Update bundled libwebp to version 1.2.0

[ChangeLog][Third-Party Code] Update bundled libwebp to version 1.2.0

Pick-to: 6.1 6.0 5.15 5.12
Change-Id: Idb95b278c613361d82ea32bd9f591fbe87bfe66f
Reviewed-by: Liang Qi <liang.qi@qt.io>

10 files changed, 207 insertions, 244 deletions

diff --git a/src/3rdparty/libwebp/src/dsp/alpha_processing.c b/src/3rdparty/libwebp/src/dsp/alpha_processing.c
index 819d139..3a27990 100644
--- a/src/3rdparty/libwebp/src/dsp/alpha_processing.c
+++ b/src/3rdparty/libwebp/src/dsp/alpha_processing.c
@@ -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);
 }
diff --git a/src/3rdparty/libwebp/src/dsp/alpha_processing_sse2.c b/src/3rdparty/libwebp/src/dsp/alpha_processing_sse2.c
index 2871c56..f6c6e0f 100644
--- a/src/3rdparty/libwebp/src/dsp/alpha_processing_sse2.c
+++ b/src/3rdparty/libwebp/src/dsp/alpha_processing_sse2.c
@@ -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
diff --git a/src/3rdparty/libwebp/src/dsp/cpu.c b/src/3rdparty/libwebp/src/dsp/cpu.c
index 0fa5b6a..4ca90d8 100644
--- a/src/3rdparty/libwebp/src/dsp/cpu.c
+++ b/src/3rdparty/libwebp/src/dsp/cpu.c
@@ -55,12 +55,18 @@ static WEBP_INLINE void GetCPUInfo(int cpu_info[4], int info_type) {
     : "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
     : "a"(info_type), "c"(0));
 }
-#elif (defined(_M_X64) || defined(_M_IX86)) && \
-      defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 150030729  // >= VS2008 SP1
+#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_IX86))
+
+#if defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 150030729  // >= VS2008 SP1
 #include <intrin.h>
 #define GetCPUInfo(info, type) __cpuidex(info, type, 0)  // set ecx=0
-#elif defined(WEBP_MSC_SSE2)
+#define WEBP_HAVE_MSC_CPUID
+#elif _MSC_VER > 1310
+#include <intrin.h>
 #define GetCPUInfo __cpuid
+#define WEBP_HAVE_MSC_CPUID
+#endif
+
 #endif
 
 // NaCl has no support for xgetbv or the raw opcode.
@@ -94,7 +100,7 @@ static WEBP_INLINE uint64_t xgetbv(void) {
 #define xgetbv() 0U  // no AVX for older x64 or unrecognized toolchains.
 #endif
 
-#if defined(__i386__) || defined(__x86_64__) || defined(WEBP_MSC_SSE2)
+#if defined(__i386__) || defined(__x86_64__) || defined(WEBP_HAVE_MSC_CPUID)
 
 // helper function for run-time detection of slow SSSE3 platforms
 static int CheckSlowModel(int info) {
@@ -179,6 +185,30 @@ static int AndroidCPUInfo(CPUFeature feature) {
   return 0;
 }
 VP8CPUInfo VP8GetCPUInfo = AndroidCPUInfo;
+#elif defined(EMSCRIPTEN) // also needs to be before generic NEON test
+// Use compile flags as an indicator of SIMD support instead of a runtime check.
+static int wasmCPUInfo(CPUFeature feature) {
+  switch (feature) {
+#ifdef WEBP_USE_SSE2
+    case kSSE2:
+      return 1;
+#endif
+#ifdef WEBP_USE_SSE41
+    case kSSE3:
+    case kSlowSSSE3:
+    case kSSE4_1:
+      return 1;
+#endif
+#ifdef WEBP_USE_NEON
+    case kNEON:
+      return 1;
+#endif
+    default:
+      break;
+  }
+  return 0;
+}
+VP8CPUInfo VP8GetCPUInfo = wasmCPUInfo;
 #elif defined(WEBP_USE_NEON)
 // define a dummy function to enable turning off NEON at runtime by setting
 // VP8DecGetCPUInfo = NULL
diff --git a/src/3rdparty/libwebp/src/dsp/dec_neon.c b/src/3rdparty/libwebp/src/dsp/dec_neon.c
index 239ec41..fa85170 100644
--- a/src/3rdparty/libwebp/src/dsp/dec_neon.c
+++ b/src/3rdparty/libwebp/src/dsp/dec_neon.c
@@ -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);
diff --git a/src/3rdparty/libwebp/src/dsp/dsp.h b/src/3rdparty/libwebp/src/dsp/dsp.h
index 7c75b26..6df48cf 100644
--- a/src/3rdparty/libwebp/src/dsp/dsp.h
+++ b/src/3rdparty/libwebp/src/dsp/dsp.h
@@ -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)) && !defined(__clang__)
 #define WEBP_MSC_SSE2  // Visual C++ SSE2 targets
@@ -63,6 +61,7 @@ extern "C" {
     (defined(_M_X64) || defined(_M_IX86)) && !defined(__clang__)
 #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);
diff --git a/src/3rdparty/libwebp/src/dsp/lossless.c b/src/3rdparty/libwebp/src/dsp/lossless.c
index aad5f43..46b220e 100644
--- a/src/3rdparty/libwebp/src/dsp/lossless.c
+++ b/src/3rdparty/libwebp/src/dsp/lossless.c
@@ -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)
 
diff --git a/src/3rdparty/libwebp/src/dsp/lossless.h b/src/3rdparty/libwebp/src/dsp/lossless.h
index f709cc8..ebd316d 100644
--- a/src/3rdparty/libwebp/src/dsp/lossless.h
+++ b/src/3rdparty/libwebp/src/dsp/lossless.h
@@ -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,
diff --git a/src/3rdparty/libwebp/src/dsp/lossless_common.h b/src/3rdparty/libwebp/src/dsp/lossless_common.h
index 9c2ebe6..96a106f 100644
--- a/src/3rdparty/libwebp/src/dsp/lossless_common.h
+++ b/src/3rdparty/libwebp/src/dsp/lossless_common.h
@@ -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
diff --git a/src/3rdparty/libwebp/src/dsp/lossless_enc.c b/src/3rdparty/libwebp/src/dsp/lossless_enc.c
index 9c36055..a0c7ab9 100644
--- a/src/3rdparty/libwebp/src/dsp/lossless_enc.c
+++ b/src/3rdparty/libwebp/src/dsp/lossless_enc.c
@@ -702,140 +702,6 @@ void VP8LHistogramAdd(const VP8LHistogram* const a,
 //------------------------------------------------------------------------------
 // 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 || LOCAL_GCC_VERSION == 0x408)
-# 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 Predictor2(uint32_t left, const uint32_t* const top) {
-  (void)left;
-  return top[0];
-}
-static uint32_t Predictor3(uint32_t left, const uint32_t* const top) {
-  (void)left;
-  return top[1];
-}
-static uint32_t Predictor4(uint32_t left, const uint32_t* const top) {
-  (void)left;
-  return top[-1];
-}
-static uint32_t Predictor5(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average3(left, top[0], top[1]);
-  return pred;
-}
-static uint32_t Predictor6(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average2(left, top[-1]);
-  return pred;
-}
-static uint32_t Predictor7(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average2(left, top[0]);
-  return pred;
-}
-static uint32_t Predictor8(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(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(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(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Select(top[0], left, top[-1]);
-  return pred;
-}
-static uint32_t Predictor12(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = ClampedAddSubtractFull(left, top[0], top[-1]);
-  return pred;
-}
-static uint32_t Predictor13(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = ClampedAddSubtractHalf(left, top[0], top[-1]);
-  return pred;
-}
-
-//------------------------------------------------------------------------------
-
 static void PredictorSub0_C(const uint32_t* in, const uint32_t* upper,
                             int num_pixels, uint32_t* out) {
   int i;
@@ -850,18 +716,33 @@ static void PredictorSub1_C(const uint32_t* in, const uint32_t* upper,
   (void)upper;
 }
 
-GENERATE_PREDICTOR_SUB(Predictor2, PredictorSub2_C)
-GENERATE_PREDICTOR_SUB(Predictor3, PredictorSub3_C)
-GENERATE_PREDICTOR_SUB(Predictor4, PredictorSub4_C)
-GENERATE_PREDICTOR_SUB(Predictor5, PredictorSub5_C)
-GENERATE_PREDICTOR_SUB(Predictor6, PredictorSub6_C)
-GENERATE_PREDICTOR_SUB(Predictor7, PredictorSub7_C)
-GENERATE_PREDICTOR_SUB(Predictor8, PredictorSub8_C)
-GENERATE_PREDICTOR_SUB(Predictor9, PredictorSub9_C)
-GENERATE_PREDICTOR_SUB(Predictor10, PredictorSub10_C)
-GENERATE_PREDICTOR_SUB(Predictor11, PredictorSub11_C)
-GENERATE_PREDICTOR_SUB(Predictor12, PredictorSub12_C)
-GENERATE_PREDICTOR_SUB(Predictor13, PredictorSub13_C)
+// It subtracts the prediction from the input pixel and stores the residual
+// in the output pixel.
+#define GENERATE_PREDICTOR_SUB(PREDICTOR_I)                                \
+static void PredictorSub##PREDICTOR_I##_C(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 =                                                  \
+        VP8LPredictor##PREDICTOR_I##_C(in[x - 1], upper + x);              \
+    out[x] = VP8LSubPixels(in[x], pred);                                   \
+  }                                                                        \
+}
+
+GENERATE_PREDICTOR_SUB(2)
+GENERATE_PREDICTOR_SUB(3)
+GENERATE_PREDICTOR_SUB(4)
+GENERATE_PREDICTOR_SUB(5)
+GENERATE_PREDICTOR_SUB(6)
+GENERATE_PREDICTOR_SUB(7)
+GENERATE_PREDICTOR_SUB(8)
+GENERATE_PREDICTOR_SUB(9)
+GENERATE_PREDICTOR_SUB(10)
+GENERATE_PREDICTOR_SUB(11)
+GENERATE_PREDICTOR_SUB(12)
+GENERATE_PREDICTOR_SUB(13)
 
 //------------------------------------------------------------------------------
 
diff --git a/src/3rdparty/libwebp/src/dsp/lossless_enc_sse2.c b/src/3rdparty/libwebp/src/dsp/lossless_enc_sse2.c
index e676f6f..90c2637 100644
--- a/src/3rdparty/libwebp/src/dsp/lossless_enc_sse2.c
+++ b/src/3rdparty/libwebp/src/dsp/lossless_enc_sse2.c
@@ -249,6 +249,7 @@ static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) {
     }                                  \
   } while (0)
 
+#if !(defined(__i386__) || defined(_M_IX86))
 static float CombinedShannonEntropy_SSE2(const int X[256], const int Y[256]) {
   int i;
   double retval = 0.;
@@ -300,6 +301,8 @@ static float CombinedShannonEntropy_SSE2(const int X[256], const int Y[256]) {
   retval += VP8LFastSLog2(sumX) + VP8LFastSLog2(sumXY);
   return (float)retval;
 }
+#endif  // !(defined(__i386__) || defined(_M_IX86))
+
 #undef ANALYZE_X_OR_Y
 #undef ANALYZE_XY
 
@@ -460,20 +463,22 @@ static void PredictorSub0_SSE2(const uint32_t* in, const uint32_t* upper,
   (void)upper;
 }
 
-#define GENERATE_PREDICTOR_1(X, IN)                                           \
-static void PredictorSub##X##_SSE2(const uint32_t* in, const uint32_t* upper, \
-                                   int num_pixels, uint32_t* out) {           \
-  int i;                                                                      \
-  for (i = 0; i + 4 <= num_pixels; i += 4) {                                  \
-    const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);              \
-    const __m128i pred = _mm_loadu_si128((const __m128i*)&(IN));              \
-    const __m128i res = _mm_sub_epi8(src, pred);                              \
-    _mm_storeu_si128((__m128i*)&out[i], res);                                 \
-  }                                                                           \
-  if (i != num_pixels) {                                                      \
-    VP8LPredictorsSub_C[(X)](in + i, upper + i, num_pixels - i, out + i);     \
-  }                                                                           \
-}
+#define GENERATE_PREDICTOR_1(X, IN)                                         \
+  static void PredictorSub##X##_SSE2(const uint32_t* const in,              \
+                                     const uint32_t* const upper,           \
+                                     int num_pixels, uint32_t* const out) { \
+    int i;                                                                  \
+    for (i = 0; i + 4 <= num_pixels; i += 4) {                              \
+      const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);          \
+      const __m128i pred = _mm_loadu_si128((const __m128i*)&(IN));          \
+      const __m128i res = _mm_sub_epi8(src, pred);                          \
+      _mm_storeu_si128((__m128i*)&out[i], res);                             \
+    }                                                                       \
+    if (i != num_pixels) {                                                  \
+      VP8LPredictorsSub_C[(X)](in + i, WEBP_OFFSET_PTR(upper, i),           \
+                               num_pixels - i, out + i);                    \
+    }                                                                       \
+  }
 
 GENERATE_PREDICTOR_1(1, in[i - 1])       // Predictor1: L
 GENERATE_PREDICTOR_1(2, upper[i])        // Predictor2: T
@@ -657,7 +662,12 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInitSSE2(void) {
   VP8LCollectColorRedTransforms = CollectColorRedTransforms_SSE2;
   VP8LAddVector = AddVector_SSE2;
   VP8LAddVectorEq = AddVectorEq_SSE2;
+  // TODO(https://crbug.com/webp/499): this function produces different results
+  // from the C code due to use of double/float resulting in output differences
+  // when compared to -noasm.
+#if !(defined(__i386__) || defined(_M_IX86))
   VP8LCombinedShannonEntropy = CombinedShannonEntropy_SSE2;
+#endif
   VP8LVectorMismatch = VectorMismatch_SSE2;
   VP8LBundleColorMap = BundleColorMap_SSE2;