зеркало из https://github.com/mozilla/gecko-dev.git
461 строка
15 KiB
C++
461 строка
15 KiB
C++
/* -*- Mode: c++; c-basic-offset: 2; indent-tabs-mode: nil; tab-width: 4; -*- */
|
|
/* 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 "GLTextureImage.h"
|
|
#include "GLContext.h"
|
|
#include "gfxContext.h"
|
|
#include "gfxPlatform.h"
|
|
#include "gfxUtils.h"
|
|
#include "gfx2DGlue.h"
|
|
#include "mozilla/gfx/2D.h"
|
|
#include "ScopedGLHelpers.h"
|
|
#include "GLUploadHelpers.h"
|
|
#include "GfxTexturesReporter.h"
|
|
|
|
#include "TextureImageEGL.h"
|
|
|
|
using namespace mozilla::gfx;
|
|
|
|
namespace mozilla {
|
|
namespace gl {
|
|
|
|
already_AddRefed<TextureImage> CreateTextureImage(
|
|
GLContext* gl, const gfx::IntSize& aSize,
|
|
TextureImage::ContentType aContentType, GLenum aWrapMode,
|
|
TextureImage::Flags aFlags, TextureImage::ImageFormat aImageFormat) {
|
|
switch (gl->GetContextType()) {
|
|
case GLContextType::EGL:
|
|
return CreateTextureImageEGL(gl, aSize, aContentType, aWrapMode, aFlags,
|
|
aImageFormat);
|
|
default: {
|
|
GLint maxTextureSize;
|
|
gl->fGetIntegerv(LOCAL_GL_MAX_TEXTURE_SIZE, &maxTextureSize);
|
|
if (aSize.width > maxTextureSize || aSize.height > maxTextureSize) {
|
|
NS_ASSERTION(aWrapMode == LOCAL_GL_CLAMP_TO_EDGE,
|
|
"Can't support wrapping with tiles!");
|
|
return CreateTiledTextureImage(gl, aSize, aContentType, aFlags,
|
|
aImageFormat);
|
|
} else {
|
|
return CreateBasicTextureImage(gl, aSize, aContentType, aWrapMode,
|
|
aFlags);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
static already_AddRefed<TextureImage> TileGenFunc(
|
|
GLContext* gl, const IntSize& aSize, TextureImage::ContentType aContentType,
|
|
TextureImage::Flags aFlags, TextureImage::ImageFormat aImageFormat) {
|
|
switch (gl->GetContextType()) {
|
|
case GLContextType::EGL:
|
|
return TileGenFuncEGL(gl, aSize, aContentType, aFlags, aImageFormat);
|
|
default:
|
|
return CreateBasicTextureImage(gl, aSize, aContentType,
|
|
LOCAL_GL_CLAMP_TO_EDGE, aFlags);
|
|
}
|
|
}
|
|
|
|
already_AddRefed<TextureImage> TextureImage::Create(
|
|
GLContext* gl, const gfx::IntSize& size,
|
|
TextureImage::ContentType contentType, GLenum wrapMode,
|
|
TextureImage::Flags flags) {
|
|
return CreateTextureImage(gl, size, contentType, wrapMode, flags);
|
|
}
|
|
|
|
bool TextureImage::UpdateFromDataSource(gfx::DataSourceSurface* aSurface,
|
|
const nsIntRegion* aDestRegion,
|
|
const gfx::IntPoint* aSrcPoint) {
|
|
nsIntRegion destRegion = aDestRegion
|
|
? *aDestRegion
|
|
: IntRect(0, 0, aSurface->GetSize().width,
|
|
aSurface->GetSize().height);
|
|
gfx::IntPoint srcPoint = aSrcPoint ? *aSrcPoint : gfx::IntPoint(0, 0);
|
|
return DirectUpdate(aSurface, destRegion, srcPoint);
|
|
}
|
|
|
|
gfx::IntRect TextureImage::GetTileRect() {
|
|
return gfx::IntRect(gfx::IntPoint(0, 0), mSize);
|
|
}
|
|
|
|
gfx::IntRect TextureImage::GetSrcTileRect() { return GetTileRect(); }
|
|
|
|
void TextureImage::UpdateUploadSize(size_t amount) {
|
|
if (mUploadSize > 0) {
|
|
GfxTexturesReporter::UpdateAmount(GfxTexturesReporter::MemoryFreed,
|
|
mUploadSize);
|
|
}
|
|
mUploadSize = amount;
|
|
GfxTexturesReporter::UpdateAmount(GfxTexturesReporter::MemoryAllocated,
|
|
mUploadSize);
|
|
}
|
|
|
|
BasicTextureImage::~BasicTextureImage() {
|
|
GLContext* ctx = mGLContext;
|
|
if (ctx->IsDestroyed() || !ctx->IsOwningThreadCurrent()) {
|
|
ctx = ctx->GetSharedContext();
|
|
}
|
|
|
|
// If we have a context, then we need to delete the texture;
|
|
// if we don't have a context (either real or shared),
|
|
// then they went away when the contex was deleted, because it
|
|
// was the only one that had access to it.
|
|
if (ctx && ctx->MakeCurrent()) {
|
|
ctx->fDeleteTextures(1, &mTexture);
|
|
}
|
|
}
|
|
|
|
void BasicTextureImage::BindTexture(GLenum aTextureUnit) {
|
|
mGLContext->fActiveTexture(aTextureUnit);
|
|
mGLContext->fBindTexture(LOCAL_GL_TEXTURE_2D, mTexture);
|
|
mGLContext->fActiveTexture(LOCAL_GL_TEXTURE0);
|
|
}
|
|
|
|
bool BasicTextureImage::DirectUpdate(
|
|
gfx::DataSourceSurface* aSurf, const nsIntRegion& aRegion,
|
|
const gfx::IntPoint& aFrom /* = gfx::IntPoint(0, 0) */) {
|
|
nsIntRegion region;
|
|
if (mTextureState == Valid) {
|
|
region = aRegion;
|
|
} else {
|
|
region = nsIntRegion(IntRect(0, 0, mSize.width, mSize.height));
|
|
}
|
|
bool needInit = mTextureState == Created;
|
|
size_t uploadSize;
|
|
|
|
mTextureFormat = UploadSurfaceToTexture(mGLContext, aSurf, region, mTexture,
|
|
mSize, &uploadSize, needInit, aFrom);
|
|
if (mTextureFormat == SurfaceFormat::UNKNOWN) {
|
|
return false;
|
|
}
|
|
|
|
if (uploadSize > 0) {
|
|
UpdateUploadSize(uploadSize);
|
|
}
|
|
mTextureState = Valid;
|
|
return true;
|
|
}
|
|
|
|
void BasicTextureImage::Resize(const gfx::IntSize& aSize) {
|
|
mGLContext->fBindTexture(LOCAL_GL_TEXTURE_2D, mTexture);
|
|
|
|
// This matches the logic in UploadImageDataToTexture so that
|
|
// we avoid mixing formats.
|
|
GLenum format;
|
|
GLenum type;
|
|
if (mGLContext->GetPreferredARGB32Format() == LOCAL_GL_BGRA) {
|
|
MOZ_ASSERT(!mGLContext->IsGLES());
|
|
format = LOCAL_GL_BGRA;
|
|
type = LOCAL_GL_UNSIGNED_INT_8_8_8_8_REV;
|
|
} else {
|
|
format = LOCAL_GL_RGBA;
|
|
type = LOCAL_GL_UNSIGNED_BYTE;
|
|
}
|
|
|
|
mGLContext->fTexImage2D(LOCAL_GL_TEXTURE_2D, 0, LOCAL_GL_RGBA, aSize.width,
|
|
aSize.height, 0, format, type, nullptr);
|
|
|
|
mTextureState = Allocated;
|
|
mSize = aSize;
|
|
}
|
|
|
|
gfx::IntSize TextureImage::GetSize() const { return mSize; }
|
|
|
|
TextureImage::TextureImage(const gfx::IntSize& aSize, GLenum aWrapMode,
|
|
ContentType aContentType, Flags aFlags)
|
|
: mSize(aSize),
|
|
mWrapMode(aWrapMode),
|
|
mContentType(aContentType),
|
|
mTextureFormat(gfx::SurfaceFormat::UNKNOWN),
|
|
mSamplingFilter(SamplingFilter::GOOD),
|
|
mFlags(aFlags),
|
|
mUploadSize(0) {}
|
|
|
|
BasicTextureImage::BasicTextureImage(GLuint aTexture, const gfx::IntSize& aSize,
|
|
GLenum aWrapMode, ContentType aContentType,
|
|
GLContext* aContext,
|
|
TextureImage::Flags aFlags)
|
|
: TextureImage(aSize, aWrapMode, aContentType, aFlags),
|
|
mTexture(aTexture),
|
|
mTextureState(Created),
|
|
mGLContext(aContext) {}
|
|
|
|
static bool WantsSmallTiles(GLContext* gl) {
|
|
// We can't use small tiles on the SGX 540, because of races in texture
|
|
// upload.
|
|
if (gl->WorkAroundDriverBugs() && gl->Renderer() == GLRenderer::SGX540)
|
|
return false;
|
|
|
|
// We should use small tiles for good performance if we can't use
|
|
// glTexSubImage2D() for some reason.
|
|
if (!CanUploadSubTextures(gl)) return true;
|
|
|
|
// Don't use small tiles otherwise. (If we implement incremental texture
|
|
// upload, then we will want to revisit this.)
|
|
return false;
|
|
}
|
|
|
|
TiledTextureImage::TiledTextureImage(GLContext* aGL, gfx::IntSize aSize,
|
|
TextureImage::ContentType aContentType,
|
|
TextureImage::Flags aFlags,
|
|
TextureImage::ImageFormat aImageFormat)
|
|
: TextureImage(aSize, LOCAL_GL_CLAMP_TO_EDGE, aContentType, aFlags),
|
|
mCurrentImage(0),
|
|
mIterationCallback(nullptr),
|
|
mIterationCallbackData(nullptr),
|
|
mTileSize(0),
|
|
mRows(0),
|
|
mColumns(0),
|
|
mGL(aGL),
|
|
mTextureState(Created),
|
|
mImageFormat(aImageFormat) {
|
|
if (!(aFlags & TextureImage::DisallowBigImage) && WantsSmallTiles(mGL)) {
|
|
mTileSize = 256;
|
|
} else {
|
|
mGL->fGetIntegerv(LOCAL_GL_MAX_TEXTURE_SIZE, (GLint*)&mTileSize);
|
|
}
|
|
if (aSize.width != 0 && aSize.height != 0) {
|
|
Resize(aSize);
|
|
}
|
|
}
|
|
|
|
TiledTextureImage::~TiledTextureImage() = default;
|
|
|
|
bool TiledTextureImage::DirectUpdate(
|
|
gfx::DataSourceSurface* aSurf, const nsIntRegion& aRegion,
|
|
const gfx::IntPoint& aFrom /* = gfx::IntPoint(0, 0) */) {
|
|
if (mSize.width == 0 || mSize.height == 0) {
|
|
return true;
|
|
}
|
|
|
|
nsIntRegion region;
|
|
|
|
if (mTextureState != Valid) {
|
|
IntRect bounds = IntRect(0, 0, mSize.width, mSize.height);
|
|
region = nsIntRegion(bounds);
|
|
} else {
|
|
region = aRegion;
|
|
}
|
|
|
|
bool result = true;
|
|
int oldCurrentImage = mCurrentImage;
|
|
BeginBigImageIteration();
|
|
do {
|
|
IntRect tileRect = GetSrcTileRect();
|
|
int xPos = tileRect.X();
|
|
int yPos = tileRect.Y();
|
|
|
|
nsIntRegion tileRegion;
|
|
tileRegion.And(region, tileRect); // intersect with tile
|
|
|
|
if (tileRegion.IsEmpty()) continue;
|
|
|
|
tileRegion.MoveBy(-xPos, -yPos); // translate into tile local space
|
|
|
|
result &= mImages[mCurrentImage]->DirectUpdate(
|
|
aSurf, tileRegion, aFrom + gfx::IntPoint(xPos, yPos));
|
|
|
|
if (mCurrentImage == mImages.Length() - 1) {
|
|
// We know we're done, but we still need to ensure that the callback
|
|
// gets called (e.g. to update the uploaded region).
|
|
NextTile();
|
|
break;
|
|
}
|
|
// Override a callback cancelling iteration if the texture wasn't valid.
|
|
// We need to force the update in that situation, or we may end up
|
|
// showing invalid/out-of-date texture data.
|
|
} while (NextTile() || (mTextureState != Valid));
|
|
mCurrentImage = oldCurrentImage;
|
|
|
|
mTextureFormat = mImages[0]->GetTextureFormat();
|
|
mTextureState = Valid;
|
|
return result;
|
|
}
|
|
|
|
void TiledTextureImage::BeginBigImageIteration() { mCurrentImage = 0; }
|
|
|
|
bool TiledTextureImage::NextTile() {
|
|
bool continueIteration = true;
|
|
|
|
if (mIterationCallback)
|
|
continueIteration =
|
|
mIterationCallback(this, mCurrentImage, mIterationCallbackData);
|
|
|
|
if (mCurrentImage + 1 < mImages.Length()) {
|
|
mCurrentImage++;
|
|
return continueIteration;
|
|
}
|
|
return false;
|
|
}
|
|
|
|
void TiledTextureImage::SetIterationCallback(
|
|
BigImageIterationCallback aCallback, void* aCallbackData) {
|
|
mIterationCallback = aCallback;
|
|
mIterationCallbackData = aCallbackData;
|
|
}
|
|
|
|
gfx::IntRect TiledTextureImage::GetTileRect() {
|
|
if (!GetTileCount()) {
|
|
return gfx::IntRect();
|
|
}
|
|
gfx::IntRect rect = mImages[mCurrentImage]->GetTileRect();
|
|
unsigned int xPos = (mCurrentImage % mColumns) * mTileSize;
|
|
unsigned int yPos = (mCurrentImage / mColumns) * mTileSize;
|
|
rect.MoveBy(xPos, yPos);
|
|
return rect;
|
|
}
|
|
|
|
gfx::IntRect TiledTextureImage::GetSrcTileRect() {
|
|
gfx::IntRect rect = GetTileRect();
|
|
const bool needsYFlip = mFlags & OriginBottomLeft;
|
|
unsigned int srcY =
|
|
needsYFlip ? mSize.height - rect.Height() - rect.Y() : rect.Y();
|
|
return gfx::IntRect(rect.X(), srcY, rect.Width(), rect.Height());
|
|
}
|
|
|
|
void TiledTextureImage::BindTexture(GLenum aTextureUnit) {
|
|
if (!GetTileCount()) {
|
|
return;
|
|
}
|
|
mImages[mCurrentImage]->BindTexture(aTextureUnit);
|
|
}
|
|
|
|
/*
|
|
* Resize, trying to reuse tiles. The reuse strategy is to decide on reuse per
|
|
* column. A tile on a column is reused if it hasn't changed size, otherwise it
|
|
* is discarded/replaced. Extra tiles on a column are pruned after iterating
|
|
* each column, and extra rows are pruned after iteration over the entire image
|
|
* finishes.
|
|
*/
|
|
void TiledTextureImage::Resize(const gfx::IntSize& aSize) {
|
|
if (mSize == aSize && mTextureState != Created) {
|
|
return;
|
|
}
|
|
|
|
// calculate rows and columns, rounding up
|
|
unsigned int columns = (aSize.width + mTileSize - 1) / mTileSize;
|
|
unsigned int rows = (aSize.height + mTileSize - 1) / mTileSize;
|
|
|
|
// Iterate over old tile-store and insert/remove tiles as necessary
|
|
int row;
|
|
unsigned int i = 0;
|
|
for (row = 0; row < (int)rows; row++) {
|
|
// If we've gone beyond how many rows there were before, set mColumns to
|
|
// zero so that we only create new tiles.
|
|
if (row >= (int)mRows) mColumns = 0;
|
|
|
|
// Similarly, if we're on the last row of old tiles and the height has
|
|
// changed, discard all tiles in that row.
|
|
// This will cause the pruning of columns not to work, but we don't need
|
|
// to worry about that, as no more tiles will be reused past this point
|
|
// anyway.
|
|
if ((row == (int)mRows - 1) && (aSize.height != mSize.height)) mColumns = 0;
|
|
|
|
int col;
|
|
for (col = 0; col < (int)columns; col++) {
|
|
IntSize size( // use tilesize first, then the remainder
|
|
(col + 1) * mTileSize > (unsigned int)aSize.width
|
|
? aSize.width % mTileSize
|
|
: mTileSize,
|
|
(row + 1) * mTileSize > (unsigned int)aSize.height
|
|
? aSize.height % mTileSize
|
|
: mTileSize);
|
|
|
|
bool replace = false;
|
|
|
|
// Check if we can re-use old tiles.
|
|
if (col < (int)mColumns) {
|
|
// Reuse an existing tile. If the tile is an end-tile and the
|
|
// width differs, replace it instead.
|
|
if (mSize.width != aSize.width) {
|
|
if (col == (int)mColumns - 1) {
|
|
// Tile at the end of the old column, replace it with
|
|
// a new one.
|
|
replace = true;
|
|
} else if (col == (int)columns - 1) {
|
|
// Tile at the end of the new column, create a new one.
|
|
} else {
|
|
// Before the last column on both the old and new sizes,
|
|
// reuse existing tile.
|
|
i++;
|
|
continue;
|
|
}
|
|
} else {
|
|
// Width hasn't changed, reuse existing tile.
|
|
i++;
|
|
continue;
|
|
}
|
|
}
|
|
|
|
// Create a new tile.
|
|
RefPtr<TextureImage> teximg =
|
|
TileGenFunc(mGL, size, mContentType, mFlags, mImageFormat);
|
|
if (replace)
|
|
mImages.ReplaceElementAt(i, teximg);
|
|
else
|
|
mImages.InsertElementAt(i, teximg);
|
|
i++;
|
|
}
|
|
|
|
// Prune any unused tiles on the end of the column.
|
|
if (row < (int)mRows) {
|
|
for (col = (int)mColumns - col; col > 0; col--) {
|
|
mImages.RemoveElementAt(i);
|
|
}
|
|
}
|
|
}
|
|
|
|
// Prune any unused tiles at the end of the store.
|
|
mImages.RemoveLastElements(mImages.Length() - i);
|
|
|
|
// Reset tile-store properties.
|
|
mRows = rows;
|
|
mColumns = columns;
|
|
mSize = aSize;
|
|
mTextureState = Allocated;
|
|
mCurrentImage = 0;
|
|
}
|
|
|
|
uint32_t TiledTextureImage::GetTileCount() { return mImages.Length(); }
|
|
|
|
already_AddRefed<TextureImage> CreateTiledTextureImage(
|
|
GLContext* aGL, const gfx::IntSize& aSize,
|
|
TextureImage::ContentType aContentType, TextureImage::Flags aFlags,
|
|
TextureImage::ImageFormat aImageFormat) {
|
|
RefPtr<TextureImage> texImage =
|
|
static_cast<TextureImage*>(new gl::TiledTextureImage(
|
|
aGL, aSize, aContentType, aFlags, aImageFormat));
|
|
return texImage.forget();
|
|
}
|
|
|
|
already_AddRefed<TextureImage> CreateBasicTextureImage(
|
|
GLContext* aGL, const gfx::IntSize& aSize,
|
|
TextureImage::ContentType aContentType, GLenum aWrapMode,
|
|
TextureImage::Flags aFlags) {
|
|
bool useNearestFilter = aFlags & TextureImage::UseNearestFilter;
|
|
if (!aGL->MakeCurrent()) {
|
|
return nullptr;
|
|
}
|
|
|
|
GLuint texture = 0;
|
|
aGL->fGenTextures(1, &texture);
|
|
|
|
ScopedBindTexture bind(aGL, texture);
|
|
|
|
GLint texfilter = useNearestFilter ? LOCAL_GL_NEAREST : LOCAL_GL_LINEAR;
|
|
aGL->fTexParameteri(LOCAL_GL_TEXTURE_2D, LOCAL_GL_TEXTURE_MIN_FILTER,
|
|
texfilter);
|
|
aGL->fTexParameteri(LOCAL_GL_TEXTURE_2D, LOCAL_GL_TEXTURE_MAG_FILTER,
|
|
texfilter);
|
|
aGL->fTexParameteri(LOCAL_GL_TEXTURE_2D, LOCAL_GL_TEXTURE_WRAP_S, aWrapMode);
|
|
aGL->fTexParameteri(LOCAL_GL_TEXTURE_2D, LOCAL_GL_TEXTURE_WRAP_T, aWrapMode);
|
|
|
|
RefPtr<BasicTextureImage> texImage = new BasicTextureImage(
|
|
texture, aSize, aWrapMode, aContentType, aGL, aFlags);
|
|
return texImage.forget();
|
|
}
|
|
|
|
} // namespace gl
|
|
} // namespace mozilla
|