fix 16b overflow in SSE2

the 'accum' variable can be larger than 15b for large
rescale values.

Assert triggered:
 src/dsp/rescaler_sse2.c:249: RescalerExportRowExpand_SSE2: Assertion `v >= 0 && v <= 255' failed.
 src/dsp/rescaler_sse2.c:350: RescalerExportRowShrink_SSE2: Assertion `v >= 0 && v <= 255' failed.

-> fall back to C implementation in this case for now

Change-Id: I7ea1cb72301cafc1459be403f6a6f4e3cbc89bb1

CMakeLists.txt

@@ -10,7 +10,7 @@ option(WEBP_BUILD_GIF2WEBP "Build the gif2webp conversion tool." OFF)
 option(WEBP_BUILD_IMG2WEBP "Build the img2webp animation tool." OFF)
 option(WEBP_BUILD_WEBPINFO "Build the webpinfo command line tool." OFF)
 option(WEBP_BUILD_WEBP_JS "Emscripten build of webp.js." OFF)
-option(WEBP_ENABLE_NEAR_LOSSLESS "Enable near-lossless encoding" ON)
+option(WEBP_NEAR_LOSSLESS "Enable near-lossless encoding" ON)
 option(WEBP_ENABLE_SWAP_16BIT_CSP "Enable byte swap for 16 bit colorspaces." OFF)
 
 if(WEBP_BUILD_WEBP_JS)
@@ -289,7 +289,8 @@ if(WEBP_BUILD_WEBP_JS)
   target_link_libraries(webp_js webpdecoder SDL)
   set(WEBP_HAVE_SDL 1)
   set_target_properties(webp_js PROPERTIES LINK_FLAGS
-      "-s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0")
+      "-s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
+       -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
   set_target_properties(webp_js PROPERTIES OUTPUT_NAME webp)
   target_compile_definitions(webp_js PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
 
@@ -299,7 +300,8 @@ if(WEBP_BUILD_WEBP_JS)
   target_link_libraries(webp_wasm webpdecoder SDL)
   set_target_properties(webp_wasm PROPERTIES LINK_FLAGS
       "-s WASM=1 -s 'BINARYEN_METHOD=\"native-wasm\"' \
-      -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0")
+       -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
+       -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
   target_compile_definitions(webp_wasm PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
 
   target_compile_definitions(webpdecoder PUBLIC EMSCRIPTEN)

ChangeLog

@@ -1,3 +1,11 @@
+2f930e08 Revert "Use proper targets for CMake."
+8165e8fb Use proper targets for CMake.
+3f157dd5 Remove some very hard TODOs.
+cd758a17 {de,}mux/Makefile.am: add missing headers
+b892b8ba makefile.unix,dist: use ascii for text output
+64a57d05 add -version option to anim_dump,anim_diff and img2webp
+fc1b8e3a webp_js: fix webp_js demo html
+15aa48d9 update ChangeLog
 e607dabc update AUTHORS
 38410c08 [CFI] Remove function pointer casts
 c57b2736 bump version to 1.0.0

README

@@ -458,6 +458,7 @@ File-level options (only used at the start of compression):
  -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)
@@ -527,6 +528,8 @@ Options:
   -max_diff <int> ..... maximum allowed difference per channel
                         between corresponding pixels in subsequent
                         frames
+  -h .................. this help
+  -version ............ print version number and exit
 
 Building:
 ---------

README.webp_js

@@ -32,7 +32,8 @@ using Emscripten and CMake.
    webp.js.mem files generated.
 
 The callable JavaScript function is WebPToSDL(), which decodes a raw WebP
-bitstream into a canvas. See webp_js/index.html for a simple usage sample.
+bitstream into a canvas. See webp_js/index.html for a simple usage sample
+(see below for instructions).
 
 Demo HTML page:
 ===============

examples/anim_diff.c

@@ -190,6 +190,8 @@ static void Help(void) {
   printf("  -max_diff <int> ..... maximum allowed difference per channel\n"
          "                        between corresponding pixels in subsequent\n"
          "                        frames\n");
+  printf("  -h .................. this help\n");
+  printf("  -version ............ print version number and exit\n");
 }
 
 int main(int argc, const char* argv[]) {
@@ -205,11 +207,6 @@ int main(int argc, const char* argv[]) {
   const char* files[2] = { NULL, NULL };
   AnimatedImage images[2];
 
-  if (argc < 3) {
-    Help();
-    return -1;
-  }
-
   for (c = 1; c < argc; ++c) {
     int parse_error = 0;
     if (!strcmp(argv[c], "-dump_frames")) {
@@ -247,6 +244,18 @@ int main(int argc, const char* argv[]) {
       } else {
         parse_error = 1;
       }
+    } else if (!strcmp(argv[c], "-h") || !strcmp(argv[c], "-help")) {
+      Help();
+      return 0;
+    } else if (!strcmp(argv[c], "-version")) {
+      int dec_version, demux_version;
+      GetAnimatedImageVersions(&dec_version, &demux_version);
+      printf("WebP Decoder version: %d.%d.%d\nWebP Demux version: %d.%d.%d\n",
+             (dec_version >> 16) & 0xff, (dec_version >> 8) & 0xff,
+             (dec_version >> 0) & 0xff,
+             (demux_version >> 16) & 0xff, (demux_version >> 8) & 0xff,
+             (demux_version >> 0) & 0xff);
+      return 0;
     } else {
       if (!got_input1) {
         files[0] = argv[c];
@@ -263,6 +272,12 @@ int main(int argc, const char* argv[]) {
       return -1;
     }
   }
+  if (argc < 3) {
+    Help();
+    return -1;
+  }
+
+
   if (!got_input2) {
     Help();
     return -1;

examples/anim_dump.c

@@ -30,6 +30,8 @@ static void Help(void) {
                                   "(default: 'dump_')\n");
   printf("  -tiff ............... save frames as TIFF\n");
   printf("  -pam ................ save frames as PAM\n");
+  printf("  -h .................. this help\n");
+  printf("  -version ............ print version number and exit\n");
 }
 
 int main(int argc, const char* argv[]) {
@@ -66,6 +68,18 @@ int main(int argc, const char* argv[]) {
     } else if (!strcmp(argv[c], "-pam")) {
       format = PAM;
       suffix = "pam";
+    } else if (!strcmp(argv[c], "-h") || !strcmp(argv[c], "-help")) {
+      Help();
+      return 0;
+    } else if (!strcmp(argv[c], "-version")) {
+      int dec_version, demux_version;
+      GetAnimatedImageVersions(&dec_version, &demux_version);
+      printf("WebP Decoder version: %d.%d.%d\nWebP Demux version: %d.%d.%d\n",
+             (dec_version >> 16) & 0xff, (dec_version >> 8) & 0xff,
+             (dec_version >> 0) & 0xff,
+             (demux_version >> 16) & 0xff, (demux_version >> 8) & 0xff,
+             (demux_version >> 0) & 0xff);
+      return 0;
     } else {
       uint32_t i;
       AnimatedImage image;

examples/anim_util.c

@@ -786,3 +786,9 @@ void GetDiffAndPSNR(const uint8_t rgba1[], const uint8_t rgba2[],
     *psnr = 4.3429448 * log(255. * 255. / sse);
   }
 }
+
+void GetAnimatedImageVersions(int* const decoder_version,
+                              int* const demux_version) {
+  *decoder_version = WebPGetDecoderVersion();
+  *demux_version = WebPGetDemuxVersion();
+}

examples/anim_util.h

@@ -56,6 +56,10 @@ void GetDiffAndPSNR(const uint8_t rgba1[], const uint8_t rgba2[],
                     uint32_t width, uint32_t height, int premultiply,
                     int* const max_diff, double* const psnr);
 
+// Return library versions used by anim_util.
+void GetAnimatedImageVersions(int* const decoder_version,
+                              int* const demux_version);
+
 #ifdef __cplusplus
 }    // extern "C"
 #endif

examples/cwebp.c

@@ -188,7 +188,7 @@ static void PrintExtraInfoLossless(const WebPPicture* const pic,
     fprintf(stderr, "File:      %s\n", file_name);
     fprintf(stderr, "Dimension: %d x %d\n", pic->width, pic->height);
     fprintf(stderr, "Output:    %d bytes (%.2f bpp)\n", stats->coded_size,
-            1.f * stats->coded_size / pic->width / pic->height);
+            8.f * stats->coded_size / pic->width / pic->height);
     PrintFullLosslessInfo(stats, "ARGB");
   }
 }
@@ -213,7 +213,7 @@ static void PrintExtraInfoLossy(const WebPPicture* const pic, int short_output,
             "           (%.2f bpp)\n",
             stats->coded_size,
             stats->PSNR[0], stats->PSNR[1], stats->PSNR[2], stats->PSNR[3],
-            1.f * stats->coded_size / pic->width / pic->height);
+            8.f * stats->coded_size / pic->width / pic->height);
     if (total > 0) {
       int totals[4] = { 0, 0, 0, 0 };
       fprintf(stderr, "block count:  intra4:     %6d  (%.2f%%)\n"

examples/img2webp.c

@@ -48,6 +48,7 @@ static void Help(void) {
   printf(" -mixed ............... use mixed lossy/lossless automatic mode\n");
   printf(" -v ................... verbose mode\n");
   printf(" -h ................... this help\n");
+  printf(" -version ............. print version number and exit\n");
   printf("\n");
 
   printf("Per-frame options (only used for subsequent images input):\n");
@@ -177,6 +178,14 @@ int main(int argc, const char* argv[]) {
       } else if (!strcmp(argv[c], "-h") || !strcmp(argv[c], "-help")) {
         Help();
         goto End;
+      } else if (!strcmp(argv[c], "-version")) {
+        const int enc_version = WebPGetEncoderVersion();
+        const int mux_version = WebPGetMuxVersion();
+        printf("WebP Encoder version: %d.%d.%d\nWebP Mux version: %d.%d.%d\n",
+               (enc_version >> 16) & 0xff, (enc_version >> 8) & 0xff,
+               enc_version & 0xff, (mux_version >> 16) & 0xff,
+               (mux_version >> 8) & 0xff, mux_version & 0xff);
+        goto End;
       } else {
         continue;
       }

extras/extras.c

@@ -48,13 +48,14 @@ int WebPImportGray(const uint8_t* gray_data, WebPPicture* pic) {
 
 int WebPImportRGB565(const uint8_t* rgb565, WebPPicture* pic) {
   int x, y;
+  uint32_t* dst;
   if (pic == NULL || rgb565 == NULL) return 0;
   pic->colorspace = WEBP_YUV420;
   pic->use_argb = 1;
   if (!WebPPictureAlloc(pic)) return 0;
+  dst = pic->argb;
   for (y = 0; y < pic->height; ++y) {
     const int width = pic->width;
-    uint32_t* dst = pic->argb + y * pic->argb_stride;
     for (x = 0; x < width; ++x) {
 #ifdef WEBP_SWAP_16BIT_CSP
       const uint32_t rg = rgb565[2 * x + 1];
@@ -70,22 +71,24 @@ int WebPImportRGB565(const uint8_t* rgb565, WebPPicture* pic) {
       r = r | (r >> 5);
       g = g | (g >> 6);
       b = b | (b >> 5);
-      dst[x] = (r << 16) | (g << 8) | b;
+      dst[x] = (0xffu << 24) | (r << 16) | (g << 8) | b;
     }
     rgb565 += 2 * width;
+    dst += pic->argb_stride;
   }
   return 1;
 }
 
 int WebPImportRGB4444(const uint8_t* rgb4444, WebPPicture* pic) {
   int x, y;
+  uint32_t* dst;
   if (pic == NULL || rgb4444 == NULL) return 0;
   pic->colorspace = WEBP_YUV420;
   pic->use_argb = 1;
   if (!WebPPictureAlloc(pic)) return 0;
+  dst = pic->argb;
   for (y = 0; y < pic->height; ++y) {
     const int width = pic->width;
-    uint32_t* dst = pic->argb + y * pic->argb_stride;
     for (x = 0; x < width; ++x) {
 #ifdef WEBP_SWAP_16BIT_CSP
       const uint32_t rg = rgb4444[2 * x + 1];
@@ -106,6 +109,7 @@ int WebPImportRGB4444(const uint8_t* rgb4444, WebPPicture* pic) {
       dst[x] = (a << 24) | (r << 16) | (g << 8) | b;
     }
     rgb4444 += 2 * width;
+    dst += pic->argb_stride;
   }
   return 1;
 }

extras/get_disto.c

@@ -293,7 +293,7 @@ int main(int argc, const char *argv[]) {
   printf("%u %.2f    %.2f %.2f %.2f %.2f [ %.2f bpp ]\n",
          (unsigned int)size1,
          disto[4], disto[0], disto[1], disto[2], disto[3],
-         1.f * size1 / pic1.width / pic1.height);
+         8.f * size1 / pic1.width / pic1.height);
 
   if (output != NULL) {
     uint8_t* data = NULL;

makefile.unix

@@ -112,7 +112,12 @@ endif
 AR = ar
 ARFLAGS = r
 CPPFLAGS = -I. -Isrc/ -Wall
-CFLAGS = -O3 -DNDEBUG $(EXTRA_FLAGS)
+ifeq ($(DEBUG), 1)
+  CFLAGS = -g
+else
+  CFLAGS = -O3 -DNDEBUG
+endif
+CFLAGS += $(EXTRA_FLAGS)
 CC = gcc
 INSTALL = install
 GROFF = /usr/bin/groff
@@ -474,7 +479,7 @@ dist: all
 	for m in man/[cdv]webp.1 man/gif2webp.1 man/webpmux.1 \
                  man/img2webp.1 man/webpinfo.1; do \
 	  basenam=$$(basename $$m .1); \
-	  $(GROFF) -t -e -man -T utf8 $$m \
+	  $(GROFF) -t -e -man -T ascii $$m \
 	    | $(COL) -bx >$(DESTDIR)/doc/$${basenam}.txt; \
 	  $(GROFF) -t -e -man -T html $$m \
 	    | $(COL) -bx >$(DESTDIR)/doc/$${basenam}.html; \

man/img2webp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH IMG2WEBP 1 "February 7, 2018"
+.TH IMG2WEBP 1 "April 3, 2018"
 .SH NAME
 img2webp \- create animated WebP file from a sequence of input images.
 .SH SYNOPSIS
@@ -53,6 +53,9 @@ Be more verbose.
 .TP
 .B \-h, \-help
 A short usage summary.
+.TP
+.B \-version
+Print the version numbers of the relevant libraries used.
 
 .SH PER-FRAME OPTIONS
 The per-frame options are applied for the images following as arguments in the

src/demux/Makefile.am

@@ -5,9 +5,12 @@ libwebpdemux_la_SOURCES =
 libwebpdemux_la_SOURCES += anim_decode.c demux.c
 
 libwebpdemuxinclude_HEADERS =
+libwebpdemuxinclude_HEADERS += ../webp/decode.h
 libwebpdemuxinclude_HEADERS += ../webp/demux.h
 libwebpdemuxinclude_HEADERS += ../webp/mux_types.h
 libwebpdemuxinclude_HEADERS += ../webp/types.h
+noinst_HEADERS =
+noinst_HEADERS += ../webp/format_constants.h
 
 libwebpdemux_la_LIBADD = ../libwebp.la
 libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:4:0

src/dsp/rescaler_sse2.c

@@ -36,7 +36,7 @@ static void LoadTwoPixels_SSE2(const uint8_t* const src, __m128i* out) {
 }
 
 // input: 8 bytes ABCDEFGH -> output: A0B0C0D0E0F0G0H0
-static void LoadHeightPixels_SSE2(const uint8_t* const src, __m128i* out) {
+static void LoadEightPixels_SSE2(const uint8_t* const src, __m128i* out) {
   const __m128i zero = _mm_setzero_si128();
   const __m128i A = _mm_loadl_epi64((const __m128i*)(src));  // ABCDEFGH
   *out = _mm_unpacklo_epi8(A, zero);
@@ -50,13 +50,15 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
   int accum = x_add;
   __m128i cur_pixels;
 
+  // SSE2 implementation only works with 16b signed arithmetic at max.
+  if (wrk->src_width < 8 || accum >= (1 << 15)) {
+    WebPRescalerImportRowExpand_C(wrk, src);
+    return;
+  }
+
   assert(!WebPRescalerInputDone(wrk));
   assert(wrk->x_expand);
   if (wrk->num_channels == 4) {
-    if (wrk->src_width < 2) {
-      WebPRescalerImportRowExpand_C(wrk, src);
-      return;
-    }
     LoadTwoPixels_SSE2(src, &cur_pixels);
     src += 4;
     while (1) {
@@ -75,11 +77,7 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
   } else {
     int left;
     const uint8_t* const src_limit = src + wrk->src_width - 8;
-    if (wrk->src_width < 8) {
+    LoadEightPixels_SSE2(src, &cur_pixels);
-      WebPRescalerImportRowExpand_C(wrk, src);
-      return;
-    }
-    LoadHeightPixels_SSE2(src, &cur_pixels);
     src += 7;
     left = 7;
     while (1) {
@@ -94,7 +92,7 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
         if (--left) {
           cur_pixels = _mm_srli_si128(cur_pixels, 2);
         } else if (src <= src_limit) {
-          LoadHeightPixels_SSE2(src, &cur_pixels);
+          LoadEightPixels_SSE2(src, &cur_pixels);
           src += 7;
           left = 7;
         } else {   // tail

src/enc/histogram_enc.c

@@ -1026,7 +1026,7 @@ int VP8LGetHistoImageSymbols(int xsize, int ysize,
     }
   }
 
-  // TODO(vikasa): Optimize HistogramRemap for low-effort compression mode also.
+  // TODO(vrabaud): Optimize HistogramRemap for low-effort compression mode.
   // Find the optimal map from original histograms to the final ones.
   HistogramRemap(orig_histo, image_histo, histogram_symbols);
 

src/enc/vp8l_enc.c

@@ -382,8 +382,7 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
       AnalyzeAndCreatePalette(pic, low_effort,
                               enc->palette_, &enc->palette_size_);
 
-  // TODO(jyrki): replace the decision to be based on an actual estimate
+  // Empirical bit sizes.
-  // of entropy, or even spatial variance of entropy.
   enc->histo_bits_ = GetHistoBits(method, use_palette,
                                   pic->width, pic->height);
   enc->transform_bits_ = GetTransformBits(method, enc->histo_bits_);
@@ -754,7 +753,6 @@ static WebPEncodingError StoreImageToBitMask(
       // Don't write the distance with the extra bits code since
       // the distance can be up to 18 bits of extra bits, and the prefix
       // 15 bits, totaling to 33, and our PutBits only supports up to 32 bits.
-      // TODO(jyrki): optimize this further.
       VP8LPrefixEncode(distance, &code, &n_bits, &bits);
       WriteHuffmanCode(bw, codes + 4, code);
       VP8LPutBits(bw, bits, n_bits);
@@ -1876,7 +1874,6 @@ int VP8LEncodeImage(const WebPConfig* const config,
   err = VP8LEncodeStream(config, picture, &bw, 1 /*use_cache*/);
   if (err != VP8_ENC_OK) goto Error;
 
-  // TODO(skal): have a fine-grained progress report in VP8LEncodeStream().
   if (!WebPReportProgress(picture, 90, &percent)) goto UserAbort;
 
   // Finish the RIFF chunk.

src/mux/Makefile.am

@@ -13,6 +13,8 @@ libwebpmuxinclude_HEADERS =
 libwebpmuxinclude_HEADERS += ../webp/mux.h
 libwebpmuxinclude_HEADERS += ../webp/mux_types.h
 libwebpmuxinclude_HEADERS += ../webp/types.h
+noinst_HEADERS =
+noinst_HEADERS += ../webp/format_constants.h
 
 libwebpmux_la_LIBADD = ../libwebp.la
 libwebpmux_la_LDFLAGS = -no-undefined -version-info 3:2:0 -lm