Bug 1691317 - Update libwebp to version 1.2.0. r=jrmuizel

Differential Revision: https://phabricator.services.mozilla.com/D104340
This commit is contained in:
Andrew Osmond 2021-02-08 16:59:10 +00:00
Родитель 2cbb8a03f1
Коммит 73c7bd09af
23 изменённых файлов: 213 добавлений и 139 удалений

Просмотреть файл

@ -6,6 +6,7 @@ Contributors:
- Christian Duvivier (cduvivier at google dot com)
- Djordje Pesut (djordje dot pesut at imgtec dot com)
- Hui Su (huisu at google dot com)
- Ingvar Stepanyan (rreverser at google dot com)
- James Zern (jzern at google dot com)
- Jan Engelhardt (jengelh at medozas dot de)
- Jehan (jehan at girinstud dot io)
@ -41,5 +42,6 @@ Contributors:
- Vikas Arora (vikasa at google dot com)
- Vincent Rabaud (vrabaud at google dot com)
- Vlad Tsyrklevich (vtsyrklevich at chromium dot org)
- Wan-Teh Chang (wtc at google dot com)
- Yang Zhang (yang dot zhang at arm dot com)
- Yannis Guyon (yguyon at google dot com)

Просмотреть файл

@ -1,5 +1,7 @@
Changes made to pristine libwebp source by mozilla.org developers.
2021/02/07 -- Synced with libwebp-1.2.0 (bug 1691317).
2020/02/26 -- Synced with libwebp-1.1.0 (bug 1618288).
2018/01/21 -- Synced with libwebp-1.0.2 (bug 1521478).

Просмотреть файл

@ -1,3 +1,15 @@
- 12/23/2020: version 1.2.0
* API changes:
- libwebp:
encode.h: add a qmin / qmax range for quality factor (cwebp adds -qrange)
* lossless encoder improvements
* SIMD support for Wasm builds
* add xcframeworkbuild.sh, supports Mac Catalyst builds
* import fuzzers from oss-fuzz & chromium (#409)
* webpmux: add an '-set loop <value>' option (#494)
* toolchain updates and bug fixes (#449, #463, #470, #475, #477, #478, #479,
#488, #491)
- 12/18/2019: version 1.1.0
* API changes:
- libwebp:

Просмотреть файл

@ -4,7 +4,7 @@
\__\__/\____/\_____/__/ ____ ___
/ _/ / \ \ / _ \/ _/
/ \_/ / / \ \ __/ \__
\____/____/\_____/_____/____/v1.1.0
\____/____/\_____/_____/____/v1.2.0
Description:
============
@ -113,7 +113,7 @@ make install
CMake:
------
With CMake, you can compile libwebp, cwebp, dwebp, gif2web, img2webp, webpinfo
With CMake, you can compile libwebp, cwebp, dwebp, gif2webp, img2webp, webpinfo
and the JS bindings.
Prerequisites:
@ -225,6 +225,7 @@ Usage:
If input size (-s) for an image is not specified, it is
assumed to be a PNG, JPEG, TIFF or WebP file.
Note: Animated PNG and WebP files are not supported.
Options:
-h / -help ............. short help
@ -254,6 +255,8 @@ Options:
-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)
-qrange <min> <max> .... specifies the permissible quality range
(default: 0 100)
-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
@ -294,6 +297,7 @@ Experimental Options:
-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
@ -341,7 +345,9 @@ 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]
Decodes the WebP image file to PNG format [Default].
Note: Animated WebP files are not supported.
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
@ -423,7 +429,7 @@ Prerequisites:
1) OpenGL & OpenGL Utility Toolkit (GLUT)
Linux:
$ sudo apt-get install freeglut3-dev mesa-common-dev
Mac + XCode:
Mac + Xcode:
- These libraries should be available in the OpenGL / GLUT frameworks.
Windows:
http://freeglut.sourceforge.net/index.php#download

Просмотреть файл

@ -1,7 +1,7 @@
 __ __ ____ ____ ____ __ __ _ __ __
/ \\/ \/ _ \/ _ \/ _ \/ \ \/ \___/_ / _\
\ / __/ _ \ __/ / / (_/ /__
\__\__/\_____/_____/__/ \__//_/\_____/__/___/v1.1.0
\__\__/\_____/_____/__/ \__//_/\_____/__/___/v1.2.0
Description:
@ -44,6 +44,7 @@ GET_OPTIONS:
SET_OPTIONS:
Set color profile/metadata:
loop LOOP_COUNT set the loop count
icc file.icc set ICC profile
exif file.exif set EXIF metadata
xmp file.xmp set XMP metadata

Просмотреть файл

@ -25,21 +25,16 @@
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;
uint8_t* const y_dst = buf->y + (size_t)io->mb_y * buf->y_stride;
uint8_t* const u_dst = buf->u + (size_t)(io->mb_y >> 1) * buf->u_stride;
uint8_t* const v_dst = buf->v + (size_t)(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);
}
WebPCopyPlane(io->y, io->y_stride, y_dst, buf->y_stride, mb_w, mb_h);
WebPCopyPlane(io->u, io->uv_stride, u_dst, buf->u_stride, uv_w, uv_h);
WebPCopyPlane(io->v, io->uv_stride, v_dst, buf->v_stride, uv_w, uv_h);
return io->mb_h;
}
@ -47,7 +42,7 @@ static int EmitYUV(const VP8Io* const io, WebPDecParams* const p) {
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;
uint8_t* const dst = buf->rgba + (size_t)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,
@ -62,7 +57,7 @@ static int EmitSampledRGB(const VP8Io* const io, WebPDecParams* const p) {
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;
uint8_t* dst = buf->rgba + (size_t)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;
@ -133,7 +128,7 @@ static int EmitAlphaYUV(const VP8Io* const io, WebPDecParams* const p,
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;
uint8_t* dst = buf->a + (size_t)io->mb_y * buf->a_stride;
int j;
(void)expected_num_lines_out;
assert(expected_num_lines_out == mb_h);
@ -186,7 +181,7 @@ static int EmitAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
(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);
const size_t 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,
@ -210,7 +205,7 @@ static int EmitAlphaRGBA4444(const VP8Io* const io, WebPDecParams* const p,
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);
const size_t 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;
@ -276,9 +271,9 @@ static int EmitRescaledYUV(const VP8Io* const io, WebPDecParams* const p) {
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;
uint8_t* const dst_a = buf->a + (size_t)p->last_y * buf->a_stride;
if (io->a != NULL) {
uint8_t* const dst_y = buf->y + p->last_y * buf->y_stride;
uint8_t* const dst_y = buf->y + (size_t)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
@ -356,7 +351,7 @@ 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;
uint8_t* dst = buf->rgba + (size_t)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.
@ -383,15 +378,15 @@ static int EmitRescaledRGB(const VP8Io* const io, WebPDecParams* const p) {
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);
io->y + (size_t)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);
const int u_lines_in = WebPRescalerImport(
p->scaler_u, uv_mb_h - uv_j, io->u + (size_t)uv_j * io->uv_stride,
io->uv_stride);
const int v_lines_in = WebPRescalerImport(
p->scaler_v, uv_mb_h - uv_j, io->v + (size_t)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;
@ -403,7 +398,7 @@ static int EmitRescaledRGB(const VP8Io* const io, WebPDecParams* const p) {
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;
uint8_t* const base_rgba = buf->rgba + (size_t)y_pos * buf->stride;
const WEBP_CSP_MODE colorspace = p->output->colorspace;
const int alpha_first =
(colorspace == MODE_ARGB || colorspace == MODE_Argb);
@ -431,7 +426,7 @@ static int ExportAlpha(WebPDecParams* const p, int y_pos, int max_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;
uint8_t* const base_rgba = buf->rgba + (size_t)y_pos * buf->stride;
#if (WEBP_SWAP_16BIT_CSP == 1)
uint8_t* alpha_dst = base_rgba;
#else
@ -470,7 +465,7 @@ static int EmitRescaledAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
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;
const int64_t row_offset = (int64_t)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);

Просмотреть файл

@ -494,13 +494,11 @@ static int GetCoeffsAlt(VP8BitReader* const br,
return 16;
}
static WEBP_TSAN_IGNORE_FUNCTION void InitGetCoeffs(void) {
if (GetCoeffs == NULL) {
if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
GetCoeffs = GetCoeffsAlt;
} else {
GetCoeffs = GetCoeffsFast;
}
WEBP_DSP_INIT_FUNC(InitGetCoeffs) {
if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
GetCoeffs = GetCoeffsAlt;
} else {
GetCoeffs = GetCoeffsFast;
}
}

Просмотреть файл

@ -31,7 +31,7 @@ extern "C" {
// version numbers
#define DEC_MAJ_VERSION 1
#define DEC_MIN_VERSION 1
#define DEC_MIN_VERSION 2
#define DEC_REV_VERSION 0
// YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).

Просмотреть файл

@ -947,7 +947,6 @@ static WEBP_INLINE void CopyBlock8b(uint8_t* const dst, int dist, int length) {
break;
default:
goto Copy;
break;
}
CopySmallPattern8b(src, dst, length, pattern);
return;

Просмотреть файл

@ -24,7 +24,7 @@
#include "src/webp/format_constants.h"
#define DMUX_MAJ_VERSION 1
#define DMUX_MIN_VERSION 1
#define DMUX_MIN_VERSION 2
#define DMUX_REV_VERSION 0
typedef struct {
@ -312,6 +312,7 @@ static ParseStatus ParseAnimationFrame(
int bits;
MemBuffer* const mem = &dmux->mem_;
Frame* frame;
size_t start_offset;
ParseStatus status =
NewFrame(mem, ANMF_CHUNK_SIZE, frame_chunk_size, &frame);
if (status != PARSE_OK) return status;
@ -332,7 +333,11 @@ static ParseStatus ParseAnimationFrame(
// Store a frame only if the animation flag is set there is some data for
// this frame is available.
start_offset = mem->start_;
status = StoreFrame(dmux->num_frames_ + 1, anmf_payload_size, mem, frame);
if (status != PARSE_ERROR && mem->start_ - start_offset > anmf_payload_size) {
status = PARSE_ERROR;
}
if (status != PARSE_ERROR && is_animation && frame->frame_num_ > 0) {
added_frame = AddFrame(dmux, frame);
if (added_frame) {

Просмотреть файл

@ -359,6 +359,11 @@ static int HasAlpha32b_C(const uint8_t* src, int length) {
return 0;
}
static void AlphaReplace_C(uint32_t* src, int length, uint32_t color) {
int x;
for (x = 0; x < length; ++x) if ((src[x] >> 24) == 0) src[x] = color;
}
//------------------------------------------------------------------------------
// Simple channel manipulations.
@ -400,6 +405,7 @@ void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
int (*WebPHasAlpha8b)(const uint8_t* src, int length);
int (*WebPHasAlpha32b)(const uint8_t* src, int length);
void (*WebPAlphaReplace)(uint32_t* src, int length, uint32_t color);
//------------------------------------------------------------------------------
// Init function
@ -428,6 +434,7 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
WebPHasAlpha8b = HasAlpha8b_C;
WebPHasAlpha32b = HasAlpha32b_C;
WebPAlphaReplace = AlphaReplace_C;
// If defined, use CPUInfo() to overwrite some pointers with faster versions.
if (VP8GetCPUInfo != NULL) {
@ -469,4 +476,5 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
assert(WebPPackRGB != NULL);
assert(WebPHasAlpha8b != NULL);
assert(WebPHasAlpha32b != NULL);
assert(WebPAlphaReplace != NULL);
}

Просмотреть файл

@ -265,6 +265,27 @@ static int HasAlpha32b_SSE2(const uint8_t* src, int length) {
return 0;
}
static void AlphaReplace_SSE2(uint32_t* src, int length, uint32_t color) {
const __m128i m_color = _mm_set1_epi32(color);
const __m128i zero = _mm_setzero_si128();
int i = 0;
for (; i + 8 <= length; i += 8) {
const __m128i a0 = _mm_loadu_si128((const __m128i*)(src + i + 0));
const __m128i a1 = _mm_loadu_si128((const __m128i*)(src + i + 4));
const __m128i b0 = _mm_srai_epi32(a0, 24);
const __m128i b1 = _mm_srai_epi32(a1, 24);
const __m128i c0 = _mm_cmpeq_epi32(b0, zero);
const __m128i c1 = _mm_cmpeq_epi32(b1, zero);
const __m128i d0 = _mm_and_si128(c0, m_color);
const __m128i d1 = _mm_and_si128(c1, m_color);
const __m128i e0 = _mm_andnot_si128(c0, a0);
const __m128i e1 = _mm_andnot_si128(c1, a1);
_mm_storeu_si128((__m128i*)(src + i + 0), _mm_or_si128(d0, e0));
_mm_storeu_si128((__m128i*)(src + i + 4), _mm_or_si128(d1, e1));
}
for (; i < length; ++i) if ((src[i] >> 24) == 0) src[i] = color;
}
// -----------------------------------------------------------------------------
// Apply alpha value to rows
@ -334,6 +355,7 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingSSE2(void) {
WebPHasAlpha8b = HasAlpha8b_SSE2;
WebPHasAlpha32b = HasAlpha32b_SSE2;
WebPAlphaReplace = AlphaReplace_SSE2;
}
#else // !WEBP_USE_SSE2

Просмотреть файл

@ -1283,12 +1283,12 @@ static void DC4_NEON(uint8_t* dst) { // DC
const uint8x8_t A = vld1_u8(dst - BPS); // top row
const uint16x4_t p0 = vpaddl_u8(A); // cascading summation of the top
const uint16x4_t p1 = vpadd_u16(p0, p0);
const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1));
const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1));
const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1));
const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1));
const uint16x8_t s0 = vaddq_u16(L0, L1);
const uint16x8_t s1 = vaddq_u16(L2, L3);
const uint8x8_t L0 = vld1_u8(dst + 0 * BPS - 1);
const uint8x8_t L1 = vld1_u8(dst + 1 * BPS - 1);
const uint8x8_t L2 = vld1_u8(dst + 2 * BPS - 1);
const uint8x8_t L3 = vld1_u8(dst + 3 * BPS - 1);
const uint16x8_t s0 = vaddl_u8(L0, L1);
const uint16x8_t s1 = vaddl_u8(L2, L3);
const uint16x8_t s01 = vaddq_u16(s0, s1);
const uint16x8_t sum = vaddq_u16(s01, vcombine_u16(p1, p1));
const uint8x8_t dc0 = vrshrn_n_u16(sum, 3); // (sum + 4) >> 3
@ -1429,8 +1429,7 @@ static WEBP_INLINE void DC8_NEON(uint8_t* dst, int do_top, int do_left) {
if (do_top) {
const uint8x8_t A = vld1_u8(dst - BPS); // top row
#if defined(__aarch64__)
const uint16x8_t B = vmovl_u8(A);
const uint16_t p2 = vaddvq_u16(B);
const uint16_t p2 = vaddlv_u8(A);
sum_top = vdupq_n_u16(p2);
#else
const uint16x4_t p0 = vpaddl_u8(A); // cascading summation of the top
@ -1441,18 +1440,18 @@ static WEBP_INLINE void DC8_NEON(uint8_t* dst, int do_top, int do_left) {
}
if (do_left) {
const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1));
const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1));
const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1));
const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1));
const uint16x8_t L4 = vmovl_u8(vld1_u8(dst + 4 * BPS - 1));
const uint16x8_t L5 = vmovl_u8(vld1_u8(dst + 5 * BPS - 1));
const uint16x8_t L6 = vmovl_u8(vld1_u8(dst + 6 * BPS - 1));
const uint16x8_t L7 = vmovl_u8(vld1_u8(dst + 7 * BPS - 1));
const uint16x8_t s0 = vaddq_u16(L0, L1);
const uint16x8_t s1 = vaddq_u16(L2, L3);
const uint16x8_t s2 = vaddq_u16(L4, L5);
const uint16x8_t s3 = vaddq_u16(L6, L7);
const uint8x8_t L0 = vld1_u8(dst + 0 * BPS - 1);
const uint8x8_t L1 = vld1_u8(dst + 1 * BPS - 1);
const uint8x8_t L2 = vld1_u8(dst + 2 * BPS - 1);
const uint8x8_t L3 = vld1_u8(dst + 3 * BPS - 1);
const uint8x8_t L4 = vld1_u8(dst + 4 * BPS - 1);
const uint8x8_t L5 = vld1_u8(dst + 5 * BPS - 1);
const uint8x8_t L6 = vld1_u8(dst + 6 * BPS - 1);
const uint8x8_t L7 = vld1_u8(dst + 7 * BPS - 1);
const uint16x8_t s0 = vaddl_u8(L0, L1);
const uint16x8_t s1 = vaddl_u8(L2, L3);
const uint16x8_t s2 = vaddl_u8(L4, L5);
const uint16x8_t s3 = vaddl_u8(L6, L7);
const uint16x8_t s01 = vaddq_u16(s0, s1);
const uint16x8_t s23 = vaddq_u16(s2, s3);
sum_left = vaddq_u16(s01, s23);
@ -1512,29 +1511,34 @@ static WEBP_INLINE void DC16_NEON(uint8_t* dst, int do_top, int do_left) {
if (do_top) {
const uint8x16_t A = vld1q_u8(dst - BPS); // top row
#if defined(__aarch64__)
const uint16_t p3 = vaddlvq_u8(A);
sum_top = vdupq_n_u16(p3);
#else
const uint16x8_t p0 = vpaddlq_u8(A); // cascading summation of the top
const uint16x4_t p1 = vadd_u16(vget_low_u16(p0), vget_high_u16(p0));
const uint16x4_t p2 = vpadd_u16(p1, p1);
const uint16x4_t p3 = vpadd_u16(p2, p2);
sum_top = vcombine_u16(p3, p3);
#endif
}
if (do_left) {
int i;
sum_left = vdupq_n_u16(0);
for (i = 0; i < 16; i += 8) {
const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + (i + 0) * BPS - 1));
const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + (i + 1) * BPS - 1));
const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + (i + 2) * BPS - 1));
const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + (i + 3) * BPS - 1));
const uint16x8_t L4 = vmovl_u8(vld1_u8(dst + (i + 4) * BPS - 1));
const uint16x8_t L5 = vmovl_u8(vld1_u8(dst + (i + 5) * BPS - 1));
const uint16x8_t L6 = vmovl_u8(vld1_u8(dst + (i + 6) * BPS - 1));
const uint16x8_t L7 = vmovl_u8(vld1_u8(dst + (i + 7) * BPS - 1));
const uint16x8_t s0 = vaddq_u16(L0, L1);
const uint16x8_t s1 = vaddq_u16(L2, L3);
const uint16x8_t s2 = vaddq_u16(L4, L5);
const uint16x8_t s3 = vaddq_u16(L6, L7);
const uint8x8_t L0 = vld1_u8(dst + (i + 0) * BPS - 1);
const uint8x8_t L1 = vld1_u8(dst + (i + 1) * BPS - 1);
const uint8x8_t L2 = vld1_u8(dst + (i + 2) * BPS - 1);
const uint8x8_t L3 = vld1_u8(dst + (i + 3) * BPS - 1);
const uint8x8_t L4 = vld1_u8(dst + (i + 4) * BPS - 1);
const uint8x8_t L5 = vld1_u8(dst + (i + 5) * BPS - 1);
const uint8x8_t L6 = vld1_u8(dst + (i + 6) * BPS - 1);
const uint8x8_t L7 = vld1_u8(dst + (i + 7) * BPS - 1);
const uint16x8_t s0 = vaddl_u8(L0, L1);
const uint16x8_t s1 = vaddl_u8(L2, L3);
const uint16x8_t s2 = vaddl_u8(L4, L5);
const uint16x8_t s3 = vaddl_u8(L6, L7);
const uint16x8_t s01 = vaddq_u16(s0, s1);
const uint16x8_t s23 = vaddq_u16(s2, s3);
const uint16x8_t sum = vaddq_u16(s01, s23);

Просмотреть файл

@ -51,9 +51,7 @@ extern "C" {
# define __has_builtin(x) 0
#endif
// for now, none of the optimizations below are available in emscripten
#if !defined(EMSCRIPTEN)
#if !defined(HAVE_CONFIG_H)
#if defined(_MSC_VER) && _MSC_VER > 1310 && \
(defined(_M_X64) || defined(_M_IX86))
#define WEBP_MSC_SSE2 // Visual C++ SSE2 targets
@ -63,6 +61,7 @@ extern "C" {
(defined(_M_X64) || defined(_M_IX86))
#define WEBP_MSC_SSE41 // Visual C++ SSE4.1 targets
#endif
#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.
@ -76,6 +75,9 @@ extern "C" {
#define WEBP_USE_SSE41
#endif
#undef WEBP_MSC_SSE41
#undef WEBP_MSC_SSE2
// 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__) || \
@ -110,8 +112,6 @@ extern "C" {
#define WEBP_USE_MSA
#endif
#endif /* EMSCRIPTEN */
#ifndef WEBP_DSP_OMIT_C_CODE
#define WEBP_DSP_OMIT_C_CODE 1
#endif
@ -193,6 +193,12 @@ extern "C" {
#endif
#endif
// If 'ptr' is NULL, returns NULL. Otherwise returns 'ptr + off'.
// Prevents undefined behavior sanitizer nullptr-with-nonzero-offset warning.
#if !defined(WEBP_OFFSET_PTR)
#define WEBP_OFFSET_PTR(ptr, off) (((ptr) == NULL) ? NULL : ((ptr) + (off)))
#endif
// Regularize the definition of WEBP_SWAP_16BIT_CSP (backward compatibility)
#if !defined(WEBP_SWAP_16BIT_CSP)
#define WEBP_SWAP_16BIT_CSP 0
@ -632,6 +638,8 @@ extern void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
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);
// replaces transparent values in src[] by 'color'.
extern void (*WebPAlphaReplace)(uint32_t* src, int length, uint32_t color);
// To be called first before using the above.
void WebPInitAlphaProcessing(void);

Просмотреть файл

@ -107,62 +107,62 @@ static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
//------------------------------------------------------------------------------
// Predictors
static uint32_t Predictor0_C(uint32_t left, const uint32_t* const top) {
uint32_t VP8LPredictor0_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) {
uint32_t VP8LPredictor1_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) {
uint32_t VP8LPredictor2_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) {
uint32_t VP8LPredictor3_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) {
uint32_t VP8LPredictor4_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) {
uint32_t VP8LPredictor5_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) {
uint32_t VP8LPredictor6_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) {
uint32_t VP8LPredictor7_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) {
uint32_t VP8LPredictor8_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) {
uint32_t VP8LPredictor9_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) {
uint32_t VP8LPredictor10_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) {
uint32_t VP8LPredictor11_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) {
uint32_t VP8LPredictor12_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) {
uint32_t VP8LPredictor13_C(uint32_t left, const uint32_t* const top) {
const uint32_t pred = ClampedAddSubtractHalf(left, top[0], top[-1]);
return pred;
}
@ -182,18 +182,18 @@ static void PredictorAdd1_C(const uint32_t* in, const uint32_t* upper,
out[i] = left = VP8LAddPixels(in[i], left);
}
}
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)
GENERATE_PREDICTOR_ADD(VP8LPredictor2_C, PredictorAdd2_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor3_C, PredictorAdd3_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor4_C, PredictorAdd4_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor5_C, PredictorAdd5_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor6_C, PredictorAdd6_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor7_C, PredictorAdd7_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor8_C, PredictorAdd8_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor9_C, PredictorAdd9_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor10_C, PredictorAdd10_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor11_C, PredictorAdd11_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor12_C, PredictorAdd12_C)
GENERATE_PREDICTOR_ADD(VP8LPredictor13_C, PredictorAdd13_C)
//------------------------------------------------------------------------------
@ -562,7 +562,6 @@ VP8LPredictorFunc VP8LPredictors[16];
// exposed plain-C implementations
VP8LPredictorAddSubFunc VP8LPredictorsAdd_C[16];
VP8LPredictorFunc VP8LPredictors_C[16];
VP8LTransformColorInverseFunc VP8LTransformColorInverse;
@ -600,8 +599,7 @@ extern void VP8LDspInitMSA(void);
} while (0);
WEBP_DSP_INIT_FUNC(VP8LDspInit) {
COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors)
COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors_C)
COPY_PREDICTOR_ARRAY(VP8LPredictor, VP8LPredictors)
COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd)
COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd_C)

Просмотреть файл

@ -30,7 +30,22 @@ extern "C" {
typedef uint32_t (*VP8LPredictorFunc)(uint32_t left, const uint32_t* const top);
extern VP8LPredictorFunc VP8LPredictors[16];
extern VP8LPredictorFunc VP8LPredictors_C[16];
uint32_t VP8LPredictor0_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor1_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor2_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor3_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor4_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor5_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor6_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor7_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor8_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor9_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor10_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor11_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor12_C(uint32_t left, const uint32_t* const top);
uint32_t VP8LPredictor13_C(uint32_t left, const uint32_t* const top);
// 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,

Просмотреть файл

@ -184,19 +184,6 @@ static void PREDICTOR_ADD(const uint32_t* in, const uint32_t* upper, \
} \
}
// 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; \
assert(upper != NULL); \
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

Просмотреть файл

@ -16,6 +16,7 @@
#include <assert.h>
#include <stdlib.h>
#include "src/webp/types.h"
#include "src/webp/encode.h"
#include "src/webp/format_constants.h"
#ifdef __cplusplus
@ -218,14 +219,19 @@ enum VP8LLZ77Type {
// Evaluates best possible backward references for specified quality.
// The input cache_bits to 'VP8LGetBackwardReferences' sets the maximum cache
// bits to use (passing 0 implies disabling the local color cache).
// The optimal cache bits is evaluated and set for the *cache_bits parameter.
// The return value is the pointer to the best of the two backward refs viz,
// refs[0] or refs[1].
VP8LBackwardRefs* VP8LGetBackwardReferences(
// The optimal cache bits is evaluated and set for the *cache_bits_best
// parameter with the matching refs_best.
// If do_no_cache == 0, refs is an array of 2 values and the best
// VP8LBackwardRefs is put in the first element.
// If do_no_cache != 0, refs is an array of 3 values and the best
// VP8LBackwardRefs is put in the first element, the best value with no-cache in
// the second element.
// In both cases, the last element is used as temporary internally.
WebPEncodingError VP8LGetBackwardReferences(
int width, int height, const uint32_t* const argb, int quality,
int low_effort, int lz77_types_to_try, int* const cache_bits,
const VP8LHashChain* const hash_chain, VP8LBackwardRefs* const refs_tmp1,
VP8LBackwardRefs* const refs_tmp2);
int low_effort, int lz77_types_to_try, int cache_bits_max, int do_no_cache,
const VP8LHashChain* const hash_chain, VP8LBackwardRefs* const refs,
int* const cache_bits_best);
#ifdef __cplusplus
}

Просмотреть файл

@ -31,7 +31,7 @@ extern "C" {
// version numbers
#define ENC_MAJ_VERSION 1
#define ENC_MIN_VERSION 1
#define ENC_MIN_VERSION 2
#define ENC_REV_VERSION 0
enum { MAX_LF_LEVELS = 64, // Maximum loop filter level
@ -505,9 +505,9 @@ int WebPPictureAllocARGB(WebPPicture* const picture, int width, int height);
// Returns false in case of error (invalid param, out-of-memory).
int WebPPictureAllocYUVA(WebPPicture* const picture, int width, int height);
// Clean-up the RGB samples under fully transparent area, to help lossless
// compressibility (no guarantee, though). Assumes that pic->use_argb is true.
void WebPCleanupTransparentAreaLossless(WebPPicture* const pic);
// Replace samples that are fully transparent by 'color' to help compressibility
// (no guarantee, though). Assumes pic->use_argb is true.
void WebPReplaceTransparentPixels(WebPPicture* const pic, uint32_t color);
//------------------------------------------------------------------------------

Просмотреть файл

@ -71,7 +71,7 @@ typedef struct {
uint32_t palette_[MAX_PALETTE_SIZE];
// Some 'scratch' (potentially large) objects.
struct VP8LBackwardRefs refs_[3]; // Backward Refs array for temporaries.
struct VP8LBackwardRefs refs_[4]; // Backward Refs array for temporaries.
VP8LHashChain hash_chain_; // HashChain data for constructing
// backward references.
} VP8LEncoder;

Просмотреть файл

@ -231,7 +231,7 @@ void WebPFree(void* ptr) {
void WebPCopyPlane(const uint8_t* src, int src_stride,
uint8_t* dst, int dst_stride, int width, int height) {
assert(src != NULL && dst != NULL);
assert(src_stride >= width && dst_stride >= width);
assert(abs(src_stride) >= width && abs(dst_stride) >= width);
while (height-- > 0) {
memcpy(dst, src, width);
src += src_stride;

Просмотреть файл

@ -453,7 +453,7 @@ struct WebPDecoderOptions {
int scaled_width, scaled_height; // final resolution
int use_threads; // if true, use multi-threaded decoding
int dithering_strength; // dithering strength (0=Off, 100=full)
int flip; // flip output vertically
int flip; // if true, flip output vertically
int alpha_dithering_strength; // alpha dithering strength in [0..100]
uint32_t pad[5]; // padding for later use

Просмотреть файл

@ -148,7 +148,8 @@ struct WebPConfig {
int use_delta_palette; // reserved for future lossless feature
int use_sharp_yuv; // if needed, use sharp (and slow) RGB->YUV conversion
uint32_t pad[2]; // padding for later use
int qmin; // minimum permissible quality factor
int qmax; // maximum permissible quality factor
};
// Enumerate some predefined settings for WebPConfig, depending on the type
@ -291,6 +292,11 @@ typedef enum WebPEncodingError {
#define WEBP_MAX_DIMENSION 16383
// Main exchange structure (input samples, output bytes, statistics)
//
// Once WebPPictureInit() has been called, it's ok to make all the INPUT fields
// (use_argb, y/u/v, argb, ...) point to user-owned data, even if
// WebPPictureAlloc() has been called. Depending on the value use_argb,
// it's guaranteed that either *argb or *y/*u/*v content will be kept untouched.
struct WebPPicture {
// INPUT
//////////////