERROR;
}
}
Skip(mem, chunk_size_padded);
} else {
status = PARSE_NEED_MORE_DATA;
}
}
}
if (mem->start_ == mem->riff_end_) {
break;
} else if (MemDataSize(mem) < CHUNK_HEADER_SIZE) {
status = PARSE_NEED_MORE_DATA;
}
} while (status == PARSE_OK);
return status;
}
static ParseStatus ParseVP8X(WebPDemuxer* const dmux) {
MemBuffer* const mem = &dmux->mem_;
uint32_t vp8x_size;
if (MemDataSize(mem) < CHUNK_HEADER_SIZE) return PARSE_NEED_MORE_DATA;
dmux->is_ext_format_ = 1;
Skip(mem, TAG_SIZE); // VP8X
vp8x_size = ReadLE32(mem);
if (vp8x_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
if (vp8x_size < VP8X_CHUNK_SIZE) return PARSE_ERROR;
vp8x_size += vp8x_size & 1;
if (SizeIsInvalid(mem, vp8x_size)) return PARSE_ERROR;
if (MemDataSize(mem) < vp8x_size) return PARSE_NEED_MORE_DATA;
dmux->feature_flags_ = ReadByte(mem);
Skip(mem, 3); // Reserved.
dmux->canvas_width_ = 1 + ReadLE24s(mem);
dmux->canvas_height_ = 1 + ReadLE24s(mem);
if (dmux->canvas_width_ * (uint64_t)dmux->canvas_height_ >= MAX_IMAGE_AREA) {
return PARSE_ERROR; // image final dimension is too large
}
Skip(mem, vp8x_size - VP8X_CHUNK_SIZE); // skip any trailing data.
dmux->state_ = WEBP_DEMUX_PARSED_HEADER;
if (SizeIsInvalid(mem, CHUNK_HEADER_SIZE)) return PARSE_ERROR;
if (MemDataSize(mem) < CHUNK_HEADER_SIZE) return PARSE_NEED_MORE_DATA;
return ParseVP8XChunks(dmux);
}
// -----------------------------------------------------------------------------
// Format validation
static int IsValidSimpleFormat(const WebPDemuxer* const dmux) {
const Frame* const frame = dmux->frames_;
if (dmux->state_ == WEBP_DEMUX_PARSING_HEADER) return 1;
if (dmux->canvas_width_ <= 0 || dmux->canvas_height_ <= 0) return 0;
if (dmux->state_ == WEBP_DEMUX_DONE && frame == NULL) return 0;
if (frame->width_ <= 0 || frame->height_ <= 0) return 0;
return 1;
}
// If 'exact' is true, check that the image resolution matches the canvas.
// If 'exact' is false, check that the x/y offsets do not exceed the canvas.
static int CheckFrameBounds(const Frame* const frame, int exact,
int canvas_width, int canvas_height) {
if (exact) {
if (frame->x_offset_ != 0 || frame->y_offset_ != 0) {
return 0;
}
if (frame->width_ != canvas_width || frame->height_ != canvas_height) {
return 0;
}
} else {
if (frame->x_offset_ < 0 || frame->y_offset_ < 0) return 0;
if (frame->width_ + frame->x_offset_ > canvas_width) return 0;
if (frame->height_ + frame->y_offset_ > canvas_height) return 0;
}
return 1;
}
static int IsValidExtendedFormat(const WebPDemuxer* const dmux) {
const int is_animation = !!(dmux->feature_flags_ & ANIMATION_FLAG);
const Frame* f = dmux->frames_;
if (dmux->state_ == WEBP_DEMUX_PARSING_HEADER) return 1;
if (dmux->canvas_width_ <= 0 || dmux->canvas_height_ <= 0) return 0;
if (dmux->loop_count_ < 0) return 0;
if (dmux->state_ == WEBP_DEMUX_DONE && dmux->frames_ == NULL) return 0;
if (dmux->feature_flags_ & ~ALL_VALID_FLAGS) return 0; // invalid bitstream
while (f != NULL) {
const int cur_frame_set = f->frame_num_;
int frame_count = 0;
// Check frame properties.
for (; f != NULL && f->frame_num_ == cur_frame_set; f = f->next_) {
const ChunkData* const image = f->img_components_;
const ChunkData* const alpha = f->img_components_ + 1;
if (!is_animation && f->frame_num_ > 1) return 0;
if (f->complete_) {
if (alpha->size_ == 0 && image->size_ == 0) return 0;
// Ensure alpha precedes image bitstream.
if (alpha->size_ > 0 && alpha->offset_ > image->offset_) {
return 0;
}
if (f->width_ <= 0 || f->height_ <= 0) return 0;
} else {
// There shouldn't be a partial frame in a complete file.
if (dmux->state_ == WEBP_DEMUX_DONE) return 0;
// Ensure alpha precedes image bitstream.
if (alpha->size_ > 0 && image->size_ > 0 &&
alpha->offset_ > image->offset_) {
return 0;
}
// There shouldn't be any frames after an incomplete one.
if (f->next_ != NULL) return 0;
}
if (f->width_ > 0 && f->height_ > 0 &&
!CheckFrameBounds(f, !is_animation,
dmux->canvas_width_, dmux->canvas_height_)) {
return 0;
}
++frame_count;
}
}
return 1;
}
// -----------------------------------------------------------------------------
// WebPDemuxer object
static void InitDemux(WebPDemuxer* const dmux, const MemBuffer* const mem) {
dmux->state_ = WEBP_DEMUX_PARSING_HEADER;
dmux->loop_count_ = 1;
dmux->bgcolor_ = 0xFFFFFFFF; // White background by default.
dmux->canvas_width_ = -1;
dmux->canvas_height_ = -1;
dmux->frames_tail_ = &dmux->frames_;
dmux->chunks_tail_ = &dmux->chunks_;
dmux->mem_ = *mem;
}
static ParseStatus CreateRawImageDemuxer(MemBuffer* const mem,
WebPDemuxer** demuxer) {
WebPBitstreamFeatures features;
const VP8StatusCode status =
WebPGetFeatures(mem->buf_, mem->buf_size_, &features);
*demuxer = NULL;
if (status != VP8_STATUS_OK) {
return (status == VP8_STATUS_NOT_ENOUGH_DATA) ? PARSE_NEED_MORE_DATA
: PARSE_ERROR;
}
{
WebPDemuxer* const dmux = (WebPDemuxer*)WebPSafeCalloc(1ULL, sizeof(*dmux));
Frame* const frame = (Frame*)WebPSafeCalloc(1ULL, sizeof(*frame));
if (dmux == NULL || frame == NULL) goto Error;
InitDemux(dmux, mem);
SetFrameInfo(0, mem->buf_size_, 1 /*frame_num*/, 1 /*complete*/, &features,
frame);
if (!AddFrame(dmux, frame)) goto Error;
dmux->state_ = WEBP_DEMUX_DONE;
dmux->canvas_width_ = frame->width_;
dmux->canvas_height_ = frame->height_;
dmux->feature_flags_ |= frame->has_alpha_ ? ALPHA_FLAG : 0;
dmux->num_frames_ = 1;
assert(IsValidSimpleFormat(dmux));
*demuxer = dmux;
return PARSE_OK;
Error:
WebPSafeFree(dmux);
WebPSafeFree(frame);
return PARSE_ERROR;
}
}
WebPDemuxer* WebPDemuxInternal(const WebPData* data, int allow_partial,
WebPDemuxState* state, int version) {
const ChunkParser* parser;
int partial;
ParseStatus status = PARSE_ERROR;
MemBuffer mem;
WebPDemuxer* dmux;
if (state != NULL) *state = WEBP_DEMUX_PARSE_ERROR;
if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_DEMUX_ABI_VERSION)) return NULL;
if (data == NULL || data->bytes == NULL || data->size == 0) return NULL;
if (!InitMemBuffer(&mem, data->bytes, data->size)) return NULL;
status = ReadHeader(&mem);
if (status != PARSE_OK) {
// If parsing of the webp file header fails attempt to handle a raw
// VP8/VP8L frame. Note 'allow_partial' is ignored in this case.
if (status == PARSE_ERROR) {
status = CreateRawImageDemuxer(&mem, &dmux);
if (status == PARSE_OK) {
if (state != NULL) *state = WEBP_DEMUX_DONE;
return dmux;
}
}
if (state != NULL) {
*state = (status == PARSE_NEED_MORE_DATA) ? WEBP_DEMUX_PARSING_HEADER
: WEBP_DEMUX_PARSE_ERROR;
}
return NULL;
}
partial = (mem.buf_size_ < mem.riff_end_);
if (!allow_partial && partial) return NULL;
dmux = (WebPDemuxer*)WebPSafeCalloc(1ULL, sizeof(*dmux));
if (dmux == NULL) return NULL;
InitDemux(dmux, &mem);
status = PARSE_ERROR;
for (parser = kMasterChunks; parser->parse != NULL; ++parser) {
if (!memcmp(parser->id, GetBuffer(&dmux->mem_), TAG_SIZE)) {
status = parser->parse(dmux);
if (status == PARSE_OK) dmux->state_ = WEBP_DEMUX_DONE;
if (status == PARSE_NEED_MORE_DATA && !partial) status = PARSE_ERROR;
if (status != PARSE_ERROR && !parser->valid(dmux)) status = PARSE_ERROR;
if (status == PARSE_ERROR) dmux->state_ = WEBP_DEMUX_PARSE_ERROR;
break;
}
}
if (state != NULL) *state = dmux->state_;
if (status == PARSE_ERROR) {
WebPDemuxDelete(dmux);
return NULL;
}
return dmux;
}
void WebPDemuxDelete(WebPDemuxer* dmux) {
Chunk* c;
Frame* f;
if (dmux == NULL) return;
for (f = dmux->frames_; f != NULL;) {
Frame* const cur_frame = f;
f = f->next_;
WebPSafeFree(cur_frame);
}
for (c = dmux->chunks_; c != NULL;) {
Chunk* const cur_chunk = c;
c = c->next_;
WebPSafeFree(cur_chunk);
}
WebPSafeFree(dmux);
}
// -----------------------------------------------------------------------------
uint32_t WebPDemuxGetI(const WebPDemuxer* dmux, WebPFormatFeature feature) {
if (dmux == NULL) return 0;
switch (feature) {
case WEBP_FF_FORMAT_FLAGS: return dmux->feature_flags_;
case WEBP_FF_CANVAS_WIDTH: return (uint32_t)dmux->canvas_width_;
case WEBP_FF_CANVAS_HEIGHT: return (uint32_t)dmux->canvas_height_;
case WEBP_FF_LOOP_COUNT: return (uint32_t)dmux->loop_count_;
case WEBP_FF_BACKGROUND_COLOR: return dmux->bgcolor_;
case WEBP_FF_FRAME_COUNT: return (uint32_t)dmux->num_frames_;
}
return 0;
}
// -----------------------------------------------------------------------------
// Frame iteration
static const Frame* GetFrame(const WebPDemuxer* const dmux, int frame_num) {
const Frame* f;
for (f = dmux->frames_; f != NULL; f = f->next_) {
if (frame_num == f->frame_num_) break;
}
return f;
}
static const uint8_t* GetFramePayload(const uint8_t* const mem_buf,
const Frame* const frame,
size_t* const data_size) {
*data_size = 0;
if (frame != NULL) {
const ChunkData* const image = frame->img_components_;
const ChunkData* const alpha = frame->img_components_ + 1;
size_t start_offset = image->offset_;
*data_size = image->size_;
// if alpha exists it precedes image, update the size allowing for
// intervening chunks.
if (alpha->size_ > 0) {
const size_t inter_size = (image->offset_ > 0)
? image->offset_ - (alpha->offset_ + alpha->size_)
: 0;
start_offset = alpha->offset_;
*data_size += alpha->size_ + inter_size;
}
return mem_buf + start_offset;
}
return NULL;
}
// Create a whole 'frame' from VP8 (+ alpha) or lossless.
static int SynthesizeFrame(const WebPDemuxer* const dmux,
const Frame* const frame,
WebPIterator* const iter) {
const uint8_t* const mem_buf = dmux->mem_.buf_;
size_t payload_size = 0;
const uint8_t* const payload = GetFramePayload(mem_buf, frame, &payload_size);
if (payload == NULL) return 0;
assert(frame != NULL);
iter->frame_num = frame->frame_num_;
iter->num_frames = dmux->num_frames_;
iter->x_offset = frame->x_offset_;
iter->y_offset = frame->y_offset_;
iter->width = frame->width_;
iter->height = frame->height_;
iter->has_alpha = frame->has_alpha_;
iter->duration = frame->duration_;
iter->dispose_method = frame->dispose_method_;
iter->blend_method = frame->blend_method_;
iter->complete = frame->complete_;
iter->fragment.bytes = payload;
iter->fragment.size = payload_size;
return 1;
}
static int SetFrame(int frame_num, WebPIterator* const iter) {
const Frame* frame;
const WebPDemuxer* const dmux = (WebPDemuxer*)iter->private_;
if (dmux == NULL || frame_num < 0) return 0;
if (frame_num > dmux->num_frames_) return 0;
if (frame_num == 0) frame_num = dmux->num_frames_;
frame = GetFrame(dmux, frame_num);
if (frame == NULL) return 0;
return SynthesizeFrame(dmux, frame, iter);
}
int WebPDemuxGetFrame(const WebPDemuxer* dmux, int frame, WebPIterator* iter) {
if (iter == NULL) return 0;
memset(iter, 0, sizeof(*iter));
iter->private_ = (void*)dmux;
return SetFrame(frame, iter);
}
int WebPDemuxNextFrame(WebPIterator* iter) {
if (iter == NULL) return 0;
return SetFrame(iter->frame_num + 1, iter);
}
int WebPDemuxPrevFrame(WebPIterator* iter) {
if (iter == NULL) return 0;
if (iter->frame_num <= 1) return 0;
return SetFrame(iter->frame_num - 1, iter);
}
void WebPDemuxReleaseIterator(WebPIterator* iter) {
(void)iter;
}
// -----------------------------------------------------------------------------
// Chunk iteration
static int ChunkCount(const WebPDemuxer* const dmux, const char fourcc[4]) {
const uint8_t* const mem_buf = dmux->mem_.buf_;
const Chunk* c;
int count = 0;
for (c = dmux->chunks_; c != NULL; c = c->next_) {
const uint8_t* const header = mem_buf + c->data_.offset_;
if (!memcmp(header, fourcc, TAG_SIZE)) ++count;
}
return count;
}
static const Chunk* GetChunk(const WebPDemuxer* const dmux,
const char fourcc[4], int chunk_num) {
const uint8_t* const mem_buf = dmux->mem_.buf_;
const Chunk* c;
int count = 0;
for (c = dmux->chunks_; c != NULL; c = c->next_) {
const uint8_t* const header = mem_buf + c->data_.offset_;
if (!memcmp(header, fourcc, TAG_SIZE)) ++count;
if (count == chunk_num) break;
}
return c;
}
static int SetChunk(const char fourcc[4], int chunk_num,
WebPChunkIterator* const iter) {
const WebPDemuxer* const dmux = (WebPDemuxer*)iter->private_;
int count;
if (dmux == NULL || fourcc == NULL || chunk_num < 0) return 0;
count = ChunkCount(dmux, fourcc);
if (count == 0) return 0;
if (chunk_num == 0) chunk_num = count;
if (chunk_num <= count) {
const uint8_t* const mem_buf = dmux->mem_.buf_;
const Chunk* const chunk = GetChunk(dmux, fourcc, chunk_num);
iter->chunk.bytes = mem_buf + chunk->data_.offset_ + CHUNK_HEADER_SIZE;
iter->chunk.size = chunk->data_.size_ - CHUNK_HEADER_SIZE;
iter->num_chunks = count;
iter->chunk_num = chunk_num;
return 1;
}
return 0;
}
int WebPDemuxGetChunk(const WebPDemuxer* dmux,
const char fourcc[4], int chunk_num,
WebPChunkIterator* iter) {
if (iter == NULL) return 0;
memset(iter, 0, sizeof(*iter));
iter->private_ = (void*)dmux;
return SetChunk(fourcc, chunk_num, iter);
}
int WebPDemuxNextChunk(WebPChunkIterator* iter) {
if (iter != NULL) {
const char* const fourcc =
(const char*)iter->chunk.bytes - CHUNK_HEADER_SIZE;
return SetChunk(fourcc, iter->chunk_num + 1, iter);
}
return 0;
}
int WebPDemuxPrevChunk(WebPChunkIterator* iter) {
if (iter != NULL && iter->chunk_num > 1) {
const char* const fourcc =
(const char*)iter->chunk.bytes - CHUNK_HEADER_SIZE;
return SetChunk(fourcc, iter->chunk_num - 1, iter);
}
return 0;
}
void WebPDemuxReleaseChunkIterator(WebPChunkIterator* iter) {
(void)iter;
}

21
media/libwebp/src/demux/moz.build Normal file
Просмотреть файл

@ -0,0 +1,21 @@
# -*- Mode: python; indent-tabs-mode: nil; tab-width: 40 -*-
# vim: set filetype=python:
# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
with Files('**'):
BUG_COMPONENT = ('Core', 'ImageLib')
SOURCES += [
'demux.c',
]
LOCAL_INCLUDES += [
'/media/libwebp',
]
FINAL_LIBRARY = 'gkmedias'
# We allow warnings for third-party code that can be updated from upstream.
AllowCompilerWarnings()

472
media/libwebp/src/dsp/alpha_processing.c Normal file
Просмотреть файл

@ -0,0 +1,472 @@
// Copyright 2013 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Utilities for processing transparent channel.
//
// Author: Skal (pascal.massimino@gmail.com)
#include <assert.h>
#include "src/dsp/dsp.h"
// Tables can be faster on some platform but incur some extra binary size (~2k).
#if !defined(USE_TABLES_FOR_ALPHA_MULT)
#define USE_TABLES_FOR_ALPHA_MULT 0 // ALTERNATE_CODE
#endif
// -----------------------------------------------------------------------------
#define MFIX 24 // 24bit fixed-point arithmetic
#define HALF ((1u << MFIX) >> 1)
#define KINV_255 ((1u << MFIX) / 255u)
static uint32_t Mult(uint8_t x, uint32_t mult) {
const uint32_t v = (x * mult + HALF) >> MFIX;
assert(v <= 255); // <- 24bit precision is enough to ensure that.
return v;
}
#if (USE_TABLES_FOR_ALPHA_MULT == 1)
static const uint32_t kMultTables[2][256] = {
{ // (255u << MFIX) / alpha
0x00000000, 0xff000000, 0x7f800000, 0x55000000, 0x3fc00000, 0x33000000,
0x2a800000, 0x246db6db, 0x1fe00000, 0x1c555555, 0x19800000, 0x172e8ba2,
0x15400000, 0x139d89d8, 0x1236db6d, 0x11000000, 0x0ff00000, 0x0f000000,
0x0e2aaaaa, 0x0d6bca1a, 0x0cc00000, 0x0c249249, 0x0b9745d1, 0x0b1642c8,
0x0aa00000, 0x0a333333, 0x09cec4ec, 0x0971c71c, 0x091b6db6, 0x08cb08d3,
0x08800000, 0x0839ce73, 0x07f80000, 0x07ba2e8b, 0x07800000, 0x07492492,
0x07155555, 0x06e45306, 0x06b5e50d, 0x0689d89d, 0x06600000, 0x063831f3,
0x06124924, 0x05ee23b8, 0x05cba2e8, 0x05aaaaaa, 0x058b2164, 0x056cefa8,
0x05500000, 0x05343eb1, 0x05199999, 0x05000000, 0x04e76276, 0x04cfb2b7,
0x04b8e38e, 0x04a2e8ba, 0x048db6db, 0x0479435e, 0x04658469, 0x045270d0,
0x04400000, 0x042e29f7, 0x041ce739, 0x040c30c3, 0x03fc0000, 0x03ec4ec4,
0x03dd1745, 0x03ce540f, 0x03c00000, 0x03b21642, 0x03a49249, 0x03976fc6,
0x038aaaaa, 0x037e3f1f, 0x03722983, 0x03666666, 0x035af286, 0x034fcace,
0x0344ec4e, 0x033a5440, 0x03300000, 0x0325ed09, 0x031c18f9, 0x0312818a,
0x03092492, 0x03000000, 0x02f711dc, 0x02ee5846, 0x02e5d174, 0x02dd7baf,
0x02d55555, 0x02cd5cd5, 0x02c590b2, 0x02bdef7b, 0x02b677d4, 0x02af286b,
0x02a80000, 0x02a0fd5c, 0x029a1f58, 0x029364d9, 0x028ccccc, 0x0286562d,
0x02800000, 0x0279c952, 0x0273b13b, 0x026db6db, 0x0267d95b, 0x026217ec,
0x025c71c7, 0x0256e62a, 0x0251745d, 0x024c1bac, 0x0246db6d, 0x0241b2f9,
0x023ca1af, 0x0237a6f4, 0x0232c234, 0x022df2df, 0x02293868, 0x02249249,
0x02200000, 0x021b810e, 0x021714fb, 0x0212bb51, 0x020e739c, 0x020a3d70,
0x02061861, 0x02020408, 0x01fe0000, 0x01fa0be8, 0x01f62762, 0x01f25213,
0x01ee8ba2, 0x01ead3ba, 0x01e72a07, 0x01e38e38, 0x01e00000, 0x01dc7f10,
0x01d90b21, 0x01d5a3e9, 0x01d24924, 0x01cefa8d, 0x01cbb7e3, 0x01c880e5,
0x01c55555, 0x01c234f7, 0x01bf1f8f, 0x01bc14e5, 0x01b914c1, 0x01b61eed,
0x01b33333, 0x01b05160, 0x01ad7943, 0x01aaaaaa, 0x01a7e567, 0x01a5294a,
0x01a27627, 0x019fcbd2, 0x019d2a20, 0x019a90e7, 0x01980000, 0x01957741,
0x0192f684, 0x01907da4, 0x018e0c7c, 0x018ba2e8, 0x018940c5, 0x0186e5f0,
0x01849249, 0x018245ae, 0x01800000, 0x017dc11f, 0x017b88ee, 0x0179574e,
0x01772c23, 0x01750750, 0x0172e8ba, 0x0170d045, 0x016ebdd7, 0x016cb157,
0x016aaaaa, 0x0168a9b9, 0x0166ae6a, 0x0164b8a7, 0x0162c859, 0x0160dd67,
0x015ef7bd, 0x015d1745, 0x015b3bea, 0x01596596, 0x01579435, 0x0155c7b4,
0x01540000, 0x01523d03, 0x01507eae, 0x014ec4ec, 0x014d0fac, 0x014b5edc,
0x0149b26c, 0x01480a4a, 0x01466666, 0x0144c6af, 0x01432b16, 0x0141938b,
0x01400000, 0x013e7063, 0x013ce4a9, 0x013b5cc0, 0x0139d89d, 0x01385830,
0x0136db6d, 0x01356246, 0x0133ecad, 0x01327a97, 0x01310bf6, 0x012fa0be,
0x012e38e3, 0x012cd459, 0x012b7315, 0x012a150a, 0x0128ba2e, 0x01276276,
0x01260dd6, 0x0124bc44, 0x01236db6, 0x01222222, 0x0120d97c, 0x011f93bc,
0x011e50d7, 0x011d10c4, 0x011bd37a, 0x011a98ef, 0x0119611a, 0x01182bf2,
0x0116f96f, 0x0115c988, 0x01149c34, 0x0113716a, 0x01124924, 0x01112358,
0x01100000, 0x010edf12, 0x010dc087, 0x010ca458, 0x010b8a7d, 0x010a72f0,
0x01095da8, 0x01084a9f, 0x010739ce, 0x01062b2e, 0x01051eb8, 0x01041465,
0x01030c30, 0x01020612, 0x01010204, 0x01000000 },
{ // alpha * KINV_255
0x00000000, 0x00010101, 0x00020202, 0x00030303, 0x00040404, 0x00050505,
0x00060606, 0x00070707, 0x00080808, 0x00090909, 0x000a0a0a, 0x000b0b0b,
0x000c0c0c, 0x000d0d0d, 0x000e0e0e, 0x000f0f0f, 0x00101010, 0x00111111,
0x00121212, 0x00131313, 0x00141414, 0x00151515, 0x00161616, 0x00171717,
0x00181818, 0x00191919, 0x001a1a1a, 0x001b1b1b, 0x001c1c1c, 0x001d1d1d,
0x001e1e1e, 0x001f1f1f, 0x00202020, 0x00212121, 0x00222222, 0x00232323,
0x00242424, 0x00252525, 0x00262626, 0x00272727, 0x00282828, 0x00292929,
0x002a2a2a, 0x002b2b2b, 0x002c2c2c, 0x002d2d2d, 0x002e2e2e, 0x002f2f2f,
0x00303030, 0x00313131, 0x00323232, 0x00333333, 0x00343434, 0x00353535,
0x00363636, 0x00373737, 0x00383838, 0x00393939, 0x003a3a3a, 0x003b3b3b,
0x003c3c3c, 0x003d3d3d, 0x003e3e3e, 0x003f3f3f, 0x00404040, 0x00414141,
0x00424242, 0x00434343, 0x00444444, 0x00454545, 0x00464646, 0x00474747,
0x00484848, 0x00494949, 0x004a4a4a, 0x004b4b4b, 0x004c4c4c, 0x004d4d4d,
0x004e4e4e, 0x004f4f4f, 0x00505050, 0x00515151, 0x00525252, 0x00535353,
0x00545454, 0x00555555, 0x00565656, 0x00575757, 0x00585858, 0x00595959,
0x005a5a5a, 0x005b5b5b, 0x005c5c5c, 0x005d5d5d, 0x005e5e5e, 0x005f5f5f,
0x00606060, 0x00616161, 0x00626262, 0x00636363, 0x00646464, 0x00656565,
0x00666666, 0x00676767, 0x00686868, 0x00696969, 0x006a6a6a, 0x006b6b6b,
0x006c6c6c, 0x006d6d6d, 0x006e6e6e, 0x006f6f6f, 0x00707070, 0x00717171,
0x00727272, 0x00737373, 0x00747474, 0x00757575, 0x00767676, 0x00777777,
0x00787878, 0x00797979, 0x007a7a7a, 0x007b7b7b, 0x007c7c7c, 0x007d7d7d,
0x007e7e7e, 0x007f7f7f, 0x00808080, 0x00818181, 0x00828282, 0x00838383,
0x00848484, 0x00858585, 0x00868686, 0x00878787, 0x00888888, 0x00898989,
0x008a8a8a, 0x008b8b8b, 0x008c8c8c, 0x008d8d8d, 0x008e8e8e, 0x008f8f8f,
0x00909090, 0x00919191, 0x00929292, 0x00939393, 0x00949494, 0x00959595,
0x00969696, 0x00979797, 0x00989898, 0x00999999, 0x009a9a9a, 0x009b9b9b,
0x009c9c9c, 0x009d9d9d, 0x009e9e9e, 0x009f9f9f, 0x00a0a0a0, 0x00a1a1a1,
0x00a2a2a2, 0x00a3a3a3, 0x00a4a4a4, 0x00a5a5a5, 0x00a6a6a6, 0x00a7a7a7,
0x00a8a8a8, 0x00a9a9a9, 0x00aaaaaa, 0x00ababab, 0x00acacac, 0x00adadad,
0x00aeaeae, 0x00afafaf, 0x00b0b0b0, 0x00b1b1b1, 0x00b2b2b2, 0x00b3b3b3,
0x00b4b4b4, 0x00b5b5b5, 0x00b6b6b6, 0x00b7b7b7, 0x00b8b8b8, 0x00b9b9b9,
0x00bababa, 0x00bbbbbb, 0x00bcbcbc, 0x00bdbdbd, 0x00bebebe, 0x00bfbfbf,
0x00c0c0c0, 0x00c1c1c1, 0x00c2c2c2, 0x00c3c3c3, 0x00c4c4c4, 0x00c5c5c5,
0x00c6c6c6, 0x00c7c7c7, 0x00c8c8c8, 0x00c9c9c9, 0x00cacaca, 0x00cbcbcb,
0x00cccccc, 0x00cdcdcd, 0x00cecece, 0x00cfcfcf, 0x00d0d0d0, 0x00d1d1d1,
0x00d2d2d2, 0x00d3d3d3, 0x00d4d4d4, 0x00d5d5d5, 0x00d6d6d6, 0x00d7d7d7,
0x00d8d8d8, 0x00d9d9d9, 0x00dadada, 0x00dbdbdb, 0x00dcdcdc, 0x00dddddd,
0x00dedede, 0x00dfdfdf, 0x00e0e0e0, 0x00e1e1e1, 0x00e2e2e2, 0x00e3e3e3,
0x00e4e4e4, 0x00e5e5e5, 0x00e6e6e6, 0x00e7e7e7, 0x00e8e8e8, 0x00e9e9e9,
0x00eaeaea, 0x00ebebeb, 0x00ececec, 0x00ededed, 0x00eeeeee, 0x00efefef,
0x00f0f0f0, 0x00f1f1f1, 0x00f2f2f2, 0x00f3f3f3, 0x00f4f4f4, 0x00f5f5f5,
0x00f6f6f6, 0x00f7f7f7, 0x00f8f8f8, 0x00f9f9f9, 0x00fafafa, 0x00fbfbfb,
0x00fcfcfc, 0x00fdfdfd, 0x00fefefe, 0x00ffffff }
};
static WEBP_INLINE uint32_t GetScale(uint32_t a, int inverse) {
return kMultTables[!inverse][a];
}
#else
static WEBP_INLINE uint32_t GetScale(uint32_t a, int inverse) {
return inverse ? (255u << MFIX) / a : a * KINV_255;
}
#endif // USE_TABLES_FOR_ALPHA_MULT
void WebPMultARGBRow_C(uint32_t* const ptr, int width, int inverse) {
int x;
for (x = 0; x < width; ++x) {
const uint32_t argb = ptr[x];
if (argb < 0xff000000u) { // alpha < 255
if (argb <= 0x00ffffffu) { // alpha == 0
ptr[x] = 0;
} else {
const uint32_t alpha = (argb >> 24) & 0xff;
const uint32_t scale = GetScale(alpha, inverse);
uint32_t out = argb & 0xff000000u;
out |= Mult(argb >> 0, scale) << 0;
out |= Mult(argb >> 8, scale) << 8;
out |= Mult(argb >> 16, scale) << 16;
ptr[x] = out;
}
}
}
}
void WebPMultRow_C(uint8_t* const ptr, const uint8_t* const alpha,
int width, int inverse) {
int x;
for (x = 0; x < width; ++x) {
const uint32_t a = alpha[x];
if (a != 255) {
if (a == 0) {
ptr[x] = 0;
} else {
const uint32_t scale = GetScale(a, inverse);
ptr[x] = Mult(ptr[x], scale);
}
}
}
}
#undef KINV_255
#undef HALF
#undef MFIX
void (*WebPMultARGBRow)(uint32_t* const ptr, int width, int inverse);
void (*WebPMultRow)(uint8_t* const ptr, const uint8_t* const alpha,
int width, int inverse);
//------------------------------------------------------------------------------
// Generic per-plane calls
void WebPMultARGBRows(uint8_t* ptr, int stride, int width, int num_rows,
int inverse) {
int n;
for (n = 0; n < num_rows; ++n) {
WebPMultARGBRow((uint32_t*)ptr, width, inverse);
ptr += stride;
}
}
void WebPMultRows(uint8_t* ptr, int stride,
const uint8_t* alpha, int alpha_stride,
int width, int num_rows, int inverse) {
int n;
for (n = 0; n < num_rows; ++n) {
WebPMultRow(ptr, alpha, width, inverse);
ptr += stride;
alpha += alpha_stride;
}
}
//------------------------------------------------------------------------------
// Premultiplied modes
// non dithered-modes
// (x * a * 32897) >> 23 is bit-wise equivalent to (int)(x * a / 255.)
// for all 8bit x or a. For bit-wise equivalence to (int)(x * a / 255. + .5),
// one can use instead: (x * a * 65793 + (1 << 23)) >> 24
#if 1 // (int)(x * a / 255.)
#define MULTIPLIER(a) ((a) * 32897U)
#define PREMULTIPLY(x, m) (((x) * (m)) >> 23)
#else // (int)(x * a / 255. + .5)
#define MULTIPLIER(a) ((a) * 65793U)
#define PREMULTIPLY(x, m) (((x) * (m) + (1U << 23)) >> 24)
#endif
#if !WEBP_NEON_OMIT_C_CODE
static void ApplyAlphaMultiply_C(uint8_t* rgba, int alpha_first,
int w, int h, int stride) {
while (h-- > 0) {
uint8_t* const rgb = rgba + (alpha_first ? 1 : 0);
const uint8_t* const alpha = rgba + (alpha_first ? 0 : 3);
int i;
for (i = 0; i < w; ++i) {
const uint32_t a = alpha[4 * i];
if (a != 0xff) {
const uint32_t mult = MULTIPLIER(a);
rgb[4 * i + 0] = PREMULTIPLY(rgb[4 * i + 0], mult);
rgb[4 * i + 1] = PREMULTIPLY(rgb[4 * i + 1], mult);
rgb[4 * i + 2] = PREMULTIPLY(rgb[4 * i + 2], mult);
}
}
rgba += stride;
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
#undef MULTIPLIER
#undef PREMULTIPLY
// rgbA4444
#define MULTIPLIER(a) ((a) * 0x1111) // 0x1111 ~= (1 << 16) / 15
static WEBP_INLINE uint8_t dither_hi(uint8_t x) {
return (x & 0xf0) | (x >> 4);
}
static WEBP_INLINE uint8_t dither_lo(uint8_t x) {
return (x & 0x0f) | (x << 4);
}
static WEBP_INLINE uint8_t multiply(uint8_t x, uint32_t m) {
return (x * m) >> 16;
}
static WEBP_INLINE void ApplyAlphaMultiply4444_C(uint8_t* rgba4444,
int w, int h, int stride,
int rg_byte_pos /* 0 or 1 */) {
while (h-- > 0) {
int i;
for (i = 0; i < w; ++i) {
const uint32_t rg = rgba4444[2 * i + rg_byte_pos];
const uint32_t ba = rgba4444[2 * i + (rg_byte_pos ^ 1)];
const uint8_t a = ba & 0x0f;
const uint32_t mult = MULTIPLIER(a);
const uint8_t r = multiply(dither_hi(rg), mult);
const uint8_t g = multiply(dither_lo(rg), mult);
const uint8_t b = multiply(dither_hi(ba), mult);
rgba4444[2 * i + rg_byte_pos] = (r & 0xf0) | ((g >> 4) & 0x0f);
rgba4444[2 * i + (rg_byte_pos ^ 1)] = (b & 0xf0) | a;
}
rgba4444 += stride;
}
}
#undef MULTIPLIER
static void ApplyAlphaMultiply_16b_C(uint8_t* rgba4444,
int w, int h, int stride) {
#if (WEBP_SWAP_16BIT_CSP == 1)
ApplyAlphaMultiply4444_C(rgba4444, w, h, stride, 1);
#else
ApplyAlphaMultiply4444_C(rgba4444, w, h, stride, 0);
#endif
}
#if !WEBP_NEON_OMIT_C_CODE
static int DispatchAlpha_C(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint8_t* dst, int dst_stride) {
uint32_t alpha_mask = 0xff;
int i, j;
for (j = 0; j < height; ++j) {
for (i = 0; i < width; ++i) {
const uint32_t alpha_value = alpha[i];
dst[4 * i] = alpha_value;
alpha_mask &= alpha_value;
}
alpha += alpha_stride;
dst += dst_stride;
}
return (alpha_mask != 0xff);
}
static void DispatchAlphaToGreen_C(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint32_t* dst, int dst_stride) {
int i, j;
for (j = 0; j < height; ++j) {
for (i = 0; i < width; ++i) {
dst[i] = alpha[i] << 8; // leave A/R/B channels zero'd.
}
alpha += alpha_stride;
dst += dst_stride;
}
}
static int ExtractAlpha_C(const uint8_t* argb, int argb_stride,
int width, int height,
uint8_t* alpha, int alpha_stride) {
uint8_t alpha_mask = 0xff;
int i, j;
for (j = 0; j < height; ++j) {
for (i = 0; i < width; ++i) {
const uint8_t alpha_value = argb[4 * i];
alpha[i] = alpha_value;
alpha_mask &= alpha_value;
}
argb += argb_stride;
alpha += alpha_stride;
}
return (alpha_mask == 0xff);
}
static void ExtractGreen_C(const uint32_t* argb, uint8_t* alpha, int size) {
int i;
for (i = 0; i < size; ++i) alpha[i] = argb[i] >> 8;
}
#endif // !WEBP_NEON_OMIT_C_CODE
//------------------------------------------------------------------------------
static int HasAlpha8b_C(const uint8_t* src, int length) {
while (length-- > 0) if (*src++ != 0xff) return 1;
return 0;
}
static int HasAlpha32b_C(const uint8_t* src, int length) {
int x;
for (x = 0; length-- > 0; x += 4) if (src[x] != 0xff) return 1;
return 0;
}
//------------------------------------------------------------------------------
// Simple channel manipulations.
static WEBP_INLINE uint32_t MakeARGB32(int a, int r, int g, int b) {
return (((uint32_t)a << 24) | (r << 16) | (g << 8) | b);
}
#ifdef WORDS_BIGENDIAN
static void PackARGB_C(const uint8_t* a, const uint8_t* r, const uint8_t* g,
const uint8_t* b, int len, uint32_t* out) {
int i;
for (i = 0; i < len; ++i) {
out[i] = MakeARGB32(a[4 * i], r[4 * i], g[4 * i], b[4 * i]);
}
}
#endif
static void PackRGB_C(const uint8_t* r, const uint8_t* g, const uint8_t* b,
int len, int step, uint32_t* out) {
int i, offset = 0;
for (i = 0; i < len; ++i) {
out[i] = MakeARGB32(0xff, r[offset], g[offset], b[offset]);
offset += step;
}
}
void (*WebPApplyAlphaMultiply)(uint8_t*, int, int, int, int);
void (*WebPApplyAlphaMultiply4444)(uint8_t*, int, int, int);
int (*WebPDispatchAlpha)(const uint8_t*, int, int, int, uint8_t*, int);
void (*WebPDispatchAlphaToGreen)(const uint8_t*, int, int, int, uint32_t*, int);
int (*WebPExtractAlpha)(const uint8_t*, int, int, int, uint8_t*, int);
void (*WebPExtractGreen)(const uint32_t* argb, uint8_t* alpha, int size);
#ifdef WORDS_BIGENDIAN
void (*WebPPackARGB)(const uint8_t* a, const uint8_t* r, const uint8_t* g,
const uint8_t* b, int, uint32_t*);
#endif
void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
int len, int step, uint32_t* out);
int (*WebPHasAlpha8b)(const uint8_t* src, int length);
int (*WebPHasAlpha32b)(const uint8_t* src, int length);
//------------------------------------------------------------------------------
// Init function
extern void WebPInitAlphaProcessingMIPSdspR2(void);
extern void WebPInitAlphaProcessingSSE2(void);
extern void WebPInitAlphaProcessingSSE41(void);
extern void WebPInitAlphaProcessingNEON(void);
WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
WebPMultARGBRow = WebPMultARGBRow_C;
WebPMultRow = WebPMultRow_C;
WebPApplyAlphaMultiply4444 = ApplyAlphaMultiply_16b_C;
#ifdef WORDS_BIGENDIAN
WebPPackARGB = PackARGB_C;
#endif
WebPPackRGB = PackRGB_C;
#if !WEBP_NEON_OMIT_C_CODE
WebPApplyAlphaMultiply = ApplyAlphaMultiply_C;
WebPDispatchAlpha = DispatchAlpha_C;
WebPDispatchAlphaToGreen = DispatchAlphaToGreen_C;
WebPExtractAlpha = ExtractAlpha_C;
WebPExtractGreen = ExtractGreen_C;
#endif
WebPHasAlpha8b = HasAlpha8b_C;
WebPHasAlpha32b = HasAlpha32b_C;
// If defined, use CPUInfo() to overwrite some pointers with faster versions.
if (VP8GetCPUInfo != NULL) {
#if defined(WEBP_USE_SSE2)
if (VP8GetCPUInfo(kSSE2)) {
WebPInitAlphaProcessingSSE2();
#if defined(WEBP_USE_SSE41)
if (VP8GetCPUInfo(kSSE4_1)) {
WebPInitAlphaProcessingSSE41();
}
#endif
}
#endif
#if defined(WEBP_USE_MIPS_DSP_R2)
if (VP8GetCPUInfo(kMIPSdspR2)) {
WebPInitAlphaProcessingMIPSdspR2();
}
#endif
}
#if defined(WEBP_USE_NEON)
if (WEBP_NEON_OMIT_C_CODE ||
(VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
WebPInitAlphaProcessingNEON();
}
#endif
assert(WebPMultARGBRow != NULL);
assert(WebPMultRow != NULL);
assert(WebPApplyAlphaMultiply != NULL);
assert(WebPApplyAlphaMultiply4444 != NULL);
assert(WebPDispatchAlpha != NULL);
assert(WebPDispatchAlphaToGreen != NULL);
assert(WebPExtractAlpha != NULL);
assert(WebPExtractGreen != NULL);
#ifdef WORDS_BIGENDIAN
assert(WebPPackARGB != NULL);
#endif
assert(WebPPackRGB != NULL);
assert(WebPHasAlpha8b != NULL);
assert(WebPHasAlpha32b != NULL);
}

228
media/libwebp/src/dsp/alpha_processing_mips_dsp_r2.c Normal file
Просмотреть файл

@ -0,0 +1,228 @@
// Copyright 2014 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Utilities for processing transparent channel.
//
// Author(s): Branimir Vasic (branimir.vasic@imgtec.com)
// Djordje Pesut (djordje.pesut@imgtec.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_MIPS_DSP_R2)
static int DispatchAlpha_MIPSdspR2(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint8_t* dst, int dst_stride) {
uint32_t alpha_mask = 0xffffffff;
int i, j, temp0;
for (j = 0; j < height; ++j) {
uint8_t* pdst = dst;
const uint8_t* palpha = alpha;
for (i = 0; i < (width >> 2); ++i) {
int temp1, temp2, temp3;
__asm__ volatile (
"ulw %[temp0], 0(%[palpha]) \n\t"
"addiu %[palpha], %[palpha], 4 \n\t"
"addiu %[pdst], %[pdst], 16 \n\t"
"srl %[temp1], %[temp0], 8 \n\t"
"srl %[temp2], %[temp0], 16 \n\t"
"srl %[temp3], %[temp0], 24 \n\t"
"and %[alpha_mask], %[alpha_mask], %[temp0] \n\t"
"sb %[temp0], -16(%[pdst]) \n\t"
"sb %[temp1], -12(%[pdst]) \n\t"
"sb %[temp2], -8(%[pdst]) \n\t"
"sb %[temp3], -4(%[pdst]) \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [palpha]"+r"(palpha), [pdst]"+r"(pdst),
[alpha_mask]"+r"(alpha_mask)
:
: "memory"
);
}
for (i = 0; i < (width & 3); ++i) {
__asm__ volatile (
"lbu %[temp0], 0(%[palpha]) \n\t"
"addiu %[palpha], %[palpha], 1 \n\t"
"sb %[temp0], 0(%[pdst]) \n\t"
"and %[alpha_mask], %[alpha_mask], %[temp0] \n\t"
"addiu %[pdst], %[pdst], 4 \n\t"
: [temp0]"=&r"(temp0), [palpha]"+r"(palpha), [pdst]"+r"(pdst),
[alpha_mask]"+r"(alpha_mask)
:
: "memory"
);
}
alpha += alpha_stride;
dst += dst_stride;
}
__asm__ volatile (
"ext %[temp0], %[alpha_mask], 0, 16 \n\t"
"srl %[alpha_mask], %[alpha_mask], 16 \n\t"
"and %[alpha_mask], %[alpha_mask], %[temp0] \n\t"
"ext %[temp0], %[alpha_mask], 0, 8 \n\t"
"srl %[alpha_mask], %[alpha_mask], 8 \n\t"
"and %[alpha_mask], %[alpha_mask], %[temp0] \n\t"
: [temp0]"=&r"(temp0), [alpha_mask]"+r"(alpha_mask)
:
);
return (alpha_mask != 0xff);
}
static void MultARGBRow_MIPSdspR2(uint32_t* const ptr, int width,
int inverse) {
int x;
const uint32_t c_00ffffff = 0x00ffffffu;
const uint32_t c_ff000000 = 0xff000000u;
const uint32_t c_8000000 = 0x00800000u;
const uint32_t c_8000080 = 0x00800080u;
for (x = 0; x < width; ++x) {
const uint32_t argb = ptr[x];
if (argb < 0xff000000u) { // alpha < 255
if (argb <= 0x00ffffffu) { // alpha == 0
ptr[x] = 0;
} else {
int temp0, temp1, temp2, temp3, alpha;
__asm__ volatile (
"srl %[alpha], %[argb], 24 \n\t"
"replv.qb %[temp0], %[alpha] \n\t"
"and %[temp0], %[temp0], %[c_00ffffff] \n\t"
"beqz %[inverse], 0f \n\t"
"divu $zero, %[c_ff000000], %[alpha] \n\t"
"mflo %[temp0] \n\t"
"0: \n\t"
"andi %[temp1], %[argb], 0xff \n\t"
"ext %[temp2], %[argb], 8, 8 \n\t"
"ext %[temp3], %[argb], 16, 8 \n\t"
"mul %[temp1], %[temp1], %[temp0] \n\t"
"mul %[temp2], %[temp2], %[temp0] \n\t"
"mul %[temp3], %[temp3], %[temp0] \n\t"
"precrq.ph.w %[temp1], %[temp2], %[temp1] \n\t"
"addu %[temp3], %[temp3], %[c_8000000] \n\t"
"addu %[temp1], %[temp1], %[c_8000080] \n\t"
"precrq.ph.w %[temp3], %[argb], %[temp3] \n\t"
"precrq.qb.ph %[temp1], %[temp3], %[temp1] \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [alpha]"=&r"(alpha)
: [inverse]"r"(inverse), [c_00ffffff]"r"(c_00ffffff),
[c_8000000]"r"(c_8000000), [c_8000080]"r"(c_8000080),
[c_ff000000]"r"(c_ff000000), [argb]"r"(argb)
: "memory", "hi", "lo"
);
ptr[x] = temp1;
}
}
}
}
#ifdef WORDS_BIGENDIAN
static void PackARGB_MIPSdspR2(const uint8_t* a, const uint8_t* r,
const uint8_t* g, const uint8_t* b, int len,
uint32_t* out) {
int temp0, temp1, temp2, temp3, offset;
const int rest = len & 1;
const uint32_t* const loop_end = out + len - rest;
const int step = 4;
__asm__ volatile (
"xor %[offset], %[offset], %[offset] \n\t"
"beq %[loop_end], %[out], 0f \n\t"
"2: \n\t"
"lbux %[temp0], %[offset](%[a]) \n\t"
"lbux %[temp1], %[offset](%[r]) \n\t"
"lbux %[temp2], %[offset](%[g]) \n\t"
"lbux %[temp3], %[offset](%[b]) \n\t"
"ins %[temp1], %[temp0], 16, 16 \n\t"
"ins %[temp3], %[temp2], 16, 16 \n\t"
"addiu %[out], %[out], 4 \n\t"
"precr.qb.ph %[temp0], %[temp1], %[temp3] \n\t"
"sw %[temp0], -4(%[out]) \n\t"
"addu %[offset], %[offset], %[step] \n\t"
"bne %[loop_end], %[out], 2b \n\t"
"0: \n\t"
"beq %[rest], $zero, 1f \n\t"
"lbux %[temp0], %[offset](%[a]) \n\t"
"lbux %[temp1], %[offset](%[r]) \n\t"
"lbux %[temp2], %[offset](%[g]) \n\t"
"lbux %[temp3], %[offset](%[b]) \n\t"
"ins %[temp1], %[temp0], 16, 16 \n\t"
"ins %[temp3], %[temp2], 16, 16 \n\t"
"precr.qb.ph %[temp0], %[temp1], %[temp3] \n\t"
"sw %[temp0], 0(%[out]) \n\t"
"1: \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [offset]"=&r"(offset), [out]"+&r"(out)
: [a]"r"(a), [r]"r"(r), [g]"r"(g), [b]"r"(b), [step]"r"(step),
[loop_end]"r"(loop_end), [rest]"r"(rest)
: "memory"
);
}
#endif // WORDS_BIGENDIAN
static void PackRGB_MIPSdspR2(const uint8_t* r, const uint8_t* g,
const uint8_t* b, int len, int step,
uint32_t* out) {
int temp0, temp1, temp2, offset;
const int rest = len & 1;
const int a = 0xff;
const uint32_t* const loop_end = out + len - rest;
__asm__ volatile (
"xor %[offset], %[offset], %[offset] \n\t"
"beq %[loop_end], %[out], 0f \n\t"
"2: \n\t"
"lbux %[temp0], %[offset](%[r]) \n\t"
"lbux %[temp1], %[offset](%[g]) \n\t"
"lbux %[temp2], %[offset](%[b]) \n\t"
"ins %[temp0], %[a], 16, 16 \n\t"
"ins %[temp2], %[temp1], 16, 16 \n\t"
"addiu %[out], %[out], 4 \n\t"
"precr.qb.ph %[temp0], %[temp0], %[temp2] \n\t"
"sw %[temp0], -4(%[out]) \n\t"
"addu %[offset], %[offset], %[step] \n\t"
"bne %[loop_end], %[out], 2b \n\t"
"0: \n\t"
"beq %[rest], $zero, 1f \n\t"
"lbux %[temp0], %[offset](%[r]) \n\t"
"lbux %[temp1], %[offset](%[g]) \n\t"
"lbux %[temp2], %[offset](%[b]) \n\t"
"ins %[temp0], %[a], 16, 16 \n\t"
"ins %[temp2], %[temp1], 16, 16 \n\t"
"precr.qb.ph %[temp0], %[temp0], %[temp2] \n\t"
"sw %[temp0], 0(%[out]) \n\t"
"1: \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[offset]"=&r"(offset), [out]"+&r"(out)
: [a]"r"(a), [r]"r"(r), [g]"r"(g), [b]"r"(b), [step]"r"(step),
[loop_end]"r"(loop_end), [rest]"r"(rest)
: "memory"
);
}
//------------------------------------------------------------------------------
// Entry point
extern void WebPInitAlphaProcessingMIPSdspR2(void);
WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingMIPSdspR2(void) {
WebPDispatchAlpha = DispatchAlpha_MIPSdspR2;
WebPMultARGBRow = MultARGBRow_MIPSdspR2;
#ifdef WORDS_BIGENDIAN
WebPPackARGB = PackARGB_MIPSdspR2;
#endif
WebPPackRGB = PackRGB_MIPSdspR2;
}
#else // !WEBP_USE_MIPS_DSP_R2
WEBP_DSP_INIT_STUB(WebPInitAlphaProcessingMIPSdspR2)
#endif // WEBP_USE_MIPS_DSP_R2

191
media/libwebp/src/dsp/alpha_processing_neon.c Normal file
Просмотреть файл

@ -0,0 +1,191 @@
// Copyright 2017 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Utilities for processing transparent channel, NEON version.
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_NEON)
#include "src/dsp/neon.h"
//------------------------------------------------------------------------------
#define MULTIPLIER(a) ((a) * 0x8081)
#define PREMULTIPLY(x, m) (((x) * (m)) >> 23)
#define MULTIPLY_BY_ALPHA(V, ALPHA, OTHER) do { \
const uint8x8_t alpha = (V).val[(ALPHA)]; \
const uint16x8_t r1 = vmull_u8((V).val[1], alpha); \
const uint16x8_t g1 = vmull_u8((V).val[2], alpha); \
const uint16x8_t b1 = vmull_u8((V).val[(OTHER)], alpha); \
/* we use: v / 255 = (v + 1 + (v >> 8)) >> 8 */ \
const uint16x8_t r2 = vsraq_n_u16(r1, r1, 8); \
const uint16x8_t g2 = vsraq_n_u16(g1, g1, 8); \
const uint16x8_t b2 = vsraq_n_u16(b1, b1, 8); \
const uint16x8_t r3 = vaddq_u16(r2, kOne); \
const uint16x8_t g3 = vaddq_u16(g2, kOne); \
const uint16x8_t b3 = vaddq_u16(b2, kOne); \
(V).val[1] = vshrn_n_u16(r3, 8); \
(V).val[2] = vshrn_n_u16(g3, 8); \
(V).val[(OTHER)] = vshrn_n_u16(b3, 8); \
} while (0)
static void ApplyAlphaMultiply_NEON(uint8_t* rgba, int alpha_first,
int w, int h, int stride) {
const uint16x8_t kOne = vdupq_n_u16(1u);
while (h-- > 0) {
uint32_t* const rgbx = (uint32_t*)rgba;
int i = 0;
if (alpha_first) {
for (; i + 8 <= w; i += 8) {
// load aaaa...|rrrr...|gggg...|bbbb...
uint8x8x4_t RGBX = vld4_u8((const uint8_t*)(rgbx + i));
MULTIPLY_BY_ALPHA(RGBX, 0, 3);
vst4_u8((uint8_t*)(rgbx + i), RGBX);
}
} else {
for (; i + 8 <= w; i += 8) {
uint8x8x4_t RGBX = vld4_u8((const uint8_t*)(rgbx + i));
MULTIPLY_BY_ALPHA(RGBX, 3, 0);
vst4_u8((uint8_t*)(rgbx + i), RGBX);
}
}
// Finish with left-overs.
for (; i < w; ++i) {
uint8_t* const rgb = rgba + (alpha_first ? 1 : 0);
const uint8_t* const alpha = rgba + (alpha_first ? 0 : 3);
const uint32_t a = alpha[4 * i];
if (a != 0xff) {
const uint32_t mult = MULTIPLIER(a);
rgb[4 * i + 0] = PREMULTIPLY(rgb[4 * i + 0], mult);
rgb[4 * i + 1] = PREMULTIPLY(rgb[4 * i + 1], mult);
rgb[4 * i + 2] = PREMULTIPLY(rgb[4 * i + 2], mult);
}
}
rgba += stride;
}
}
#undef MULTIPLY_BY_ALPHA
#undef MULTIPLIER
#undef PREMULTIPLY
//------------------------------------------------------------------------------
static int DispatchAlpha_NEON(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint8_t* dst, int dst_stride) {
uint32_t alpha_mask = 0xffffffffu;
uint8x8_t mask8 = vdup_n_u8(0xff);
uint32_t tmp[2];
int i, j;
for (j = 0; j < height; ++j) {
// We don't know if alpha is first or last in dst[] (depending on rgbA/Argb
// mode). So we must be sure dst[4*i + 8 - 1] is writable for the store.
// Hence the test with 'width - 1' instead of just 'width'.
for (i = 0; i + 8 <= width - 1; i += 8) {
uint8x8x4_t rgbX = vld4_u8((const uint8_t*)(dst + 4 * i));
const uint8x8_t alphas = vld1_u8(alpha + i);
rgbX.val[0] = alphas;
vst4_u8((uint8_t*)(dst + 4 * i), rgbX);
mask8 = vand_u8(mask8, alphas);
}
for (; i < width; ++i) {
const uint32_t alpha_value = alpha[i];
dst[4 * i] = alpha_value;
alpha_mask &= alpha_value;
}
alpha += alpha_stride;
dst += dst_stride;
}
vst1_u8((uint8_t*)tmp, mask8);
alpha_mask &= tmp[0];
alpha_mask &= tmp[1];
return (alpha_mask != 0xffffffffu);
}
static void DispatchAlphaToGreen_NEON(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint32_t* dst, int dst_stride) {
int i, j;
uint8x8x4_t greens; // leave A/R/B channels zero'd.
greens.val[0] = vdup_n_u8(0);
greens.val[2] = vdup_n_u8(0);
greens.val[3] = vdup_n_u8(0);
for (j = 0; j < height; ++j) {
for (i = 0; i + 8 <= width; i += 8) {
greens.val[1] = vld1_u8(alpha + i);
vst4_u8((uint8_t*)(dst + i), greens);
}
for (; i < width; ++i) dst[i] = alpha[i] << 8;
alpha += alpha_stride;
dst += dst_stride;
}
}
static int ExtractAlpha_NEON(const uint8_t* argb, int argb_stride,
int width, int height,
uint8_t* alpha, int alpha_stride) {
uint32_t alpha_mask = 0xffffffffu;
uint8x8_t mask8 = vdup_n_u8(0xff);
uint32_t tmp[2];
int i, j;
for (j = 0; j < height; ++j) {
// We don't know if alpha is first or last in dst[] (depending on rgbA/Argb
// mode). So we must be sure dst[4*i + 8 - 1] is writable for the store.
// Hence the test with 'width - 1' instead of just 'width'.
for (i = 0; i + 8 <= width - 1; i += 8) {
const uint8x8x4_t rgbX = vld4_u8((const uint8_t*)(argb + 4 * i));
const uint8x8_t alphas = rgbX.val[0];
vst1_u8((uint8_t*)(alpha + i), alphas);
mask8 = vand_u8(mask8, alphas);
}
for (; i < width; ++i) {
alpha[i] = argb[4 * i];
alpha_mask &= alpha[i];
}
argb += argb_stride;
alpha += alpha_stride;
}
vst1_u8((uint8_t*)tmp, mask8);
alpha_mask &= tmp[0];
alpha_mask &= tmp[1];
return (alpha_mask == 0xffffffffu);
}
static void ExtractGreen_NEON(const uint32_t* argb,
uint8_t* alpha, int size) {
int i;
for (i = 0; i + 16 <= size; i += 16) {
const uint8x16x4_t rgbX = vld4q_u8((const uint8_t*)(argb + i));
const uint8x16_t greens = rgbX.val[1];
vst1q_u8(alpha + i, greens);
}
for (; i < size; ++i) alpha[i] = (argb[i] >> 8) & 0xff;
}
//------------------------------------------------------------------------------
extern void WebPInitAlphaProcessingNEON(void);
WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingNEON(void) {
WebPApplyAlphaMultiply = ApplyAlphaMultiply_NEON;
WebPDispatchAlpha = DispatchAlpha_NEON;
WebPDispatchAlphaToGreen = DispatchAlphaToGreen_NEON;
WebPExtractAlpha = ExtractAlpha_NEON;
WebPExtractGreen = ExtractGreen_NEON;
}
#else // !WEBP_USE_NEON
WEBP_DSP_INIT_STUB(WebPInitAlphaProcessingNEON)
#endif // WEBP_USE_NEON

343
media/libwebp/src/dsp/alpha_processing_sse2.c Normal file
Просмотреть файл

@ -0,0 +1,343 @@
// Copyright 2014 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Utilities for processing transparent channel.
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_SSE2)
#include <emmintrin.h>
//------------------------------------------------------------------------------
static int DispatchAlpha_SSE2(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint8_t* dst, int dst_stride) {
// alpha_and stores an 'and' operation of all the alpha[] values. The final
// value is not 0xff if any of the alpha[] is not equal to 0xff.
uint32_t alpha_and = 0xff;
int i, j;
const __m128i zero = _mm_setzero_si128();
const __m128i rgb_mask = _mm_set1_epi32(0xffffff00u); // to preserve RGB
const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u);
__m128i all_alphas = all_0xff;
// We must be able to access 3 extra bytes after the last written byte
// 'dst[4 * width - 4]', because we don't know if alpha is the first or the
// last byte of the quadruplet.
const int limit = (width - 1) & ~7;
for (j = 0; j < height; ++j) {
__m128i* out = (__m128i*)dst;
for (i = 0; i < limit; i += 8) {
// load 8 alpha bytes
const __m128i a0 = _mm_loadl_epi64((const __m128i*)&alpha[i]);
const __m128i a1 = _mm_unpacklo_epi8(a0, zero);
const __m128i a2_lo = _mm_unpacklo_epi16(a1, zero);
const __m128i a2_hi = _mm_unpackhi_epi16(a1, zero);
// load 8 dst pixels (32 bytes)
const __m128i b0_lo = _mm_loadu_si128(out + 0);
const __m128i b0_hi = _mm_loadu_si128(out + 1);
// mask dst alpha values
const __m128i b1_lo = _mm_and_si128(b0_lo, rgb_mask);
const __m128i b1_hi = _mm_and_si128(b0_hi, rgb_mask);
// combine
const __m128i b2_lo = _mm_or_si128(b1_lo, a2_lo);
const __m128i b2_hi = _mm_or_si128(b1_hi, a2_hi);
// store
_mm_storeu_si128(out + 0, b2_lo);
_mm_storeu_si128(out + 1, b2_hi);
// accumulate eight alpha 'and' in parallel
all_alphas = _mm_and_si128(all_alphas, a0);
out += 2;
}
for (; i < width; ++i) {
const uint32_t alpha_value = alpha[i];
dst[4 * i] = alpha_value;
alpha_and &= alpha_value;
}
alpha += alpha_stride;
dst += dst_stride;
}
// Combine the eight alpha 'and' into a 8-bit mask.
alpha_and &= _mm_movemask_epi8(_mm_cmpeq_epi8(all_alphas, all_0xff));
return (alpha_and != 0xff);
}
static void DispatchAlphaToGreen_SSE2(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint32_t* dst, int dst_stride) {
int i, j;
const __m128i zero = _mm_setzero_si128();
const int limit = width & ~15;
for (j = 0; j < height; ++j) {
for (i = 0; i < limit; i += 16) { // process 16 alpha bytes
const __m128i a0 = _mm_loadu_si128((const __m128i*)&alpha[i]);
const __m128i a1 = _mm_unpacklo_epi8(zero, a0); // note the 'zero' first!
const __m128i b1 = _mm_unpackhi_epi8(zero, a0);
const __m128i a2_lo = _mm_unpacklo_epi16(a1, zero);
const __m128i b2_lo = _mm_unpacklo_epi16(b1, zero);
const __m128i a2_hi = _mm_unpackhi_epi16(a1, zero);
const __m128i b2_hi = _mm_unpackhi_epi16(b1, zero);
_mm_storeu_si128((__m128i*)&dst[i + 0], a2_lo);
_mm_storeu_si128((__m128i*)&dst[i + 4], a2_hi);
_mm_storeu_si128((__m128i*)&dst[i + 8], b2_lo);
_mm_storeu_si128((__m128i*)&dst[i + 12], b2_hi);
}
for (; i < width; ++i) dst[i] = alpha[i] << 8;
alpha += alpha_stride;
dst += dst_stride;
}
}
static int ExtractAlpha_SSE2(const uint8_t* argb, int argb_stride,
int width, int height,
uint8_t* alpha, int alpha_stride) {
// alpha_and stores an 'and' operation of all the alpha[] values. The final
// value is not 0xff if any of the alpha[] is not equal to 0xff.
uint32_t alpha_and = 0xff;
int i, j;
const __m128i a_mask = _mm_set1_epi32(0xffu); // to preserve alpha
const __m128i all_0xff = _mm_set_epi32(0, 0, ~0u, ~0u);
__m128i all_alphas = all_0xff;
// We must be able to access 3 extra bytes after the last written byte
// 'src[4 * width - 4]', because we don't know if alpha is the first or the
// last byte of the quadruplet.
const int limit = (width - 1) & ~7;
for (j = 0; j < height; ++j) {
const __m128i* src = (const __m128i*)argb;
for (i = 0; i < limit; i += 8) {
// load 32 argb bytes
const __m128i a0 = _mm_loadu_si128(src + 0);
const __m128i a1 = _mm_loadu_si128(src + 1);
const __m128i b0 = _mm_and_si128(a0, a_mask);
const __m128i b1 = _mm_and_si128(a1, a_mask);
const __m128i c0 = _mm_packs_epi32(b0, b1);
const __m128i d0 = _mm_packus_epi16(c0, c0);
// store
_mm_storel_epi64((__m128i*)&alpha[i], d0);
// accumulate eight alpha 'and' in parallel
all_alphas = _mm_and_si128(all_alphas, d0);
src += 2;
}
for (; i < width; ++i) {
const uint32_t alpha_value = argb[4 * i];
alpha[i] = alpha_value;
alpha_and &= alpha_value;
}
argb += argb_stride;
alpha += alpha_stride;
}
// Combine the eight alpha 'and' into a 8-bit mask.
alpha_and &= _mm_movemask_epi8(_mm_cmpeq_epi8(all_alphas, all_0xff));
return (alpha_and == 0xff);
}
//------------------------------------------------------------------------------
// Non-dither premultiplied modes
#define MULTIPLIER(a) ((a) * 0x8081)
#define PREMULTIPLY(x, m) (((x) * (m)) >> 23)
// We can't use a 'const int' for the SHUFFLE value, because it has to be an
// immediate in the _mm_shufflexx_epi16() instruction. We really need a macro.
// We use: v / 255 = (v * 0x8081) >> 23, where v = alpha * {r,g,b} is a 16bit
// value.
#define APPLY_ALPHA(RGBX, SHUFFLE) do { \
const __m128i argb0 = _mm_loadu_si128((const __m128i*)&(RGBX)); \
const __m128i argb1_lo = _mm_unpacklo_epi8(argb0, zero); \
const __m128i argb1_hi = _mm_unpackhi_epi8(argb0, zero); \
const __m128i alpha0_lo = _mm_or_si128(argb1_lo, kMask); \
const __m128i alpha0_hi = _mm_or_si128(argb1_hi, kMask); \
const __m128i alpha1_lo = _mm_shufflelo_epi16(alpha0_lo, SHUFFLE); \
const __m128i alpha1_hi = _mm_shufflelo_epi16(alpha0_hi, SHUFFLE); \
const __m128i alpha2_lo = _mm_shufflehi_epi16(alpha1_lo, SHUFFLE); \
const __m128i alpha2_hi = _mm_shufflehi_epi16(alpha1_hi, SHUFFLE); \
/* alpha2 = [ff a0 a0 a0][ff a1 a1 a1] */ \
const __m128i A0_lo = _mm_mullo_epi16(alpha2_lo, argb1_lo); \
const __m128i A0_hi = _mm_mullo_epi16(alpha2_hi, argb1_hi); \
const __m128i A1_lo = _mm_mulhi_epu16(A0_lo, kMult); \
const __m128i A1_hi = _mm_mulhi_epu16(A0_hi, kMult); \
const __m128i A2_lo = _mm_srli_epi16(A1_lo, 7); \
const __m128i A2_hi = _mm_srli_epi16(A1_hi, 7); \
const __m128i A3 = _mm_packus_epi16(A2_lo, A2_hi); \
_mm_storeu_si128((__m128i*)&(RGBX), A3); \
} while (0)
static void ApplyAlphaMultiply_SSE2(uint8_t* rgba, int alpha_first,
int w, int h, int stride) {
const __m128i zero = _mm_setzero_si128();
const __m128i kMult = _mm_set1_epi16(0x8081u);
const __m128i kMask = _mm_set_epi16(0, 0xff, 0xff, 0, 0, 0xff, 0xff, 0);
const int kSpan = 4;
while (h-- > 0) {
uint32_t* const rgbx = (uint32_t*)rgba;
int i;
if (!alpha_first) {
for (i = 0; i + kSpan <= w; i += kSpan) {
APPLY_ALPHA(rgbx[i], _MM_SHUFFLE(2, 3, 3, 3));
}
} else {
for (i = 0; i + kSpan <= w; i += kSpan) {
APPLY_ALPHA(rgbx[i], _MM_SHUFFLE(0, 0, 0, 1));
}
}
// Finish with left-overs.
for (; i < w; ++i) {
uint8_t* const rgb = rgba + (alpha_first ? 1 : 0);
const uint8_t* const alpha = rgba + (alpha_first ? 0 : 3);
const uint32_t a = alpha[4 * i];
if (a != 0xff) {
const uint32_t mult = MULTIPLIER(a);
rgb[4 * i + 0] = PREMULTIPLY(rgb[4 * i + 0], mult);
rgb[4 * i + 1] = PREMULTIPLY(rgb[4 * i + 1], mult);
rgb[4 * i + 2] = PREMULTIPLY(rgb[4 * i + 2], mult);
}
}
rgba += stride;
}
}
#undef MULTIPLIER
#undef PREMULTIPLY
//------------------------------------------------------------------------------
// Alpha detection
static int HasAlpha8b_SSE2(const uint8_t* src, int length) {
const __m128i all_0xff = _mm_set1_epi8(0xff);
int i = 0;
for (; i + 16 <= length; i += 16) {
const __m128i v = _mm_loadu_si128((const __m128i*)(src + i));
const __m128i bits = _mm_cmpeq_epi8(v, all_0xff);
const int mask = _mm_movemask_epi8(bits);
if (mask != 0xffff) return 1;
}
for (; i < length; ++i) if (src[i] != 0xff) return 1;
return 0;
}
static int HasAlpha32b_SSE2(const uint8_t* src, int length) {
const __m128i alpha_mask = _mm_set1_epi32(0xff);
const __m128i all_0xff = _mm_set1_epi8(0xff);
int i = 0;
// We don't know if we can access the last 3 bytes after the last alpha
// value 'src[4 * length - 4]' (because we don't know if alpha is the first
// or the last byte of the quadruplet). Hence the '-3' protection below.
length = length * 4 - 3; // size in bytes
for (; i + 64 <= length; i += 64) {
const __m128i a0 = _mm_loadu_si128((const __m128i*)(src + i + 0));
const __m128i a1 = _mm_loadu_si128((const __m128i*)(src + i + 16));
const __m128i a2 = _mm_loadu_si128((const __m128i*)(src + i + 32));
const __m128i a3 = _mm_loadu_si128((const __m128i*)(src + i + 48));
const __m128i b0 = _mm_and_si128(a0, alpha_mask);
const __m128i b1 = _mm_and_si128(a1, alpha_mask);
const __m128i b2 = _mm_and_si128(a2, alpha_mask);
const __m128i b3 = _mm_and_si128(a3, alpha_mask);
const __m128i c0 = _mm_packs_epi32(b0, b1);
const __m128i c1 = _mm_packs_epi32(b2, b3);
const __m128i d = _mm_packus_epi16(c0, c1);
const __m128i bits = _mm_cmpeq_epi8(d, all_0xff);
const int mask = _mm_movemask_epi8(bits);
if (mask != 0xffff) return 1;
}
for (; i + 32 <= length; i += 32) {
const __m128i a0 = _mm_loadu_si128((const __m128i*)(src + i + 0));
const __m128i a1 = _mm_loadu_si128((const __m128i*)(src + i + 16));
const __m128i b0 = _mm_and_si128(a0, alpha_mask);
const __m128i b1 = _mm_and_si128(a1, alpha_mask);
const __m128i c = _mm_packs_epi32(b0, b1);
const __m128i d = _mm_packus_epi16(c, c);
const __m128i bits = _mm_cmpeq_epi8(d, all_0xff);
const int mask = _mm_movemask_epi8(bits);
if (mask != 0xffff) return 1;
}
for (; i <= length; i += 4) if (src[i] != 0xff) return 1;
return 0;
}
// -----------------------------------------------------------------------------
// Apply alpha value to rows
static void MultARGBRow_SSE2(uint32_t* const ptr, int width, int inverse) {
int x = 0;
if (!inverse) {
const int kSpan = 2;
const __m128i zero = _mm_setzero_si128();
const __m128i k128 = _mm_set1_epi16(128);
const __m128i kMult = _mm_set1_epi16(0x0101);
const __m128i kMask = _mm_set_epi16(0, 0xff, 0, 0, 0, 0xff, 0, 0);
for (x = 0; x + kSpan <= width; x += kSpan) {
// To compute 'result = (int)(a * x / 255. + .5)', we use:
// tmp = a * v + 128, result = (tmp * 0x0101u) >> 16
const __m128i A0 = _mm_loadl_epi64((const __m128i*)&ptr[x]);
const __m128i A1 = _mm_unpacklo_epi8(A0, zero);
const __m128i A2 = _mm_or_si128(A1, kMask);
const __m128i A3 = _mm_shufflelo_epi16(A2, _MM_SHUFFLE(2, 3, 3, 3));
const __m128i A4 = _mm_shufflehi_epi16(A3, _MM_SHUFFLE(2, 3, 3, 3));
// here, A4 = [ff a0 a0 a0][ff a1 a1 a1]
const __m128i A5 = _mm_mullo_epi16(A4, A1);
const __m128i A6 = _mm_add_epi16(A5, k128);
const __m128i A7 = _mm_mulhi_epu16(A6, kMult);
const __m128i A10 = _mm_packus_epi16(A7, zero);
_mm_storel_epi64((__m128i*)&ptr[x], A10);
}
}
width -= x;
if (width > 0) WebPMultARGBRow_C(ptr + x, width, inverse);
}
static void MultRow_SSE2(uint8_t* const ptr, const uint8_t* const alpha,
int width, int inverse) {
int x = 0;
if (!inverse) {
const __m128i zero = _mm_setzero_si128();
const __m128i k128 = _mm_set1_epi16(128);
const __m128i kMult = _mm_set1_epi16(0x0101);
for (x = 0; x + 8 <= width; x += 8) {
const __m128i v0 = _mm_loadl_epi64((__m128i*)&ptr[x]);
const __m128i a0 = _mm_loadl_epi64((const __m128i*)&alpha[x]);
const __m128i v1 = _mm_unpacklo_epi8(v0, zero);
const __m128i a1 = _mm_unpacklo_epi8(a0, zero);
const __m128i v2 = _mm_mullo_epi16(v1, a1);
const __m128i v3 = _mm_add_epi16(v2, k128);
const __m128i v4 = _mm_mulhi_epu16(v3, kMult);
const __m128i v5 = _mm_packus_epi16(v4, zero);
_mm_storel_epi64((__m128i*)&ptr[x], v5);
}
}
width -= x;
if (width > 0) WebPMultRow_C(ptr + x, alpha + x, width, inverse);
}
//------------------------------------------------------------------------------
// Entry point
extern void WebPInitAlphaProcessingSSE2(void);
WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingSSE2(void) {
WebPMultARGBRow = MultARGBRow_SSE2;
WebPMultRow = MultRow_SSE2;
WebPApplyAlphaMultiply = ApplyAlphaMultiply_SSE2;
WebPDispatchAlpha = DispatchAlpha_SSE2;
WebPDispatchAlphaToGreen = DispatchAlphaToGreen_SSE2;
WebPExtractAlpha = ExtractAlpha_SSE2;
WebPHasAlpha8b = HasAlpha8b_SSE2;
WebPHasAlpha32b = HasAlpha32b_SSE2;
}
#else // !WEBP_USE_SSE2
WEBP_DSP_INIT_STUB(WebPInitAlphaProcessingSSE2)
#endif // WEBP_USE_SSE2

92
media/libwebp/src/dsp/alpha_processing_sse41.c Normal file
Просмотреть файл

@ -0,0 +1,92 @@
// Copyright 2015 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Utilities for processing transparent channel, SSE4.1 variant.
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_SSE41)
#include <smmintrin.h>
//------------------------------------------------------------------------------
static int ExtractAlpha_SSE41(const uint8_t* argb, int argb_stride,
int width, int height,
uint8_t* alpha, int alpha_stride) {
// alpha_and stores an 'and' operation of all the alpha[] values. The final
// value is not 0xff if any of the alpha[] is not equal to 0xff.
uint32_t alpha_and = 0xff;
int i, j;
const __m128i all_0xff = _mm_set1_epi32(~0u);
__m128i all_alphas = all_0xff;
// We must be able to access 3 extra bytes after the last written byte
// 'src[4 * width - 4]', because we don't know if alpha is the first or the
// last byte of the quadruplet.
const int limit = (width - 1) & ~15;
const __m128i kCstAlpha0 = _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1,
-1, -1, -1, -1, 12, 8, 4, 0);
const __m128i kCstAlpha1 = _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1,
12, 8, 4, 0, -1, -1, -1, -1);
const __m128i kCstAlpha2 = _mm_set_epi8(-1, -1, -1, -1, 12, 8, 4, 0,
-1, -1, -1, -1, -1, -1, -1, -1);
const __m128i kCstAlpha3 = _mm_set_epi8(12, 8, 4, 0, -1, -1, -1, -1,
-1, -1, -1, -1, -1, -1, -1, -1);
for (j = 0; j < height; ++j) {
const __m128i* src = (const __m128i*)argb;
for (i = 0; i < limit; i += 16) {
// load 64 argb bytes
const __m128i a0 = _mm_loadu_si128(src + 0);
const __m128i a1 = _mm_loadu_si128(src + 1);
const __m128i a2 = _mm_loadu_si128(src + 2);
const __m128i a3 = _mm_loadu_si128(src + 3);
const __m128i b0 = _mm_shuffle_epi8(a0, kCstAlpha0);
const __m128i b1 = _mm_shuffle_epi8(a1, kCstAlpha1);
const __m128i b2 = _mm_shuffle_epi8(a2, kCstAlpha2);
const __m128i b3 = _mm_shuffle_epi8(a3, kCstAlpha3);
const __m128i c0 = _mm_or_si128(b0, b1);
const __m128i c1 = _mm_or_si128(b2, b3);
const __m128i d0 = _mm_or_si128(c0, c1);
// store
_mm_storeu_si128((__m128i*)&alpha[i], d0);
// accumulate sixteen alpha 'and' in parallel
all_alphas = _mm_and_si128(all_alphas, d0);
src += 4;
}
for (; i < width; ++i) {
const uint32_t alpha_value = argb[4 * i];
alpha[i] = alpha_value;
alpha_and &= alpha_value;
}
argb += argb_stride;
alpha += alpha_stride;
}
// Combine the sixteen alpha 'and' into an 8-bit mask.
alpha_and |= 0xff00u; // pretend the upper bits [8..15] were tested ok.
alpha_and &= _mm_movemask_epi8(_mm_cmpeq_epi8(all_alphas, all_0xff));
return (alpha_and == 0xffffu);
}
//------------------------------------------------------------------------------
// Entry point
extern void WebPInitAlphaProcessingSSE41(void);
WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingSSE41(void) {
WebPExtractAlpha = ExtractAlpha_SSE41;
}
#else // !WEBP_USE_SSE41
WEBP_DSP_INIT_STUB(WebPInitAlphaProcessingSSE41)
#endif // WEBP_USE_SSE41

194
media/libwebp/src/dsp/common_sse2.h Normal file
Просмотреть файл

@ -0,0 +1,194 @@
// Copyright 2016 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// SSE2 code common to several files.
//
// Author: Vincent Rabaud (vrabaud@google.com)
#ifndef WEBP_DSP_COMMON_SSE2_H_
#define WEBP_DSP_COMMON_SSE2_H_
#ifdef __cplusplus
extern "C" {
#endif
#if defined(WEBP_USE_SSE2)
#include <emmintrin.h>
//------------------------------------------------------------------------------
// Quite useful macro for debugging. Left here for convenience.
#if 0
#include <stdio.h>
static WEBP_INLINE void PrintReg(const __m128i r, const char* const name,
int size) {
int n;
union {
__m128i r;
uint8_t i8[16];
uint16_t i16[8];
uint32_t i32[4];
uint64_t i64[2];
} tmp;
tmp.r = r;
fprintf(stderr, "%s\t: ", name);
if (size == 8) {
for (n = 0; n < 16; ++n) fprintf(stderr, "%.2x ", tmp.i8[n]);
} else if (size == 16) {
for (n = 0; n < 8; ++n) fprintf(stderr, "%.4x ", tmp.i16[n]);
} else if (size == 32) {
for (n = 0; n < 4; ++n) fprintf(stderr, "%.8x ", tmp.i32[n]);
} else {
for (n = 0; n < 2; ++n) fprintf(stderr, "%.16lx ", tmp.i64[n]);
}
fprintf(stderr, "\n");
}
#endif
//------------------------------------------------------------------------------
// Math functions.
// Return the sum of all the 8b in the register.
static WEBP_INLINE int VP8HorizontalAdd8b(const __m128i* const a) {
const __m128i zero = _mm_setzero_si128();
const __m128i sad8x2 = _mm_sad_epu8(*a, zero);
// sum the two sads: sad8x2[0:1] + sad8x2[8:9]
const __m128i sum = _mm_add_epi32(sad8x2, _mm_shuffle_epi32(sad8x2, 2));
return _mm_cvtsi128_si32(sum);
}
// Transpose two 4x4 16b matrices horizontally stored in registers.
static WEBP_INLINE void VP8Transpose_2_4x4_16b(
const __m128i* const in0, const __m128i* const in1,
const __m128i* const in2, const __m128i* const in3, __m128i* const out0,
__m128i* const out1, __m128i* const out2, __m128i* const out3) {
// Transpose the two 4x4.
// a00 a01 a02 a03 b00 b01 b02 b03
// a10 a11 a12 a13 b10 b11 b12 b13
// a20 a21 a22 a23 b20 b21 b22 b23
// a30 a31 a32 a33 b30 b31 b32 b33
const __m128i transpose0_0 = _mm_unpacklo_epi16(*in0, *in1);
const __m128i transpose0_1 = _mm_unpacklo_epi16(*in2, *in3);
const __m128i transpose0_2 = _mm_unpackhi_epi16(*in0, *in1);
const __m128i transpose0_3 = _mm_unpackhi_epi16(*in2, *in3);
// a00 a10 a01 a11 a02 a12 a03 a13
// a20 a30 a21 a31 a22 a32 a23 a33
// b00 b10 b01 b11 b02 b12 b03 b13
// b20 b30 b21 b31 b22 b32 b23 b33
const __m128i transpose1_0 = _mm_unpacklo_epi32(transpose0_0, transpose0_1);
const __m128i transpose1_1 = _mm_unpacklo_epi32(transpose0_2, transpose0_3);
const __m128i transpose1_2 = _mm_unpackhi_epi32(transpose0_0, transpose0_1);
const __m128i transpose1_3 = _mm_unpackhi_epi32(transpose0_2, transpose0_3);
// a00 a10 a20 a30 a01 a11 a21 a31
// b00 b10 b20 b30 b01 b11 b21 b31
// a02 a12 a22 a32 a03 a13 a23 a33
// b02 b12 a22 b32 b03 b13 b23 b33
*out0 = _mm_unpacklo_epi64(transpose1_0, transpose1_1);
*out1 = _mm_unpackhi_epi64(transpose1_0, transpose1_1);
*out2 = _mm_unpacklo_epi64(transpose1_2, transpose1_3);
*out3 = _mm_unpackhi_epi64(transpose1_2, transpose1_3);
// a00 a10 a20 a30 b00 b10 b20 b30
// a01 a11 a21 a31 b01 b11 b21 b31
// a02 a12 a22 a32 b02 b12 b22 b32
// a03 a13 a23 a33 b03 b13 b23 b33
}
//------------------------------------------------------------------------------
// Channel mixing.
// Function used several times in VP8PlanarTo24b.
// It samples the in buffer as follows: one every two unsigned char is stored
// at the beginning of the buffer, while the other half is stored at the end.
#define VP8PlanarTo24bHelper(IN, OUT) \
do { \
const __m128i v_mask = _mm_set1_epi16(0x00ff); \
/* Take one every two upper 8b values.*/ \
(OUT##0) = _mm_packus_epi16(_mm_and_si128((IN##0), v_mask), \
_mm_and_si128((IN##1), v_mask)); \
(OUT##1) = _mm_packus_epi16(_mm_and_si128((IN##2), v_mask), \
_mm_and_si128((IN##3), v_mask)); \
(OUT##2) = _mm_packus_epi16(_mm_and_si128((IN##4), v_mask), \
_mm_and_si128((IN##5), v_mask)); \
/* Take one every two lower 8b values.*/ \
(OUT##3) = _mm_packus_epi16(_mm_srli_epi16((IN##0), 8), \
_mm_srli_epi16((IN##1), 8)); \
(OUT##4) = _mm_packus_epi16(_mm_srli_epi16((IN##2), 8), \
_mm_srli_epi16((IN##3), 8)); \
(OUT##5) = _mm_packus_epi16(_mm_srli_epi16((IN##4), 8), \
_mm_srli_epi16((IN##5), 8)); \
} while (0)
// Pack the planar buffers
// rrrr... rrrr... gggg... gggg... bbbb... bbbb....
// triplet by triplet in the output buffer rgb as rgbrgbrgbrgb ...
static WEBP_INLINE void VP8PlanarTo24b_SSE2(
__m128i* const in0, __m128i* const in1, __m128i* const in2,
__m128i* const in3, __m128i* const in4, __m128i* const in5) {
// The input is 6 registers of sixteen 8b but for the sake of explanation,
// let's take 6 registers of four 8b values.
// To pack, we will keep taking one every two 8b integer and move it
// around as follows:
// Input:
// r0r1r2r3 | r4r5r6r7 | g0g1g2g3 | g4g5g6g7 | b0b1b2b3 | b4b5b6b7
// Split the 6 registers in two sets of 3 registers: the first set as the even
// 8b bytes, the second the odd ones:
// r0r2r4r6 | g0g2g4g6 | b0b2b4b6 | r1r3r5r7 | g1g3g5g7 | b1b3b5b7
// Repeat the same permutations twice more:
// r0r4g0g4 | b0b4r1r5 | g1g5b1b5 | r2r6g2g6 | b2b6r3r7 | g3g7b3b7
// r0g0b0r1 | g1b1r2g2 | b2r3g3b3 | r4g4b4r5 | g5b5r6g6 | b6r7g7b7
__m128i tmp0, tmp1, tmp2, tmp3, tmp4, tmp5;
VP8PlanarTo24bHelper(*in, tmp);
VP8PlanarTo24bHelper(tmp, *in);
VP8PlanarTo24bHelper(*in, tmp);
// We need to do it two more times than the example as we have sixteen bytes.
{
__m128i out0, out1, out2, out3, out4, out5;
VP8PlanarTo24bHelper(tmp, out);
VP8PlanarTo24bHelper(out, *in);
}
}
#undef VP8PlanarTo24bHelper
// Convert four packed four-channel buffers like argbargbargbargb... into the
// split channels aaaaa ... rrrr ... gggg .... bbbbb ......
static WEBP_INLINE void VP8L32bToPlanar_SSE2(__m128i* const in0,
__m128i* const in1,
__m128i* const in2,
__m128i* const in3) {
// Column-wise transpose.
const __m128i A0 = _mm_unpacklo_epi8(*in0, *in1);
const __m128i A1 = _mm_unpackhi_epi8(*in0, *in1);
const __m128i A2 = _mm_unpacklo_epi8(*in2, *in3);
const __m128i A3 = _mm_unpackhi_epi8(*in2, *in3);
const __m128i B0 = _mm_unpacklo_epi8(A0, A1);
const __m128i B1 = _mm_unpackhi_epi8(A0, A1);
const __m128i B2 = _mm_unpacklo_epi8(A2, A3);
const __m128i B3 = _mm_unpackhi_epi8(A2, A3);
// C0 = g7 g6 ... g1 g0 | b7 b6 ... b1 b0
// C1 = a7 a6 ... a1 a0 | r7 r6 ... r1 r0
const __m128i C0 = _mm_unpacklo_epi8(B0, B1);
const __m128i C1 = _mm_unpackhi_epi8(B0, B1);
const __m128i C2 = _mm_unpacklo_epi8(B2, B3);
const __m128i C3 = _mm_unpackhi_epi8(B2, B3);
// Gather the channels.
*in0 = _mm_unpackhi_epi64(C1, C3);
*in1 = _mm_unpacklo_epi64(C1, C3);
*in2 = _mm_unpackhi_epi64(C0, C2);
*in3 = _mm_unpacklo_epi64(C0, C2);
}
#endif // WEBP_USE_SSE2
#ifdef __cplusplus
} // extern "C"
#endif
#endif // WEBP_DSP_COMMON_SSE2_H_

132
media/libwebp/src/dsp/common_sse41.h Normal file
Просмотреть файл

@ -0,0 +1,132 @@
// Copyright 2016 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// SSE4 code common to several files.
//
// Author: Vincent Rabaud (vrabaud@google.com)
#ifndef WEBP_DSP_COMMON_SSE41_H_
#define WEBP_DSP_COMMON_SSE41_H_
#ifdef __cplusplus
extern "C" {
#endif
#if defined(WEBP_USE_SSE41)
#include <smmintrin.h>
//------------------------------------------------------------------------------
// Channel mixing.
// Shuffles the input buffer as A0 0 0 A1 0 0 A2 ...
#define WEBP_SSE41_SHUFF(OUT, IN0, IN1) \
OUT##0 = _mm_shuffle_epi8(*IN0, shuff0); \
OUT##1 = _mm_shuffle_epi8(*IN0, shuff1); \
OUT##2 = _mm_shuffle_epi8(*IN0, shuff2); \
OUT##3 = _mm_shuffle_epi8(*IN1, shuff0); \
OUT##4 = _mm_shuffle_epi8(*IN1, shuff1); \
OUT##5 = _mm_shuffle_epi8(*IN1, shuff2);
// Pack the planar buffers
// rrrr... rrrr... gggg... gggg... bbbb... bbbb....
// triplet by triplet in the output buffer rgb as rgbrgbrgbrgb ...
static WEBP_INLINE void VP8PlanarTo24b_SSE41(
__m128i* const in0, __m128i* const in1, __m128i* const in2,
__m128i* const in3, __m128i* const in4, __m128i* const in5) {
__m128i R0, R1, R2, R3, R4, R5;
__m128i G0, G1, G2, G3, G4, G5;
__m128i B0, B1, B2, B3, B4, B5;
// Process R.
{
const __m128i shuff0 = _mm_set_epi8(
5, -1, -1, 4, -1, -1, 3, -1, -1, 2, -1, -1, 1, -1, -1, 0);
const __m128i shuff1 = _mm_set_epi8(
-1, 10, -1, -1, 9, -1, -1, 8, -1, -1, 7, -1, -1, 6, -1, -1);
const __m128i shuff2 = _mm_set_epi8(
-1, -1, 15, -1, -1, 14, -1, -1, 13, -1, -1, 12, -1, -1, 11, -1);
WEBP_SSE41_SHUFF(R, in0, in1)
}
// Process G.
{
// Same as before, just shifted to the left by one and including the right
// padding.
const __m128i shuff0 = _mm_set_epi8(
-1, -1, 4, -1, -1, 3, -1, -1, 2, -1, -1, 1, -1, -1, 0, -1);
const __m128i shuff1 = _mm_set_epi8(
10, -1, -1, 9, -1, -1, 8, -1, -1, 7, -1, -1, 6, -1, -1, 5);
const __m128i shuff2 = _mm_set_epi8(
-1, 15, -1, -1, 14, -1, -1, 13, -1, -1, 12, -1, -1, 11, -1, -1);
WEBP_SSE41_SHUFF(G, in2, in3)
}
// Process B.
{
const __m128i shuff0 = _mm_set_epi8(
-1, 4, -1, -1, 3, -1, -1, 2, -1, -1, 1, -1, -1, 0, -1, -1);
const __m128i shuff1 = _mm_set_epi8(
-1, -1, 9, -1, -1, 8, -1, -1, 7, -1, -1, 6, -1, -1, 5, -1);
const __m128i shuff2 = _mm_set_epi8(
15, -1, -1, 14, -1, -1, 13, -1, -1, 12, -1, -1, 11, -1, -1, 10);
WEBP_SSE41_SHUFF(B, in4, in5)
}
// OR the different channels.
{
const __m128i RG0 = _mm_or_si128(R0, G0);
const __m128i RG1 = _mm_or_si128(R1, G1);
const __m128i RG2 = _mm_or_si128(R2, G2);
const __m128i RG3 = _mm_or_si128(R3, G3);
const __m128i RG4 = _mm_or_si128(R4, G4);
const __m128i RG5 = _mm_or_si128(R5, G5);
*in0 = _mm_or_si128(RG0, B0);
*in1 = _mm_or_si128(RG1, B1);
*in2 = _mm_or_si128(RG2, B2);
*in3 = _mm_or_si128(RG3, B3);
*in4 = _mm_or_si128(RG4, B4);
*in5 = _mm_or_si128(RG5, B5);
}
}
#undef WEBP_SSE41_SHUFF
// Convert four packed four-channel buffers like argbargbargbargb... into the
// split channels aaaaa ... rrrr ... gggg .... bbbbb ......
static WEBP_INLINE void VP8L32bToPlanar_SSE41(__m128i* const in0,
__m128i* const in1,
__m128i* const in2,
__m128i* const in3) {
// aaaarrrrggggbbbb
const __m128i shuff0 =
_mm_set_epi8(15, 11, 7, 3, 14, 10, 6, 2, 13, 9, 5, 1, 12, 8, 4, 0);
const __m128i A0 = _mm_shuffle_epi8(*in0, shuff0);
const __m128i A1 = _mm_shuffle_epi8(*in1, shuff0);
const __m128i A2 = _mm_shuffle_epi8(*in2, shuff0);
const __m128i A3 = _mm_shuffle_epi8(*in3, shuff0);
// A0A1R0R1
// G0G1B0B1
// A2A3R2R3
// G0G1B0B1
const __m128i B0 = _mm_unpacklo_epi32(A0, A1);
const __m128i B1 = _mm_unpackhi_epi32(A0, A1);
const __m128i B2 = _mm_unpacklo_epi32(A2, A3);
const __m128i B3 = _mm_unpackhi_epi32(A2, A3);
*in3 = _mm_unpacklo_epi64(B0, B2);
*in2 = _mm_unpackhi_epi64(B0, B2);
*in1 = _mm_unpacklo_epi64(B1, B3);
*in0 = _mm_unpackhi_epi64(B1, B3);
}
#endif // WEBP_USE_SSE41
#ifdef __cplusplus
} // extern "C"
#endif
#endif // WEBP_DSP_COMMON_SSE41_H_

887
media/libwebp/src/dsp/dec.c Normal file
Просмотреть файл

@ -0,0 +1,887 @@
// Copyright 2010 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Speed-critical decoding functions, default plain-C implementations.
//
// Author: Skal (pascal.massimino@gmail.com)
#include <assert.h>
#include "src/dsp/dsp.h"
#include "src/dec/vp8i_dec.h"
#include "src/utils/utils.h"
//------------------------------------------------------------------------------
static WEBP_INLINE uint8_t clip_8b(int v) {
return (!(v & ~0xff)) ? v : (v < 0) ? 0 : 255;
}
//------------------------------------------------------------------------------
// Transforms (Paragraph 14.4)
#define STORE(x, y, v) \
dst[(x) + (y) * BPS] = clip_8b(dst[(x) + (y) * BPS] + ((v) >> 3))
#define STORE2(y, dc, d, c) do { \
const int DC = (dc); \
STORE(0, y, DC + (d)); \
STORE(1, y, DC + (c)); \
STORE(2, y, DC - (c)); \
STORE(3, y, DC - (d)); \
} while (0)
#define MUL1(a) ((((a) * 20091) >> 16) + (a))
#define MUL2(a) (((a) * 35468) >> 16)
#if !WEBP_NEON_OMIT_C_CODE
static void TransformOne_C(const int16_t* in, uint8_t* dst) {
int C[4 * 4], *tmp;
int i;
tmp = C;
for (i = 0; i < 4; ++i) { // vertical pass
const int a = in[0] + in[8]; // [-4096, 4094]
const int b = in[0] - in[8]; // [-4095, 4095]
const int c = MUL2(in[4]) - MUL1(in[12]); // [-3783, 3783]
const int d = MUL1(in[4]) + MUL2(in[12]); // [-3785, 3781]
tmp[0] = a + d; // [-7881, 7875]
tmp[1] = b + c; // [-7878, 7878]
tmp[2] = b - c; // [-7878, 7878]
tmp[3] = a - d; // [-7877, 7879]
tmp += 4;
in++;
}
// Each pass is expanding the dynamic range by ~3.85 (upper bound).
// The exact value is (2. + (20091 + 35468) / 65536).
// After the second pass, maximum interval is [-3794, 3794], assuming
// an input in [-2048, 2047] interval. We then need to add a dst value
// in the [0, 255] range.
// In the worst case scenario, the input to clip_8b() can be as large as
// [-60713, 60968].
tmp = C;
for (i = 0; i < 4; ++i) { // horizontal pass
const int dc = tmp[0] + 4;
const int a = dc + tmp[8];
const int b = dc - tmp[8];
const int c = MUL2(tmp[4]) - MUL1(tmp[12]);
const int d = MUL1(tmp[4]) + MUL2(tmp[12]);
STORE(0, 0, a + d);
STORE(1, 0, b + c);
STORE(2, 0, b - c);
STORE(3, 0, a - d);
tmp++;
dst += BPS;
}
}
// Simplified transform when only in[0], in[1] and in[4] are non-zero
static void TransformAC3_C(const int16_t* in, uint8_t* dst) {
const int a = in[0] + 4;
const int c4 = MUL2(in[4]);
const int d4 = MUL1(in[4]);
const int c1 = MUL2(in[1]);
const int d1 = MUL1(in[1]);
STORE2(0, a + d4, d1, c1);
STORE2(1, a + c4, d1, c1);
STORE2(2, a - c4, d1, c1);
STORE2(3, a - d4, d1, c1);
}
#undef MUL1
#undef MUL2
#undef STORE2
static void TransformTwo_C(const int16_t* in, uint8_t* dst, int do_two) {
TransformOne_C(in, dst);
if (do_two) {
TransformOne_C(in + 16, dst + 4);
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
static void TransformUV_C(const int16_t* in, uint8_t* dst) {
VP8Transform(in + 0 * 16, dst, 1);
VP8Transform(in + 2 * 16, dst + 4 * BPS, 1);
}
#if !WEBP_NEON_OMIT_C_CODE
static void TransformDC_C(const int16_t* in, uint8_t* dst) {
const int DC = in[0] + 4;
int i, j;
for (j = 0; j < 4; ++j) {
for (i = 0; i < 4; ++i) {
STORE(i, j, DC);
}
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
static void TransformDCUV_C(const int16_t* in, uint8_t* dst) {
if (in[0 * 16]) VP8TransformDC(in + 0 * 16, dst);
if (in[1 * 16]) VP8TransformDC(in + 1 * 16, dst + 4);
if (in[2 * 16]) VP8TransformDC(in + 2 * 16, dst + 4 * BPS);
if (in[3 * 16]) VP8TransformDC(in + 3 * 16, dst + 4 * BPS + 4);
}
#undef STORE
//------------------------------------------------------------------------------
// Paragraph 14.3
#if !WEBP_NEON_OMIT_C_CODE
static void TransformWHT_C(const int16_t* in, int16_t* out) {
int tmp[16];
int i;
for (i = 0; i < 4; ++i) {
const int a0 = in[0 + i] + in[12 + i];
const int a1 = in[4 + i] + in[ 8 + i];
const int a2 = in[4 + i] - in[ 8 + i];
const int a3 = in[0 + i] - in[12 + i];
tmp[0 + i] = a0 + a1;
tmp[8 + i] = a0 - a1;
tmp[4 + i] = a3 + a2;
tmp[12 + i] = a3 - a2;
}
for (i = 0; i < 4; ++i) {
const int dc = tmp[0 + i * 4] + 3; // w/ rounder
const int a0 = dc + tmp[3 + i * 4];
const int a1 = tmp[1 + i * 4] + tmp[2 + i * 4];
const int a2 = tmp[1 + i * 4] - tmp[2 + i * 4];
const int a3 = dc - tmp[3 + i * 4];
out[ 0] = (a0 + a1) >> 3;
out[16] = (a3 + a2) >> 3;
out[32] = (a0 - a1) >> 3;
out[48] = (a3 - a2) >> 3;
out += 64;
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
void (*VP8TransformWHT)(const int16_t* in, int16_t* out);
//------------------------------------------------------------------------------
// Intra predictions
#define DST(x, y) dst[(x) + (y) * BPS]
#if !WEBP_NEON_OMIT_C_CODE
static WEBP_INLINE void TrueMotion(uint8_t* dst, int size) {
const uint8_t* top = dst - BPS;
const uint8_t* const clip0 = VP8kclip1 - top[-1];
int y;
for (y = 0; y < size; ++y) {
const uint8_t* const clip = clip0 + dst[-1];
int x;
for (x = 0; x < size; ++x) {
dst[x] = clip[top[x]];
}
dst += BPS;
}
}
static void TM4_C(uint8_t* dst) { TrueMotion(dst, 4); }
static void TM8uv_C(uint8_t* dst) { TrueMotion(dst, 8); }
static void TM16_C(uint8_t* dst) { TrueMotion(dst, 16); }
//------------------------------------------------------------------------------
// 16x16
static void VE16_C(uint8_t* dst) { // vertical
int j;
for (j = 0; j < 16; ++j) {
memcpy(dst + j * BPS, dst - BPS, 16);
}
}
static void HE16_C(uint8_t* dst) { // horizontal
int j;
for (j = 16; j > 0; --j) {
memset(dst, dst[-1], 16);
dst += BPS;
}
}
static WEBP_INLINE void Put16(int v, uint8_t* dst) {
int j;
for (j = 0; j < 16; ++j) {
memset(dst + j * BPS, v, 16);
}
}
static void DC16_C(uint8_t* dst) { // DC
int DC = 16;
int j;
for (j = 0; j < 16; ++j) {
DC += dst[-1 + j * BPS] + dst[j - BPS];
}
Put16(DC >> 5, dst);
}
static void DC16NoTop_C(uint8_t* dst) { // DC with top samples not available
int DC = 8;
int j;
for (j = 0; j < 16; ++j) {
DC += dst[-1 + j * BPS];
}
Put16(DC >> 4, dst);
}
static void DC16NoLeft_C(uint8_t* dst) { // DC with left samples not available
int DC = 8;
int i;
for (i = 0; i < 16; ++i) {
DC += dst[i - BPS];
}
Put16(DC >> 4, dst);
}
static void DC16NoTopLeft_C(uint8_t* dst) { // DC with no top and left samples
Put16(0x80, dst);
}
#endif // !WEBP_NEON_OMIT_C_CODE
VP8PredFunc VP8PredLuma16[NUM_B_DC_MODES];
//------------------------------------------------------------------------------
// 4x4
#define AVG3(a, b, c) ((uint8_t)(((a) + 2 * (b) + (c) + 2) >> 2))
#define AVG2(a, b) (((a) + (b) + 1) >> 1)
#if !WEBP_NEON_OMIT_C_CODE
static void VE4_C(uint8_t* dst) { // vertical
const uint8_t* top = dst - BPS;
const uint8_t vals[4] = {
AVG3(top[-1], top[0], top[1]),
AVG3(top[ 0], top[1], top[2]),
AVG3(top[ 1], top[2], top[3]),
AVG3(top[ 2], top[3], top[4])
};
int i;
for (i = 0; i < 4; ++i) {
memcpy(dst + i * BPS, vals, sizeof(vals));
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
static void HE4_C(uint8_t* dst) { // horizontal
const int A = dst[-1 - BPS];
const int B = dst[-1];
const int C = dst[-1 + BPS];
const int D = dst[-1 + 2 * BPS];
const int E = dst[-1 + 3 * BPS];
WebPUint32ToMem(dst + 0 * BPS, 0x01010101U * AVG3(A, B, C));
WebPUint32ToMem(dst + 1 * BPS, 0x01010101U * AVG3(B, C, D));
WebPUint32ToMem(dst + 2 * BPS, 0x01010101U * AVG3(C, D, E));
WebPUint32ToMem(dst + 3 * BPS, 0x01010101U * AVG3(D, E, E));
}
#if !WEBP_NEON_OMIT_C_CODE
static void DC4_C(uint8_t* dst) { // DC
uint32_t dc = 4;
int i;
for (i = 0; i < 4; ++i) dc += dst[i - BPS] + dst[-1 + i * BPS];
dc >>= 3;
for (i = 0; i < 4; ++i) memset(dst + i * BPS, dc, 4);
}
static void RD4_C(uint8_t* dst) { // Down-right
const int I = dst[-1 + 0 * BPS];
const int J = dst[-1 + 1 * BPS];
const int K = dst[-1 + 2 * BPS];
const int L = dst[-1 + 3 * BPS];
const int X = dst[-1 - BPS];
const int A = dst[0 - BPS];
const int B = dst[1 - BPS];
const int C = dst[2 - BPS];
const int D = dst[3 - BPS];
DST(0, 3) = AVG3(J, K, L);
DST(1, 3) = DST(0, 2) = AVG3(I, J, K);
DST(2, 3) = DST(1, 2) = DST(0, 1) = AVG3(X, I, J);
DST(3, 3) = DST(2, 2) = DST(1, 1) = DST(0, 0) = AVG3(A, X, I);
DST(3, 2) = DST(2, 1) = DST(1, 0) = AVG3(B, A, X);
DST(3, 1) = DST(2, 0) = AVG3(C, B, A);
DST(3, 0) = AVG3(D, C, B);
}
static void LD4_C(uint8_t* dst) { // Down-Left
const int A = dst[0 - BPS];
const int B = dst[1 - BPS];
const int C = dst[2 - BPS];
const int D = dst[3 - BPS];
const int E = dst[4 - BPS];
const int F = dst[5 - BPS];
const int G = dst[6 - BPS];
const int H = dst[7 - BPS];
DST(0, 0) = AVG3(A, B, C);
DST(1, 0) = DST(0, 1) = AVG3(B, C, D);
DST(2, 0) = DST(1, 1) = DST(0, 2) = AVG3(C, D, E);
DST(3, 0) = DST(2, 1) = DST(1, 2) = DST(0, 3) = AVG3(D, E, F);
DST(3, 1) = DST(2, 2) = DST(1, 3) = AVG3(E, F, G);
DST(3, 2) = DST(2, 3) = AVG3(F, G, H);
DST(3, 3) = AVG3(G, H, H);
}
#endif // !WEBP_NEON_OMIT_C_CODE
static void VR4_C(uint8_t* dst) { // Vertical-Right
const int I = dst[-1 + 0 * BPS];
const int J = dst[-1 + 1 * BPS];
const int K = dst[-1 + 2 * BPS];
const int X = dst[-1 - BPS];
const int A = dst[0 - BPS];
const int B = dst[1 - BPS];
const int C = dst[2 - BPS];
const int D = dst[3 - BPS];
DST(0, 0) = DST(1, 2) = AVG2(X, A);
DST(1, 0) = DST(2, 2) = AVG2(A, B);
DST(2, 0) = DST(3, 2) = AVG2(B, C);
DST(3, 0) = AVG2(C, D);
DST(0, 3) = AVG3(K, J, I);
DST(0, 2) = AVG3(J, I, X);
DST(0, 1) = DST(1, 3) = AVG3(I, X, A);
DST(1, 1) = DST(2, 3) = AVG3(X, A, B);
DST(2, 1) = DST(3, 3) = AVG3(A, B, C);
DST(3, 1) = AVG3(B, C, D);
}
static void VL4_C(uint8_t* dst) { // Vertical-Left
const int A = dst[0 - BPS];
const int B = dst[1 - BPS];
const int C = dst[2 - BPS];
const int D = dst[3 - BPS];
const int E = dst[4 - BPS];
const int F = dst[5 - BPS];
const int G = dst[6 - BPS];
const int H = dst[7 - BPS];
DST(0, 0) = AVG2(A, B);
DST(1, 0) = DST(0, 2) = AVG2(B, C);
DST(2, 0) = DST(1, 2) = AVG2(C, D);
DST(3, 0) = DST(2, 2) = AVG2(D, E);
DST(0, 1) = AVG3(A, B, C);
DST(1, 1) = DST(0, 3) = AVG3(B, C, D);
DST(2, 1) = DST(1, 3) = AVG3(C, D, E);
DST(3, 1) = DST(2, 3) = AVG3(D, E, F);
DST(3, 2) = AVG3(E, F, G);
DST(3, 3) = AVG3(F, G, H);
}
static void HU4_C(uint8_t* dst) { // Horizontal-Up
const int I = dst[-1 + 0 * BPS];
const int J = dst[-1 + 1 * BPS];
const int K = dst[-1 + 2 * BPS];
const int L = dst[-1 + 3 * BPS];
DST(0, 0) = AVG2(I, J);
DST(2, 0) = DST(0, 1) = AVG2(J, K);
DST(2, 1) = DST(0, 2) = AVG2(K, L);
DST(1, 0) = AVG3(I, J, K);
DST(3, 0) = DST(1, 1) = AVG3(J, K, L);
DST(3, 1) = DST(1, 2) = AVG3(K, L, L);
DST(3, 2) = DST(2, 2) =
DST(0, 3) = DST(1, 3) = DST(2, 3) = DST(3, 3) = L;
}
static void HD4_C(uint8_t* dst) { // Horizontal-Down
const int I = dst[-1 + 0 * BPS];
const int J = dst[-1 + 1 * BPS];
const int K = dst[-1 + 2 * BPS];
const int L = dst[-1 + 3 * BPS];
const int X = dst[-1 - BPS];
const int A = dst[0 - BPS];
const int B = dst[1 - BPS];
const int C = dst[2 - BPS];
DST(0, 0) = DST(2, 1) = AVG2(I, X);
DST(0, 1) = DST(2, 2) = AVG2(J, I);
DST(0, 2) = DST(2, 3) = AVG2(K, J);
DST(0, 3) = AVG2(L, K);
DST(3, 0) = AVG3(A, B, C);
DST(2, 0) = AVG3(X, A, B);
DST(1, 0) = DST(3, 1) = AVG3(I, X, A);
DST(1, 1) = DST(3, 2) = AVG3(J, I, X);
DST(1, 2) = DST(3, 3) = AVG3(K, J, I);
DST(1, 3) = AVG3(L, K, J);
}
#undef DST
#undef AVG3
#undef AVG2
VP8PredFunc VP8PredLuma4[NUM_BMODES];
//------------------------------------------------------------------------------
// Chroma
#if !WEBP_NEON_OMIT_C_CODE
static void VE8uv_C(uint8_t* dst) { // vertical
int j;
for (j = 0; j < 8; ++j) {
memcpy(dst + j * BPS, dst - BPS, 8);
}
}
static void HE8uv_C(uint8_t* dst) { // horizontal
int j;
for (j = 0; j < 8; ++j) {
memset(dst, dst[-1], 8);
dst += BPS;
}
}
// helper for chroma-DC predictions
static WEBP_INLINE void Put8x8uv(uint8_t value, uint8_t* dst) {
int j;
for (j = 0; j < 8; ++j) {
memset(dst + j * BPS, value, 8);
}
}
static void DC8uv_C(uint8_t* dst) { // DC
int dc0 = 8;
int i;
for (i = 0; i < 8; ++i) {
dc0 += dst[i - BPS] + dst[-1 + i * BPS];
}
Put8x8uv(dc0 >> 4, dst);
}
static void DC8uvNoLeft_C(uint8_t* dst) { // DC with no left samples
int dc0 = 4;
int i;
for (i = 0; i < 8; ++i) {
dc0 += dst[i - BPS];
}
Put8x8uv(dc0 >> 3, dst);
}
static void DC8uvNoTop_C(uint8_t* dst) { // DC with no top samples
int dc0 = 4;
int i;
for (i = 0; i < 8; ++i) {
dc0 += dst[-1 + i * BPS];
}
Put8x8uv(dc0 >> 3, dst);
}
static void DC8uvNoTopLeft_C(uint8_t* dst) { // DC with nothing
Put8x8uv(0x80, dst);
}
#endif // !WEBP_NEON_OMIT_C_CODE
VP8PredFunc VP8PredChroma8[NUM_B_DC_MODES];
//------------------------------------------------------------------------------
// Edge filtering functions
#if !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
// 4 pixels in, 2 pixels out
static WEBP_INLINE void DoFilter2_C(uint8_t* p, int step) {
const int p1 = p[-2*step], p0 = p[-step], q0 = p[0], q1 = p[step];
const int a = 3 * (q0 - p0) + VP8ksclip1[p1 - q1]; // in [-893,892]
const int a1 = VP8ksclip2[(a + 4) >> 3]; // in [-16,15]
const int a2 = VP8ksclip2[(a + 3) >> 3];
p[-step] = VP8kclip1[p0 + a2];
p[ 0] = VP8kclip1[q0 - a1];
}
// 4 pixels in, 4 pixels out
static WEBP_INLINE void DoFilter4_C(uint8_t* p, int step) {
const int p1 = p[-2*step], p0 = p[-step], q0 = p[0], q1 = p[step];
const int a = 3 * (q0 - p0);
const int a1 = VP8ksclip2[(a + 4) >> 3];
const int a2 = VP8ksclip2[(a + 3) >> 3];
const int a3 = (a1 + 1) >> 1;
p[-2*step] = VP8kclip1[p1 + a3];
p[- step] = VP8kclip1[p0 + a2];
p[ 0] = VP8kclip1[q0 - a1];
p[ step] = VP8kclip1[q1 - a3];
}
// 6 pixels in, 6 pixels out
static WEBP_INLINE void DoFilter6_C(uint8_t* p, int step) {
const int p2 = p[-3*step], p1 = p[-2*step], p0 = p[-step];
const int q0 = p[0], q1 = p[step], q2 = p[2*step];
const int a = VP8ksclip1[3 * (q0 - p0) + VP8ksclip1[p1 - q1]];
// a is in [-128,127], a1 in [-27,27], a2 in [-18,18] and a3 in [-9,9]
const int a1 = (27 * a + 63) >> 7; // eq. to ((3 * a + 7) * 9) >> 7
const int a2 = (18 * a + 63) >> 7; // eq. to ((2 * a + 7) * 9) >> 7
const int a3 = (9 * a + 63) >> 7; // eq. to ((1 * a + 7) * 9) >> 7
p[-3*step] = VP8kclip1[p2 + a3];
p[-2*step] = VP8kclip1[p1 + a2];
p[- step] = VP8kclip1[p0 + a1];
p[ 0] = VP8kclip1[q0 - a1];
p[ step] = VP8kclip1[q1 - a2];
p[ 2*step] = VP8kclip1[q2 - a3];
}
static WEBP_INLINE int Hev(const uint8_t* p, int step, int thresh) {
const int p1 = p[-2*step], p0 = p[-step], q0 = p[0], q1 = p[step];
return (VP8kabs0[p1 - p0] > thresh) || (VP8kabs0[q1 - q0] > thresh);
}
#endif // !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
#if !WEBP_NEON_OMIT_C_CODE
static WEBP_INLINE int NeedsFilter_C(const uint8_t* p, int step, int t) {
const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step];
return ((4 * VP8kabs0[p0 - q0] + VP8kabs0[p1 - q1]) <= t);
}
#endif // !WEBP_NEON_OMIT_C_CODE
#if !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
static WEBP_INLINE int NeedsFilter2_C(const uint8_t* p,
int step, int t, int it) {
const int p3 = p[-4 * step], p2 = p[-3 * step], p1 = p[-2 * step];
const int p0 = p[-step], q0 = p[0];
const int q1 = p[step], q2 = p[2 * step], q3 = p[3 * step];
if ((4 * VP8kabs0[p0 - q0] + VP8kabs0[p1 - q1]) > t) return 0;
return VP8kabs0[p3 - p2] <= it && VP8kabs0[p2 - p1] <= it &&
VP8kabs0[p1 - p0] <= it && VP8kabs0[q3 - q2] <= it &&
VP8kabs0[q2 - q1] <= it && VP8kabs0[q1 - q0] <= it;
}
#endif // !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
//------------------------------------------------------------------------------
// Simple In-loop filtering (Paragraph 15.2)
#if !WEBP_NEON_OMIT_C_CODE
static void SimpleVFilter16_C(uint8_t* p, int stride, int thresh) {
int i;
const int thresh2 = 2 * thresh + 1;
for (i = 0; i < 16; ++i) {
if (NeedsFilter_C(p + i, stride, thresh2)) {
DoFilter2_C(p + i, stride);
}
}
}
static void SimpleHFilter16_C(uint8_t* p, int stride, int thresh) {
int i;
const int thresh2 = 2 * thresh + 1;
for (i = 0; i < 16; ++i) {
if (NeedsFilter_C(p + i * stride, 1, thresh2)) {
DoFilter2_C(p + i * stride, 1);
}
}
}
static void SimpleVFilter16i_C(uint8_t* p, int stride, int thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4 * stride;
SimpleVFilter16_C(p, stride, thresh);
}
}
static void SimpleHFilter16i_C(uint8_t* p, int stride, int thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4;
SimpleHFilter16_C(p, stride, thresh);
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
//------------------------------------------------------------------------------
// Complex In-loop filtering (Paragraph 15.3)
#if !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
static WEBP_INLINE void FilterLoop26_C(uint8_t* p,
int hstride, int vstride, int size,
int thresh, int ithresh,
int hev_thresh) {
const int thresh2 = 2 * thresh + 1;
while (size-- > 0) {
if (NeedsFilter2_C(p, hstride, thresh2, ithresh)) {
if (Hev(p, hstride, hev_thresh)) {
DoFilter2_C(p, hstride);
} else {
DoFilter6_C(p, hstride);
}
}
p += vstride;
}
}
static WEBP_INLINE void FilterLoop24_C(uint8_t* p,
int hstride, int vstride, int size,
int thresh, int ithresh,
int hev_thresh) {
const int thresh2 = 2 * thresh + 1;
while (size-- > 0) {
if (NeedsFilter2_C(p, hstride, thresh2, ithresh)) {
if (Hev(p, hstride, hev_thresh)) {
DoFilter2_C(p, hstride);
} else {
DoFilter4_C(p, hstride);
}
}
p += vstride;
}
}
#endif // !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
#if !WEBP_NEON_OMIT_C_CODE
// on macroblock edges
static void VFilter16_C(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26_C(p, stride, 1, 16, thresh, ithresh, hev_thresh);
}
static void HFilter16_C(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26_C(p, 1, stride, 16, thresh, ithresh, hev_thresh);
}
// on three inner edges
static void VFilter16i_C(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4 * stride;
FilterLoop24_C(p, stride, 1, 16, thresh, ithresh, hev_thresh);
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
#if !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
static void HFilter16i_C(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4;
FilterLoop24_C(p, 1, stride, 16, thresh, ithresh, hev_thresh);
}
}
#endif // !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
#if !WEBP_NEON_OMIT_C_CODE
// 8-pixels wide variant, for chroma filtering
static void VFilter8_C(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26_C(u, stride, 1, 8, thresh, ithresh, hev_thresh);
FilterLoop26_C(v, stride, 1, 8, thresh, ithresh, hev_thresh);
}
#endif // !WEBP_NEON_OMIT_C_CODE
#if !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
static void HFilter8_C(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26_C(u, 1, stride, 8, thresh, ithresh, hev_thresh);
FilterLoop26_C(v, 1, stride, 8, thresh, ithresh, hev_thresh);
}
#endif // !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
#if !WEBP_NEON_OMIT_C_CODE
static void VFilter8i_C(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop24_C(u + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh);
FilterLoop24_C(v + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh);
}
#endif // !WEBP_NEON_OMIT_C_CODE
#if !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
static void HFilter8i_C(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop24_C(u + 4, 1, stride, 8, thresh, ithresh, hev_thresh);
FilterLoop24_C(v + 4, 1, stride, 8, thresh, ithresh, hev_thresh);
}
#endif // !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
//------------------------------------------------------------------------------
static void DitherCombine8x8_C(const uint8_t* dither, uint8_t* dst,
int dst_stride) {
int i, j;
for (j = 0; j < 8; ++j) {
for (i = 0; i < 8; ++i) {
const int delta0 = dither[i] - VP8_DITHER_AMP_CENTER;
const int delta1 =
(delta0 + VP8_DITHER_DESCALE_ROUNDER) >> VP8_DITHER_DESCALE;
dst[i] = clip_8b((int)dst[i] + delta1);
}
dst += dst_stride;
dither += 8;
}
}
//------------------------------------------------------------------------------
VP8DecIdct2 VP8Transform;
VP8DecIdct VP8TransformAC3;
VP8DecIdct VP8TransformUV;
VP8DecIdct VP8TransformDC;
VP8DecIdct VP8TransformDCUV;
VP8LumaFilterFunc VP8VFilter16;
VP8LumaFilterFunc VP8HFilter16;
VP8ChromaFilterFunc VP8VFilter8;
VP8ChromaFilterFunc VP8HFilter8;
VP8LumaFilterFunc VP8VFilter16i;
VP8LumaFilterFunc VP8HFilter16i;
VP8ChromaFilterFunc VP8VFilter8i;
VP8ChromaFilterFunc VP8HFilter8i;
VP8SimpleFilterFunc VP8SimpleVFilter16;
VP8SimpleFilterFunc VP8SimpleHFilter16;
VP8SimpleFilterFunc VP8SimpleVFilter16i;
VP8SimpleFilterFunc VP8SimpleHFilter16i;
void (*VP8DitherCombine8x8)(const uint8_t* dither, uint8_t* dst,
int dst_stride);
extern void VP8DspInitSSE2(void);
extern void VP8DspInitSSE41(void);
extern void VP8DspInitNEON(void);
extern void VP8DspInitMIPS32(void);
extern void VP8DspInitMIPSdspR2(void);
extern void VP8DspInitMSA(void);
WEBP_DSP_INIT_FUNC(VP8DspInit) {
VP8InitClipTables();
#if !WEBP_NEON_OMIT_C_CODE
VP8TransformWHT = TransformWHT_C;
VP8Transform = TransformTwo_C;
VP8TransformDC = TransformDC_C;
VP8TransformAC3 = TransformAC3_C;
#endif
VP8TransformUV = TransformUV_C;
VP8TransformDCUV = TransformDCUV_C;
#if !WEBP_NEON_OMIT_C_CODE
VP8VFilter16 = VFilter16_C;
VP8VFilter16i = VFilter16i_C;
VP8HFilter16 = HFilter16_C;
VP8VFilter8 = VFilter8_C;
VP8VFilter8i = VFilter8i_C;
VP8SimpleVFilter16 = SimpleVFilter16_C;
VP8SimpleHFilter16 = SimpleHFilter16_C;
VP8SimpleVFilter16i = SimpleVFilter16i_C;
VP8SimpleHFilter16i = SimpleHFilter16i_C;
#endif
#if !WEBP_NEON_OMIT_C_CODE || WEBP_NEON_WORK_AROUND_GCC
VP8HFilter16i = HFilter16i_C;
VP8HFilter8 = HFilter8_C;
VP8HFilter8i = HFilter8i_C;
#endif
#if !WEBP_NEON_OMIT_C_CODE
VP8PredLuma4[0] = DC4_C;
VP8PredLuma4[1] = TM4_C;
VP8PredLuma4[2] = VE4_C;
VP8PredLuma4[4] = RD4_C;
VP8PredLuma4[6] = LD4_C;
#endif
VP8PredLuma4[3] = HE4_C;
VP8PredLuma4[5] = VR4_C;
VP8PredLuma4[7] = VL4_C;
VP8PredLuma4[8] = HD4_C;
VP8PredLuma4[9] = HU4_C;
#if !WEBP_NEON_OMIT_C_CODE
VP8PredLuma16[0] = DC16_C;
VP8PredLuma16[1] = TM16_C;
VP8PredLuma16[2] = VE16_C;
VP8PredLuma16[3] = HE16_C;
VP8PredLuma16[4] = DC16NoTop_C;
VP8PredLuma16[5] = DC16NoLeft_C;
VP8PredLuma16[6] = DC16NoTopLeft_C;
VP8PredChroma8[0] = DC8uv_C;
VP8PredChroma8[1] = TM8uv_C;
VP8PredChroma8[2] = VE8uv_C;
VP8PredChroma8[3] = HE8uv_C;
VP8PredChroma8[4] = DC8uvNoTop_C;
VP8PredChroma8[5] = DC8uvNoLeft_C;
VP8PredChroma8[6] = DC8uvNoTopLeft_C;
#endif
VP8DitherCombine8x8 = DitherCombine8x8_C;
// If defined, use CPUInfo() to overwrite some pointers with faster versions.
if (VP8GetCPUInfo != NULL) {
#if defined(WEBP_USE_SSE2)
if (VP8GetCPUInfo(kSSE2)) {
VP8DspInitSSE2();
#if defined(WEBP_USE_SSE41)
if (VP8GetCPUInfo(kSSE4_1)) {
VP8DspInitSSE41();
}
#endif
}
#endif
#if defined(WEBP_USE_MIPS32)
if (VP8GetCPUInfo(kMIPS32)) {
VP8DspInitMIPS32();
}
#endif
#if defined(WEBP_USE_MIPS_DSP_R2)
if (VP8GetCPUInfo(kMIPSdspR2)) {
VP8DspInitMIPSdspR2();
}
#endif
#if defined(WEBP_USE_MSA)
if (VP8GetCPUInfo(kMSA)) {
VP8DspInitMSA();
}
#endif
}
#if defined(WEBP_USE_NEON)
if (WEBP_NEON_OMIT_C_CODE ||
(VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
VP8DspInitNEON();
}
#endif
assert(VP8TransformWHT != NULL);
assert(VP8Transform != NULL);
assert(VP8TransformDC != NULL);
assert(VP8TransformAC3 != NULL);
assert(VP8TransformUV != NULL);
assert(VP8TransformDCUV != NULL);
assert(VP8VFilter16 != NULL);
assert(VP8HFilter16 != NULL);
assert(VP8VFilter8 != NULL);
assert(VP8HFilter8 != NULL);
assert(VP8VFilter16i != NULL);
assert(VP8HFilter16i != NULL);
assert(VP8VFilter8i != NULL);
assert(VP8HFilter8i != NULL);
assert(VP8SimpleVFilter16 != NULL);
assert(VP8SimpleHFilter16 != NULL);
assert(VP8SimpleVFilter16i != NULL);
assert(VP8SimpleHFilter16i != NULL);
assert(VP8PredLuma4[0] != NULL);
assert(VP8PredLuma4[1] != NULL);
assert(VP8PredLuma4[2] != NULL);
assert(VP8PredLuma4[3] != NULL);
assert(VP8PredLuma4[4] != NULL);
assert(VP8PredLuma4[5] != NULL);
assert(VP8PredLuma4[6] != NULL);
assert(VP8PredLuma4[7] != NULL);
assert(VP8PredLuma4[8] != NULL);
assert(VP8PredLuma4[9] != NULL);
assert(VP8PredLuma16[0] != NULL);
assert(VP8PredLuma16[1] != NULL);
assert(VP8PredLuma16[2] != NULL);
assert(VP8PredLuma16[3] != NULL);
assert(VP8PredLuma16[4] != NULL);
assert(VP8PredLuma16[5] != NULL);
assert(VP8PredLuma16[6] != NULL);
assert(VP8PredChroma8[0] != NULL);
assert(VP8PredChroma8[1] != NULL);
assert(VP8PredChroma8[2] != NULL);
assert(VP8PredChroma8[3] != NULL);
assert(VP8PredChroma8[4] != NULL);
assert(VP8PredChroma8[5] != NULL);
assert(VP8PredChroma8[6] != NULL);
assert(VP8DitherCombine8x8 != NULL);
}

369
media/libwebp/src/dsp/dec_clip_tables.c Normal file
Просмотреть файл

@ -0,0 +1,369 @@
// Copyright 2014 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Clipping tables for filtering
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dsp/dsp.h"
// define to 0 to have run-time table initialization
#if !defined(USE_STATIC_TABLES)
#define USE_STATIC_TABLES 1 // ALTERNATE_CODE
#endif
#if (USE_STATIC_TABLES == 1)
static const uint8_t abs0[255 + 255 + 1] = {
0xff, 0xfe, 0xfd, 0xfc, 0xfb, 0xfa, 0xf9, 0xf8, 0xf7, 0xf6, 0xf5, 0xf4,
0xf3, 0xf2, 0xf1, 0xf0, 0xef, 0xee, 0xed, 0xec, 0xeb, 0xea, 0xe9, 0xe8,
0xe7, 0xe6, 0xe5, 0xe4, 0xe3, 0xe2, 0xe1, 0xe0, 0xdf, 0xde, 0xdd, 0xdc,
0xdb, 0xda, 0xd9, 0xd8, 0xd7, 0xd6, 0xd5, 0xd4, 0xd3, 0xd2, 0xd1, 0xd0,
0xcf, 0xce, 0xcd, 0xcc, 0xcb, 0xca, 0xc9, 0xc8, 0xc7, 0xc6, 0xc5, 0xc4,
0xc3, 0xc2, 0xc1, 0xc0, 0xbf, 0xbe, 0xbd, 0xbc, 0xbb, 0xba, 0xb9, 0xb8,
0xb7, 0xb6, 0xb5, 0xb4, 0xb3, 0xb2, 0xb1, 0xb0, 0xaf, 0xae, 0xad, 0xac,
0xab, 0xaa, 0xa9, 0xa8, 0xa7, 0xa6, 0xa5, 0xa4, 0xa3, 0xa2, 0xa1, 0xa0,
0x9f, 0x9e, 0x9d, 0x9c, 0x9b, 0x9a, 0x99, 0x98, 0x97, 0x96, 0x95, 0x94,
0x93, 0x92, 0x91, 0x90, 0x8f, 0x8e, 0x8d, 0x8c, 0x8b, 0x8a, 0x89, 0x88,
0x87, 0x86, 0x85, 0x84, 0x83, 0x82, 0x81, 0x80, 0x7f, 0x7e, 0x7d, 0x7c,
0x7b, 0x7a, 0x79, 0x78, 0x77, 0x76, 0x75, 0x74, 0x73, 0x72, 0x71, 0x70,
0x6f, 0x6e, 0x6d, 0x6c, 0x6b, 0x6a, 0x69, 0x68, 0x67, 0x66, 0x65, 0x64,
0x63, 0x62, 0x61, 0x60, 0x5f, 0x5e, 0x5d, 0x5c, 0x5b, 0x5a, 0x59, 0x58,
0x57, 0x56, 0x55, 0x54, 0x53, 0x52, 0x51, 0x50, 0x4f, 0x4e, 0x4d, 0x4c,
0x4b, 0x4a, 0x49, 0x48, 0x47, 0x46, 0x45, 0x44, 0x43, 0x42, 0x41, 0x40,
0x3f, 0x3e, 0x3d, 0x3c, 0x3b, 0x3a, 0x39, 0x38, 0x37, 0x36, 0x35, 0x34,
0x33, 0x32, 0x31, 0x30, 0x2f, 0x2e, 0x2d, 0x2c, 0x2b, 0x2a, 0x29, 0x28,
0x27, 0x26, 0x25, 0x24, 0x23, 0x22, 0x21, 0x20, 0x1f, 0x1e, 0x1d, 0x1c,
0x1b, 0x1a, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10,
0x0f, 0x0e, 0x0d, 0x0c, 0x0b, 0x0a, 0x09, 0x08, 0x07, 0x06, 0x05, 0x04,
0x03, 0x02, 0x01, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08,
0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x10, 0x11, 0x12, 0x13, 0x14,
0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20,
0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c,
0x2d, 0x2e, 0x2f, 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38,
0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, 0x40, 0x41, 0x42, 0x43, 0x44,
0x45, 0x46, 0x47, 0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f, 0x50,
0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, 0x5b, 0x5c,
0x5d, 0x5e, 0x5f, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72, 0x73, 0x74,
0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80,
0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8a, 0x8b, 0x8c,
0x8d, 0x8e, 0x8f, 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1, 0xa2, 0xa3, 0xa4,
0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf, 0xb0,
0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xbb, 0xbc,
0xbd, 0xbe, 0xbf, 0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7, 0xc8,
0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf, 0xd0, 0xd1, 0xd2, 0xd3, 0xd4,
0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf, 0xe0,
0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec,
0xed, 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff
};
static const uint8_t sclip1[1020 + 1020 + 1] = {
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80,
0x80, 0x80, 0x80, 0x80, 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x90, 0x91, 0x92, 0x93,
0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f,
0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xab,
0xac, 0xad, 0xae, 0xaf, 0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7,
0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf, 0xc0, 0xc1, 0xc2, 0xc3,
0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf,
0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xdb,
0xdc, 0xdd, 0xde, 0xdf, 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7,
0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3,
0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0a, 0x0b,
0x0c, 0x0d, 0x0e, 0x0f, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20, 0x21, 0x22, 0x23,
0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f,
0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b,
0x3c, 0x3d, 0x3e, 0x3f, 0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47,
0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f, 0x50, 0x51, 0x52, 0x53,
0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f,
0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b,
0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f,
0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f
};
static const uint8_t sclip2[112 + 112 + 1] = {
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0,
0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb,
0xfc, 0xfd, 0xfe, 0xff, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f
};
static const uint8_t clip1[255 + 511 + 1] = {
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08,
0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x10, 0x11, 0x12, 0x13, 0x14,
0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20,
0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c,
0x2d, 0x2e, 0x2f, 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38,
0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, 0x40, 0x41, 0x42, 0x43, 0x44,
0x45, 0x46, 0x47, 0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f, 0x50,
0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a, 0x5b, 0x5c,
0x5d, 0x5e, 0x5f, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72, 0x73, 0x74,
0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80,
0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8a, 0x8b, 0x8c,
0x8d, 0x8e, 0x8f, 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1, 0xa2, 0xa3, 0xa4,
0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf, 0xb0,
0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xbb, 0xbc,
0xbd, 0xbe, 0xbf, 0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7, 0xc8,
0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf, 0xd0, 0xd1, 0xd2, 0xd3, 0xd4,
0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf, 0xe0,
0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec,
0xed, 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
};
#else
// uninitialized tables
static uint8_t abs0[255 + 255 + 1];
static int8_t sclip1[1020 + 1020 + 1];
static int8_t sclip2[112 + 112 + 1];
static uint8_t clip1[255 + 511 + 1];
// We declare this variable 'volatile' to prevent instruction reordering
// and make sure it's set to true _last_ (so as to be thread-safe)
static volatile int tables_ok = 0;
#endif // USE_STATIC_TABLES
const int8_t* const VP8ksclip1 = (const int8_t*)&sclip1[1020];
const int8_t* const VP8ksclip2 = (const int8_t*)&sclip2[112];
const uint8_t* const VP8kclip1 = &clip1[255];
const uint8_t* const VP8kabs0 = &abs0[255];
WEBP_TSAN_IGNORE_FUNCTION void VP8InitClipTables(void) {
#if (USE_STATIC_TABLES == 0)
int i;
if (!tables_ok) {
for (i = -255; i <= 255; ++i) {
abs0[255 + i] = (i < 0) ? -i : i;
}
for (i = -1020; i <= 1020; ++i) {
sclip1[1020 + i] = (i < -128) ? -128 : (i > 127) ? 127 : i;
}
for (i = -112; i <= 112; ++i) {
sclip2[112 + i] = (i < -16) ? -16 : (i > 15) ? 15 : i;
}
for (i = -255; i <= 255 + 255; ++i) {
clip1[255 + i] = (i < 0) ? 0 : (i > 255) ? 255 : i;
}
tables_ok = 1;
}
#endif // USE_STATIC_TABLES
}

587
media/libwebp/src/dsp/dec_mips32.c Normal file
Просмотреть файл

@ -0,0 +1,587 @@
// Copyright 2014 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// MIPS version of dsp functions
//
// Author(s): Djordje Pesut (djordje.pesut@imgtec.com)
// Jovan Zelincevic (jovan.zelincevic@imgtec.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_MIPS32)
#include "src/dsp/mips_macro.h"
static const int kC1 = 20091 + (1 << 16);
static const int kC2 = 35468;
static WEBP_INLINE int abs_mips32(int x) {
const int sign = x >> 31;
return (x ^ sign) - sign;
}
// 4 pixels in, 2 pixels out
static WEBP_INLINE void do_filter2(uint8_t* p, int step) {
const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step];
const int a = 3 * (q0 - p0) + VP8ksclip1[p1 - q1];
const int a1 = VP8ksclip2[(a + 4) >> 3];
const int a2 = VP8ksclip2[(a + 3) >> 3];
p[-step] = VP8kclip1[p0 + a2];
p[ 0] = VP8kclip1[q0 - a1];
}
// 4 pixels in, 4 pixels out
static WEBP_INLINE void do_filter4(uint8_t* p, int step) {
const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step];
const int a = 3 * (q0 - p0);
const int a1 = VP8ksclip2[(a + 4) >> 3];
const int a2 = VP8ksclip2[(a + 3) >> 3];
const int a3 = (a1 + 1) >> 1;
p[-2 * step] = VP8kclip1[p1 + a3];
p[- step] = VP8kclip1[p0 + a2];
p[ 0] = VP8kclip1[q0 - a1];
p[ step] = VP8kclip1[q1 - a3];
}
// 6 pixels in, 6 pixels out
static WEBP_INLINE void do_filter6(uint8_t* p, int step) {
const int p2 = p[-3 * step], p1 = p[-2 * step], p0 = p[-step];
const int q0 = p[0], q1 = p[step], q2 = p[2 * step];
const int a = VP8ksclip1[3 * (q0 - p0) + VP8ksclip1[p1 - q1]];
// a is in [-128,127], a1 in [-27,27], a2 in [-18,18] and a3 in [-9,9]
const int a1 = (27 * a + 63) >> 7; // eq. to ((3 * a + 7) * 9) >> 7
const int a2 = (18 * a + 63) >> 7; // eq. to ((2 * a + 7) * 9) >> 7
const int a3 = (9 * a + 63) >> 7; // eq. to ((1 * a + 7) * 9) >> 7
p[-3 * step] = VP8kclip1[p2 + a3];
p[-2 * step] = VP8kclip1[p1 + a2];
p[- step] = VP8kclip1[p0 + a1];
p[ 0] = VP8kclip1[q0 - a1];
p[ step] = VP8kclip1[q1 - a2];
p[ 2 * step] = VP8kclip1[q2 - a3];
}
static WEBP_INLINE int hev(const uint8_t* p, int step, int thresh) {
const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step];
return (abs_mips32(p1 - p0) > thresh) || (abs_mips32(q1 - q0) > thresh);
}
static WEBP_INLINE int needs_filter(const uint8_t* p, int step, int t) {
const int p1 = p[-2 * step], p0 = p[-step], q0 = p[0], q1 = p[step];
return ((4 * abs_mips32(p0 - q0) + abs_mips32(p1 - q1)) <= t);
}
static WEBP_INLINE int needs_filter2(const uint8_t* p,
int step, int t, int it) {
const int p3 = p[-4 * step], p2 = p[-3 * step];
const int p1 = p[-2 * step], p0 = p[-step];
const int q0 = p[0], q1 = p[step], q2 = p[2 * step], q3 = p[3 * step];
if ((4 * abs_mips32(p0 - q0) + abs_mips32(p1 - q1)) > t) {
return 0;
}
return abs_mips32(p3 - p2) <= it && abs_mips32(p2 - p1) <= it &&
abs_mips32(p1 - p0) <= it && abs_mips32(q3 - q2) <= it &&
abs_mips32(q2 - q1) <= it && abs_mips32(q1 - q0) <= it;
}
static WEBP_INLINE void FilterLoop26(uint8_t* p,
int hstride, int vstride, int size,
int thresh, int ithresh, int hev_thresh) {
const int thresh2 = 2 * thresh + 1;
while (size-- > 0) {
if (needs_filter2(p, hstride, thresh2, ithresh)) {
if (hev(p, hstride, hev_thresh)) {
do_filter2(p, hstride);
} else {
do_filter6(p, hstride);
}
}
p += vstride;
}
}
static WEBP_INLINE void FilterLoop24(uint8_t* p,
int hstride, int vstride, int size,
int thresh, int ithresh, int hev_thresh) {
const int thresh2 = 2 * thresh + 1;
while (size-- > 0) {
if (needs_filter2(p, hstride, thresh2, ithresh)) {
if (hev(p, hstride, hev_thresh)) {
do_filter2(p, hstride);
} else {
do_filter4(p, hstride);
}
}
p += vstride;
}
}
// on macroblock edges
static void VFilter16(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(p, stride, 1, 16, thresh, ithresh, hev_thresh);
}
static void HFilter16(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(p, 1, stride, 16, thresh, ithresh, hev_thresh);
}
// 8-pixels wide variant, for chroma filtering
static void VFilter8(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(u, stride, 1, 8, thresh, ithresh, hev_thresh);
FilterLoop26(v, stride, 1, 8, thresh, ithresh, hev_thresh);
}
static void HFilter8(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(u, 1, stride, 8, thresh, ithresh, hev_thresh);
FilterLoop26(v, 1, stride, 8, thresh, ithresh, hev_thresh);
}
static void VFilter8i(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop24(u + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh);
FilterLoop24(v + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh);
}
static void HFilter8i(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop24(u + 4, 1, stride, 8, thresh, ithresh, hev_thresh);
FilterLoop24(v + 4, 1, stride, 8, thresh, ithresh, hev_thresh);
}
// on three inner edges
static void VFilter16i(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4 * stride;
FilterLoop24(p, stride, 1, 16, thresh, ithresh, hev_thresh);
}
}
static void HFilter16i(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4;
FilterLoop24(p, 1, stride, 16, thresh, ithresh, hev_thresh);
}
}
//------------------------------------------------------------------------------
// Simple In-loop filtering (Paragraph 15.2)
static void SimpleVFilter16(uint8_t* p, int stride, int thresh) {
int i;
const int thresh2 = 2 * thresh + 1;
for (i = 0; i < 16; ++i) {
if (needs_filter(p + i, stride, thresh2)) {
do_filter2(p + i, stride);
}
}
}
static void SimpleHFilter16(uint8_t* p, int stride, int thresh) {
int i;
const int thresh2 = 2 * thresh + 1;
for (i = 0; i < 16; ++i) {
if (needs_filter(p + i * stride, 1, thresh2)) {
do_filter2(p + i * stride, 1);
}
}
}
static void SimpleVFilter16i(uint8_t* p, int stride, int thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4 * stride;
SimpleVFilter16(p, stride, thresh);
}
}
static void SimpleHFilter16i(uint8_t* p, int stride, int thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4;
SimpleHFilter16(p, stride, thresh);
}
}
static void TransformOne(const int16_t* in, uint8_t* dst) {
int temp0, temp1, temp2, temp3, temp4;
int temp5, temp6, temp7, temp8, temp9;
int temp10, temp11, temp12, temp13, temp14;
int temp15, temp16, temp17, temp18;
int16_t* p_in = (int16_t*)in;
// loops unrolled and merged to avoid usage of tmp buffer
// and to reduce number of stalls. MUL macro is written
// in assembler and inlined
__asm__ volatile(
"lh %[temp0], 0(%[in]) \n\t"
"lh %[temp8], 16(%[in]) \n\t"
"lh %[temp4], 8(%[in]) \n\t"
"lh %[temp12], 24(%[in]) \n\t"
"addu %[temp16], %[temp0], %[temp8] \n\t"
"subu %[temp0], %[temp0], %[temp8] \n\t"
"mul %[temp8], %[temp4], %[kC2] \n\t"
"mul %[temp17], %[temp12], %[kC1] \n\t"
"mul %[temp4], %[temp4], %[kC1] \n\t"
"mul %[temp12], %[temp12], %[kC2] \n\t"
"lh %[temp1], 2(%[in]) \n\t"
"lh %[temp5], 10(%[in]) \n\t"
"lh %[temp9], 18(%[in]) \n\t"
"lh %[temp13], 26(%[in]) \n\t"
"sra %[temp8], %[temp8], 16 \n\t"
"sra %[temp17], %[temp17], 16 \n\t"
"sra %[temp4], %[temp4], 16 \n\t"
"sra %[temp12], %[temp12], 16 \n\t"
"lh %[temp2], 4(%[in]) \n\t"
"lh %[temp6], 12(%[in]) \n\t"
"lh %[temp10], 20(%[in]) \n\t"
"lh %[temp14], 28(%[in]) \n\t"
"subu %[temp17], %[temp8], %[temp17] \n\t"
"addu %[temp4], %[temp4], %[temp12] \n\t"
"addu %[temp8], %[temp16], %[temp4] \n\t"
"subu %[temp4], %[temp16], %[temp4] \n\t"
"addu %[temp16], %[temp1], %[temp9] \n\t"
"subu %[temp1], %[temp1], %[temp9] \n\t"
"lh %[temp3], 6(%[in]) \n\t"
"lh %[temp7], 14(%[in]) \n\t"
"lh %[temp11], 22(%[in]) \n\t"
"lh %[temp15], 30(%[in]) \n\t"
"addu %[temp12], %[temp0], %[temp17] \n\t"
"subu %[temp0], %[temp0], %[temp17] \n\t"
"mul %[temp9], %[temp5], %[kC2] \n\t"
"mul %[temp17], %[temp13], %[kC1] \n\t"
"mul %[temp5], %[temp5], %[kC1] \n\t"
"mul %[temp13], %[temp13], %[kC2] \n\t"
"sra %[temp9], %[temp9], 16 \n\t"
"sra %[temp17], %[temp17], 16 \n\t"
"subu %[temp17], %[temp9], %[temp17] \n\t"
"sra %[temp5], %[temp5], 16 \n\t"
"sra %[temp13], %[temp13], 16 \n\t"
"addu %[temp5], %[temp5], %[temp13] \n\t"
"addu %[temp13], %[temp1], %[temp17] \n\t"
"subu %[temp1], %[temp1], %[temp17] \n\t"
"mul %[temp17], %[temp14], %[kC1] \n\t"
"mul %[temp14], %[temp14], %[kC2] \n\t"
"addu %[temp9], %[temp16], %[temp5] \n\t"
"subu %[temp5], %[temp16], %[temp5] \n\t"
"addu %[temp16], %[temp2], %[temp10] \n\t"
"subu %[temp2], %[temp2], %[temp10] \n\t"
"mul %[temp10], %[temp6], %[kC2] \n\t"
"mul %[temp6], %[temp6], %[kC1] \n\t"
"sra %[temp17], %[temp17], 16 \n\t"
"sra %[temp14], %[temp14], 16 \n\t"
"sra %[temp10], %[temp10], 16 \n\t"
"sra %[temp6], %[temp6], 16 \n\t"
"subu %[temp17], %[temp10], %[temp17] \n\t"
"addu %[temp6], %[temp6], %[temp14] \n\t"
"addu %[temp10], %[temp16], %[temp6] \n\t"
"subu %[temp6], %[temp16], %[temp6] \n\t"
"addu %[temp14], %[temp2], %[temp17] \n\t"
"subu %[temp2], %[temp2], %[temp17] \n\t"
"mul %[temp17], %[temp15], %[kC1] \n\t"
"mul %[temp15], %[temp15], %[kC2] \n\t"
"addu %[temp16], %[temp3], %[temp11] \n\t"
"subu %[temp3], %[temp3], %[temp11] \n\t"
"mul %[temp11], %[temp7], %[kC2] \n\t"
"mul %[temp7], %[temp7], %[kC1] \n\t"
"addiu %[temp8], %[temp8], 4 \n\t"
"addiu %[temp12], %[temp12], 4 \n\t"
"addiu %[temp0], %[temp0], 4 \n\t"
"addiu %[temp4], %[temp4], 4 \n\t"
"sra %[temp17], %[temp17], 16 \n\t"
"sra %[temp15], %[temp15], 16 \n\t"
"sra %[temp11], %[temp11], 16 \n\t"
"sra %[temp7], %[temp7], 16 \n\t"
"subu %[temp17], %[temp11], %[temp17] \n\t"
"addu %[temp7], %[temp7], %[temp15] \n\t"
"addu %[temp15], %[temp3], %[temp17] \n\t"
"subu %[temp3], %[temp3], %[temp17] \n\t"
"addu %[temp11], %[temp16], %[temp7] \n\t"
"subu %[temp7], %[temp16], %[temp7] \n\t"
"addu %[temp16], %[temp8], %[temp10] \n\t"
"subu %[temp8], %[temp8], %[temp10] \n\t"
"mul %[temp10], %[temp9], %[kC2] \n\t"
"mul %[temp17], %[temp11], %[kC1] \n\t"
"mul %[temp9], %[temp9], %[kC1] \n\t"
"mul %[temp11], %[temp11], %[kC2] \n\t"
"sra %[temp10], %[temp10], 16 \n\t"
"sra %[temp17], %[temp17], 16 \n\t"
"sra %[temp9], %[temp9], 16 \n\t"
"sra %[temp11], %[temp11], 16 \n\t"
"subu %[temp17], %[temp10], %[temp17] \n\t"
"addu %[temp11], %[temp9], %[temp11] \n\t"
"addu %[temp10], %[temp12], %[temp14] \n\t"
"subu %[temp12], %[temp12], %[temp14] \n\t"
"mul %[temp14], %[temp13], %[kC2] \n\t"
"mul %[temp9], %[temp15], %[kC1] \n\t"
"mul %[temp13], %[temp13], %[kC1] \n\t"
"mul %[temp15], %[temp15], %[kC2] \n\t"
"sra %[temp14], %[temp14], 16 \n\t"
"sra %[temp9], %[temp9], 16 \n\t"
"sra %[temp13], %[temp13], 16 \n\t"
"sra %[temp15], %[temp15], 16 \n\t"
"subu %[temp9], %[temp14], %[temp9] \n\t"
"addu %[temp15], %[temp13], %[temp15] \n\t"
"addu %[temp14], %[temp0], %[temp2] \n\t"
"subu %[temp0], %[temp0], %[temp2] \n\t"
"mul %[temp2], %[temp1], %[kC2] \n\t"
"mul %[temp13], %[temp3], %[kC1] \n\t"
"mul %[temp1], %[temp1], %[kC1] \n\t"
"mul %[temp3], %[temp3], %[kC2] \n\t"
"sra %[temp2], %[temp2], 16 \n\t"
"sra %[temp13], %[temp13], 16 \n\t"
"sra %[temp1], %[temp1], 16 \n\t"
"sra %[temp3], %[temp3], 16 \n\t"
"subu %[temp13], %[temp2], %[temp13] \n\t"
"addu %[temp3], %[temp1], %[temp3] \n\t"
"addu %[temp2], %[temp4], %[temp6] \n\t"
"subu %[temp4], %[temp4], %[temp6] \n\t"
"mul %[temp6], %[temp5], %[kC2] \n\t"
"mul %[temp1], %[temp7], %[kC1] \n\t"
"mul %[temp5], %[temp5], %[kC1] \n\t"
"mul %[temp7], %[temp7], %[kC2] \n\t"
"sra %[temp6], %[temp6], 16 \n\t"
"sra %[temp1], %[temp1], 16 \n\t"
"sra %[temp5], %[temp5], 16 \n\t"
"sra %[temp7], %[temp7], 16 \n\t"
"subu %[temp1], %[temp6], %[temp1] \n\t"
"addu %[temp7], %[temp5], %[temp7] \n\t"
"addu %[temp5], %[temp16], %[temp11] \n\t"
"subu %[temp16], %[temp16], %[temp11] \n\t"
"addu %[temp11], %[temp8], %[temp17] \n\t"
"subu %[temp8], %[temp8], %[temp17] \n\t"
"sra %[temp5], %[temp5], 3 \n\t"
"sra %[temp16], %[temp16], 3 \n\t"
"sra %[temp11], %[temp11], 3 \n\t"
"sra %[temp8], %[temp8], 3 \n\t"
"addu %[temp17], %[temp10], %[temp15] \n\t"
"subu %[temp10], %[temp10], %[temp15] \n\t"
"addu %[temp15], %[temp12], %[temp9] \n\t"
"subu %[temp12], %[temp12], %[temp9] \n\t"
"sra %[temp17], %[temp17], 3 \n\t"
"sra %[temp10], %[temp10], 3 \n\t"
"sra %[temp15], %[temp15], 3 \n\t"
"sra %[temp12], %[temp12], 3 \n\t"
"addu %[temp9], %[temp14], %[temp3] \n\t"
"subu %[temp14], %[temp14], %[temp3] \n\t"
"addu %[temp3], %[temp0], %[temp13] \n\t"
"subu %[temp0], %[temp0], %[temp13] \n\t"
"sra %[temp9], %[temp9], 3 \n\t"
"sra %[temp14], %[temp14], 3 \n\t"
"sra %[temp3], %[temp3], 3 \n\t"
"sra %[temp0], %[temp0], 3 \n\t"
"addu %[temp13], %[temp2], %[temp7] \n\t"
"subu %[temp2], %[temp2], %[temp7] \n\t"
"addu %[temp7], %[temp4], %[temp1] \n\t"
"subu %[temp4], %[temp4], %[temp1] \n\t"
"sra %[temp13], %[temp13], 3 \n\t"
"sra %[temp2], %[temp2], 3 \n\t"
"sra %[temp7], %[temp7], 3 \n\t"
"sra %[temp4], %[temp4], 3 \n\t"
"addiu %[temp6], $zero, 255 \n\t"
"lbu %[temp1], 0+0*" XSTR(BPS) "(%[dst]) \n\t"
"addu %[temp1], %[temp1], %[temp5] \n\t"
"sra %[temp5], %[temp1], 8 \n\t"
"sra %[temp18], %[temp1], 31 \n\t"
"beqz %[temp5], 1f \n\t"
"xor %[temp1], %[temp1], %[temp1] \n\t"
"movz %[temp1], %[temp6], %[temp18] \n\t"
"1: \n\t"
"lbu %[temp18], 1+0*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp1], 0+0*" XSTR(BPS) "(%[dst]) \n\t"
"addu %[temp18], %[temp18], %[temp11] \n\t"
"sra %[temp11], %[temp18], 8 \n\t"
"sra %[temp1], %[temp18], 31 \n\t"
"beqz %[temp11], 2f \n\t"
"xor %[temp18], %[temp18], %[temp18] \n\t"
"movz %[temp18], %[temp6], %[temp1] \n\t"
"2: \n\t"
"lbu %[temp1], 2+0*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp18], 1+0*" XSTR(BPS) "(%[dst]) \n\t"
"addu %[temp1], %[temp1], %[temp8] \n\t"
"sra %[temp8], %[temp1], 8 \n\t"
"sra %[temp18], %[temp1], 31 \n\t"
"beqz %[temp8], 3f \n\t"
"xor %[temp1], %[temp1], %[temp1] \n\t"
"movz %[temp1], %[temp6], %[temp18] \n\t"
"3: \n\t"
"lbu %[temp18], 3+0*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp1], 2+0*" XSTR(BPS) "(%[dst]) \n\t"
"addu %[temp18], %[temp18], %[temp16] \n\t"
"sra %[temp16], %[temp18], 8 \n\t"
"sra %[temp1], %[temp18], 31 \n\t"
"beqz %[temp16], 4f \n\t"
"xor %[temp18], %[temp18], %[temp18] \n\t"
"movz %[temp18], %[temp6], %[temp1] \n\t"
"4: \n\t"
"sb %[temp18], 3+0*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp5], 0+1*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp8], 1+1*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp11], 2+1*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp16], 3+1*" XSTR(BPS) "(%[dst]) \n\t"
"addu %[temp5], %[temp5], %[temp17] \n\t"
"addu %[temp8], %[temp8], %[temp15] \n\t"
"addu %[temp11], %[temp11], %[temp12] \n\t"
"addu %[temp16], %[temp16], %[temp10] \n\t"
"sra %[temp18], %[temp5], 8 \n\t"
"sra %[temp1], %[temp5], 31 \n\t"
"beqz %[temp18], 5f \n\t"
"xor %[temp5], %[temp5], %[temp5] \n\t"
"movz %[temp5], %[temp6], %[temp1] \n\t"
"5: \n\t"
"sra %[temp18], %[temp8], 8 \n\t"
"sra %[temp1], %[temp8], 31 \n\t"
"beqz %[temp18], 6f \n\t"
"xor %[temp8], %[temp8], %[temp8] \n\t"
"movz %[temp8], %[temp6], %[temp1] \n\t"
"6: \n\t"
"sra %[temp18], %[temp11], 8 \n\t"
"sra %[temp1], %[temp11], 31 \n\t"
"sra %[temp17], %[temp16], 8 \n\t"
"sra %[temp15], %[temp16], 31 \n\t"
"beqz %[temp18], 7f \n\t"
"xor %[temp11], %[temp11], %[temp11] \n\t"
"movz %[temp11], %[temp6], %[temp1] \n\t"
"7: \n\t"
"beqz %[temp17], 8f \n\t"
"xor %[temp16], %[temp16], %[temp16] \n\t"
"movz %[temp16], %[temp6], %[temp15] \n\t"
"8: \n\t"
"sb %[temp5], 0+1*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp8], 1+1*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp11], 2+1*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp16], 3+1*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp5], 0+2*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp8], 1+2*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp11], 2+2*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp16], 3+2*" XSTR(BPS) "(%[dst]) \n\t"
"addu %[temp5], %[temp5], %[temp9] \n\t"
"addu %[temp8], %[temp8], %[temp3] \n\t"
"addu %[temp11], %[temp11], %[temp0] \n\t"
"addu %[temp16], %[temp16], %[temp14] \n\t"
"sra %[temp18], %[temp5], 8 \n\t"
"sra %[temp1], %[temp5], 31 \n\t"
"sra %[temp17], %[temp8], 8 \n\t"
"sra %[temp15], %[temp8], 31 \n\t"
"sra %[temp12], %[temp11], 8 \n\t"
"sra %[temp10], %[temp11], 31 \n\t"
"sra %[temp9], %[temp16], 8 \n\t"
"sra %[temp3], %[temp16], 31 \n\t"
"beqz %[temp18], 9f \n\t"
"xor %[temp5], %[temp5], %[temp5] \n\t"
"movz %[temp5], %[temp6], %[temp1] \n\t"
"9: \n\t"
"beqz %[temp17], 10f \n\t"
"xor %[temp8], %[temp8], %[temp8] \n\t"
"movz %[temp8], %[temp6], %[temp15] \n\t"
"10: \n\t"
"beqz %[temp12], 11f \n\t"
"xor %[temp11], %[temp11], %[temp11] \n\t"
"movz %[temp11], %[temp6], %[temp10] \n\t"
"11: \n\t"
"beqz %[temp9], 12f \n\t"
"xor %[temp16], %[temp16], %[temp16] \n\t"
"movz %[temp16], %[temp6], %[temp3] \n\t"
"12: \n\t"
"sb %[temp5], 0+2*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp8], 1+2*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp11], 2+2*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp16], 3+2*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp5], 0+3*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp8], 1+3*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp11], 2+3*" XSTR(BPS) "(%[dst]) \n\t"
"lbu %[temp16], 3+3*" XSTR(BPS) "(%[dst]) \n\t"
"addu %[temp5], %[temp5], %[temp13] \n\t"
"addu %[temp8], %[temp8], %[temp7] \n\t"
"addu %[temp11], %[temp11], %[temp4] \n\t"
"addu %[temp16], %[temp16], %[temp2] \n\t"
"sra %[temp18], %[temp5], 8 \n\t"
"sra %[temp1], %[temp5], 31 \n\t"
"sra %[temp17], %[temp8], 8 \n\t"
"sra %[temp15], %[temp8], 31 \n\t"
"sra %[temp12], %[temp11], 8 \n\t"
"sra %[temp10], %[temp11], 31 \n\t"
"sra %[temp9], %[temp16], 8 \n\t"
"sra %[temp3], %[temp16], 31 \n\t"
"beqz %[temp18], 13f \n\t"
"xor %[temp5], %[temp5], %[temp5] \n\t"
"movz %[temp5], %[temp6], %[temp1] \n\t"
"13: \n\t"
"beqz %[temp17], 14f \n\t"
"xor %[temp8], %[temp8], %[temp8] \n\t"
"movz %[temp8], %[temp6], %[temp15] \n\t"
"14: \n\t"
"beqz %[temp12], 15f \n\t"
"xor %[temp11], %[temp11], %[temp11] \n\t"
"movz %[temp11], %[temp6], %[temp10] \n\t"
"15: \n\t"
"beqz %[temp9], 16f \n\t"
"xor %[temp16], %[temp16], %[temp16] \n\t"
"movz %[temp16], %[temp6], %[temp3] \n\t"
"16: \n\t"
"sb %[temp5], 0+3*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp8], 1+3*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp11], 2+3*" XSTR(BPS) "(%[dst]) \n\t"
"sb %[temp16], 3+3*" XSTR(BPS) "(%[dst]) \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8),
[temp9]"=&r"(temp9), [temp10]"=&r"(temp10), [temp11]"=&r"(temp11),
[temp12]"=&r"(temp12), [temp13]"=&r"(temp13), [temp14]"=&r"(temp14),
[temp15]"=&r"(temp15), [temp16]"=&r"(temp16), [temp17]"=&r"(temp17),
[temp18]"=&r"(temp18)
: [in]"r"(p_in), [kC1]"r"(kC1), [kC2]"r"(kC2), [dst]"r"(dst)
: "memory", "hi", "lo"
);
}
static void TransformTwo(const int16_t* in, uint8_t* dst, int do_two) {
TransformOne(in, dst);
if (do_two) {
TransformOne(in + 16, dst + 4);
}
}
//------------------------------------------------------------------------------
// Entry point
extern void VP8DspInitMIPS32(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8DspInitMIPS32(void) {
VP8InitClipTables();
VP8Transform = TransformTwo;
VP8VFilter16 = VFilter16;
VP8HFilter16 = HFilter16;
VP8VFilter8 = VFilter8;
VP8HFilter8 = HFilter8;
VP8VFilter16i = VFilter16i;
VP8HFilter16i = HFilter16i;
VP8VFilter8i = VFilter8i;
VP8HFilter8i = HFilter8i;
VP8SimpleVFilter16 = SimpleVFilter16;
VP8SimpleHFilter16 = SimpleHFilter16;
VP8SimpleVFilter16i = SimpleVFilter16i;
VP8SimpleHFilter16i = SimpleHFilter16i;
}
#else // !WEBP_USE_MIPS32
WEBP_DSP_INIT_STUB(VP8DspInitMIPS32)
#endif // WEBP_USE_MIPS32

994
media/libwebp/src/dsp/dec_mips_dsp_r2.c Normal file
Просмотреть файл

@ -0,0 +1,994 @@
// Copyright 2014 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// MIPS version of dsp functions
//
// Author(s): Djordje Pesut (djordje.pesut@imgtec.com)
// Jovan Zelincevic (jovan.zelincevic@imgtec.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_MIPS_DSP_R2)
#include "src/dsp/mips_macro.h"
static const int kC1 = 20091 + (1 << 16);
static const int kC2 = 35468;
#define MUL(a, b) (((a) * (b)) >> 16)
static void TransformDC(const int16_t* in, uint8_t* dst) {
int temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8, temp9, temp10;
__asm__ volatile (
LOAD_WITH_OFFSET_X4(temp1, temp2, temp3, temp4, dst,
0, 0, 0, 0,
0, 1, 2, 3,
BPS)
"lh %[temp5], 0(%[in]) \n\t"
"addiu %[temp5], %[temp5], 4 \n\t"
"ins %[temp5], %[temp5], 16, 16 \n\t"
"shra.ph %[temp5], %[temp5], 3 \n\t"
CONVERT_2_BYTES_TO_HALF(temp6, temp7, temp8, temp9, temp10, temp1, temp2,
temp3, temp1, temp2, temp3, temp4)
STORE_SAT_SUM_X2(temp6, temp7, temp8, temp9, temp10, temp1, temp2, temp3,
temp5, temp5, temp5, temp5, temp5, temp5, temp5, temp5,
dst, 0, 1, 2, 3, BPS)
OUTPUT_EARLY_CLOBBER_REGS_10()
: [in]"r"(in), [dst]"r"(dst)
: "memory"
);
}
static void TransformAC3(const int16_t* in, uint8_t* dst) {
const int a = in[0] + 4;
int c4 = MUL(in[4], kC2);
const int d4 = MUL(in[4], kC1);
const int c1 = MUL(in[1], kC2);
const int d1 = MUL(in[1], kC1);
int temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8, temp9;
int temp10, temp11, temp12, temp13, temp14, temp15, temp16, temp17, temp18;
__asm__ volatile (
"ins %[c4], %[d4], 16, 16 \n\t"
"replv.ph %[temp1], %[a] \n\t"
"replv.ph %[temp4], %[d1] \n\t"
ADD_SUB_HALVES(temp2, temp3, temp1, c4)
"replv.ph %[temp5], %[c1] \n\t"
SHIFT_R_SUM_X2(temp1, temp6, temp7, temp8, temp2, temp9, temp10, temp4,
temp2, temp2, temp3, temp3, temp4, temp5, temp4, temp5)
LOAD_WITH_OFFSET_X4(temp3, temp5, temp11, temp12, dst,
0, 0, 0, 0,
0, 1, 2, 3,
BPS)
CONVERT_2_BYTES_TO_HALF(temp13, temp14, temp3, temp15, temp5, temp16,
temp11, temp17, temp3, temp5, temp11, temp12)
PACK_2_HALVES_TO_WORD(temp12, temp18, temp7, temp6, temp1, temp8, temp2,
temp4, temp7, temp6, temp10, temp9)
STORE_SAT_SUM_X2(temp13, temp14, temp3, temp15, temp5, temp16, temp11,
temp17, temp12, temp18, temp1, temp8, temp2, temp4,
temp7, temp6, dst, 0, 1, 2, 3, BPS)
OUTPUT_EARLY_CLOBBER_REGS_18(),
[c4]"+&r"(c4)
: [dst]"r"(dst), [a]"r"(a), [d1]"r"(d1), [d4]"r"(d4), [c1]"r"(c1)
: "memory"
);
}
static void TransformOne(const int16_t* in, uint8_t* dst) {
int temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8, temp9;
int temp10, temp11, temp12, temp13, temp14, temp15, temp16, temp17, temp18;
__asm__ volatile (
"ulw %[temp1], 0(%[in]) \n\t"
"ulw %[temp2], 16(%[in]) \n\t"
LOAD_IN_X2(temp5, temp6, 24, 26)
ADD_SUB_HALVES(temp3, temp4, temp1, temp2)
LOAD_IN_X2(temp1, temp2, 8, 10)
MUL_SHIFT_SUM(temp7, temp8, temp9, temp10, temp11, temp12, temp13, temp14,
temp10, temp8, temp9, temp7, temp1, temp2, temp5, temp6,
temp13, temp11, temp14, temp12)
INSERT_HALF_X2(temp8, temp7, temp10, temp9)
"ulw %[temp17], 4(%[in]) \n\t"
"ulw %[temp18], 20(%[in]) \n\t"
ADD_SUB_HALVES(temp1, temp2, temp3, temp8)
ADD_SUB_HALVES(temp5, temp6, temp4, temp7)
ADD_SUB_HALVES(temp7, temp8, temp17, temp18)
LOAD_IN_X2(temp17, temp18, 12, 14)
LOAD_IN_X2(temp9, temp10, 28, 30)
MUL_SHIFT_SUM(temp11, temp12, temp13, temp14, temp15, temp16, temp4, temp17,
temp12, temp14, temp11, temp13, temp17, temp18, temp9, temp10,
temp15, temp4, temp16, temp17)
INSERT_HALF_X2(temp11, temp12, temp13, temp14)
ADD_SUB_HALVES(temp17, temp8, temp8, temp11)
ADD_SUB_HALVES(temp3, temp4, temp7, temp12)
// horizontal
SRA_16(temp9, temp10, temp11, temp12, temp1, temp2, temp5, temp6)
INSERT_HALF_X2(temp1, temp6, temp5, temp2)
SRA_16(temp13, temp14, temp15, temp16, temp3, temp4, temp17, temp8)
"repl.ph %[temp2], 0x4 \n\t"
INSERT_HALF_X2(temp3, temp8, temp17, temp4)
"addq.ph %[temp1], %[temp1], %[temp2] \n\t"
"addq.ph %[temp6], %[temp6], %[temp2] \n\t"
ADD_SUB_HALVES(temp2, temp4, temp1, temp3)
ADD_SUB_HALVES(temp5, temp7, temp6, temp8)
MUL_SHIFT_SUM(temp1, temp3, temp6, temp8, temp9, temp13, temp17, temp18,
temp3, temp13, temp1, temp9, temp9, temp13, temp11, temp15,
temp6, temp17, temp8, temp18)
MUL_SHIFT_SUM(temp6, temp8, temp18, temp17, temp11, temp15, temp12, temp16,
temp8, temp15, temp6, temp11, temp12, temp16, temp10, temp14,
temp18, temp12, temp17, temp16)
INSERT_HALF_X2(temp1, temp3, temp9, temp13)
INSERT_HALF_X2(temp6, temp8, temp11, temp15)
SHIFT_R_SUM_X2(temp9, temp10, temp11, temp12, temp13, temp14, temp15,
temp16, temp2, temp4, temp5, temp7, temp3, temp1, temp8,
temp6)
PACK_2_HALVES_TO_WORD(temp1, temp2, temp3, temp4, temp9, temp12, temp13,
temp16, temp11, temp10, temp15, temp14)
LOAD_WITH_OFFSET_X4(temp10, temp11, temp14, temp15, dst,
0, 0, 0, 0,
0, 1, 2, 3,
BPS)
CONVERT_2_BYTES_TO_HALF(temp5, temp6, temp7, temp8, temp17, temp18, temp10,
temp11, temp10, temp11, temp14, temp15)
STORE_SAT_SUM_X2(temp5, temp6, temp7, temp8, temp17, temp18, temp10, temp11,
temp9, temp12, temp1, temp2, temp13, temp16, temp3, temp4,
dst, 0, 1, 2, 3, BPS)
OUTPUT_EARLY_CLOBBER_REGS_18()
: [dst]"r"(dst), [in]"r"(in), [kC1]"r"(kC1), [kC2]"r"(kC2)
: "memory", "hi", "lo"
);
}
static void TransformTwo(const int16_t* in, uint8_t* dst, int do_two) {
TransformOne(in, dst);
if (do_two) {
TransformOne(in + 16, dst + 4);
}
}
static WEBP_INLINE void FilterLoop26(uint8_t* p,
int hstride, int vstride, int size,
int thresh, int ithresh, int hev_thresh) {
const int thresh2 = 2 * thresh + 1;
int temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8, temp9;
int temp10, temp11, temp12, temp13, temp14, temp15;
__asm__ volatile (
".set push \n\t"
".set noreorder \n\t"
"1: \n\t"
"negu %[temp1], %[hstride] \n\t"
"addiu %[size], %[size], -1 \n\t"
"sll %[temp2], %[hstride], 1 \n\t"
"sll %[temp3], %[temp1], 1 \n\t"
"addu %[temp4], %[temp2], %[hstride] \n\t"
"addu %[temp5], %[temp3], %[temp1] \n\t"
"lbu %[temp7], 0(%[p]) \n\t"
"sll %[temp6], %[temp3], 1 \n\t"
"lbux %[temp8], %[temp5](%[p]) \n\t"
"lbux %[temp9], %[temp3](%[p]) \n\t"
"lbux %[temp10], %[temp1](%[p]) \n\t"
"lbux %[temp11], %[temp6](%[p]) \n\t"
"lbux %[temp12], %[hstride](%[p]) \n\t"
"lbux %[temp13], %[temp2](%[p]) \n\t"
"lbux %[temp14], %[temp4](%[p]) \n\t"
"subu %[temp1], %[temp10], %[temp7] \n\t"
"subu %[temp2], %[temp9], %[temp12] \n\t"
"absq_s.w %[temp3], %[temp1] \n\t"
"absq_s.w %[temp4], %[temp2] \n\t"
"negu %[temp1], %[temp1] \n\t"
"sll %[temp3], %[temp3], 2 \n\t"
"addu %[temp15], %[temp3], %[temp4] \n\t"
"subu %[temp3], %[temp15], %[thresh2] \n\t"
"sll %[temp6], %[temp1], 1 \n\t"
"bgtz %[temp3], 3f \n\t"
" subu %[temp4], %[temp11], %[temp8] \n\t"
"absq_s.w %[temp4], %[temp4] \n\t"
"shll_s.w %[temp2], %[temp2], 24 \n\t"
"subu %[temp4], %[temp4], %[ithresh] \n\t"
"bgtz %[temp4], 3f \n\t"
" subu %[temp3], %[temp8], %[temp9] \n\t"
"absq_s.w %[temp3], %[temp3] \n\t"
"subu %[temp3], %[temp3], %[ithresh] \n\t"
"bgtz %[temp3], 3f \n\t"
" subu %[temp5], %[temp9], %[temp10] \n\t"
"absq_s.w %[temp3], %[temp5] \n\t"
"absq_s.w %[temp5], %[temp5] \n\t"
"subu %[temp3], %[temp3], %[ithresh] \n\t"
"bgtz %[temp3], 3f \n\t"
" subu %[temp3], %[temp14], %[temp13] \n\t"
"absq_s.w %[temp3], %[temp3] \n\t"
"slt %[temp5], %[hev_thresh], %[temp5] \n\t"
"subu %[temp3], %[temp3], %[ithresh] \n\t"
"bgtz %[temp3], 3f \n\t"
" subu %[temp3], %[temp13], %[temp12] \n\t"
"absq_s.w %[temp3], %[temp3] \n\t"
"sra %[temp4], %[temp2], 24 \n\t"
"subu %[temp3], %[temp3], %[ithresh] \n\t"
"bgtz %[temp3], 3f \n\t"
" subu %[temp15], %[temp12], %[temp7] \n\t"
"absq_s.w %[temp3], %[temp15] \n\t"
"absq_s.w %[temp15], %[temp15] \n\t"
"subu %[temp3], %[temp3], %[ithresh] \n\t"
"bgtz %[temp3], 3f \n\t"
" slt %[temp15], %[hev_thresh], %[temp15] \n\t"
"addu %[temp3], %[temp6], %[temp1] \n\t"
"or %[temp2], %[temp5], %[temp15] \n\t"
"addu %[temp5], %[temp4], %[temp3] \n\t"
"beqz %[temp2], 4f \n\t"
" shra_r.w %[temp1], %[temp5], 3 \n\t"
"addiu %[temp2], %[temp5], 3 \n\t"
"sra %[temp2], %[temp2], 3 \n\t"
"shll_s.w %[temp1], %[temp1], 27 \n\t"
"shll_s.w %[temp2], %[temp2], 27 \n\t"
"subu %[temp3], %[p], %[hstride] \n\t"
"sra %[temp1], %[temp1], 27 \n\t"
"sra %[temp2], %[temp2], 27 \n\t"
"subu %[temp1], %[temp7], %[temp1] \n\t"
"addu %[temp2], %[temp10], %[temp2] \n\t"
"lbux %[temp2], %[temp2](%[VP8kclip1]) \n\t"
"lbux %[temp1], %[temp1](%[VP8kclip1]) \n\t"
"sb %[temp2], 0(%[temp3]) \n\t"
"j 3f \n\t"
" sb %[temp1], 0(%[p]) \n\t"
"4: \n\t"
"shll_s.w %[temp5], %[temp5], 24 \n\t"
"subu %[temp14], %[p], %[hstride] \n\t"
"subu %[temp11], %[temp14], %[hstride] \n\t"
"sra %[temp6], %[temp5], 24 \n\t"
"sll %[temp1], %[temp6], 3 \n\t"
"subu %[temp15], %[temp11], %[hstride] \n\t"
"addu %[temp2], %[temp6], %[temp1] \n\t"
"sll %[temp3], %[temp2], 1 \n\t"
"addu %[temp4], %[temp3], %[temp2] \n\t"
"addiu %[temp2], %[temp2], 63 \n\t"
"addiu %[temp3], %[temp3], 63 \n\t"
"addiu %[temp4], %[temp4], 63 \n\t"
"sra %[temp2], %[temp2], 7 \n\t"
"sra %[temp3], %[temp3], 7 \n\t"
"sra %[temp4], %[temp4], 7 \n\t"
"addu %[temp1], %[temp8], %[temp2] \n\t"
"addu %[temp5], %[temp9], %[temp3] \n\t"
"addu %[temp6], %[temp10], %[temp4] \n\t"
"subu %[temp8], %[temp7], %[temp4] \n\t"
"subu %[temp7], %[temp12], %[temp3] \n\t"
"addu %[temp10], %[p], %[hstride] \n\t"
"subu %[temp9], %[temp13], %[temp2] \n\t"
"addu %[temp12], %[temp10], %[hstride] \n\t"
"lbux %[temp2], %[temp1](%[VP8kclip1]) \n\t"
"lbux %[temp3], %[temp5](%[VP8kclip1]) \n\t"
"lbux %[temp4], %[temp6](%[VP8kclip1]) \n\t"
"lbux %[temp5], %[temp8](%[VP8kclip1]) \n\t"
"lbux %[temp6], %[temp7](%[VP8kclip1]) \n\t"
"lbux %[temp8], %[temp9](%[VP8kclip1]) \n\t"
"sb %[temp2], 0(%[temp15]) \n\t"
"sb %[temp3], 0(%[temp11]) \n\t"
"sb %[temp4], 0(%[temp14]) \n\t"
"sb %[temp5], 0(%[p]) \n\t"
"sb %[temp6], 0(%[temp10]) \n\t"
"sb %[temp8], 0(%[temp12]) \n\t"
"3: \n\t"
"bgtz %[size], 1b \n\t"
" addu %[p], %[p], %[vstride] \n\t"
".set pop \n\t"
: [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),[temp3]"=&r"(temp3),
[temp4]"=&r"(temp4), [temp5]"=&r"(temp5), [temp6]"=&r"(temp6),
[temp7]"=&r"(temp7),[temp8]"=&r"(temp8),[temp9]"=&r"(temp9),
[temp10]"=&r"(temp10),[temp11]"=&r"(temp11),[temp12]"=&r"(temp12),
[temp13]"=&r"(temp13),[temp14]"=&r"(temp14),[temp15]"=&r"(temp15),
[size]"+&r"(size), [p]"+&r"(p)
: [hstride]"r"(hstride), [thresh2]"r"(thresh2),
[ithresh]"r"(ithresh),[vstride]"r"(vstride), [hev_thresh]"r"(hev_thresh),
[VP8kclip1]"r"(VP8kclip1)
: "memory"
);
}
static WEBP_INLINE void FilterLoop24(uint8_t* p,
int hstride, int vstride, int size,
int thresh, int ithresh, int hev_thresh) {
int p0, q0, p1, q1, p2, q2, p3, q3;
int step1, step2, temp1, temp2, temp3, temp4;
uint8_t* pTemp0;
uint8_t* pTemp1;
const int thresh2 = 2 * thresh + 1;
__asm__ volatile (
".set push \n\t"
".set noreorder \n\t"
"bltz %[size], 3f \n\t"
" nop \n\t"
"2: \n\t"
"negu %[step1], %[hstride] \n\t"
"lbu %[q0], 0(%[p]) \n\t"
"lbux %[p0], %[step1](%[p]) \n\t"
"subu %[step1], %[step1], %[hstride] \n\t"
"lbux %[q1], %[hstride](%[p]) \n\t"
"subu %[temp1], %[p0], %[q0] \n\t"
"lbux %[p1], %[step1](%[p]) \n\t"
"addu %[step2], %[hstride], %[hstride] \n\t"
"absq_s.w %[temp2], %[temp1] \n\t"
"subu %[temp3], %[p1], %[q1] \n\t"
"absq_s.w %[temp4], %[temp3] \n\t"
"sll %[temp2], %[temp2], 2 \n\t"
"addu %[temp2], %[temp2], %[temp4] \n\t"
"subu %[temp4], %[temp2], %[thresh2] \n\t"
"subu %[step1], %[step1], %[hstride] \n\t"
"bgtz %[temp4], 0f \n\t"
" lbux %[p2], %[step1](%[p]) \n\t"
"subu %[step1], %[step1], %[hstride] \n\t"
"lbux %[q2], %[step2](%[p]) \n\t"
"lbux %[p3], %[step1](%[p]) \n\t"
"subu %[temp4], %[p2], %[p1] \n\t"
"addu %[step2], %[step2], %[hstride] \n\t"
"subu %[temp2], %[p3], %[p2] \n\t"
"absq_s.w %[temp4], %[temp4] \n\t"
"absq_s.w %[temp2], %[temp2] \n\t"
"lbux %[q3], %[step2](%[p]) \n\t"
"subu %[temp4], %[temp4], %[ithresh] \n\t"
"negu %[temp1], %[temp1] \n\t"
"bgtz %[temp4], 0f \n\t"
" subu %[temp2], %[temp2], %[ithresh] \n\t"
"subu %[p3], %[p1], %[p0] \n\t"
"bgtz %[temp2], 0f \n\t"
" absq_s.w %[p3], %[p3] \n\t"
"subu %[temp4], %[q3], %[q2] \n\t"
"subu %[pTemp0], %[p], %[hstride] \n\t"
"absq_s.w %[temp4], %[temp4] \n\t"
"subu %[temp2], %[p3], %[ithresh] \n\t"
"sll %[step1], %[temp1], 1 \n\t"
"bgtz %[temp2], 0f \n\t"
" subu %[temp4], %[temp4], %[ithresh] \n\t"
"subu %[temp2], %[q2], %[q1] \n\t"
"bgtz %[temp4], 0f \n\t"
" absq_s.w %[temp2], %[temp2] \n\t"
"subu %[q3], %[q1], %[q0] \n\t"
"absq_s.w %[q3], %[q3] \n\t"
"subu %[temp2], %[temp2], %[ithresh] \n\t"
"addu %[temp1], %[temp1], %[step1] \n\t"
"bgtz %[temp2], 0f \n\t"
" subu %[temp4], %[q3], %[ithresh] \n\t"
"slt %[p3], %[hev_thresh], %[p3] \n\t"
"bgtz %[temp4], 0f \n\t"
" slt %[q3], %[hev_thresh], %[q3] \n\t"
"or %[q3], %[q3], %[p3] \n\t"
"bgtz %[q3], 1f \n\t"
" shra_r.w %[temp2], %[temp1], 3 \n\t"
"addiu %[temp1], %[temp1], 3 \n\t"
"sra %[temp1], %[temp1], 3 \n\t"
"shll_s.w %[temp2], %[temp2], 27 \n\t"
"shll_s.w %[temp1], %[temp1], 27 \n\t"
"addu %[pTemp1], %[p], %[hstride] \n\t"
"sra %[temp2], %[temp2], 27 \n\t"
"sra %[temp1], %[temp1], 27 \n\t"
"addiu %[step1], %[temp2], 1 \n\t"
"sra %[step1], %[step1], 1 \n\t"
"addu %[p0], %[p0], %[temp1] \n\t"
"addu %[p1], %[p1], %[step1] \n\t"
"subu %[q0], %[q0], %[temp2] \n\t"
"subu %[q1], %[q1], %[step1] \n\t"
"lbux %[temp2], %[p0](%[VP8kclip1]) \n\t"
"lbux %[temp3], %[q0](%[VP8kclip1]) \n\t"
"lbux %[temp4], %[q1](%[VP8kclip1]) \n\t"
"sb %[temp2], 0(%[pTemp0]) \n\t"
"lbux %[temp1], %[p1](%[VP8kclip1]) \n\t"
"subu %[pTemp0], %[pTemp0], %[hstride] \n\t"
"sb %[temp3], 0(%[p]) \n\t"
"sb %[temp4], 0(%[pTemp1]) \n\t"
"j 0f \n\t"
" sb %[temp1], 0(%[pTemp0]) \n\t"
"1: \n\t"
"shll_s.w %[temp3], %[temp3], 24 \n\t"
"sra %[temp3], %[temp3], 24 \n\t"
"addu %[temp1], %[temp1], %[temp3] \n\t"
"shra_r.w %[temp2], %[temp1], 3 \n\t"
"addiu %[temp1], %[temp1], 3 \n\t"
"shll_s.w %[temp2], %[temp2], 27 \n\t"
"sra %[temp1], %[temp1], 3 \n\t"
"shll_s.w %[temp1], %[temp1], 27 \n\t"
"sra %[temp2], %[temp2], 27 \n\t"
"sra %[temp1], %[temp1], 27 \n\t"
"addu %[p0], %[p0], %[temp1] \n\t"
"subu %[q0], %[q0], %[temp2] \n\t"
"lbux %[temp1], %[p0](%[VP8kclip1]) \n\t"
"lbux %[temp2], %[q0](%[VP8kclip1]) \n\t"
"sb %[temp2], 0(%[p]) \n\t"
"sb %[temp1], 0(%[pTemp0]) \n\t"
"0: \n\t"
"subu %[size], %[size], 1 \n\t"
"bgtz %[size], 2b \n\t"
" addu %[p], %[p], %[vstride] \n\t"
"3: \n\t"
".set pop \n\t"
: [p0]"=&r"(p0), [q0]"=&r"(q0), [p1]"=&r"(p1), [q1]"=&r"(q1),
[p2]"=&r"(p2), [q2]"=&r"(q2), [p3]"=&r"(p3), [q3]"=&r"(q3),
[step2]"=&r"(step2), [step1]"=&r"(step1), [temp1]"=&r"(temp1),
[temp2]"=&r"(temp2), [temp3]"=&r"(temp3), [temp4]"=&r"(temp4),
[pTemp0]"=&r"(pTemp0), [pTemp1]"=&r"(pTemp1), [p]"+&r"(p),
[size]"+&r"(size)
: [vstride]"r"(vstride), [ithresh]"r"(ithresh),
[hev_thresh]"r"(hev_thresh), [hstride]"r"(hstride),
[VP8kclip1]"r"(VP8kclip1), [thresh2]"r"(thresh2)
: "memory"
);
}
// on macroblock edges
static void VFilter16(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(p, stride, 1, 16, thresh, ithresh, hev_thresh);
}
static void HFilter16(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(p, 1, stride, 16, thresh, ithresh, hev_thresh);
}
// 8-pixels wide variant, for chroma filtering
static void VFilter8(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(u, stride, 1, 8, thresh, ithresh, hev_thresh);
FilterLoop26(v, stride, 1, 8, thresh, ithresh, hev_thresh);
}
static void HFilter8(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop26(u, 1, stride, 8, thresh, ithresh, hev_thresh);
FilterLoop26(v, 1, stride, 8, thresh, ithresh, hev_thresh);
}
// on three inner edges
static void VFilter16i(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4 * stride;
FilterLoop24(p, stride, 1, 16, thresh, ithresh, hev_thresh);
}
}
static void HFilter16i(uint8_t* p, int stride,
int thresh, int ithresh, int hev_thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4;
FilterLoop24(p, 1, stride, 16, thresh, ithresh, hev_thresh);
}
}
static void VFilter8i(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop24(u + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh);
FilterLoop24(v + 4 * stride, stride, 1, 8, thresh, ithresh, hev_thresh);
}
static void HFilter8i(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_thresh) {
FilterLoop24(u + 4, 1, stride, 8, thresh, ithresh, hev_thresh);
FilterLoop24(v + 4, 1, stride, 8, thresh, ithresh, hev_thresh);
}
#undef MUL
//------------------------------------------------------------------------------
// Simple In-loop filtering (Paragraph 15.2)
static void SimpleVFilter16(uint8_t* p, int stride, int thresh) {
int i;
const int thresh2 = 2 * thresh + 1;
int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8;
uint8_t* p1 = p - stride;
__asm__ volatile (
".set push \n\t"
".set noreorder \n\t"
"li %[i], 16 \n\t"
"0: \n\t"
"negu %[temp4], %[stride] \n\t"
"sll %[temp5], %[temp4], 1 \n\t"
"lbu %[temp2], 0(%[p]) \n\t"
"lbux %[temp3], %[stride](%[p]) \n\t"
"lbux %[temp1], %[temp4](%[p]) \n\t"
"lbux %[temp0], %[temp5](%[p]) \n\t"
"subu %[temp7], %[temp1], %[temp2] \n\t"
"subu %[temp6], %[temp0], %[temp3] \n\t"
"absq_s.w %[temp4], %[temp7] \n\t"
"absq_s.w %[temp5], %[temp6] \n\t"
"sll %[temp4], %[temp4], 2 \n\t"
"subu %[temp5], %[temp5], %[thresh2] \n\t"
"addu %[temp5], %[temp4], %[temp5] \n\t"
"negu %[temp8], %[temp7] \n\t"
"bgtz %[temp5], 1f \n\t"
" addiu %[i], %[i], -1 \n\t"
"sll %[temp4], %[temp8], 1 \n\t"
"shll_s.w %[temp5], %[temp6], 24 \n\t"
"addu %[temp3], %[temp4], %[temp8] \n\t"
"sra %[temp5], %[temp5], 24 \n\t"
"addu %[temp3], %[temp3], %[temp5] \n\t"
"addiu %[temp7], %[temp3], 3 \n\t"
"sra %[temp7], %[temp7], 3 \n\t"
"shra_r.w %[temp8], %[temp3], 3 \n\t"
"shll_s.w %[temp0], %[temp7], 27 \n\t"
"shll_s.w %[temp4], %[temp8], 27 \n\t"
"sra %[temp0], %[temp0], 27 \n\t"
"sra %[temp4], %[temp4], 27 \n\t"
"addu %[temp7], %[temp1], %[temp0] \n\t"
"subu %[temp2], %[temp2], %[temp4] \n\t"
"lbux %[temp3], %[temp7](%[VP8kclip1]) \n\t"
"lbux %[temp4], %[temp2](%[VP8kclip1]) \n\t"
"sb %[temp3], 0(%[p1]) \n\t"
"sb %[temp4], 0(%[p]) \n\t"
"1: \n\t"
"addiu %[p1], %[p1], 1 \n\t"
"bgtz %[i], 0b \n\t"
" addiu %[p], %[p], 1 \n\t"
" .set pop \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8),
[p]"+&r"(p), [i]"=&r"(i), [p1]"+&r"(p1)
: [stride]"r"(stride), [VP8kclip1]"r"(VP8kclip1), [thresh2]"r"(thresh2)
: "memory"
);
}
// TEMP0 = SRC[A + A1 * BPS]
// TEMP1 = SRC[B + B1 * BPS]
// TEMP2 = SRC[C + C1 * BPS]
// TEMP3 = SRC[D + D1 * BPS]
#define LOAD_4_BYTES(TEMP0, TEMP1, TEMP2, TEMP3, \
A, A1, B, B1, C, C1, D, D1, SRC) \
"lbu %[" #TEMP0 "], " #A "+" #A1 "*" XSTR(BPS) "(%[" #SRC "]) \n\t" \
"lbu %[" #TEMP1 "], " #B "+" #B1 "*" XSTR(BPS) "(%[" #SRC "]) \n\t" \
"lbu %[" #TEMP2 "], " #C "+" #C1 "*" XSTR(BPS) "(%[" #SRC "]) \n\t" \
"lbu %[" #TEMP3 "], " #D "+" #D1 "*" XSTR(BPS) "(%[" #SRC "]) \n\t" \
static void SimpleHFilter16(uint8_t* p, int stride, int thresh) {
int i;
const int thresh2 = 2 * thresh + 1;
int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7, temp8;
__asm__ volatile (
".set push \n\t"
".set noreorder \n\t"
"li %[i], 16 \n\t"
"0: \n\t"
LOAD_4_BYTES(temp0, temp1, temp2, temp3, -2, 0, -1, 0, 0, 0, 1, 0, p)
"subu %[temp7], %[temp1], %[temp2] \n\t"
"subu %[temp6], %[temp0], %[temp3] \n\t"
"absq_s.w %[temp4], %[temp7] \n\t"
"absq_s.w %[temp5], %[temp6] \n\t"
"sll %[temp4], %[temp4], 2 \n\t"
"addu %[temp5], %[temp4], %[temp5] \n\t"
"subu %[temp5], %[temp5], %[thresh2] \n\t"
"negu %[temp8], %[temp7] \n\t"
"bgtz %[temp5], 1f \n\t"
" addiu %[i], %[i], -1 \n\t"
"sll %[temp4], %[temp8], 1 \n\t"
"shll_s.w %[temp5], %[temp6], 24 \n\t"
"addu %[temp3], %[temp4], %[temp8] \n\t"
"sra %[temp5], %[temp5], 24 \n\t"
"addu %[temp3], %[temp3], %[temp5] \n\t"
"addiu %[temp7], %[temp3], 3 \n\t"
"sra %[temp7], %[temp7], 3 \n\t"
"shra_r.w %[temp8], %[temp3], 3 \n\t"
"shll_s.w %[temp0], %[temp7], 27 \n\t"
"shll_s.w %[temp4], %[temp8], 27 \n\t"
"sra %[temp0], %[temp0], 27 \n\t"
"sra %[temp4], %[temp4], 27 \n\t"
"addu %[temp7], %[temp1], %[temp0] \n\t"
"subu %[temp2], %[temp2], %[temp4] \n\t"
"lbux %[temp3], %[temp7](%[VP8kclip1]) \n\t"
"lbux %[temp4], %[temp2](%[VP8kclip1]) \n\t"
"sb %[temp3], -1(%[p]) \n\t"
"sb %[temp4], 0(%[p]) \n\t"
"1: \n\t"
"bgtz %[i], 0b \n\t"
" addu %[p], %[p], %[stride] \n\t"
".set pop \n\t"
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8),
[p]"+&r"(p), [i]"=&r"(i)
: [stride]"r"(stride), [VP8kclip1]"r"(VP8kclip1), [thresh2]"r"(thresh2)
: "memory"
);
}
static void SimpleVFilter16i(uint8_t* p, int stride, int thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4 * stride;
SimpleVFilter16(p, stride, thresh);
}
}
static void SimpleHFilter16i(uint8_t* p, int stride, int thresh) {
int k;
for (k = 3; k > 0; --k) {
p += 4;
SimpleHFilter16(p, stride, thresh);
}
}
// DST[A * BPS] = TEMP0
// DST[B + C * BPS] = TEMP1
#define STORE_8_BYTES(TEMP0, TEMP1, A, B, C, DST) \
"usw %[" #TEMP0 "], " #A "*" XSTR(BPS) "(%[" #DST "]) \n\t" \
"usw %[" #TEMP1 "], " #B "+" #C "*" XSTR(BPS) "(%[" #DST "]) \n\t"
static void VE4(uint8_t* dst) { // vertical
const uint8_t* top = dst - BPS;
int temp0, temp1, temp2, temp3, temp4, temp5, temp6;
__asm__ volatile (
"ulw %[temp0], -1(%[top]) \n\t"
"ulh %[temp1], 3(%[top]) \n\t"
"preceu.ph.qbr %[temp2], %[temp0] \n\t"
"preceu.ph.qbl %[temp3], %[temp0] \n\t"
"preceu.ph.qbr %[temp4], %[temp1] \n\t"
"packrl.ph %[temp5], %[temp3], %[temp2] \n\t"
"packrl.ph %[temp6], %[temp4], %[temp3] \n\t"
"shll.ph %[temp5], %[temp5], 1 \n\t"
"shll.ph %[temp6], %[temp6], 1 \n\t"
"addq.ph %[temp2], %[temp5], %[temp2] \n\t"
"addq.ph %[temp6], %[temp6], %[temp4] \n\t"
"addq.ph %[temp2], %[temp2], %[temp3] \n\t"
"addq.ph %[temp6], %[temp6], %[temp3] \n\t"
"shra_r.ph %[temp2], %[temp2], 2 \n\t"
"shra_r.ph %[temp6], %[temp6], 2 \n\t"
"precr.qb.ph %[temp4], %[temp6], %[temp2] \n\t"
STORE_8_BYTES(temp4, temp4, 0, 0, 1, dst)
STORE_8_BYTES(temp4, temp4, 2, 0, 3, dst)
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6)
: [top]"r"(top), [dst]"r"(dst)
: "memory"
);
}
static void DC4(uint8_t* dst) { // DC
int temp0, temp1, temp2, temp3, temp4;
__asm__ volatile (
"ulw %[temp0], -1*" XSTR(BPS) "(%[dst]) \n\t"
LOAD_4_BYTES(temp1, temp2, temp3, temp4, -1, 0, -1, 1, -1, 2, -1, 3, dst)
"ins %[temp1], %[temp2], 8, 8 \n\t"
"ins %[temp1], %[temp3], 16, 8 \n\t"
"ins %[temp1], %[temp4], 24, 8 \n\t"
"raddu.w.qb %[temp0], %[temp0] \n\t"
"raddu.w.qb %[temp1], %[temp1] \n\t"
"addu %[temp0], %[temp0], %[temp1] \n\t"
"shra_r.w %[temp0], %[temp0], 3 \n\t"
"replv.qb %[temp0], %[temp0] \n\t"
STORE_8_BYTES(temp0, temp0, 0, 0, 1, dst)
STORE_8_BYTES(temp0, temp0, 2, 0, 3, dst)
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4)
: [dst]"r"(dst)
: "memory"
);
}
static void RD4(uint8_t* dst) { // Down-right
int temp0, temp1, temp2, temp3, temp4;
int temp5, temp6, temp7, temp8;
__asm__ volatile (
LOAD_4_BYTES(temp0, temp1, temp2, temp3, -1, 0, -1, 1, -1, 2, -1, 3, dst)
"ulw %[temp7], -1-" XSTR(BPS) "(%[dst]) \n\t"
"ins %[temp1], %[temp0], 16, 16 \n\t"
"preceu.ph.qbr %[temp5], %[temp7] \n\t"
"ins %[temp2], %[temp1], 16, 16 \n\t"
"preceu.ph.qbl %[temp4], %[temp7] \n\t"
"ins %[temp3], %[temp2], 16, 16 \n\t"
"shll.ph %[temp2], %[temp2], 1 \n\t"
"addq.ph %[temp3], %[temp3], %[temp1] \n\t"
"packrl.ph %[temp6], %[temp5], %[temp1] \n\t"
"addq.ph %[temp3], %[temp3], %[temp2] \n\t"
"addq.ph %[temp1], %[temp1], %[temp5] \n\t"
"shll.ph %[temp6], %[temp6], 1 \n\t"
"addq.ph %[temp1], %[temp1], %[temp6] \n\t"
"packrl.ph %[temp0], %[temp4], %[temp5] \n\t"
"addq.ph %[temp8], %[temp5], %[temp4] \n\t"
"shra_r.ph %[temp3], %[temp3], 2 \n\t"
"shll.ph %[temp0], %[temp0], 1 \n\t"
"shra_r.ph %[temp1], %[temp1], 2 \n\t"
"addq.ph %[temp8], %[temp0], %[temp8] \n\t"
"lbu %[temp5], 3-" XSTR(BPS) "(%[dst]) \n\t"
"precrq.ph.w %[temp7], %[temp7], %[temp7] \n\t"
"shra_r.ph %[temp8], %[temp8], 2 \n\t"
"ins %[temp7], %[temp5], 0, 8 \n\t"
"precr.qb.ph %[temp2], %[temp1], %[temp3] \n\t"
"raddu.w.qb %[temp4], %[temp7] \n\t"
"precr.qb.ph %[temp6], %[temp8], %[temp1] \n\t"
"shra_r.w %[temp4], %[temp4], 2 \n\t"
STORE_8_BYTES(temp2, temp6, 3, 0, 1, dst)
"prepend %[temp2], %[temp8], 8 \n\t"
"prepend %[temp6], %[temp4], 8 \n\t"
STORE_8_BYTES(temp2, temp6, 2, 0, 0, dst)
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8)
: [dst]"r"(dst)
: "memory"
);
}
// TEMP0 = SRC[A * BPS]
// TEMP1 = SRC[B + C * BPS]
#define LOAD_8_BYTES(TEMP0, TEMP1, A, B, C, SRC) \
"ulw %[" #TEMP0 "], " #A "*" XSTR(BPS) "(%[" #SRC "]) \n\t" \
"ulw %[" #TEMP1 "], " #B "+" #C "*" XSTR(BPS) "(%[" #SRC "]) \n\t"
static void LD4(uint8_t* dst) { // Down-Left
int temp0, temp1, temp2, temp3, temp4;
int temp5, temp6, temp7, temp8, temp9;
__asm__ volatile (
LOAD_8_BYTES(temp0, temp1, -1, 4, -1, dst)
"preceu.ph.qbl %[temp2], %[temp0] \n\t"
"preceu.ph.qbr %[temp3], %[temp0] \n\t"
"preceu.ph.qbr %[temp4], %[temp1] \n\t"
"preceu.ph.qbl %[temp5], %[temp1] \n\t"
"packrl.ph %[temp6], %[temp2], %[temp3] \n\t"
"packrl.ph %[temp7], %[temp4], %[temp2] \n\t"
"packrl.ph %[temp8], %[temp5], %[temp4] \n\t"
"shll.ph %[temp6], %[temp6], 1 \n\t"
"addq.ph %[temp9], %[temp2], %[temp6] \n\t"
"shll.ph %[temp7], %[temp7], 1 \n\t"
"addq.ph %[temp9], %[temp9], %[temp3] \n\t"
"shll.ph %[temp8], %[temp8], 1 \n\t"
"shra_r.ph %[temp9], %[temp9], 2 \n\t"
"addq.ph %[temp3], %[temp4], %[temp7] \n\t"
"addq.ph %[temp0], %[temp5], %[temp8] \n\t"
"addq.ph %[temp3], %[temp3], %[temp2] \n\t"
"addq.ph %[temp0], %[temp0], %[temp4] \n\t"
"shra_r.ph %[temp3], %[temp3], 2 \n\t"
"shra_r.ph %[temp0], %[temp0], 2 \n\t"
"srl %[temp1], %[temp1], 24 \n\t"
"sll %[temp1], %[temp1], 1 \n\t"
"raddu.w.qb %[temp5], %[temp5] \n\t"
"precr.qb.ph %[temp9], %[temp3], %[temp9] \n\t"
"precr.qb.ph %[temp3], %[temp0], %[temp3] \n\t"
"addu %[temp1], %[temp1], %[temp5] \n\t"
"shra_r.w %[temp1], %[temp1], 2 \n\t"
STORE_8_BYTES(temp9, temp3, 0, 0, 2, dst)
"prepend %[temp9], %[temp0], 8 \n\t"
"prepend %[temp3], %[temp1], 8 \n\t"
STORE_8_BYTES(temp9, temp3, 1, 0, 3, dst)
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8),
[temp9]"=&r"(temp9)
: [dst]"r"(dst)
: "memory"
);
}
//------------------------------------------------------------------------------
// Chroma
static void DC8uv(uint8_t* dst) { // DC
int temp0, temp1, temp2, temp3, temp4;
int temp5, temp6, temp7, temp8, temp9;
__asm__ volatile (
LOAD_8_BYTES(temp0, temp1, -1, 4, -1, dst)
LOAD_4_BYTES(temp2, temp3, temp4, temp5, -1, 0, -1, 1, -1, 2, -1, 3, dst)
LOAD_4_BYTES(temp6, temp7, temp8, temp9, -1, 4, -1, 5, -1, 6, -1, 7, dst)
"raddu.w.qb %[temp0], %[temp0] \n\t"
"raddu.w.qb %[temp1], %[temp1] \n\t"
"addu %[temp2], %[temp2], %[temp3] \n\t"
"addu %[temp4], %[temp4], %[temp5] \n\t"
"addu %[temp6], %[temp6], %[temp7] \n\t"
"addu %[temp8], %[temp8], %[temp9] \n\t"
"addu %[temp0], %[temp0], %[temp1] \n\t"
"addu %[temp2], %[temp2], %[temp4] \n\t"
"addu %[temp6], %[temp6], %[temp8] \n\t"
"addu %[temp0], %[temp0], %[temp2] \n\t"
"addu %[temp0], %[temp0], %[temp6] \n\t"
"shra_r.w %[temp0], %[temp0], 4 \n\t"
"replv.qb %[temp0], %[temp0] \n\t"
STORE_8_BYTES(temp0, temp0, 0, 4, 0, dst)
STORE_8_BYTES(temp0, temp0, 1, 4, 1, dst)
STORE_8_BYTES(temp0, temp0, 2, 4, 2, dst)
STORE_8_BYTES(temp0, temp0, 3, 4, 3, dst)
STORE_8_BYTES(temp0, temp0, 4, 4, 4, dst)
STORE_8_BYTES(temp0, temp0, 5, 4, 5, dst)
STORE_8_BYTES(temp0, temp0, 6, 4, 6, dst)
STORE_8_BYTES(temp0, temp0, 7, 4, 7, dst)
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8),
[temp9]"=&r"(temp9)
: [dst]"r"(dst)
: "memory"
);
}
static void DC8uvNoLeft(uint8_t* dst) { // DC with no left samples
int temp0, temp1;
__asm__ volatile (
LOAD_8_BYTES(temp0, temp1, -1, 4, -1, dst)
"raddu.w.qb %[temp0], %[temp0] \n\t"
"raddu.w.qb %[temp1], %[temp1] \n\t"
"addu %[temp0], %[temp0], %[temp1] \n\t"
"shra_r.w %[temp0], %[temp0], 3 \n\t"
"replv.qb %[temp0], %[temp0] \n\t"
STORE_8_BYTES(temp0, temp0, 0, 4, 0, dst)
STORE_8_BYTES(temp0, temp0, 1, 4, 1, dst)
STORE_8_BYTES(temp0, temp0, 2, 4, 2, dst)
STORE_8_BYTES(temp0, temp0, 3, 4, 3, dst)
STORE_8_BYTES(temp0, temp0, 4, 4, 4, dst)
STORE_8_BYTES(temp0, temp0, 5, 4, 5, dst)
STORE_8_BYTES(temp0, temp0, 6, 4, 6, dst)
STORE_8_BYTES(temp0, temp0, 7, 4, 7, dst)
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1)
: [dst]"r"(dst)
: "memory"
);
}
static void DC8uvNoTop(uint8_t* dst) { // DC with no top samples
int temp0, temp1, temp2, temp3, temp4;
int temp5, temp6, temp7, temp8;
__asm__ volatile (
LOAD_4_BYTES(temp2, temp3, temp4, temp5, -1, 0, -1, 1, -1, 2, -1, 3, dst)
LOAD_4_BYTES(temp6, temp7, temp8, temp1, -1, 4, -1, 5, -1, 6, -1, 7, dst)
"addu %[temp2], %[temp2], %[temp3] \n\t"
"addu %[temp4], %[temp4], %[temp5] \n\t"
"addu %[temp6], %[temp6], %[temp7] \n\t"
"addu %[temp8], %[temp8], %[temp1] \n\t"
"addu %[temp2], %[temp2], %[temp4] \n\t"
"addu %[temp6], %[temp6], %[temp8] \n\t"
"addu %[temp0], %[temp6], %[temp2] \n\t"
"shra_r.w %[temp0], %[temp0], 3 \n\t"
"replv.qb %[temp0], %[temp0] \n\t"
STORE_8_BYTES(temp0, temp0, 0, 4, 0, dst)
STORE_8_BYTES(temp0, temp0, 1, 4, 1, dst)
STORE_8_BYTES(temp0, temp0, 2, 4, 2, dst)
STORE_8_BYTES(temp0, temp0, 3, 4, 3, dst)
STORE_8_BYTES(temp0, temp0, 4, 4, 4, dst)
STORE_8_BYTES(temp0, temp0, 5, 4, 5, dst)
STORE_8_BYTES(temp0, temp0, 6, 4, 6, dst)
STORE_8_BYTES(temp0, temp0, 7, 4, 7, dst)
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2),
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5),
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [temp8]"=&r"(temp8)
: [dst]"r"(dst)
: "memory"
);
}
#undef LOAD_8_BYTES
#undef STORE_8_BYTES
#undef LOAD_4_BYTES
#define CLIPPING(SIZE) \
"preceu.ph.qbl %[temp2], %[temp0] \n\t" \
"preceu.ph.qbr %[temp0], %[temp0] \n\t" \
".if " #SIZE " == 8 \n\t" \
"preceu.ph.qbl %[temp3], %[temp1] \n\t" \
"preceu.ph.qbr %[temp1], %[temp1] \n\t" \
".endif \n\t" \
"addu.ph %[temp2], %[temp2], %[dst_1] \n\t" \
"addu.ph %[temp0], %[temp0], %[dst_1] \n\t" \
".if " #SIZE " == 8 \n\t" \
"addu.ph %[temp3], %[temp3], %[dst_1] \n\t" \
"addu.ph %[temp1], %[temp1], %[dst_1] \n\t" \
".endif \n\t" \
"shll_s.ph %[temp2], %[temp2], 7 \n\t" \
"shll_s.ph %[temp0], %[temp0], 7 \n\t" \
".if " #SIZE " == 8 \n\t" \
"shll_s.ph %[temp3], %[temp3], 7 \n\t" \
"shll_s.ph %[temp1], %[temp1], 7 \n\t" \
".endif \n\t" \
"precrqu_s.qb.ph %[temp0], %[temp2], %[temp0] \n\t" \
".if " #SIZE " == 8 \n\t" \
"precrqu_s.qb.ph %[temp1], %[temp3], %[temp1] \n\t" \
".endif \n\t"
#define CLIP_8B_TO_DST(DST, TOP, SIZE) do { \
int dst_1 = ((int)(DST)[-1] << 16) + (DST)[-1]; \
int temp0, temp1, temp2, temp3; \
__asm__ volatile ( \
".if " #SIZE " < 8 \n\t" \
"ulw %[temp0], 0(%[top]) \n\t" \
"subu.ph %[dst_1], %[dst_1], %[top_1] \n\t" \
CLIPPING(4) \
"usw %[temp0], 0(%[dst]) \n\t" \
".else \n\t" \
"ulw %[temp0], 0(%[top]) \n\t" \
"ulw %[temp1], 4(%[top]) \n\t" \
"subu.ph %[dst_1], %[dst_1], %[top_1] \n\t" \
CLIPPING(8) \
"usw %[temp0], 0(%[dst]) \n\t" \
"usw %[temp1], 4(%[dst]) \n\t" \
".if " #SIZE " == 16 \n\t" \
"ulw %[temp0], 8(%[top]) \n\t" \
"ulw %[temp1], 12(%[top]) \n\t" \
CLIPPING(8) \
"usw %[temp0], 8(%[dst]) \n\t" \
"usw %[temp1], 12(%[dst]) \n\t" \
".endif \n\t" \
".endif \n\t" \
: [dst_1]"+&r"(dst_1), [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), \
[temp2]"=&r"(temp2), [temp3]"=&r"(temp3) \
: [top_1]"r"(top_1), [top]"r"((TOP)), [dst]"r"((DST)) \
: "memory" \
); \
} while (0)
#define CLIP_TO_DST(DST, SIZE) do { \
int y; \
const uint8_t* top = (DST) - BPS; \
const int top_1 = ((int)top[-1] << 16) + top[-1]; \
for (y = 0; y < (SIZE); ++y) { \
CLIP_8B_TO_DST((DST), top, (SIZE)); \
(DST) += BPS; \
} \
} while (0)
#define TRUE_MOTION(DST, SIZE) \
static void TrueMotion##SIZE(uint8_t* (DST)) { \
CLIP_TO_DST((DST), (SIZE)); \
}
TRUE_MOTION(dst, 4)
TRUE_MOTION(dst, 8)
TRUE_MOTION(dst, 16)
#undef TRUE_MOTION
#undef CLIP_TO_DST
#undef CLIP_8B_TO_DST
#undef CLIPPING
//------------------------------------------------------------------------------
// Entry point
extern void VP8DspInitMIPSdspR2(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8DspInitMIPSdspR2(void) {
VP8TransformDC = TransformDC;
VP8TransformAC3 = TransformAC3;
VP8Transform = TransformTwo;
VP8VFilter16 = VFilter16;
VP8HFilter16 = HFilter16;
VP8VFilter8 = VFilter8;
VP8HFilter8 = HFilter8;
VP8VFilter16i = VFilter16i;
VP8HFilter16i = HFilter16i;
VP8VFilter8i = VFilter8i;
VP8HFilter8i = HFilter8i;
VP8SimpleVFilter16 = SimpleVFilter16;
VP8SimpleHFilter16 = SimpleHFilter16;
VP8SimpleVFilter16i = SimpleVFilter16i;
VP8SimpleHFilter16i = SimpleHFilter16i;
VP8PredLuma4[0] = DC4;
VP8PredLuma4[1] = TrueMotion4;
VP8PredLuma4[2] = VE4;
VP8PredLuma4[4] = RD4;
VP8PredLuma4[6] = LD4;
VP8PredChroma8[0] = DC8uv;
VP8PredChroma8[1] = TrueMotion8;
VP8PredChroma8[4] = DC8uvNoTop;
VP8PredChroma8[5] = DC8uvNoLeft;
VP8PredLuma16[1] = TrueMotion16;
}
#else // !WEBP_USE_MIPS_DSP_R2
WEBP_DSP_INIT_STUB(VP8DspInitMIPSdspR2)
#endif // WEBP_USE_MIPS_DSP_R2

1020
media/libwebp/src/dsp/dec_msa.c Normal file
Просмотреть файл

Разница между файлами не показана из-за своего большого размера Загрузить разницу

1652
media/libwebp/src/dsp/dec_neon.c Normal file
Просмотреть файл

Разница между файлами не показана из-за своего большого размера Загрузить разницу

1227
media/libwebp/src/dsp/dec_sse2.c Normal file
Просмотреть файл

Разница между файлами не показана из-за своего большого размера Загрузить разницу

46
media/libwebp/src/dsp/dec_sse41.c Normal file
Просмотреть файл

@ -0,0 +1,46 @@
// Copyright 2015 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// SSE4 version of some decoding functions.
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_SSE41)
#include <smmintrin.h>
#include "src/dec/vp8i_dec.h"
#include "src/utils/utils.h"
static void HE16_SSE41(uint8_t* dst) { // horizontal
int j;
const __m128i kShuffle3 = _mm_set1_epi8(3);
for (j = 16; j > 0; --j) {
const __m128i in = _mm_cvtsi32_si128(WebPMemToUint32(dst - 4));
const __m128i values = _mm_shuffle_epi8(in, kShuffle3);
_mm_storeu_si128((__m128i*)dst, values);
dst += BPS;
}
}
//------------------------------------------------------------------------------
// Entry point
extern void VP8DspInitSSE41(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8DspInitSSE41(void) {
VP8PredLuma16[3] = HE16_SSE41;
}
#else // !WEBP_USE_SSE41
WEBP_DSP_INIT_STUB(VP8DspInitSSE41)
#endif // WEBP_USE_SSE41

682
media/libwebp/src/dsp/dsp.h Normal file
Просмотреть файл

@ -0,0 +1,682 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Speed-critical functions.
//
// Author: Skal (pascal.massimino@gmail.com)
#ifndef WEBP_DSP_DSP_H_
#define WEBP_DSP_DSP_H_
#ifdef HAVE_CONFIG_H
#include "src/webp/config.h"
#endif
#include "src/webp/types.h"
#ifdef __cplusplus
extern "C" {
#endif
#define BPS 32 // this is the common stride for enc/dec
//------------------------------------------------------------------------------
// CPU detection
#if defined(__GNUC__)
# define LOCAL_GCC_VERSION ((__GNUC__ << 8) | __GNUC_MINOR__)
# define LOCAL_GCC_PREREQ(maj, min) \
(LOCAL_GCC_VERSION >= (((maj) << 8) | (min)))
#else
# define LOCAL_GCC_VERSION 0
# define LOCAL_GCC_PREREQ(maj, min) 0
#endif
#if defined(__clang__)
# define LOCAL_CLANG_VERSION ((__clang_major__ << 8) | __clang_minor__)
# define LOCAL_CLANG_PREREQ(maj, min) \
(LOCAL_CLANG_VERSION >= (((maj) << 8) | (min)))
#else
# define LOCAL_CLANG_VERSION 0
# define LOCAL_CLANG_PREREQ(maj, min) 0
#endif
#ifndef __has_builtin
# define __has_builtin(x) 0
#endif
// for now, none of the optimizations below are available in emscripten
#if !defined(EMSCRIPTEN)
#if defined(_MSC_VER) && _MSC_VER > 1310 && \
(defined(_M_X64) || defined(_M_IX86))
#define WEBP_MSC_SSE2 // Visual C++ SSE2 targets
#endif
#if defined(_MSC_VER) && _MSC_VER >= 1500 && \
(defined(_M_X64) || defined(_M_IX86))
#define WEBP_MSC_SSE41 // Visual C++ SSE4.1 targets
#endif
// WEBP_HAVE_* are used to indicate the presence of the instruction set in dsp
// files without intrinsics, allowing the corresponding Init() to be called.
// Files containing intrinsics will need to be built targeting the instruction
// set so should succeed on one of the earlier tests.
#if defined(__SSE2__) || defined(WEBP_MSC_SSE2) || defined(WEBP_HAVE_SSE2)
#define WEBP_USE_SSE2
#endif
#if defined(__SSE4_1__) || defined(WEBP_MSC_SSE41) || defined(WEBP_HAVE_SSE41)
#define WEBP_USE_SSE41
#endif
#if defined(__AVX2__) || defined(WEBP_HAVE_AVX2)
#define WEBP_USE_AVX2
#endif
// The intrinsics currently cause compiler errors with arm-nacl-gcc and the
// inline assembly would need to be modified for use with Native Client.
#if (defined(__ARM_NEON__) || \
defined(__aarch64__) || defined(WEBP_HAVE_NEON)) && \
!defined(__native_client__)
#define WEBP_USE_NEON
#endif
#if !defined(WEBP_USE_NEON) && defined(__ANDROID__) && \
defined(__ARM_ARCH_7A__) && defined(HAVE_CPU_FEATURES_H)
#define WEBP_ANDROID_NEON // Android targets that may have NEON
#define WEBP_USE_NEON
#endif
#if defined(_MSC_VER) && _MSC_VER >= 1700 && defined(_M_ARM)
#define WEBP_USE_NEON
#define WEBP_USE_INTRINSICS
#endif
#if defined(__mips__) && !defined(__mips64) && \
defined(__mips_isa_rev) && (__mips_isa_rev >= 1) && (__mips_isa_rev < 6)
#define WEBP_USE_MIPS32
#if (__mips_isa_rev >= 2)
#define WEBP_USE_MIPS32_R2
#if defined(__mips_dspr2) || (defined(__mips_dsp_rev) && __mips_dsp_rev >= 2)
#define WEBP_USE_MIPS_DSP_R2
#endif
#endif
#endif
#if defined(__mips_msa) && defined(__mips_isa_rev) && (__mips_isa_rev >= 5)
#define WEBP_USE_MSA
#endif
#endif /* EMSCRIPTEN */
#ifndef WEBP_DSP_OMIT_C_CODE
#define WEBP_DSP_OMIT_C_CODE 1
#endif
#if (defined(__aarch64__) || defined(__ARM_NEON__)) && WEBP_DSP_OMIT_C_CODE
#define WEBP_NEON_OMIT_C_CODE 1
#else
#define WEBP_NEON_OMIT_C_CODE 0
#endif
#if !(LOCAL_CLANG_PREREQ(3,8) || LOCAL_GCC_PREREQ(4,8) || defined(__aarch64__))
#define WEBP_NEON_WORK_AROUND_GCC 1
#else
#define WEBP_NEON_WORK_AROUND_GCC 0
#endif
// This macro prevents thread_sanitizer from reporting known concurrent writes.
#define WEBP_TSAN_IGNORE_FUNCTION
#if defined(__has_feature)
#if __has_feature(thread_sanitizer)
#undef WEBP_TSAN_IGNORE_FUNCTION
#define WEBP_TSAN_IGNORE_FUNCTION __attribute__((no_sanitize_thread))
#endif
#endif
#if defined(WEBP_USE_THREAD) && !defined(_WIN32)
#include <pthread.h> // NOLINT
#define WEBP_DSP_INIT(func) do { \
static volatile VP8CPUInfo func ## _last_cpuinfo_used = \
(VP8CPUInfo)&func ## _last_cpuinfo_used; \
static pthread_mutex_t func ## _lock = PTHREAD_MUTEX_INITIALIZER; \
if (pthread_mutex_lock(&func ## _lock)) break; \
if (func ## _last_cpuinfo_used != VP8GetCPUInfo) func(); \
func ## _last_cpuinfo_used = VP8GetCPUInfo; \
(void)pthread_mutex_unlock(&func ## _lock); \
} while (0)
#else // !(defined(WEBP_USE_THREAD) && !defined(_WIN32))
#define WEBP_DSP_INIT(func) do { \
static volatile VP8CPUInfo func ## _last_cpuinfo_used = \
(VP8CPUInfo)&func ## _last_cpuinfo_used; \
if (func ## _last_cpuinfo_used == VP8GetCPUInfo) break; \
func(); \
func ## _last_cpuinfo_used = VP8GetCPUInfo; \
} while (0)
#endif // defined(WEBP_USE_THREAD) && !defined(_WIN32)
// Defines an Init + helper function that control multiple initialization of
// function pointers / tables.
/* Usage:
WEBP_DSP_INIT_FUNC(InitFunc) {
...function body
}
*/
#define WEBP_DSP_INIT_FUNC(name) \
static WEBP_TSAN_IGNORE_FUNCTION void name ## _body(void); \
WEBP_TSAN_IGNORE_FUNCTION void name(void) { \
WEBP_DSP_INIT(name ## _body); \
} \
static WEBP_TSAN_IGNORE_FUNCTION void name ## _body(void)
#define WEBP_UBSAN_IGNORE_UNDEF
#define WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
#if defined(__clang__) && defined(__has_attribute)
#if __has_attribute(no_sanitize)
// This macro prevents the undefined behavior sanitizer from reporting
// failures. This is only meant to silence unaligned loads on platforms that
// are known to support them.
#undef WEBP_UBSAN_IGNORE_UNDEF
#define WEBP_UBSAN_IGNORE_UNDEF \
__attribute__((no_sanitize("undefined")))
// This macro prevents the undefined behavior sanitizer from reporting
// failures related to unsigned integer overflows. This is only meant to
// silence cases where this well defined behavior is expected.
#undef WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
#define WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW \
__attribute__((no_sanitize("unsigned-integer-overflow")))
#endif
#endif
// Regularize the definition of WEBP_SWAP_16BIT_CSP (backward compatibility)
#if !defined(WEBP_SWAP_16BIT_CSP)
#define WEBP_SWAP_16BIT_CSP 0
#endif
// some endian fix (e.g.: mips-gcc doesn't define __BIG_ENDIAN__)
#if !defined(WORDS_BIGENDIAN) && \
(defined(__BIG_ENDIAN__) || defined(_M_PPC) || \
(defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)))
#define WORDS_BIGENDIAN
#endif
typedef enum {
kSSE2,
kSSE3,
kSlowSSSE3, // special feature for slow SSSE3 architectures
kSSE4_1,
kAVX,
kAVX2,
kNEON,
kMIPS32,
kMIPSdspR2,
kMSA
} CPUFeature;
// returns true if the CPU supports the feature.
typedef int (*VP8CPUInfo)(CPUFeature feature);
WEBP_EXTERN VP8CPUInfo VP8GetCPUInfo;
//------------------------------------------------------------------------------
// Init stub generator
// Defines an init function stub to ensure each module exposes a symbol,
// avoiding a compiler warning.
#define WEBP_DSP_INIT_STUB(func) \
extern void func(void); \
void func(void) {}
//------------------------------------------------------------------------------
// Encoding
// Transforms
// VP8Idct: Does one of two inverse transforms. If do_two is set, the transforms
// will be done for (ref, in, dst) and (ref + 4, in + 16, dst + 4).
typedef void (*VP8Idct)(const uint8_t* ref, const int16_t* in, uint8_t* dst,
int do_two);
typedef void (*VP8Fdct)(const uint8_t* src, const uint8_t* ref, int16_t* out);
typedef void (*VP8WHT)(const int16_t* in, int16_t* out);
extern VP8Idct VP8ITransform;
extern VP8Fdct VP8FTransform;
extern VP8Fdct VP8FTransform2; // performs two transforms at a time
extern VP8WHT VP8FTransformWHT;
// Predictions
// *dst is the destination block. *top and *left can be NULL.
typedef void (*VP8IntraPreds)(uint8_t *dst, const uint8_t* left,
const uint8_t* top);
typedef void (*VP8Intra4Preds)(uint8_t *dst, const uint8_t* top);
extern VP8Intra4Preds VP8EncPredLuma4;
extern VP8IntraPreds VP8EncPredLuma16;
extern VP8IntraPreds VP8EncPredChroma8;
typedef int (*VP8Metric)(const uint8_t* pix, const uint8_t* ref);
extern VP8Metric VP8SSE16x16, VP8SSE16x8, VP8SSE8x8, VP8SSE4x4;
typedef int (*VP8WMetric)(const uint8_t* pix, const uint8_t* ref,
const uint16_t* const weights);
// The weights for VP8TDisto4x4 and VP8TDisto16x16 contain a row-major
// 4 by 4 symmetric matrix.
extern VP8WMetric VP8TDisto4x4, VP8TDisto16x16;
// Compute the average (DC) of four 4x4 blocks.
// Each sub-4x4 block #i sum is stored in dc[i].
typedef void (*VP8MeanMetric)(const uint8_t* ref, uint32_t dc[4]);
extern VP8MeanMetric VP8Mean16x4;
typedef void (*VP8BlockCopy)(const uint8_t* src, uint8_t* dst);
extern VP8BlockCopy VP8Copy4x4;
extern VP8BlockCopy VP8Copy16x8;
// Quantization
struct VP8Matrix; // forward declaration
typedef int (*VP8QuantizeBlock)(int16_t in[16], int16_t out[16],
const struct VP8Matrix* const mtx);
// Same as VP8QuantizeBlock, but quantizes two consecutive blocks.
typedef int (*VP8Quantize2Blocks)(int16_t in[32], int16_t out[32],
const struct VP8Matrix* const mtx);
extern VP8QuantizeBlock VP8EncQuantizeBlock;
extern VP8Quantize2Blocks VP8EncQuantize2Blocks;
// specific to 2nd transform:
typedef int (*VP8QuantizeBlockWHT)(int16_t in[16], int16_t out[16],
const struct VP8Matrix* const mtx);
extern VP8QuantizeBlockWHT VP8EncQuantizeBlockWHT;
extern const int VP8DspScan[16 + 4 + 4];
// Collect histogram for susceptibility calculation.
#define MAX_COEFF_THRESH 31 // size of histogram used by CollectHistogram.
typedef struct {
// We only need to store max_value and last_non_zero, not the distribution.
int max_value;
int last_non_zero;
} VP8Histogram;
typedef void (*VP8CHisto)(const uint8_t* ref, const uint8_t* pred,
int start_block, int end_block,
VP8Histogram* const histo);
extern VP8CHisto VP8CollectHistogram;
// General-purpose util function to help VP8CollectHistogram().
void VP8SetHistogramData(const int distribution[MAX_COEFF_THRESH + 1],
VP8Histogram* const histo);
// must be called before using any of the above
void VP8EncDspInit(void);
//------------------------------------------------------------------------------
// cost functions (encoding)
extern const uint16_t VP8EntropyCost[256]; // 8bit fixed-point log(p)
// approximate cost per level:
extern const uint16_t VP8LevelFixedCosts[2047 /*MAX_LEVEL*/ + 1];
extern const uint8_t VP8EncBands[16 + 1];
struct VP8Residual;
typedef void (*VP8SetResidualCoeffsFunc)(const int16_t* const coeffs,
struct VP8Residual* const res);
extern VP8SetResidualCoeffsFunc VP8SetResidualCoeffs;
// Cost calculation function.
typedef int (*VP8GetResidualCostFunc)(int ctx0,
const struct VP8Residual* const res);
extern VP8GetResidualCostFunc VP8GetResidualCost;
// must be called before anything using the above
void VP8EncDspCostInit(void);
//------------------------------------------------------------------------------
// SSIM / PSNR utils
// struct for accumulating statistical moments
typedef struct {
uint32_t w; // sum(w_i) : sum of weights
uint32_t xm, ym; // sum(w_i * x_i), sum(w_i * y_i)
uint32_t xxm, xym, yym; // sum(w_i * x_i * x_i), etc.
} VP8DistoStats;
// Compute the final SSIM value
// The non-clipped version assumes stats->w = (2 * VP8_SSIM_KERNEL + 1)^2.
double VP8SSIMFromStats(const VP8DistoStats* const stats);
double VP8SSIMFromStatsClipped(const VP8DistoStats* const stats);
#define VP8_SSIM_KERNEL 3 // total size of the kernel: 2 * VP8_SSIM_KERNEL + 1
typedef double (*VP8SSIMGetClippedFunc)(const uint8_t* src1, int stride1,
const uint8_t* src2, int stride2,
int xo, int yo, // center position
int W, int H); // plane dimension
#if !defined(WEBP_REDUCE_SIZE)
// This version is called with the guarantee that you can load 8 bytes and
// 8 rows at offset src1 and src2
typedef double (*VP8SSIMGetFunc)(const uint8_t* src1, int stride1,
const uint8_t* src2, int stride2);
extern VP8SSIMGetFunc VP8SSIMGet; // unclipped / unchecked
extern VP8SSIMGetClippedFunc VP8SSIMGetClipped; // with clipping
#endif
#if !defined(WEBP_DISABLE_STATS)
typedef uint32_t (*VP8AccumulateSSEFunc)(const uint8_t* src1,
const uint8_t* src2, int len);
extern VP8AccumulateSSEFunc VP8AccumulateSSE;
#endif
// must be called before using any of the above directly
void VP8SSIMDspInit(void);
//------------------------------------------------------------------------------
// Decoding
typedef void (*VP8DecIdct)(const int16_t* coeffs, uint8_t* dst);
// when doing two transforms, coeffs is actually int16_t[2][16].
typedef void (*VP8DecIdct2)(const int16_t* coeffs, uint8_t* dst, int do_two);
extern VP8DecIdct2 VP8Transform;
extern VP8DecIdct VP8TransformAC3;
extern VP8DecIdct VP8TransformUV;
extern VP8DecIdct VP8TransformDC;
extern VP8DecIdct VP8TransformDCUV;
extern VP8WHT VP8TransformWHT;
// *dst is the destination block, with stride BPS. Boundary samples are
// assumed accessible when needed.
typedef void (*VP8PredFunc)(uint8_t* dst);
extern VP8PredFunc VP8PredLuma16[/* NUM_B_DC_MODES */];
extern VP8PredFunc VP8PredChroma8[/* NUM_B_DC_MODES */];
extern VP8PredFunc VP8PredLuma4[/* NUM_BMODES */];
// clipping tables (for filtering)
extern const int8_t* const VP8ksclip1; // clips [-1020, 1020] to [-128, 127]
extern const int8_t* const VP8ksclip2; // clips [-112, 112] to [-16, 15]
extern const uint8_t* const VP8kclip1; // clips [-255,511] to [0,255]
extern const uint8_t* const VP8kabs0; // abs(x) for x in [-255,255]
// must be called first
void VP8InitClipTables(void);
// simple filter (only for luma)
typedef void (*VP8SimpleFilterFunc)(uint8_t* p, int stride, int thresh);
extern VP8SimpleFilterFunc VP8SimpleVFilter16;
extern VP8SimpleFilterFunc VP8SimpleHFilter16;
extern VP8SimpleFilterFunc VP8SimpleVFilter16i; // filter 3 inner edges
extern VP8SimpleFilterFunc VP8SimpleHFilter16i;
// regular filter (on both macroblock edges and inner edges)
typedef void (*VP8LumaFilterFunc)(uint8_t* luma, int stride,
int thresh, int ithresh, int hev_t);
typedef void (*VP8ChromaFilterFunc)(uint8_t* u, uint8_t* v, int stride,
int thresh, int ithresh, int hev_t);
// on outer edge
extern VP8LumaFilterFunc VP8VFilter16;
extern VP8LumaFilterFunc VP8HFilter16;
extern VP8ChromaFilterFunc VP8VFilter8;
extern VP8ChromaFilterFunc VP8HFilter8;
// on inner edge
extern VP8LumaFilterFunc VP8VFilter16i; // filtering 3 inner edges altogether
extern VP8LumaFilterFunc VP8HFilter16i;
extern VP8ChromaFilterFunc VP8VFilter8i; // filtering u and v altogether
extern VP8ChromaFilterFunc VP8HFilter8i;
// Dithering. Combines dithering values (centered around 128) with dst[],
// according to: dst[] = clip(dst[] + (((dither[]-128) + 8) >> 4)
#define VP8_DITHER_DESCALE 4
#define VP8_DITHER_DESCALE_ROUNDER (1 << (VP8_DITHER_DESCALE - 1))
#define VP8_DITHER_AMP_BITS 7
#define VP8_DITHER_AMP_CENTER (1 << VP8_DITHER_AMP_BITS)
extern void (*VP8DitherCombine8x8)(const uint8_t* dither, uint8_t* dst,
int dst_stride);
// must be called before anything using the above
void VP8DspInit(void);
//------------------------------------------------------------------------------
// WebP I/O
#define FANCY_UPSAMPLING // undefined to remove fancy upsampling support
// Convert a pair of y/u/v lines together to the output rgb/a colorspace.
// bottom_y can be NULL if only one line of output is needed (at top/bottom).
typedef void (*WebPUpsampleLinePairFunc)(
const uint8_t* top_y, const uint8_t* bottom_y,
const uint8_t* top_u, const uint8_t* top_v,
const uint8_t* cur_u, const uint8_t* cur_v,
uint8_t* top_dst, uint8_t* bottom_dst, int len);
#ifdef FANCY_UPSAMPLING
// Fancy upsampling functions to convert YUV to RGB(A) modes
extern WebPUpsampleLinePairFunc WebPUpsamplers[/* MODE_LAST */];
#endif // FANCY_UPSAMPLING
// Per-row point-sampling methods.
typedef void (*WebPSamplerRowFunc)(const uint8_t* y,
const uint8_t* u, const uint8_t* v,
uint8_t* dst, int len);
// Generic function to apply 'WebPSamplerRowFunc' to the whole plane:
void WebPSamplerProcessPlane(const uint8_t* y, int y_stride,
const uint8_t* u, const uint8_t* v, int uv_stride,
uint8_t* dst, int dst_stride,
int width, int height, WebPSamplerRowFunc func);
// Sampling functions to convert rows of YUV to RGB(A)
extern WebPSamplerRowFunc WebPSamplers[/* MODE_LAST */];
// General function for converting two lines of ARGB or RGBA.
// 'alpha_is_last' should be true if 0xff000000 is stored in memory as
// as 0x00, 0x00, 0x00, 0xff (little endian).
WebPUpsampleLinePairFunc WebPGetLinePairConverter(int alpha_is_last);
// YUV444->RGB converters
typedef void (*WebPYUV444Converter)(const uint8_t* y,
const uint8_t* u, const uint8_t* v,
uint8_t* dst, int len);
extern WebPYUV444Converter WebPYUV444Converters[/* MODE_LAST */];
// Must be called before using the WebPUpsamplers[] (and for premultiplied
// colorspaces like rgbA, rgbA4444, etc)
void WebPInitUpsamplers(void);
// Must be called before using WebPSamplers[]
void WebPInitSamplers(void);
// Must be called before using WebPYUV444Converters[]
void WebPInitYUV444Converters(void);
//------------------------------------------------------------------------------
// ARGB -> YUV converters
// Convert ARGB samples to luma Y.
extern void (*WebPConvertARGBToY)(const uint32_t* argb, uint8_t* y, int width);
// Convert ARGB samples to U/V with downsampling. do_store should be '1' for
// even lines and '0' for odd ones. 'src_width' is the original width, not
// the U/V one.
extern void (*WebPConvertARGBToUV)(const uint32_t* argb, uint8_t* u, uint8_t* v,
int src_width, int do_store);
// Convert a row of accumulated (four-values) of rgba32 toward U/V
extern void (*WebPConvertRGBA32ToUV)(const uint16_t* rgb,
uint8_t* u, uint8_t* v, int width);
// Convert RGB or BGR to Y
extern void (*WebPConvertRGB24ToY)(const uint8_t* rgb, uint8_t* y, int width);
extern void (*WebPConvertBGR24ToY)(const uint8_t* bgr, uint8_t* y, int width);
// used for plain-C fallback.
extern void WebPConvertARGBToUV_C(const uint32_t* argb, uint8_t* u, uint8_t* v,
int src_width, int do_store);
extern void WebPConvertRGBA32ToUV_C(const uint16_t* rgb,
uint8_t* u, uint8_t* v, int width);
// utilities for accurate RGB->YUV conversion
extern uint64_t (*WebPSharpYUVUpdateY)(const uint16_t* src, const uint16_t* ref,
uint16_t* dst, int len);
extern void (*WebPSharpYUVUpdateRGB)(const int16_t* src, const int16_t* ref,
int16_t* dst, int len);
extern void (*WebPSharpYUVFilterRow)(const int16_t* A, const int16_t* B,
int len,
const uint16_t* best_y, uint16_t* out);
// Must be called before using the above.
void WebPInitConvertARGBToYUV(void);
//------------------------------------------------------------------------------
// Rescaler
struct WebPRescaler;
// Import a row of data and save its contribution in the rescaler.
// 'channel' denotes the channel number to be imported. 'Expand' corresponds to
// the wrk->x_expand case. Otherwise, 'Shrink' is to be used.
typedef void (*WebPRescalerImportRowFunc)(struct WebPRescaler* const wrk,
const uint8_t* src);
extern WebPRescalerImportRowFunc WebPRescalerImportRowExpand;
extern WebPRescalerImportRowFunc WebPRescalerImportRowShrink;
// Export one row (starting at x_out position) from rescaler.
// 'Expand' corresponds to the wrk->y_expand case.
// Otherwise 'Shrink' is to be used
typedef void (*WebPRescalerExportRowFunc)(struct WebPRescaler* const wrk);
extern WebPRescalerExportRowFunc WebPRescalerExportRowExpand;
extern WebPRescalerExportRowFunc WebPRescalerExportRowShrink;
// Plain-C implementation, as fall-back.
extern void WebPRescalerImportRowExpand_C(struct WebPRescaler* const wrk,
const uint8_t* src);
extern void WebPRescalerImportRowShrink_C(struct WebPRescaler* const wrk,
const uint8_t* src);
extern void WebPRescalerExportRowExpand_C(struct WebPRescaler* const wrk);
extern void WebPRescalerExportRowShrink_C(struct WebPRescaler* const wrk);
// Main entry calls:
extern void WebPRescalerImportRow(struct WebPRescaler* const wrk,
const uint8_t* src);
// Export one row (starting at x_out position) from rescaler.
extern void WebPRescalerExportRow(struct WebPRescaler* const wrk);
// Must be called first before using the above.
void WebPRescalerDspInit(void);
//------------------------------------------------------------------------------
// Utilities for processing transparent channel.
// Apply alpha pre-multiply on an rgba, bgra or argb plane of size w * h.
// alpha_first should be 0 for argb, 1 for rgba or bgra (where alpha is last).
extern void (*WebPApplyAlphaMultiply)(
uint8_t* rgba, int alpha_first, int w, int h, int stride);
// Same, buf specifically for RGBA4444 format
extern void (*WebPApplyAlphaMultiply4444)(
uint8_t* rgba4444, int w, int h, int stride);
// Dispatch the values from alpha[] plane to the ARGB destination 'dst'.
// Returns true if alpha[] plane has non-trivial values different from 0xff.
extern int (*WebPDispatchAlpha)(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint8_t* dst, int dst_stride);
// Transfer packed 8b alpha[] values to green channel in dst[], zero'ing the
// A/R/B values. 'dst_stride' is the stride for dst[] in uint32_t units.
extern void (*WebPDispatchAlphaToGreen)(const uint8_t* alpha, int alpha_stride,
int width, int height,
uint32_t* dst, int dst_stride);
// Extract the alpha values from 32b values in argb[] and pack them into alpha[]
// (this is the opposite of WebPDispatchAlpha).
// Returns true if there's only trivial 0xff alpha values.
extern int (*WebPExtractAlpha)(const uint8_t* argb, int argb_stride,
int width, int height,
uint8_t* alpha, int alpha_stride);
// Extract the green values from 32b values in argb[] and pack them into alpha[]
// (this is the opposite of WebPDispatchAlphaToGreen).
extern void (*WebPExtractGreen)(const uint32_t* argb, uint8_t* alpha, int size);
// Pre-Multiply operation transforms x into x * A / 255 (where x=Y,R,G or B).
// Un-Multiply operation transforms x into x * 255 / A.
// Pre-Multiply or Un-Multiply (if 'inverse' is true) argb values in a row.
extern void (*WebPMultARGBRow)(uint32_t* const ptr, int width, int inverse);
// Same a WebPMultARGBRow(), but for several rows.
void WebPMultARGBRows(uint8_t* ptr, int stride, int width, int num_rows,
int inverse);
// Same for a row of single values, with side alpha values.
extern void (*WebPMultRow)(uint8_t* const ptr, const uint8_t* const alpha,
int width, int inverse);
// Same a WebPMultRow(), but for several 'num_rows' rows.
void WebPMultRows(uint8_t* ptr, int stride,
const uint8_t* alpha, int alpha_stride,
int width, int num_rows, int inverse);
// Plain-C versions, used as fallback by some implementations.
void WebPMultRow_C(uint8_t* const ptr, const uint8_t* const alpha,
int width, int inverse);
void WebPMultARGBRow_C(uint32_t* const ptr, int width, int inverse);
#ifdef WORDS_BIGENDIAN
// ARGB packing function: a/r/g/b input is rgba or bgra order.
extern void (*WebPPackARGB)(const uint8_t* a, const uint8_t* r,
const uint8_t* g, const uint8_t* b, int len,
uint32_t* out);
#endif
// RGB packing function. 'step' can be 3 or 4. r/g/b input is rgb or bgr order.
extern void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
int len, int step, uint32_t* out);
// This function returns true if src[i] contains a value different from 0xff.
extern int (*WebPHasAlpha8b)(const uint8_t* src, int length);
// This function returns true if src[4*i] contains a value different from 0xff.
extern int (*WebPHasAlpha32b)(const uint8_t* src, int length);
// To be called first before using the above.
void WebPInitAlphaProcessing(void);
//------------------------------------------------------------------------------
// Filter functions
typedef enum { // Filter types.
WEBP_FILTER_NONE = 0,
WEBP_FILTER_HORIZONTAL,
WEBP_FILTER_VERTICAL,
WEBP_FILTER_GRADIENT,
WEBP_FILTER_LAST = WEBP_FILTER_GRADIENT + 1, // end marker
WEBP_FILTER_BEST, // meta-types
WEBP_FILTER_FAST
} WEBP_FILTER_TYPE;
typedef void (*WebPFilterFunc)(const uint8_t* in, int width, int height,
int stride, uint8_t* out);
// In-place un-filtering.
// Warning! 'prev_line' pointer can be equal to 'cur_line' or 'preds'.
typedef void (*WebPUnfilterFunc)(const uint8_t* prev_line, const uint8_t* preds,
uint8_t* cur_line, int width);
// Filter the given data using the given predictor.
// 'in' corresponds to a 2-dimensional pixel array of size (stride * height)
// in raster order.
// 'stride' is number of bytes per scan line (with possible padding).
// 'out' should be pre-allocated.
extern WebPFilterFunc WebPFilters[WEBP_FILTER_LAST];
// In-place reconstruct the original data from the given filtered data.
// The reconstruction will be done for 'num_rows' rows starting from 'row'
// (assuming rows upto 'row - 1' are already reconstructed).
extern WebPUnfilterFunc WebPUnfilters[WEBP_FILTER_LAST];
// To be called first before using the above.
void VP8FiltersInit(void);
#ifdef __cplusplus
} // extern "C"
#endif
#endif /* WEBP_DSP_DSP_H_ */

287
media/libwebp/src/dsp/filters.c Normal file
Просмотреть файл

@ -0,0 +1,287 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Spatial prediction using various filters
//
// Author: Urvang (urvang@google.com)
#include "src/dsp/dsp.h"
#include <assert.h>
#include <stdlib.h>
#include <string.h>
//------------------------------------------------------------------------------
// Helpful macro.
# define SANITY_CHECK(in, out) \
assert((in) != NULL); \
assert((out) != NULL); \
assert(width > 0); \
assert(height > 0); \
assert(stride >= width); \
assert(row >= 0 && num_rows > 0 && row + num_rows <= height); \
(void)height; // Silence unused warning.
#if !WEBP_NEON_OMIT_C_CODE
static WEBP_INLINE void PredictLine_C(const uint8_t* src, const uint8_t* pred,
uint8_t* dst, int length, int inverse) {
int i;
if (inverse) {
for (i = 0; i < length; ++i) dst[i] = src[i] + pred[i];
} else {
for (i = 0; i < length; ++i) dst[i] = src[i] - pred[i];
}
}
//------------------------------------------------------------------------------
// Horizontal filter.
static WEBP_INLINE void DoHorizontalFilter_C(const uint8_t* in,
int width, int height, int stride,
int row, int num_rows,
int inverse, uint8_t* out) {
const uint8_t* preds;
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
preds = inverse ? out : in;
if (row == 0) {
// Leftmost pixel is the same as input for topmost scanline.
out[0] = in[0];
PredictLine_C(in + 1, preds, out + 1, width - 1, inverse);
row = 1;
preds += stride;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
// Leftmost pixel is predicted from above.
PredictLine_C(in, preds - stride, out, 1, inverse);
PredictLine_C(in + 1, preds, out + 1, width - 1, inverse);
++row;
preds += stride;
in += stride;
out += stride;
}
}
//------------------------------------------------------------------------------
// Vertical filter.
static WEBP_INLINE void DoVerticalFilter_C(const uint8_t* in,
int width, int height, int stride,
int row, int num_rows,
int inverse, uint8_t* out) {
const uint8_t* preds;
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
preds = inverse ? out : in;
if (row == 0) {
// Very first top-left pixel is copied.
out[0] = in[0];
// Rest of top scan-line is left-predicted.
PredictLine_C(in + 1, preds, out + 1, width - 1, inverse);
row = 1;
in += stride;
out += stride;
} else {
// We are starting from in-between. Make sure 'preds' points to prev row.
preds -= stride;
}
// Filter line-by-line.
while (row < last_row) {
PredictLine_C(in, preds, out, width, inverse);
++row;
preds += stride;
in += stride;
out += stride;
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
//------------------------------------------------------------------------------
// Gradient filter.
static WEBP_INLINE int GradientPredictor_C(uint8_t a, uint8_t b, uint8_t c) {
const int g = a + b - c;
return ((g & ~0xff) == 0) ? g : (g < 0) ? 0 : 255; // clip to 8bit
}
#if !WEBP_NEON_OMIT_C_CODE
static WEBP_INLINE void DoGradientFilter_C(const uint8_t* in,
int width, int height, int stride,
int row, int num_rows,
int inverse, uint8_t* out) {
const uint8_t* preds;
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
preds = inverse ? out : in;
// left prediction for top scan-line
if (row == 0) {
out[0] = in[0];
PredictLine_C(in + 1, preds, out + 1, width - 1, inverse);
row = 1;
preds += stride;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
int w;
// leftmost pixel: predict from above.
PredictLine_C(in, preds - stride, out, 1, inverse);
for (w = 1; w < width; ++w) {
const int pred = GradientPredictor_C(preds[w - 1],
preds[w - stride],
preds[w - stride - 1]);
out[w] = in[w] + (inverse ? pred : -pred);
}
++row;
preds += stride;
in += stride;
out += stride;
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
#undef SANITY_CHECK
//------------------------------------------------------------------------------
#if !WEBP_NEON_OMIT_C_CODE
static void HorizontalFilter_C(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoHorizontalFilter_C(data, width, height, stride, 0, height, 0,
filtered_data);
}
static void VerticalFilter_C(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoVerticalFilter_C(data, width, height, stride, 0, height, 0, filtered_data);
}
static void GradientFilter_C(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoGradientFilter_C(data, width, height, stride, 0, height, 0, filtered_data);
}
#endif // !WEBP_NEON_OMIT_C_CODE
//------------------------------------------------------------------------------
static void HorizontalUnfilter_C(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
uint8_t pred = (prev == NULL) ? 0 : prev[0];
int i;
for (i = 0; i < width; ++i) {
out[i] = pred + in[i];
pred = out[i];
}
}
#if !WEBP_NEON_OMIT_C_CODE
static void VerticalUnfilter_C(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_C(NULL, in, out, width);
} else {
int i;
for (i = 0; i < width; ++i) out[i] = prev[i] + in[i];
}
}
#endif // !WEBP_NEON_OMIT_C_CODE
static void GradientUnfilter_C(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_C(NULL, in, out, width);
} else {
uint8_t top = prev[0], top_left = top, left = top;
int i;
for (i = 0; i < width; ++i) {
top = prev[i]; // need to read this first, in case prev==out
left = in[i] + GradientPredictor_C(left, top, top_left);
top_left = top;
out[i] = left;
}
}
}
//------------------------------------------------------------------------------
// Init function
WebPFilterFunc WebPFilters[WEBP_FILTER_LAST];
WebPUnfilterFunc WebPUnfilters[WEBP_FILTER_LAST];
extern void VP8FiltersInitMIPSdspR2(void);
extern void VP8FiltersInitMSA(void);
extern void VP8FiltersInitNEON(void);
extern void VP8FiltersInitSSE2(void);
WEBP_DSP_INIT_FUNC(VP8FiltersInit) {
WebPUnfilters[WEBP_FILTER_NONE] = NULL;
#if !WEBP_NEON_OMIT_C_CODE
WebPUnfilters[WEBP_FILTER_HORIZONTAL] = HorizontalUnfilter_C;
WebPUnfilters[WEBP_FILTER_VERTICAL] = VerticalUnfilter_C;
#endif
WebPUnfilters[WEBP_FILTER_GRADIENT] = GradientUnfilter_C;
WebPFilters[WEBP_FILTER_NONE] = NULL;
#if !WEBP_NEON_OMIT_C_CODE
WebPFilters[WEBP_FILTER_HORIZONTAL] = HorizontalFilter_C;
WebPFilters[WEBP_FILTER_VERTICAL] = VerticalFilter_C;
WebPFilters[WEBP_FILTER_GRADIENT] = GradientFilter_C;
#endif
if (VP8GetCPUInfo != NULL) {
#if defined(WEBP_USE_SSE2)
if (VP8GetCPUInfo(kSSE2)) {
VP8FiltersInitSSE2();
}
#endif
#if defined(WEBP_USE_MIPS_DSP_R2)
if (VP8GetCPUInfo(kMIPSdspR2)) {
VP8FiltersInitMIPSdspR2();
}
#endif
#if defined(WEBP_USE_MSA)
if (VP8GetCPUInfo(kMSA)) {
VP8FiltersInitMSA();
}
#endif
}
#if defined(WEBP_USE_NEON)
if (WEBP_NEON_OMIT_C_CODE ||
(VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
VP8FiltersInitNEON();
}
#endif
assert(WebPUnfilters[WEBP_FILTER_HORIZONTAL] != NULL);
assert(WebPUnfilters[WEBP_FILTER_VERTICAL] != NULL);
assert(WebPUnfilters[WEBP_FILTER_GRADIENT] != NULL);
assert(WebPFilters[WEBP_FILTER_HORIZONTAL] != NULL);
assert(WebPFilters[WEBP_FILTER_VERTICAL] != NULL);
assert(WebPFilters[WEBP_FILTER_GRADIENT] != NULL);
}

402
media/libwebp/src/dsp/filters_mips_dsp_r2.c Normal file
Просмотреть файл

@ -0,0 +1,402 @@
// Copyright 2014 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Spatial prediction using various filters
//
// Author(s): Branimir Vasic (branimir.vasic@imgtec.com)
// Djordje Pesut (djordje.pesut@imgtec.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_MIPS_DSP_R2)
#include "src/dsp/dsp.h"
#include <assert.h>
#include <stdlib.h>
#include <string.h>
//------------------------------------------------------------------------------
// Helpful macro.
# define SANITY_CHECK(in, out) \
assert(in != NULL); \
assert(out != NULL); \
assert(width > 0); \
assert(height > 0); \
assert(stride >= width); \
assert(row >= 0 && num_rows > 0 && row + num_rows <= height); \
(void)height; // Silence unused warning.
#define DO_PREDICT_LINE(SRC, DST, LENGTH, INVERSE) do { \
const uint8_t* psrc = (uint8_t*)(SRC); \
uint8_t* pdst = (uint8_t*)(DST); \
const int ilength = (int)(LENGTH); \
int temp0, temp1, temp2, temp3, temp4, temp5, temp6; \
__asm__ volatile ( \
".set push \n\t" \
".set noreorder \n\t" \
"srl %[temp0], %[length], 2 \n\t" \
"beqz %[temp0], 4f \n\t" \
" andi %[temp6], %[length], 3 \n\t" \
".if " #INVERSE " \n\t" \
"1: \n\t" \
"lbu %[temp1], -1(%[dst]) \n\t" \
"lbu %[temp2], 0(%[src]) \n\t" \
"lbu %[temp3], 1(%[src]) \n\t" \
"lbu %[temp4], 2(%[src]) \n\t" \
"lbu %[temp5], 3(%[src]) \n\t" \
"addu %[temp1], %[temp1], %[temp2] \n\t" \
"addu %[temp2], %[temp1], %[temp3] \n\t" \
"addu %[temp3], %[temp2], %[temp4] \n\t" \
"addu %[temp4], %[temp3], %[temp5] \n\t" \
"sb %[temp1], 0(%[dst]) \n\t" \
"sb %[temp2], 1(%[dst]) \n\t" \
"sb %[temp3], 2(%[dst]) \n\t" \
"sb %[temp4], 3(%[dst]) \n\t" \
"addiu %[src], %[src], 4 \n\t" \
"addiu %[temp0], %[temp0], -1 \n\t" \
"bnez %[temp0], 1b \n\t" \
" addiu %[dst], %[dst], 4 \n\t" \
".else \n\t" \
"1: \n\t" \
"ulw %[temp1], -1(%[src]) \n\t" \
"ulw %[temp2], 0(%[src]) \n\t" \
"addiu %[src], %[src], 4 \n\t" \
"addiu %[temp0], %[temp0], -1 \n\t" \
"subu.qb %[temp3], %[temp2], %[temp1] \n\t" \
"usw %[temp3], 0(%[dst]) \n\t" \
"bnez %[temp0], 1b \n\t" \
" addiu %[dst], %[dst], 4 \n\t" \
".endif \n\t" \
"4: \n\t" \
"beqz %[temp6], 3f \n\t" \
" nop \n\t" \
"2: \n\t" \
"lbu %[temp2], 0(%[src]) \n\t" \
".if " #INVERSE " \n\t" \
"lbu %[temp1], -1(%[dst]) \n\t" \
"addu %[temp3], %[temp1], %[temp2] \n\t" \
".else \n\t" \
"lbu %[temp1], -1(%[src]) \n\t" \
"subu %[temp3], %[temp1], %[temp2] \n\t" \
".endif \n\t" \
"addiu %[src], %[src], 1 \n\t" \
"sb %[temp3], 0(%[dst]) \n\t" \
"addiu %[temp6], %[temp6], -1 \n\t" \
"bnez %[temp6], 2b \n\t" \
" addiu %[dst], %[dst], 1 \n\t" \
"3: \n\t" \
".set pop \n\t" \
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), \
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), \
[temp6]"=&r"(temp6), [dst]"+&r"(pdst), [src]"+&r"(psrc) \
: [length]"r"(ilength) \
: "memory" \
); \
} while (0)
static WEBP_INLINE void PredictLine_MIPSdspR2(const uint8_t* src, uint8_t* dst,
int length) {
DO_PREDICT_LINE(src, dst, length, 0);
}
#define DO_PREDICT_LINE_VERTICAL(SRC, PRED, DST, LENGTH, INVERSE) do { \
const uint8_t* psrc = (uint8_t*)(SRC); \
const uint8_t* ppred = (uint8_t*)(PRED); \
uint8_t* pdst = (uint8_t*)(DST); \
const int ilength = (int)(LENGTH); \
int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; \
__asm__ volatile ( \
".set push \n\t" \
".set noreorder \n\t" \
"srl %[temp0], %[length], 0x3 \n\t" \
"beqz %[temp0], 4f \n\t" \
" andi %[temp7], %[length], 0x7 \n\t" \
"1: \n\t" \
"ulw %[temp1], 0(%[src]) \n\t" \
"ulw %[temp2], 0(%[pred]) \n\t" \
"ulw %[temp3], 4(%[src]) \n\t" \
"ulw %[temp4], 4(%[pred]) \n\t" \
"addiu %[src], %[src], 8 \n\t" \
".if " #INVERSE " \n\t" \
"addu.qb %[temp5], %[temp1], %[temp2] \n\t" \
"addu.qb %[temp6], %[temp3], %[temp4] \n\t" \
".else \n\t" \
"subu.qb %[temp5], %[temp1], %[temp2] \n\t" \
"subu.qb %[temp6], %[temp3], %[temp4] \n\t" \
".endif \n\t" \
"addiu %[pred], %[pred], 8 \n\t" \
"usw %[temp5], 0(%[dst]) \n\t" \
"usw %[temp6], 4(%[dst]) \n\t" \
"addiu %[temp0], %[temp0], -1 \n\t" \
"bnez %[temp0], 1b \n\t" \
" addiu %[dst], %[dst], 8 \n\t" \
"4: \n\t" \
"beqz %[temp7], 3f \n\t" \
" nop \n\t" \
"2: \n\t" \
"lbu %[temp1], 0(%[src]) \n\t" \
"lbu %[temp2], 0(%[pred]) \n\t" \
"addiu %[src], %[src], 1 \n\t" \
"addiu %[pred], %[pred], 1 \n\t" \
".if " #INVERSE " \n\t" \
"addu %[temp3], %[temp1], %[temp2] \n\t" \
".else \n\t" \
"subu %[temp3], %[temp1], %[temp2] \n\t" \
".endif \n\t" \
"sb %[temp3], 0(%[dst]) \n\t" \
"addiu %[temp7], %[temp7], -1 \n\t" \
"bnez %[temp7], 2b \n\t" \
" addiu %[dst], %[dst], 1 \n\t" \
"3: \n\t" \
".set pop \n\t" \
: [temp0]"=&r"(temp0), [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), \
[temp3]"=&r"(temp3), [temp4]"=&r"(temp4), [temp5]"=&r"(temp5), \
[temp6]"=&r"(temp6), [temp7]"=&r"(temp7), [pred]"+&r"(ppred), \
[dst]"+&r"(pdst), [src]"+&r"(psrc) \
: [length]"r"(ilength) \
: "memory" \
); \
} while (0)
#define PREDICT_LINE_ONE_PASS(SRC, PRED, DST) do { \
int temp1, temp2, temp3; \
__asm__ volatile ( \
"lbu %[temp1], 0(%[src]) \n\t" \
"lbu %[temp2], 0(%[pred]) \n\t" \
"subu %[temp3], %[temp1], %[temp2] \n\t" \
"sb %[temp3], 0(%[dst]) \n\t" \
: [temp1]"=&r"(temp1), [temp2]"=&r"(temp2), [temp3]"=&r"(temp3) \
: [pred]"r"((PRED)), [dst]"r"((DST)), [src]"r"((SRC)) \
: "memory" \
); \
} while (0)
//------------------------------------------------------------------------------
// Horizontal filter.
#define FILTER_LINE_BY_LINE do { \
while (row < last_row) { \
PREDICT_LINE_ONE_PASS(in, preds - stride, out); \
DO_PREDICT_LINE(in + 1, out + 1, width - 1, 0); \
++row; \
preds += stride; \
in += stride; \
out += stride; \
} \
} while (0)
static WEBP_INLINE void DoHorizontalFilter_MIPSdspR2(const uint8_t* in,
int width, int height,
int stride,
int row, int num_rows,
uint8_t* out) {
const uint8_t* preds;
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
preds = in;
if (row == 0) {
// Leftmost pixel is the same as input for topmost scanline.
out[0] = in[0];
PredictLine_MIPSdspR2(in + 1, out + 1, width - 1);
row = 1;
preds += stride;
in += stride;
out += stride;
}
// Filter line-by-line.
FILTER_LINE_BY_LINE;
}
#undef FILTER_LINE_BY_LINE
static void HorizontalFilter_MIPSdspR2(const uint8_t* data,
int width, int height,
int stride, uint8_t* filtered_data) {
DoHorizontalFilter_MIPSdspR2(data, width, height, stride, 0, height,
filtered_data);
}
//------------------------------------------------------------------------------
// Vertical filter.
#define FILTER_LINE_BY_LINE do { \
while (row < last_row) { \
DO_PREDICT_LINE_VERTICAL(in, preds, out, width, 0); \
++row; \
preds += stride; \
in += stride; \
out += stride; \
} \
} while (0)
static WEBP_INLINE void DoVerticalFilter_MIPSdspR2(const uint8_t* in,
int width, int height,
int stride,
int row, int num_rows,
uint8_t* out) {
const uint8_t* preds;
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
preds = in;
if (row == 0) {
// Very first top-left pixel is copied.
out[0] = in[0];
// Rest of top scan-line is left-predicted.
PredictLine_MIPSdspR2(in + 1, out + 1, width - 1);
row = 1;
in += stride;
out += stride;
} else {
// We are starting from in-between. Make sure 'preds' points to prev row.
preds -= stride;
}
// Filter line-by-line.
FILTER_LINE_BY_LINE;
}
#undef FILTER_LINE_BY_LINE
static void VerticalFilter_MIPSdspR2(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoVerticalFilter_MIPSdspR2(data, width, height, stride, 0, height,
filtered_data);
}
//------------------------------------------------------------------------------
// Gradient filter.
static int GradientPredictor_MIPSdspR2(uint8_t a, uint8_t b, uint8_t c) {
int temp0;
__asm__ volatile (
"addu %[temp0], %[a], %[b] \n\t"
"subu %[temp0], %[temp0], %[c] \n\t"
"shll_s.w %[temp0], %[temp0], 23 \n\t"
"precrqu_s.qb.ph %[temp0], %[temp0], $zero \n\t"
"srl %[temp0], %[temp0], 24 \n\t"
: [temp0]"=&r"(temp0)
: [a]"r"(a),[b]"r"(b),[c]"r"(c)
);
return temp0;
}
#define FILTER_LINE_BY_LINE(PREDS, OPERATION) do { \
while (row < last_row) { \
int w; \
PREDICT_LINE_ONE_PASS(in, PREDS - stride, out); \
for (w = 1; w < width; ++w) { \
const int pred = GradientPredictor_MIPSdspR2(PREDS[w - 1], \
PREDS[w - stride], \
PREDS[w - stride - 1]); \
out[w] = in[w] OPERATION pred; \
} \
++row; \
in += stride; \
out += stride; \
} \
} while (0)
static void DoGradientFilter_MIPSdspR2(const uint8_t* in,
int width, int height, int stride,
int row, int num_rows, uint8_t* out) {
const uint8_t* preds;
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
preds = in;
// left prediction for top scan-line
if (row == 0) {
out[0] = in[0];
PredictLine_MIPSdspR2(in + 1, out + 1, width - 1);
row = 1;
preds += stride;
in += stride;
out += stride;
}
// Filter line-by-line.
FILTER_LINE_BY_LINE(in, -);
}
#undef FILTER_LINE_BY_LINE
static void GradientFilter_MIPSdspR2(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoGradientFilter_MIPSdspR2(data, width, height, stride, 0, height,
filtered_data);
}
//------------------------------------------------------------------------------
static void HorizontalUnfilter_MIPSdspR2(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
out[0] = in[0] + (prev == NULL ? 0 : prev[0]);
DO_PREDICT_LINE(in + 1, out + 1, width - 1, 1);
}
static void VerticalUnfilter_MIPSdspR2(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_MIPSdspR2(NULL, in, out, width);
} else {
DO_PREDICT_LINE_VERTICAL(in, prev, out, width, 1);
}
}
static void GradientUnfilter_MIPSdspR2(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_MIPSdspR2(NULL, in, out, width);
} else {
uint8_t top = prev[0], top_left = top, left = top;
int i;
for (i = 0; i < width; ++i) {
top = prev[i]; // need to read this first, in case prev==dst
left = in[i] + GradientPredictor_MIPSdspR2(left, top, top_left);
top_left = top;
out[i] = left;
}
}
}
#undef DO_PREDICT_LINE_VERTICAL
#undef PREDICT_LINE_ONE_PASS
#undef DO_PREDICT_LINE
#undef SANITY_CHECK
//------------------------------------------------------------------------------
// Entry point
extern void VP8FiltersInitMIPSdspR2(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8FiltersInitMIPSdspR2(void) {
WebPUnfilters[WEBP_FILTER_HORIZONTAL] = HorizontalUnfilter_MIPSdspR2;
WebPUnfilters[WEBP_FILTER_VERTICAL] = VerticalUnfilter_MIPSdspR2;
WebPUnfilters[WEBP_FILTER_GRADIENT] = GradientUnfilter_MIPSdspR2;
WebPFilters[WEBP_FILTER_HORIZONTAL] = HorizontalFilter_MIPSdspR2;
WebPFilters[WEBP_FILTER_VERTICAL] = VerticalFilter_MIPSdspR2;
WebPFilters[WEBP_FILTER_GRADIENT] = GradientFilter_MIPSdspR2;
}
#else // !WEBP_USE_MIPS_DSP_R2
WEBP_DSP_INIT_STUB(VP8FiltersInitMIPSdspR2)
#endif // WEBP_USE_MIPS_DSP_R2

202
media/libwebp/src/dsp/filters_msa.c Normal file
Просмотреть файл

@ -0,0 +1,202 @@
// Copyright 2016 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// MSA variant of alpha filters
//
// Author: Prashant Patil (prashant.patil@imgtec.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_MSA)
#include "src/dsp/msa_macro.h"
#include <assert.h>
static WEBP_INLINE void PredictLineInverse0(const uint8_t* src,
const uint8_t* pred,
uint8_t* dst, int length) {
v16u8 src0, pred0, dst0;
assert(length >= 0);
while (length >= 32) {
v16u8 src1, pred1, dst1;
LD_UB2(src, 16, src0, src1);
LD_UB2(pred, 16, pred0, pred1);
SUB2(src0, pred0, src1, pred1, dst0, dst1);
ST_UB2(dst0, dst1, dst, 16);
src += 32;
pred += 32;
dst += 32;
length -= 32;
}
if (length > 0) {
int i;
if (length >= 16) {
src0 = LD_UB(src);
pred0 = LD_UB(pred);
dst0 = src0 - pred0;
ST_UB(dst0, dst);
src += 16;
pred += 16;
dst += 16;
length -= 16;
}
for (i = 0; i < length; i++) {
dst[i] = src[i] - pred[i];
}
}
}
//------------------------------------------------------------------------------
// Helpful macro.
#define SANITY_CHECK(in, out) \
assert(in != NULL); \
assert(out != NULL); \
assert(width > 0); \
assert(height > 0); \
assert(stride >= width);
//------------------------------------------------------------------------------
// Horrizontal filter
static void HorizontalFilter_MSA(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
const uint8_t* preds = data;
const uint8_t* in = data;
uint8_t* out = filtered_data;
int row = 1;
SANITY_CHECK(in, out);
// Leftmost pixel is the same as input for topmost scanline.
out[0] = in[0];
PredictLineInverse0(in + 1, preds, out + 1, width - 1);
preds += stride;
in += stride;
out += stride;
// Filter line-by-line.
while (row < height) {
// Leftmost pixel is predicted from above.
PredictLineInverse0(in, preds - stride, out, 1);
PredictLineInverse0(in + 1, preds, out + 1, width - 1);
++row;
preds += stride;
in += stride;
out += stride;
}
}
//------------------------------------------------------------------------------
// Gradient filter
static WEBP_INLINE void PredictLineGradient(const uint8_t* pinput,
const uint8_t* ppred,
uint8_t* poutput, int stride,
int size) {
int w;
const v16i8 zero = { 0 };
while (size >= 16) {
v16u8 pred0, dst0;
v8i16 a0, a1, b0, b1, c0, c1;
const v16u8 tmp0 = LD_UB(ppred - 1);
const v16u8 tmp1 = LD_UB(ppred - stride);
const v16u8 tmp2 = LD_UB(ppred - stride - 1);
const v16u8 src0 = LD_UB(pinput);
ILVRL_B2_SH(zero, tmp0, a0, a1);
ILVRL_B2_SH(zero, tmp1, b0, b1);
ILVRL_B2_SH(zero, tmp2, c0, c1);
ADD2(a0, b0, a1, b1, a0, a1);
SUB2(a0, c0, a1, c1, a0, a1);
CLIP_SH2_0_255(a0, a1);
pred0 = (v16u8)__msa_pckev_b((v16i8)a1, (v16i8)a0);
dst0 = src0 - pred0;
ST_UB(dst0, poutput);
ppred += 16;
pinput += 16;
poutput += 16;
size -= 16;
}
for (w = 0; w < size; ++w) {
const int pred = ppred[w - 1] + ppred[w - stride] - ppred[w - stride - 1];
poutput[w] = pinput[w] - (pred < 0 ? 0 : pred > 255 ? 255 : pred);
}
}
static void GradientFilter_MSA(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
const uint8_t* in = data;
const uint8_t* preds = data;
uint8_t* out = filtered_data;
int row = 1;
SANITY_CHECK(in, out);
// left prediction for top scan-line
out[0] = in[0];
PredictLineInverse0(in + 1, preds, out + 1, width - 1);
preds += stride;
in += stride;
out += stride;
// Filter line-by-line.
while (row < height) {
out[0] = in[0] - preds[- stride];
PredictLineGradient(preds + 1, in + 1, out + 1, stride, width - 1);
++row;
preds += stride;
in += stride;
out += stride;
}
}
//------------------------------------------------------------------------------
// Vertical filter
static void VerticalFilter_MSA(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
const uint8_t* in = data;
const uint8_t* preds = data;
uint8_t* out = filtered_data;
int row = 1;
SANITY_CHECK(in, out);
// Very first top-left pixel is copied.
out[0] = in[0];
// Rest of top scan-line is left-predicted.
PredictLineInverse0(in + 1, preds, out + 1, width - 1);
in += stride;
out += stride;
// Filter line-by-line.
while (row < height) {
PredictLineInverse0(in, preds, out, width);
++row;
preds += stride;
in += stride;
out += stride;
}
}
#undef SANITY_CHECK
//------------------------------------------------------------------------------
// Entry point
extern void VP8FiltersInitMSA(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8FiltersInitMSA(void) {
WebPFilters[WEBP_FILTER_HORIZONTAL] = HorizontalFilter_MSA;
WebPFilters[WEBP_FILTER_VERTICAL] = VerticalFilter_MSA;
WebPFilters[WEBP_FILTER_GRADIENT] = GradientFilter_MSA;
}
#else // !WEBP_USE_MSA
WEBP_DSP_INIT_STUB(VP8FiltersInitMSA)
#endif // WEBP_USE_MSA

329
media/libwebp/src/dsp/filters_neon.c Normal file
Просмотреть файл

@ -0,0 +1,329 @@
// Copyright 2017 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// NEON variant of alpha filters
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_NEON)
#include <assert.h>
#include "src/dsp/neon.h"
//------------------------------------------------------------------------------
// Helpful macros.
# define SANITY_CHECK(in, out) \
assert(in != NULL); \
assert(out != NULL); \
assert(width > 0); \
assert(height > 0); \
assert(stride >= width); \
assert(row >= 0 && num_rows > 0 && row + num_rows <= height); \
(void)height; // Silence unused warning.
// load eight u8 and widen to s16
#define U8_TO_S16(A) vreinterpretq_s16_u16(vmovl_u8(A))
#define LOAD_U8_TO_S16(A) U8_TO_S16(vld1_u8(A))
// shift left or right by N byte, inserting zeros
#define SHIFT_RIGHT_N_Q(A, N) vextq_u8((A), zero, (N))
#define SHIFT_LEFT_N_Q(A, N) vextq_u8(zero, (A), (16 - (N)) % 16)
// rotate left by N bytes
#define ROTATE_LEFT_N(A, N) vext_u8((A), (A), (N))
// rotate right by N bytes
#define ROTATE_RIGHT_N(A, N) vext_u8((A), (A), (8 - (N)) % 8)
static void PredictLine_NEON(const uint8_t* src, const uint8_t* pred,
uint8_t* dst, int length) {
int i;
assert(length >= 0);
for (i = 0; i + 16 <= length; i += 16) {
const uint8x16_t A = vld1q_u8(&src[i]);
const uint8x16_t B = vld1q_u8(&pred[i]);
const uint8x16_t C = vsubq_u8(A, B);
vst1q_u8(&dst[i], C);
}
for (; i < length; ++i) dst[i] = src[i] - pred[i];
}
// Special case for left-based prediction (when preds==dst-1 or preds==src-1).
static void PredictLineLeft_NEON(const uint8_t* src, uint8_t* dst, int length) {
PredictLine_NEON(src, src - 1, dst, length);
}
//------------------------------------------------------------------------------
// Horizontal filter.
static WEBP_INLINE void DoHorizontalFilter_NEON(const uint8_t* in,
int width, int height,
int stride,
int row, int num_rows,
uint8_t* out) {
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
if (row == 0) {
// Leftmost pixel is the same as input for topmost scanline.
out[0] = in[0];
PredictLineLeft_NEON(in + 1, out + 1, width - 1);
row = 1;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
// Leftmost pixel is predicted from above.
out[0] = in[0] - in[-stride];
PredictLineLeft_NEON(in + 1, out + 1, width - 1);
++row;
in += stride;
out += stride;
}
}
static void HorizontalFilter_NEON(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoHorizontalFilter_NEON(data, width, height, stride, 0, height,
filtered_data);
}
//------------------------------------------------------------------------------
// Vertical filter.
static WEBP_INLINE void DoVerticalFilter_NEON(const uint8_t* in,
int width, int height, int stride,
int row, int num_rows,
uint8_t* out) {
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
if (row == 0) {
// Very first top-left pixel is copied.
out[0] = in[0];
// Rest of top scan-line is left-predicted.
PredictLineLeft_NEON(in + 1, out + 1, width - 1);
row = 1;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
PredictLine_NEON(in, in - stride, out, width);
++row;
in += stride;
out += stride;
}
}
static void VerticalFilter_NEON(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoVerticalFilter_NEON(data, width, height, stride, 0, height,
filtered_data);
}
//------------------------------------------------------------------------------
// Gradient filter.
static WEBP_INLINE int GradientPredictor_C(uint8_t a, uint8_t b, uint8_t c) {
const int g = a + b - c;
return ((g & ~0xff) == 0) ? g : (g < 0) ? 0 : 255; // clip to 8bit
}
static void GradientPredictDirect_NEON(const uint8_t* const row,
const uint8_t* const top,
uint8_t* const out, int length) {
int i;
for (i = 0; i + 8 <= length; i += 8) {
const uint8x8_t A = vld1_u8(&row[i - 1]);
const uint8x8_t B = vld1_u8(&top[i + 0]);
const int16x8_t C = vreinterpretq_s16_u16(vaddl_u8(A, B));
const int16x8_t D = LOAD_U8_TO_S16(&top[i - 1]);
const uint8x8_t E = vqmovun_s16(vsubq_s16(C, D));
const uint8x8_t F = vld1_u8(&row[i + 0]);
vst1_u8(&out[i], vsub_u8(F, E));
}
for (; i < length; ++i) {
out[i] = row[i] - GradientPredictor_C(row[i - 1], top[i], top[i - 1]);
}
}
static WEBP_INLINE void DoGradientFilter_NEON(const uint8_t* in,
int width, int height,
int stride,
int row, int num_rows,
uint8_t* out) {
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
// left prediction for top scan-line
if (row == 0) {
out[0] = in[0];
PredictLineLeft_NEON(in + 1, out + 1, width - 1);
row = 1;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
out[0] = in[0] - in[-stride];
GradientPredictDirect_NEON(in + 1, in + 1 - stride, out + 1, width - 1);
++row;
in += stride;
out += stride;
}
}
static void GradientFilter_NEON(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoGradientFilter_NEON(data, width, height, stride, 0, height,
filtered_data);
}
#undef SANITY_CHECK
//------------------------------------------------------------------------------
// Inverse transforms
static void HorizontalUnfilter_NEON(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
int i;
const uint8x16_t zero = vdupq_n_u8(0);
uint8x16_t last;
out[0] = in[0] + (prev == NULL ? 0 : prev[0]);
if (width <= 1) return;
last = vsetq_lane_u8(out[0], zero, 0);
for (i = 1; i + 16 <= width; i += 16) {
const uint8x16_t A0 = vld1q_u8(&in[i]);
const uint8x16_t A1 = vaddq_u8(A0, last);
const uint8x16_t A2 = SHIFT_LEFT_N_Q(A1, 1);
const uint8x16_t A3 = vaddq_u8(A1, A2);
const uint8x16_t A4 = SHIFT_LEFT_N_Q(A3, 2);
const uint8x16_t A5 = vaddq_u8(A3, A4);
const uint8x16_t A6 = SHIFT_LEFT_N_Q(A5, 4);
const uint8x16_t A7 = vaddq_u8(A5, A6);
const uint8x16_t A8 = SHIFT_LEFT_N_Q(A7, 8);
const uint8x16_t A9 = vaddq_u8(A7, A8);
vst1q_u8(&out[i], A9);
last = SHIFT_RIGHT_N_Q(A9, 15);
}
for (; i < width; ++i) out[i] = in[i] + out[i - 1];
}
static void VerticalUnfilter_NEON(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_NEON(NULL, in, out, width);
} else {
int i;
assert(width >= 0);
for (i = 0; i + 16 <= width; i += 16) {
const uint8x16_t A = vld1q_u8(&in[i]);
const uint8x16_t B = vld1q_u8(&prev[i]);
const uint8x16_t C = vaddq_u8(A, B);
vst1q_u8(&out[i], C);
}
for (; i < width; ++i) out[i] = in[i] + prev[i];
}
}
// GradientUnfilter_NEON is correct but slower than the C-version,
// at least on ARM64. For armv7, it's a wash.
// So best is to disable it for now, but keep the idea around...
#if !defined(USE_GRADIENT_UNFILTER)
#define USE_GRADIENT_UNFILTER 0 // ALTERNATE_CODE
#endif
#if (USE_GRADIENT_UNFILTER == 1)
#define GRAD_PROCESS_LANE(L) do { \
const uint8x8_t tmp1 = ROTATE_RIGHT_N(pred, 1); /* rotate predictor in */ \
const int16x8_t tmp2 = vaddq_s16(BC, U8_TO_S16(tmp1)); \
const uint8x8_t delta = vqmovun_s16(tmp2); \
pred = vadd_u8(D, delta); \
out = vext_u8(out, ROTATE_LEFT_N(pred, (L)), 1); \
} while (0)
static void GradientPredictInverse_NEON(const uint8_t* const in,
const uint8_t* const top,
uint8_t* const row, int length) {
if (length > 0) {
int i;
uint8x8_t pred = vdup_n_u8(row[-1]); // left sample
uint8x8_t out = vdup_n_u8(0);
for (i = 0; i + 8 <= length; i += 8) {
const int16x8_t B = LOAD_U8_TO_S16(&top[i + 0]);
const int16x8_t C = LOAD_U8_TO_S16(&top[i - 1]);
const int16x8_t BC = vsubq_s16(B, C); // unclipped gradient basis B - C
const uint8x8_t D = vld1_u8(&in[i]); // base input
GRAD_PROCESS_LANE(0);
GRAD_PROCESS_LANE(1);
GRAD_PROCESS_LANE(2);
GRAD_PROCESS_LANE(3);
GRAD_PROCESS_LANE(4);
GRAD_PROCESS_LANE(5);
GRAD_PROCESS_LANE(6);
GRAD_PROCESS_LANE(7);
vst1_u8(&row[i], out);
}
for (; i < length; ++i) {
row[i] = in[i] + GradientPredictor_C(row[i - 1], top[i], top[i - 1]);
}
}
}
#undef GRAD_PROCESS_LANE
static void GradientUnfilter_NEON(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_NEON(NULL, in, out, width);
} else {
out[0] = in[0] + prev[0]; // predict from above
GradientPredictInverse_NEON(in + 1, prev + 1, out + 1, width - 1);
}
}
#endif // USE_GRADIENT_UNFILTER
//------------------------------------------------------------------------------
// Entry point
extern void VP8FiltersInitNEON(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8FiltersInitNEON(void) {
WebPUnfilters[WEBP_FILTER_HORIZONTAL] = HorizontalUnfilter_NEON;
WebPUnfilters[WEBP_FILTER_VERTICAL] = VerticalUnfilter_NEON;
#if (USE_GRADIENT_UNFILTER == 1)
WebPUnfilters[WEBP_FILTER_GRADIENT] = GradientUnfilter_NEON;
#endif
WebPFilters[WEBP_FILTER_HORIZONTAL] = HorizontalFilter_NEON;
WebPFilters[WEBP_FILTER_VERTICAL] = VerticalFilter_NEON;
WebPFilters[WEBP_FILTER_GRADIENT] = GradientFilter_NEON;
}
#else // !WEBP_USE_NEON
WEBP_DSP_INIT_STUB(VP8FiltersInitNEON)
#endif // WEBP_USE_NEON

333
media/libwebp/src/dsp/filters_sse2.c Normal file
Просмотреть файл

@ -0,0 +1,333 @@
// Copyright 2015 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// SSE2 variant of alpha filters
//
// Author: Skal (pascal.massimino@gmail.com)
#include "src/dsp/dsp.h"
#if defined(WEBP_USE_SSE2)
#include <assert.h>
#include <emmintrin.h>
#include <stdlib.h>
#include <string.h>
//------------------------------------------------------------------------------
// Helpful macro.
# define SANITY_CHECK(in, out) \
assert((in) != NULL); \
assert((out) != NULL); \
assert(width > 0); \
assert(height > 0); \
assert(stride >= width); \
assert(row >= 0 && num_rows > 0 && row + num_rows <= height); \
(void)height; // Silence unused warning.
static void PredictLineTop_SSE2(const uint8_t* src, const uint8_t* pred,
uint8_t* dst, int length) {
int i;
const int max_pos = length & ~31;
assert(length >= 0);
for (i = 0; i < max_pos; i += 32) {
const __m128i A0 = _mm_loadu_si128((const __m128i*)&src[i + 0]);
const __m128i A1 = _mm_loadu_si128((const __m128i*)&src[i + 16]);
const __m128i B0 = _mm_loadu_si128((const __m128i*)&pred[i + 0]);
const __m128i B1 = _mm_loadu_si128((const __m128i*)&pred[i + 16]);
const __m128i C0 = _mm_sub_epi8(A0, B0);
const __m128i C1 = _mm_sub_epi8(A1, B1);
_mm_storeu_si128((__m128i*)&dst[i + 0], C0);
_mm_storeu_si128((__m128i*)&dst[i + 16], C1);
}
for (; i < length; ++i) dst[i] = src[i] - pred[i];
}
// Special case for left-based prediction (when preds==dst-1 or preds==src-1).
static void PredictLineLeft_SSE2(const uint8_t* src, uint8_t* dst, int length) {
int i;
const int max_pos = length & ~31;
assert(length >= 0);
for (i = 0; i < max_pos; i += 32) {
const __m128i A0 = _mm_loadu_si128((const __m128i*)(src + i + 0 ));
const __m128i B0 = _mm_loadu_si128((const __m128i*)(src + i + 0 - 1));
const __m128i A1 = _mm_loadu_si128((const __m128i*)(src + i + 16 ));
const __m128i B1 = _mm_loadu_si128((const __m128i*)(src + i + 16 - 1));
const __m128i C0 = _mm_sub_epi8(A0, B0);
const __m128i C1 = _mm_sub_epi8(A1, B1);
_mm_storeu_si128((__m128i*)(dst + i + 0), C0);
_mm_storeu_si128((__m128i*)(dst + i + 16), C1);
}
for (; i < length; ++i) dst[i] = src[i] - src[i - 1];
}
//------------------------------------------------------------------------------
// Horizontal filter.
static WEBP_INLINE void DoHorizontalFilter_SSE2(const uint8_t* in,
int width, int height,
int stride,
int row, int num_rows,
uint8_t* out) {
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
if (row == 0) {
// Leftmost pixel is the same as input for topmost scanline.
out[0] = in[0];
PredictLineLeft_SSE2(in + 1, out + 1, width - 1);
row = 1;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
// Leftmost pixel is predicted from above.
out[0] = in[0] - in[-stride];
PredictLineLeft_SSE2(in + 1, out + 1, width - 1);
++row;
in += stride;
out += stride;
}
}
//------------------------------------------------------------------------------
// Vertical filter.
static WEBP_INLINE void DoVerticalFilter_SSE2(const uint8_t* in,
int width, int height, int stride,
int row, int num_rows,
uint8_t* out) {
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
if (row == 0) {
// Very first top-left pixel is copied.
out[0] = in[0];
// Rest of top scan-line is left-predicted.
PredictLineLeft_SSE2(in + 1, out + 1, width - 1);
row = 1;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
PredictLineTop_SSE2(in, in - stride, out, width);
++row;
in += stride;
out += stride;
}
}
//------------------------------------------------------------------------------
// Gradient filter.
static WEBP_INLINE int GradientPredictor_SSE2(uint8_t a, uint8_t b, uint8_t c) {
const int g = a + b - c;
return ((g & ~0xff) == 0) ? g : (g < 0) ? 0 : 255; // clip to 8bit
}
static void GradientPredictDirect_SSE2(const uint8_t* const row,
const uint8_t* const top,
uint8_t* const out, int length) {
const int max_pos = length & ~7;
int i;
const __m128i zero = _mm_setzero_si128();
for (i = 0; i < max_pos; i += 8) {
const __m128i A0 = _mm_loadl_epi64((const __m128i*)&row[i - 1]);
const __m128i B0 = _mm_loadl_epi64((const __m128i*)&top[i]);
const __m128i C0 = _mm_loadl_epi64((const __m128i*)&top[i - 1]);
const __m128i D = _mm_loadl_epi64((const __m128i*)&row[i]);
const __m128i A1 = _mm_unpacklo_epi8(A0, zero);
const __m128i B1 = _mm_unpacklo_epi8(B0, zero);
const __m128i C1 = _mm_unpacklo_epi8(C0, zero);
const __m128i E = _mm_add_epi16(A1, B1);
const __m128i F = _mm_sub_epi16(E, C1);
const __m128i G = _mm_packus_epi16(F, zero);
const __m128i H = _mm_sub_epi8(D, G);
_mm_storel_epi64((__m128i*)(out + i), H);
}
for (; i < length; ++i) {
out[i] = row[i] - GradientPredictor_SSE2(row[i - 1], top[i], top[i - 1]);
}
}
static WEBP_INLINE void DoGradientFilter_SSE2(const uint8_t* in,
int width, int height, int stride,
int row, int num_rows,
uint8_t* out) {
const size_t start_offset = row * stride;
const int last_row = row + num_rows;
SANITY_CHECK(in, out);
in += start_offset;
out += start_offset;
// left prediction for top scan-line
if (row == 0) {
out[0] = in[0];
PredictLineLeft_SSE2(in + 1, out + 1, width - 1);
row = 1;
in += stride;
out += stride;
}
// Filter line-by-line.
while (row < last_row) {
out[0] = in[0] - in[-stride];
GradientPredictDirect_SSE2(in + 1, in + 1 - stride, out + 1, width - 1);
++row;
in += stride;
out += stride;
}
}
#undef SANITY_CHECK
//------------------------------------------------------------------------------
static void HorizontalFilter_SSE2(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoHorizontalFilter_SSE2(data, width, height, stride, 0, height,
filtered_data);
}
static void VerticalFilter_SSE2(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoVerticalFilter_SSE2(data, width, height, stride, 0, height, filtered_data);
}
static void GradientFilter_SSE2(const uint8_t* data, int width, int height,
int stride, uint8_t* filtered_data) {
DoGradientFilter_SSE2(data, width, height, stride, 0, height, filtered_data);
}
//------------------------------------------------------------------------------
// Inverse transforms
static void HorizontalUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
int i;
__m128i last;
out[0] = in[0] + (prev == NULL ? 0 : prev[0]);
if (width <= 1) return;
last = _mm_set_epi32(0, 0, 0, out[0]);
for (i = 1; i + 8 <= width; i += 8) {
const __m128i A0 = _mm_loadl_epi64((const __m128i*)(in + i));
const __m128i A1 = _mm_add_epi8(A0, last);
const __m128i A2 = _mm_slli_si128(A1, 1);
const __m128i A3 = _mm_add_epi8(A1, A2);
const __m128i A4 = _mm_slli_si128(A3, 2);
const __m128i A5 = _mm_add_epi8(A3, A4);
const __m128i A6 = _mm_slli_si128(A5, 4);
const __m128i A7 = _mm_add_epi8(A5, A6);
_mm_storel_epi64((__m128i*)(out + i), A7);
last = _mm_srli_epi64(A7, 56);
}
for (; i < width; ++i) out[i] = in[i] + out[i - 1];
}
static void VerticalUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_SSE2(NULL, in, out, width);
} else {
int i;
const int max_pos = width & ~31;
assert(width >= 0);
for (i = 0; i < max_pos; i += 32) {
const __m128i A0 = _mm_loadu_si128((const __m128i*)&in[i + 0]);
const __m128i A1 = _mm_loadu_si128((const __m128i*)&in[i + 16]);
const __m128i B0 = _mm_loadu_si128((const __m128i*)&prev[i + 0]);
const __m128i B1 = _mm_loadu_si128((const __m128i*)&prev[i + 16]);
const __m128i C0 = _mm_add_epi8(A0, B0);
const __m128i C1 = _mm_add_epi8(A1, B1);
_mm_storeu_si128((__m128i*)&out[i + 0], C0);
_mm_storeu_si128((__m128i*)&out[i + 16], C1);
}
for (; i < width; ++i) out[i] = in[i] + prev[i];
}
}
static void GradientPredictInverse_SSE2(const uint8_t* const in,
const uint8_t* const top,
uint8_t* const row, int length) {
if (length > 0) {
int i;
const int max_pos = length & ~7;
const __m128i zero = _mm_setzero_si128();
__m128i A = _mm_set_epi32(0, 0, 0, row[-1]); // left sample
for (i = 0; i < max_pos; i += 8) {
const __m128i tmp0 = _mm_loadl_epi64((const __m128i*)&top[i]);
const __m128i tmp1 = _mm_loadl_epi64((const __m128i*)&top[i - 1]);
const __m128i B = _mm_unpacklo_epi8(tmp0, zero);
const __m128i C = _mm_unpacklo_epi8(tmp1, zero);
const __m128i D = _mm_loadl_epi64((const __m128i*)&in[i]); // base input
const __m128i E = _mm_sub_epi16(B, C); // unclipped gradient basis B - C
__m128i out = zero; // accumulator for output
__m128i mask_hi = _mm_set_epi32(0, 0, 0, 0xff);
int k = 8;
while (1) {
const __m128i tmp3 = _mm_add_epi16(A, E); // delta = A + B - C
const __m128i tmp4 = _mm_packus_epi16(tmp3, zero); // saturate delta
const __m128i tmp5 = _mm_add_epi8(tmp4, D); // add to in[]
A = _mm_and_si128(tmp5, mask_hi); // 1-complement clip
out = _mm_or_si128(out, A); // accumulate output
if (--k == 0) break;
A = _mm_slli_si128(A, 1); // rotate left sample
mask_hi = _mm_slli_si128(mask_hi, 1); // rotate mask
A = _mm_unpacklo_epi8(A, zero); // convert 8b->16b
}
A = _mm_srli_si128(A, 7); // prepare left sample for next iteration
_mm_storel_epi64((__m128i*)&row[i], out);
}
for (; i < length; ++i) {
row[i] = in[i] + GradientPredictor_SSE2(row[i - 1], top[i], top[i - 1]);
}
}
}
static void GradientUnfilter_SSE2(const uint8_t* prev, const uint8_t* in,
uint8_t* out, int width) {
if (prev == NULL) {
HorizontalUnfilter_SSE2(NULL, in, out, width);
} else {
out[0] = in[0] + prev[0]; // predict from above
GradientPredictInverse_SSE2(in + 1, prev + 1, out + 1, width - 1);
}
}
//------------------------------------------------------------------------------
// Entry point
extern void VP8FiltersInitSSE2(void);
WEBP_TSAN_IGNORE_FUNCTION void VP8FiltersInitSSE2(void) {
WebPUnfilters[WEBP_FILTER_HORIZONTAL] = HorizontalUnfilter_SSE2;
WebPUnfilters[WEBP_FILTER_VERTICAL] = VerticalUnfilter_SSE2;
WebPUnfilters[WEBP_FILTER_GRADIENT] = GradientUnfilter_SSE2;
WebPFilters[WEBP_FILTER_HORIZONTAL] = HorizontalFilter_SSE2;
WebPFilters[WEBP_FILTER_VERTICAL] = VerticalFilter_SSE2;
WebPFilters[WEBP_FILTER_GRADIENT] = GradientFilter_SSE2;
}
#else // !WEBP_USE_SSE2
WEBP_DSP_INIT_STUB(VP8FiltersInitSSE2)
#endif // WEBP_USE_SSE2

659
media/libwebp/src/dsp/lossless.c Normal file
Просмотреть файл

@ -0,0 +1,659 @@
// Copyright 2012 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Image transforms and color space conversion methods for lossless decoder.
//
// Authors: Vikas Arora (vikaas.arora@gmail.com)
// Jyrki Alakuijala (jyrki@google.com)
// Urvang Joshi (urvang@google.com)
#include "src/dsp/dsp.h"
#include <assert.h>
#include <math.h>
#include <stdlib.h>
#include "src/dec/vp8li_dec.h"
#include "src/utils/endian_inl_utils.h"
#include "src/dsp/lossless.h"
#include "src/dsp/lossless_common.h"
#define MAX_DIFF_COST (1e30f)
//------------------------------------------------------------------------------
// Image transforms.
static WEBP_INLINE uint32_t Average2(uint32_t a0, uint32_t a1) {
return (((a0 ^ a1) & 0xfefefefeu) >> 1) + (a0 & a1);
}
static WEBP_INLINE uint32_t Average3(uint32_t a0, uint32_t a1, uint32_t a2) {
return Average2(Average2(a0, a2), a1);
}
static WEBP_INLINE uint32_t Average4(uint32_t a0, uint32_t a1,
uint32_t a2, uint32_t a3) {
return Average2(Average2(a0, a1), Average2(a2, a3));
}
static WEBP_INLINE uint32_t Clip255(uint32_t a) {
if (a < 256) {
return a;
}
// return 0, when a is a negative integer.
// return 255, when a is positive.
return ~a >> 24;
}
static WEBP_INLINE int AddSubtractComponentFull(int a, int b, int c) {
return Clip255(a + b - c);
}
static WEBP_INLINE uint32_t ClampedAddSubtractFull(uint32_t c0, uint32_t c1,
uint32_t c2) {
const int a = AddSubtractComponentFull(c0 >> 24, c1 >> 24, c2 >> 24);
const int r = AddSubtractComponentFull((c0 >> 16) & 0xff,
(c1 >> 16) & 0xff,
(c2 >> 16) & 0xff);
const int g = AddSubtractComponentFull((c0 >> 8) & 0xff,
(c1 >> 8) & 0xff,
(c2 >> 8) & 0xff);
const int b = AddSubtractComponentFull(c0 & 0xff, c1 & 0xff, c2 & 0xff);
return ((uint32_t)a << 24) | (r << 16) | (g << 8) | b;
}
static WEBP_INLINE int AddSubtractComponentHalf(int a, int b) {
return Clip255(a + (a - b) / 2);
}
static WEBP_INLINE uint32_t ClampedAddSubtractHalf(uint32_t c0, uint32_t c1,
uint32_t c2) {
const uint32_t ave = Average2(c0, c1);
const int a = AddSubtractComponentHalf(ave >> 24, c2 >> 24);
const int r = AddSubtractComponentHalf((ave >> 16) & 0xff, (c2 >> 16) & 0xff);
const int g = AddSubtractComponentHalf((ave >> 8) & 0xff, (c2 >> 8) & 0xff);
const int b = AddSubtractComponentHalf((ave >> 0) & 0xff, (c2 >> 0) & 0xff);
return ((uint32_t)a << 24) | (r << 16) | (g << 8) | b;
}
// gcc <= 4.9 on ARM generates incorrect code in Select() when Sub3() is
// inlined.
#if defined(__arm__) && LOCAL_GCC_VERSION <= 0x409
# define LOCAL_INLINE __attribute__ ((noinline))
#else
# define LOCAL_INLINE WEBP_INLINE
#endif
static LOCAL_INLINE int Sub3(int a, int b, int c) {
const int pb = b - c;
const int pa = a - c;
return abs(pb) - abs(pa);
}
#undef LOCAL_INLINE
static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
const int pa_minus_pb =
Sub3((a >> 24) , (b >> 24) , (c >> 24) ) +
Sub3((a >> 16) & 0xff, (b >> 16) & 0xff, (c >> 16) & 0xff) +
Sub3((a >> 8) & 0xff, (b >> 8) & 0xff, (c >> 8) & 0xff) +
Sub3((a ) & 0xff, (b ) & 0xff, (c ) & 0xff);
return (pa_minus_pb <= 0) ? a : b;
}
//------------------------------------------------------------------------------
// Predictors
static uint32_t Predictor0_C(uint32_t left, const uint32_t* const top) {
(void)top;
(void)left;
return ARGB_BLACK;
}
static uint32_t Predictor1_C(uint32_t left, const uint32_t* const top) {
(void)top;
return left;
}
static uint32_t Predictor2_C(uint32_t left, const uint32_t* const top) {
(void)left;
return top[0];
}
static uint32_t Predictor3_C(uint32_t left, const uint32_t* const top) {
(void)left;
return top[1];
}
static uint32_t Predictor4_C(uint32_t left, const uint32_t* const top) {
(void)left;
return top[-1];
}
static uint32_t Predictor5_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = Average3(left, top[0], top[1]);
return pred;
}
static uint32_t Predictor6_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = Average2(left, top[-1]);
return pred;
}
static uint32_t Predictor7_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = Average2(left, top[0]);
return pred;
}
static uint32_t Predictor8_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = Average2(top[-1], top[0]);
(void)left;
return pred;
}
static uint32_t Predictor9_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = Average2(top[0], top[1]);
(void)left;
return pred;
}
static uint32_t Predictor10_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = Average4(left, top[-1], top[0], top[1]);
return pred;
}
static uint32_t Predictor11_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = Select(top[0], left, top[-1]);
return pred;
}
static uint32_t Predictor12_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = ClampedAddSubtractFull(left, top[0], top[-1]);
return pred;
}
static uint32_t Predictor13_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = ClampedAddSubtractHalf(left, top[0], top[-1]);
return pred;
}
GENERATE_PREDICTOR_ADD(Predictor0_C, PredictorAdd0_C)
static void PredictorAdd1_C(const uint32_t* in, const uint32_t* upper,
int num_pixels, uint32_t* out) {
int i;
uint32_t left = out[-1];
for (i = 0; i < num_pixels; ++i) {
out[i] = left = VP8LAddPixels(in[i], left);
}
(void)upper;
}
GENERATE_PREDICTOR_ADD(Predictor2_C, PredictorAdd2_C)
GENERATE_PREDICTOR_ADD(Predictor3_C, PredictorAdd3_C)
GENERATE_PREDICTOR_ADD(Predictor4_C, PredictorAdd4_C)
GENERATE_PREDICTOR_ADD(Predictor5_C, PredictorAdd5_C)
GENERATE_PREDICTOR_ADD(Predictor6_C, PredictorAdd6_C)
GENERATE_PREDICTOR_ADD(Predictor7_C, PredictorAdd7_C)
GENERATE_PREDICTOR_ADD(Predictor8_C, PredictorAdd8_C)
GENERATE_PREDICTOR_ADD(Predictor9_C, PredictorAdd9_C)
GENERATE_PREDICTOR_ADD(Predictor10_C, PredictorAdd10_C)
GENERATE_PREDICTOR_ADD(Predictor11_C, PredictorAdd11_C)
GENERATE_PREDICTOR_ADD(Predictor12_C, PredictorAdd12_C)
GENERATE_PREDICTOR_ADD(Predictor13_C, PredictorAdd13_C)
//------------------------------------------------------------------------------
// Inverse prediction.
static void PredictorInverseTransform_C(const VP8LTransform* const transform,
int y_start, int y_end,
const uint32_t* in, uint32_t* out) {
const int width = transform->xsize_;
if (y_start == 0) { // First Row follows the L (mode=1) mode.
PredictorAdd0_C(in, NULL, 1, out);
PredictorAdd1_C(in + 1, NULL, width - 1, out + 1);
in += width;
out += width;
++y_start;
}
{
int y = y_start;
const int tile_width = 1 << transform->bits_;
const int mask = tile_width - 1;
const int tiles_per_row = VP8LSubSampleSize(width, transform->bits_);
const uint32_t* pred_mode_base =
transform->data_ + (y >> transform->bits_) * tiles_per_row;
while (y < y_end) {
const uint32_t* pred_mode_src = pred_mode_base;
int x = 1;
// First pixel follows the T (mode=2) mode.
PredictorAdd2_C(in, out - width, 1, out);
// .. the rest:
while (x < width) {
const VP8LPredictorAddSubFunc pred_func =
VP8LPredictorsAdd[((*pred_mode_src++) >> 8) & 0xf];
int x_end = (x & ~mask) + tile_width;
if (x_end > width) x_end = width;
pred_func(in + x, out + x - width, x_end - x, out + x);
x = x_end;
}
in += width;
out += width;
++y;
if ((y & mask) == 0) { // Use the same mask, since tiles are squares.
pred_mode_base += tiles_per_row;
}
}
}
}
// Add green to blue and red channels (i.e. perform the inverse transform of
// 'subtract green').
void VP8LAddGreenToBlueAndRed_C(const uint32_t* src, int num_pixels,
uint32_t* dst) {
int i;
for (i = 0; i < num_pixels; ++i) {
const uint32_t argb = src[i];
const uint32_t green = ((argb >> 8) & 0xff);
uint32_t red_blue = (argb & 0x00ff00ffu);
red_blue += (green << 16) | green;
red_blue &= 0x00ff00ffu;
dst[i] = (argb & 0xff00ff00u) | red_blue;
}
}
static WEBP_INLINE int ColorTransformDelta(int8_t color_pred,
int8_t color) {
return ((int)color_pred * color) >> 5;
}
static WEBP_INLINE void ColorCodeToMultipliers(uint32_t color_code,
VP8LMultipliers* const m) {
m->green_to_red_ = (color_code >> 0) & 0xff;
m->green_to_blue_ = (color_code >> 8) & 0xff;
m->red_to_blue_ = (color_code >> 16) & 0xff;
}
void VP8LTransformColorInverse_C(const VP8LMultipliers* const m,
const uint32_t* src, int num_pixels,
uint32_t* dst) {
int i;
for (i = 0; i < num_pixels; ++i) {
const uint32_t argb = src[i];
const uint32_t green = argb >> 8;
const uint32_t red = argb >> 16;
int new_red = red & 0xff;
int new_blue = argb & 0xff;
new_red += ColorTransformDelta(m->green_to_red_, green);
new_red &= 0xff;
new_blue += ColorTransformDelta(m->green_to_blue_, green);
new_blue += ColorTransformDelta(m->red_to_blue_, new_red);
new_blue &= 0xff;
dst[i] = (argb & 0xff00ff00u) | (new_red << 16) | (new_blue);
}
}
// Color space inverse transform.
static void ColorSpaceInverseTransform_C(const VP8LTransform* const transform,
int y_start, int y_end,
const uint32_t* src, uint32_t* dst) {
const int width = transform->xsize_;
const int tile_width = 1 << transform->bits_;
const int mask = tile_width - 1;
const int safe_width = width & ~mask;
const int remaining_width = width - safe_width;
const int tiles_per_row = VP8LSubSampleSize(width, transform->bits_);
int y = y_start;
const uint32_t* pred_row =
transform->data_ + (y >> transform->bits_) * tiles_per_row;
while (y < y_end) {
const uint32_t* pred = pred_row;
VP8LMultipliers m = { 0, 0, 0 };
const uint32_t* const src_safe_end = src + safe_width;
const uint32_t* const src_end = src + width;
while (src < src_safe_end) {
ColorCodeToMultipliers(*pred++, &m);
VP8LTransformColorInverse(&m, src, tile_width, dst);
src += tile_width;
dst += tile_width;
}
if (src < src_end) { // Left-overs using C-version.
ColorCodeToMultipliers(*pred++, &m);
VP8LTransformColorInverse(&m, src, remaining_width, dst);
src += remaining_width;
dst += remaining_width;
}
++y;
if ((y & mask) == 0) pred_row += tiles_per_row;
}
}
// Separate out pixels packed together using pixel-bundling.
// We define two methods for ARGB data (uint32_t) and alpha-only data (uint8_t).
#define COLOR_INDEX_INVERSE(FUNC_NAME, F_NAME, STATIC_DECL, TYPE, BIT_SUFFIX, \
GET_INDEX, GET_VALUE) \
static void F_NAME(const TYPE* src, const uint32_t* const color_map, \
TYPE* dst, int y_start, int y_end, int width) { \
int y; \
for (y = y_start; y < y_end; ++y) { \
int x; \
for (x = 0; x < width; ++x) { \
*dst++ = GET_VALUE(color_map[GET_INDEX(*src++)]); \
} \
} \
} \
STATIC_DECL void FUNC_NAME(const VP8LTransform* const transform, \
int y_start, int y_end, const TYPE* src, \
TYPE* dst) { \
int y; \
const int bits_per_pixel = 8 >> transform->bits_; \
const int width = transform->xsize_; \
const uint32_t* const color_map = transform->data_; \
if (bits_per_pixel < 8) { \
const int pixels_per_byte = 1 << transform->bits_; \
const int count_mask = pixels_per_byte - 1; \
const uint32_t bit_mask = (1 << bits_per_pixel) - 1; \
for (y = y_start; y < y_end; ++y) { \
uint32_t packed_pixels = 0; \
int x; \
for (x = 0; x < width; ++x) { \
/* We need to load fresh 'packed_pixels' once every */ \
/* 'pixels_per_byte' increments of x. Fortunately, pixels_per_byte */ \
/* is a power of 2, so can just use a mask for that, instead of */ \
/* decrementing a counter. */ \
if ((x & count_mask) == 0) packed_pixels = GET_INDEX(*src++); \
*dst++ = GET_VALUE(color_map[packed_pixels & bit_mask]); \
packed_pixels >>= bits_per_pixel; \
} \
} \
} else { \
VP8LMapColor##BIT_SUFFIX(src, color_map, dst, y_start, y_end, width); \
} \
}
COLOR_INDEX_INVERSE(ColorIndexInverseTransform_C, MapARGB_C, static,
uint32_t, 32b, VP8GetARGBIndex, VP8GetARGBValue)
COLOR_INDEX_INVERSE(VP8LColorIndexInverseTransformAlpha, MapAlpha_C, ,
uint8_t, 8b, VP8GetAlphaIndex, VP8GetAlphaValue)
#undef COLOR_INDEX_INVERSE
void VP8LInverseTransform(const VP8LTransform* const transform,
int row_start, int row_end,
const uint32_t* const in, uint32_t* const out) {
const int width = transform->xsize_;
assert(row_start < row_end);
assert(row_end <= transform->ysize_);
switch (transform->type_) {
case SUBTRACT_GREEN:
VP8LAddGreenToBlueAndRed(in, (row_end - row_start) * width, out);
break;
case PREDICTOR_TRANSFORM:
PredictorInverseTransform_C(transform, row_start, row_end, in, out);
if (row_end != transform->ysize_) {
// The last predicted row in this iteration will be the top-pred row
// for the first row in next iteration.
memcpy(out - width, out + (row_end - row_start - 1) * width,
width * sizeof(*out));
}
break;
case CROSS_COLOR_TRANSFORM:
ColorSpaceInverseTransform_C(transform, row_start, row_end, in, out);
break;
case COLOR_INDEXING_TRANSFORM:
if (in == out && transform->bits_ > 0) {
// Move packed pixels to the end of unpacked region, so that unpacking
// can occur seamlessly.
// Also, note that this is the only transform that applies on
// the effective width of VP8LSubSampleSize(xsize_, bits_). All other
// transforms work on effective width of xsize_.
const int out_stride = (row_end - row_start) * width;
const int in_stride = (row_end - row_start) *
VP8LSubSampleSize(transform->xsize_, transform->bits_);
uint32_t* const src = out + out_stride - in_stride;
memmove(src, out, in_stride * sizeof(*src));
ColorIndexInverseTransform_C(transform, row_start, row_end, src, out);
} else {
ColorIndexInverseTransform_C(transform, row_start, row_end, in, out);
}
break;
}
}
//------------------------------------------------------------------------------
// Color space conversion.
static int is_big_endian(void) {
static const union {
uint16_t w;
uint8_t b[2];
} tmp = { 1 };
return (tmp.b[0] != 1);
}
void VP8LConvertBGRAToRGB_C(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint32_t* const src_end = src + num_pixels;
while (src < src_end) {
const uint32_t argb = *src++;
*dst++ = (argb >> 16) & 0xff;
*dst++ = (argb >> 8) & 0xff;
*dst++ = (argb >> 0) & 0xff;
}
}
void VP8LConvertBGRAToRGBA_C(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint32_t* const src_end = src + num_pixels;
while (src < src_end) {
const uint32_t argb = *src++;
*dst++ = (argb >> 16) & 0xff;
*dst++ = (argb >> 8) & 0xff;
*dst++ = (argb >> 0) & 0xff;
*dst++ = (argb >> 24) & 0xff;
}
}
void VP8LConvertBGRAToRGBA4444_C(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint32_t* const src_end = src + num_pixels;
while (src < src_end) {
const uint32_t argb = *src++;
const uint8_t rg = ((argb >> 16) & 0xf0) | ((argb >> 12) & 0xf);
const uint8_t ba = ((argb >> 0) & 0xf0) | ((argb >> 28) & 0xf);
#if (WEBP_SWAP_16BIT_CSP == 1)
*dst++ = ba;
*dst++ = rg;
#else
*dst++ = rg;
*dst++ = ba;
#endif
}
}
void VP8LConvertBGRAToRGB565_C(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint32_t* const src_end = src + num_pixels;
while (src < src_end) {
const uint32_t argb = *src++;
const uint8_t rg = ((argb >> 16) & 0xf8) | ((argb >> 13) & 0x7);
const uint8_t gb = ((argb >> 5) & 0xe0) | ((argb >> 3) & 0x1f);
#if (WEBP_SWAP_16BIT_CSP == 1)
*dst++ = gb;
*dst++ = rg;
#else
*dst++ = rg;
*dst++ = gb;
#endif
}
}
void VP8LConvertBGRAToBGR_C(const uint32_t* src,
int num_pixels, uint8_t* dst) {
const uint32_t* const src_end = src + num_pixels;
while (src < src_end) {
const uint32_t argb = *src++;
*dst++ = (argb >> 0) & 0xff;
*dst++ = (argb >> 8) & 0xff;
*dst++ = (argb >> 16) & 0xff;
}
}
static void CopyOrSwap(const uint32_t* src, int num_pixels, uint8_t* dst,
int swap_on_big_endian) {
if (is_big_endian() == swap_on_big_endian) {
const uint32_t* const src_end = src + num_pixels;
while (src < src_end) {
const uint32_t argb = *src++;
WebPUint32ToMem(dst, BSwap32(argb));
dst += sizeof(argb);
}
} else {
memcpy(dst, src, num_pixels * sizeof(*src));
}
}
void VP8LConvertFromBGRA(const uint32_t* const in_data, int num_pixels,
WEBP_CSP_MODE out_colorspace, uint8_t* const rgba) {
switch (out_colorspace) {
case MODE_RGB:
VP8LConvertBGRAToRGB(in_data, num_pixels, rgba);
break;
case MODE_RGBA:
VP8LConvertBGRAToRGBA(in_data, num_pixels, rgba);
break;
case MODE_rgbA:
VP8LConvertBGRAToRGBA(in_data, num_pixels, rgba);
WebPApplyAlphaMultiply(rgba, 0, num_pixels, 1, 0);
break;
case MODE_BGR:
VP8LConvertBGRAToBGR(in_data, num_pixels, rgba);
break;
case MODE_BGRA:
CopyOrSwap(in_data, num_pixels, rgba, 1);
break;
case MODE_bgrA:
CopyOrSwap(in_data, num_pixels, rgba, 1);
WebPApplyAlphaMultiply(rgba, 0, num_pixels, 1, 0);
break;
case MODE_ARGB:
CopyOrSwap(in_data, num_pixels, rgba, 0);
break;
case MODE_Argb:
CopyOrSwap(in_data, num_pixels, rgba, 0);
WebPApplyAlphaMultiply(rgba, 1, num_pixels, 1, 0);
break;
case MODE_RGBA_4444:
VP8LConvertBGRAToRGBA4444(in_data, num_pixels, rgba);
break;
case MODE_rgbA_4444:
VP8LConvertBGRAToRGBA4444(in_data, num_pixels, rgba);
WebPApplyAlphaMultiply4444(rgba, num_pixels, 1, 0);
break;
case MODE_RGB_565:
VP8LConvertBGRAToRGB565(in_data, num_pixels, rgba);
break;
default:
assert(0); // Code flow should not reach here.
}
}
//------------------------------------------------------------------------------
VP8LProcessDecBlueAndRedFunc VP8LAddGreenToBlueAndRed;
VP8LPredictorAddSubFunc VP8LPredictorsAdd[16];
VP8LPredictorFunc VP8LPredictors[16];
// exposed plain-C implementations
VP8LPredictorAddSubFunc VP8LPredictorsAdd_C[16];
VP8LPredictorFunc VP8LPredictors_C[16];
VP8LTransformColorInverseFunc VP8LTransformColorInverse;
VP8LConvertFunc VP8LConvertBGRAToRGB;
VP8LConvertFunc VP8LConvertBGRAToRGBA;
VP8LConvertFunc VP8LConvertBGRAToRGBA4444;
VP8LConvertFunc VP8LConvertBGRAToRGB565;
VP8LConvertFunc VP8LConvertBGRAToBGR;
VP8LMapARGBFunc VP8LMapColor32b;
VP8LMapAlphaFunc VP8LMapColor8b;
extern void VP8LDspInitSSE2(void);
extern void VP8LDspInitNEON(void);
extern void VP8LDspInitMIPSdspR2(void);
extern void VP8LDspInitMSA(void);
#define COPY_PREDICTOR_ARRAY(IN, OUT) do { \
(OUT)[0] = IN##0_C; \
(OUT)[1] = IN##1_C; \
(OUT)[2] = IN##2_C; \
(OUT)[3] = IN##3_C; \
(OUT)[4] = IN##4_C; \
(OUT)[5] = IN##5_C; \
(OUT)[6] = IN##6_C; \
(OUT)[7] = IN##7_C; \
(OUT)[8] = IN##8_C; \
(OUT)[9] = IN##9_C; \
(OUT)[10] = IN##10_C; \
(OUT)[11] = IN##11_C; \
(OUT)[12] = IN##12_C; \
(OUT)[13] = IN##13_C; \
(OUT)[14] = IN##0_C; /* <- padding security sentinels*/ \
(OUT)[15] = IN##0_C; \
} while (0);
WEBP_DSP_INIT_FUNC(VP8LDspInit) {
COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors)
COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors_C)
COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd)
COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd_C)
#if !WEBP_NEON_OMIT_C_CODE
VP8LAddGreenToBlueAndRed = VP8LAddGreenToBlueAndRed_C;
VP8LTransformColorInverse = VP8LTransformColorInverse_C;
VP8LConvertBGRAToRGBA = VP8LConvertBGRAToRGBA_C;
VP8LConvertBGRAToRGB = VP8LConvertBGRAToRGB_C;
VP8LConvertBGRAToBGR = VP8LConvertBGRAToBGR_C;
#endif
VP8LConvertBGRAToRGBA4444 = VP8LConvertBGRAToRGBA4444_C;
VP8LConvertBGRAToRGB565 = VP8LConvertBGRAToRGB565_C;
VP8LMapColor32b = MapARGB_C;
VP8LMapColor8b = MapAlpha_C;
// If defined, use CPUInfo() to overwrite some pointers with faster versions.
if (VP8GetCPUInfo != NULL) {
#if defined(WEBP_USE_SSE2)
if (VP8GetCPUInfo(kSSE2)) {
VP8LDspInitSSE2();
}
#endif
#if defined(WEBP_USE_MIPS_DSP_R2)
if (VP8GetCPUInfo(kMIPSdspR2)) {
VP8LDspInitMIPSdspR2();
}
#endif
#if defined(WEBP_USE_MSA)
if (VP8GetCPUInfo(kMSA)) {
VP8LDspInitMSA();
}
#endif
}
#if defined(WEBP_USE_NEON)
if (WEBP_NEON_OMIT_C_CODE ||
(VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
VP8LDspInitNEON();
}
#endif
assert(VP8LAddGreenToBlueAndRed != NULL);
assert(VP8LTransformColorInverse != NULL);
assert(VP8LConvertBGRAToRGBA != NULL);
assert(VP8LConvertBGRAToRGB != NULL);
assert(VP8LConvertBGRAToBGR != NULL);
assert(VP8LConvertBGRAToRGBA4444 != NULL);
assert(VP8LConvertBGRAToRGB565 != NULL);
assert(VP8LMapColor32b != NULL);
assert(VP8LMapColor8b != NULL);
}
#undef COPY_PREDICTOR_ARRAY
//------------------------------------------------------------------------------

225
media/libwebp/src/dsp/lossless.h Normal file
Просмотреть файл

@ -0,0 +1,225 @@
// Copyright 2012 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Image transforms and color space conversion methods for lossless decoder.
//
// Authors: Vikas Arora (vikaas.arora@gmail.com)
// Jyrki Alakuijala (jyrki@google.com)
#ifndef WEBP_DSP_LOSSLESS_H_
#define WEBP_DSP_LOSSLESS_H_
#include "src/webp/types.h"
#include "src/webp/decode.h"
#include "src/enc/histogram_enc.h"
#include "src/utils/utils.h"
#ifdef __cplusplus
extern "C" {
#endif
//------------------------------------------------------------------------------
// Decoding
typedef uint32_t (*VP8LPredictorFunc)(uint32_t left, const uint32_t* const top);
extern VP8LPredictorFunc VP8LPredictors[16];
extern VP8LPredictorFunc VP8LPredictors_C[16];
// These Add/Sub function expects upper[-1] and out[-1] to be readable.
typedef void (*VP8LPredictorAddSubFunc)(const uint32_t* in,
const uint32_t* upper, int num_pixels,
uint32_t* out);
extern VP8LPredictorAddSubFunc VP8LPredictorsAdd[16];
extern VP8LPredictorAddSubFunc VP8LPredictorsAdd_C[16];
typedef void (*VP8LProcessDecBlueAndRedFunc)(const uint32_t* src,
int num_pixels, uint32_t* dst);
extern VP8LProcessDecBlueAndRedFunc VP8LAddGreenToBlueAndRed;
typedef struct {
// Note: the members are uint8_t, so that any negative values are
// automatically converted to "mod 256" values.
uint8_t green_to_red_;
uint8_t green_to_blue_;
uint8_t red_to_blue_;
} VP8LMultipliers;
typedef void (*VP8LTransformColorInverseFunc)(const VP8LMultipliers* const m,
const uint32_t* src,
int num_pixels, uint32_t* dst);
extern VP8LTransformColorInverseFunc VP8LTransformColorInverse;
struct VP8LTransform; // Defined in dec/vp8li.h.
// Performs inverse transform of data given transform information, start and end
// rows. Transform will be applied to rows [row_start, row_end[.
// The *in and *out pointers refer to source and destination data respectively
// corresponding to the intermediate row (row_start).
void VP8LInverseTransform(const struct VP8LTransform* const transform,
int row_start, int row_end,
const uint32_t* const in, uint32_t* const out);
// Color space conversion.
typedef void (*VP8LConvertFunc)(const uint32_t* src, int num_pixels,
uint8_t* dst);
extern VP8LConvertFunc VP8LConvertBGRAToRGB;
extern VP8LConvertFunc VP8LConvertBGRAToRGBA;
extern VP8LConvertFunc VP8LConvertBGRAToRGBA4444;
extern VP8LConvertFunc VP8LConvertBGRAToRGB565;
extern VP8LConvertFunc VP8LConvertBGRAToBGR;
// Converts from BGRA to other color spaces.
void VP8LConvertFromBGRA(const uint32_t* const in_data, int num_pixels,
WEBP_CSP_MODE out_colorspace, uint8_t* const rgba);
typedef void (*VP8LMapARGBFunc)(const uint32_t* src,
const uint32_t* const color_map,
uint32_t* dst, int y_start,
int y_end, int width);
typedef void (*VP8LMapAlphaFunc)(const uint8_t* src,
const uint32_t* const color_map,
uint8_t* dst, int y_start,
int y_end, int width);
extern VP8LMapARGBFunc VP8LMapColor32b;
extern VP8LMapAlphaFunc VP8LMapColor8b;
// Similar to the static method ColorIndexInverseTransform() that is part of
// lossless.c, but used only for alpha decoding. It takes uint8_t (rather than
// uint32_t) arguments for 'src' and 'dst'.
void VP8LColorIndexInverseTransformAlpha(
const struct VP8LTransform* const transform, int y_start, int y_end,
const uint8_t* src, uint8_t* dst);
// Expose some C-only fallback functions
void VP8LTransformColorInverse_C(const VP8LMultipliers* const m,
const uint32_t* src, int num_pixels,
uint32_t* dst);
void VP8LConvertBGRAToRGB_C(const uint32_t* src, int num_pixels, uint8_t* dst);
void VP8LConvertBGRAToRGBA_C(const uint32_t* src, int num_pixels, uint8_t* dst);
void VP8LConvertBGRAToRGBA4444_C(const uint32_t* src,
int num_pixels, uint8_t* dst);
void VP8LConvertBGRAToRGB565_C(const uint32_t* src,
int num_pixels, uint8_t* dst);
void VP8LConvertBGRAToBGR_C(const uint32_t* src, int num_pixels, uint8_t* dst);
void VP8LAddGreenToBlueAndRed_C(const uint32_t* src, int num_pixels,
uint32_t* dst);
// Must be called before calling any of the above methods.
void VP8LDspInit(void);
//------------------------------------------------------------------------------
// Encoding
typedef void (*VP8LProcessEncBlueAndRedFunc)(uint32_t* dst, int num_pixels);
extern VP8LProcessEncBlueAndRedFunc VP8LSubtractGreenFromBlueAndRed;
typedef void (*VP8LTransformColorFunc)(const VP8LMultipliers* const m,
uint32_t* dst, int num_pixels);
extern VP8LTransformColorFunc VP8LTransformColor;
typedef void (*VP8LCollectColorBlueTransformsFunc)(
const uint32_t* argb, int stride,
int tile_width, int tile_height,
int green_to_blue, int red_to_blue, int histo[]);
extern VP8LCollectColorBlueTransformsFunc VP8LCollectColorBlueTransforms;
typedef void (*VP8LCollectColorRedTransformsFunc)(
const uint32_t* argb, int stride,
int tile_width, int tile_height,
int green_to_red, int histo[]);
extern VP8LCollectColorRedTransformsFunc VP8LCollectColorRedTransforms;
// Expose some C-only fallback functions
void VP8LTransformColor_C(const VP8LMultipliers* const m,
uint32_t* data, int num_pixels);
void VP8LSubtractGreenFromBlueAndRed_C(uint32_t* argb_data, int num_pixels);
void VP8LCollectColorRedTransforms_C(const uint32_t* argb, int stride,
int tile_width, int tile_height,
int green_to_red, int histo[]);
void VP8LCollectColorBlueTransforms_C(const uint32_t* argb, int stride,
int tile_width, int tile_height,
int green_to_blue, int red_to_blue,
int histo[]);
extern VP8LPredictorAddSubFunc VP8LPredictorsSub[16];
extern VP8LPredictorAddSubFunc VP8LPredictorsSub_C[16];
// -----------------------------------------------------------------------------
// Huffman-cost related functions.
typedef double (*VP8LCostFunc)(const uint32_t* population, int length);
typedef double (*VP8LCostCombinedFunc)(const uint32_t* X, const uint32_t* Y,
int length);
typedef float (*VP8LCombinedShannonEntropyFunc)(const int X[256],
const int Y[256]);
extern VP8LCostFunc VP8LExtraCost;
extern VP8LCostCombinedFunc VP8LExtraCostCombined;
extern VP8LCombinedShannonEntropyFunc VP8LCombinedShannonEntropy;
typedef struct { // small struct to hold counters
int counts[2]; // index: 0=zero steak, 1=non-zero streak
int streaks[2][2]; // [zero/non-zero][streak<3 / streak>=3]
} VP8LStreaks;
typedef struct { // small struct to hold bit entropy results
double entropy; // entropy
uint32_t sum; // sum of the population
int nonzeros; // number of non-zero elements in the population
uint32_t max_val; // maximum value in the population
uint32_t nonzero_code; // index of the last non-zero in the population
} VP8LBitEntropy;
void VP8LBitEntropyInit(VP8LBitEntropy* const entropy);
// Get the combined symbol bit entropy and Huffman cost stats for the
// distributions 'X' and 'Y'. Those results can then be refined according to
// codec specific heuristics.
typedef void (*VP8LGetCombinedEntropyUnrefinedFunc)(
const uint32_t X[], const uint32_t Y[], int length,
VP8LBitEntropy* const bit_entropy, VP8LStreaks* const stats);
extern VP8LGetCombinedEntropyUnrefinedFunc VP8LGetCombinedEntropyUnrefined;
// Get the entropy for the distribution 'X'.
typedef void (*VP8LGetEntropyUnrefinedFunc)(const uint32_t X[], int length,
VP8LBitEntropy* const bit_entropy,
VP8LStreaks* const stats);
extern VP8LGetEntropyUnrefinedFunc VP8LGetEntropyUnrefined;
void VP8LBitsEntropyUnrefined(const uint32_t* const array, int n,
VP8LBitEntropy* const entropy);
typedef void (*VP8LHistogramAddFunc)(const VP8LHistogram* const a,
const VP8LHistogram* const b,
VP8LHistogram* const out);
extern VP8LHistogramAddFunc VP8LHistogramAdd;
// -----------------------------------------------------------------------------
// PrefixEncode()
typedef int (*VP8LVectorMismatchFunc)(const uint32_t* const array1,
const uint32_t* const array2, int length);
// Returns the first index where array1 and array2 are different.
extern VP8LVectorMismatchFunc VP8LVectorMismatch;
typedef void (*VP8LBundleColorMapFunc)(const uint8_t* const row, int width,
int xbits, uint32_t* dst);
extern VP8LBundleColorMapFunc VP8LBundleColorMap;
void VP8LBundleColorMap_C(const uint8_t* const row, int width, int xbits,
uint32_t* dst);
// Must be called before calling any of the above methods.
void VP8LEncDspInit(void);
//------------------------------------------------------------------------------
#ifdef __cplusplus
} // extern "C"
#endif
#endif // WEBP_DSP_LOSSLESS_H_

202
media/libwebp/src/dsp/lossless_common.h Normal file
Просмотреть файл

@ -0,0 +1,202 @@
// Copyright 2012 Google Inc. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the COPYING file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
// -----------------------------------------------------------------------------
//
// Image transforms and color space conversion methods for lossless decoder.
//
// Authors: Vikas Arora (vikaas.arora@gmail.com)
// Jyrki Alakuijala (jyrki@google.com)
// Vincent Rabaud (vrabaud@google.com)
#ifndef WEBP_DSP_LOSSLESS_COMMON_H_
#define WEBP_DSP_LOSSLESS_COMMON_H_
#include "src/webp/types.h"
#include "src/utils/utils.h"
#ifdef __cplusplus
extern "C" {
#endif
//------------------------------------------------------------------------------
// Decoding
// color mapping related functions.
static WEBP_INLINE uint32_t VP8GetARGBIndex(uint32_t idx) {
return (idx >> 8) & 0xff;
}
static WEBP_INLINE uint8_t VP8GetAlphaIndex(uint8_t idx) {
return idx;
}
static WEBP_INLINE uint32_t VP8GetARGBValue(uint32_t val) {
return val;
}
static WEBP_INLINE uint8_t VP8GetAlphaValue(uint32_t val) {
return (val >> 8) & 0xff;
}
//------------------------------------------------------------------------------
// Misc methods.
// Computes sampled size of 'size' when sampling using 'sampling bits'.
static WEBP_INLINE uint32_t VP8LSubSampleSize(uint32_t size,
uint32_t sampling_bits) {
return (size + (1 << sampling_bits) - 1) >> sampling_bits;
}
// Converts near lossless quality into max number of bits shaved off.
static WEBP_INLINE int VP8LNearLosslessBits(int near_lossless_quality) {
// 100 -> 0
// 80..99 -> 1
// 60..79 -> 2
// 40..59 -> 3
// 20..39 -> 4
// 0..19 -> 5
return 5 - near_lossless_quality / 20;
}
// -----------------------------------------------------------------------------
// Faster logarithm for integers. Small values use a look-up table.
// The threshold till approximate version of log_2 can be used.
// Practically, we can get rid of the call to log() as the two values match to
// very high degree (the ratio of these two is 0.99999x).
// Keeping a high threshold for now.
#define APPROX_LOG_WITH_CORRECTION_MAX 65536
#define APPROX_LOG_MAX 4096
#define LOG_2_RECIPROCAL 1.44269504088896338700465094007086
#define LOG_LOOKUP_IDX_MAX 256
extern const float kLog2Table[LOG_LOOKUP_IDX_MAX];
extern const float kSLog2Table[LOG_LOOKUP_IDX_MAX];
typedef float (*VP8LFastLog2SlowFunc)(uint32_t v);
extern VP8LFastLog2SlowFunc VP8LFastLog2Slow;
extern VP8LFastLog2SlowFunc VP8LFastSLog2Slow;
static WEBP_INLINE float VP8LFastLog2(uint32_t v) {
return (v < LOG_LOOKUP_IDX_MAX) ? kLog2Table[v] : VP8LFastLog2Slow(v);
}
// Fast calculation of v * log2(v) for integer input.
static WEBP_INLINE float VP8LFastSLog2(uint32_t v) {
return (v < LOG_LOOKUP_IDX_MAX) ? kSLog2Table[v] : VP8LFastSLog2Slow(v);
}
// -----------------------------------------------------------------------------
// PrefixEncode()
// Splitting of distance and length codes into prefixes and
// extra bits. The prefixes are encoded with an entropy code
// while the extra bits are stored just as normal bits.
static WEBP_INLINE void VP8LPrefixEncodeBitsNoLUT(int distance, int* const code,
int* const extra_bits) {
const int highest_bit = BitsLog2Floor(--distance);
const int second_highest_bit = (distance >> (highest_bit - 1)) & 1;
*extra_bits = highest_bit - 1;
*code = 2 * highest_bit + second_highest_bit;
}
static WEBP_INLINE void VP8LPrefixEncodeNoLUT(int distance, int* const code,
int* const extra_bits,
int* const extra_bits_value) {
const int highest_bit = BitsLog2Floor(--distance);
const int second_highest_bit = (distance >> (highest_bit - 1)) & 1;
*extra_bits = highest_bit - 1;
*extra_bits_value = distance & ((1 << *extra_bits) - 1);
*code = 2 * highest_bit + second_highest_bit;
}
#define PREFIX_LOOKUP_IDX_MAX 512
typedef struct {
int8_t code_;
int8_t extra_bits_;
} VP8LPrefixCode;
// These tables are derived using VP8LPrefixEncodeNoLUT.
extern const VP8LPrefixCode kPrefixEncodeCode[PREFIX_LOOKUP_IDX_MAX];
extern const uint8_t kPrefixEncodeExtraBitsValue[PREFIX_LOOKUP_IDX_MAX];
static WEBP_INLINE void VP8LPrefixEncodeBits(int distance, int* const code,
int* const extra_bits) {
if (distance < PREFIX_LOOKUP_IDX_MAX) {
const VP8LPrefixCode prefix_code = kPrefixEncodeCode[distance];
*code = prefix_code.code_;
*extra_bits = prefix_code.extra_bits_;
} else {
VP8LPrefixEncodeBitsNoLUT(distance, code, extra_bits);
}
}
static WEBP_INLINE void VP8LPrefixEncode(int distance, int* const code,
int* const extra_bits,
int* const extra_bits_value) {
if (distance < PREFIX_LOOKUP_IDX_MAX) {
const VP8LPrefixCode prefix_code = kPrefixEncodeCode[distance];
*code = prefix_code.code_;
*extra_bits = prefix_code.extra_bits_;
*extra_bits_value = kPrefixEncodeExtraBitsValue[distance];
} else {
VP8LPrefixEncodeNoLUT(distance, code, extra_bits, extra_bits_value);
}
}
// Sum of each component, mod 256.
static WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW WEBP_INLINE
uint32_t VP8LAddPixels(uint32_t a, uint32_t b) {
const uint32_t alpha_and_green = (a & 0xff00ff00u) + (b & 0xff00ff00u);
const uint32_t red_and_blue = (a & 0x00ff00ffu) + (b & 0x00ff00ffu);
return (alpha_and_green & 0xff00ff00u) | (red_and_blue & 0x00ff00ffu);
}
// Difference of each component, mod 256.
static WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW WEBP_INLINE
uint32_t VP8LSubPixels(uint32_t a, uint32_t b) {
const uint32_t alpha_and_green =
0x00ff00ffu + (a & 0xff00ff00u) - (b & 0xff00ff00u);
const uint32_t red_and_blue =
0xff00ff00u + (a & 0x00ff00ffu) - (b & 0x00ff00ffu);
return (alpha_and_green & 0xff00ff00u) | (red_and_blue & 0x00ff00ffu);
}
//------------------------------------------------------------------------------
// Transform-related functions use din both encoding and decoding.
// Macros used to create a batch predictor that iteratively uses a
// one-pixel predictor.
// The predictor is added to the output pixel (which
// is therefore considered as a residual) to get the final prediction.
#define GENERATE_PREDICTOR_ADD(PREDICTOR, PREDICTOR_ADD) \
static void PREDICTOR_ADD(const uint32_t* in, const uint32_t* upper, \
int num_pixels, uint32_t* out) { \
int x; \
for (x = 0; x < num_pixels; ++x) { \
const uint32_t pred = (PREDICTOR)(out[x - 1], upper + x); \
out[x] = VP8LAddPixels(in[x], pred); \
} \
}
// It subtracts the prediction from the input pixel and stores the residual
// in the output pixel.
#define GENERATE_PREDICTOR_SUB(PREDICTOR, PREDICTOR_SUB) \
static void PREDICTOR_SUB(const uint32_t* in, const uint32_t* upper, \
int num_pixels, uint32_t* out) { \
int x; \
for (x = 0; x < num_pixels; ++x) { \
const uint32_t pred = (PREDICTOR)(in[x - 1], upper + x); \
out[x] = VP8LSubPixels(in[x], pred); \
} \
}
#ifdef __cplusplus
} // extern "C"
#endif
#endif // WEBP_DSP_LOSSLESS_COMMON_H_

Некоторые файлы не были показаны из-за слишком большого количества измененных файлов Показать больше