yuy2 to rgb gcc versions

source/row_gcc.cc

 // VERSION 2
 /*
  *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS. All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
@@ skipping 1306 matching lines @@
   );
 }
 
 #if defined(HAS_I422TOARGBROW_SSSE3) || defined(HAS_I422TOARGBROW_AVX2)
 
 // Read 8 UV from 411
 #define READYUV444                                                             \
     "movq       " MEMACCESS([u_buf]) ",%%xmm0                   \n"            \
     MEMOPREG(movq, 0x00, [u_buf], [v_buf], 1, xmm1)                            \
     "lea        " MEMLEA(0x8, [u_buf]) ",%[u_buf]               \n"            \
-    "punpcklbw  %%xmm1,%%xmm0                                   \n"
+    "punpcklbw  %%xmm1,%%xmm0                                   \n"            \
+    "movq       " MEMACCESS([y_buf]) ",%%xmm4                   \n"            \
+    "punpcklbw  %%xmm4,%%xmm4                                   \n"            \
+    "lea        " MEMLEA(0x8, [y_buf]) ",%[y_buf]               \n"
 
 // Read 4 UV from 422, upsample to 8 UV
 #define READYUV422                                                             \
     "movd       " MEMACCESS([u_buf]) ",%%xmm0                   \n"            \
     MEMOPREG(movd, 0x00, [u_buf], [v_buf], 1, xmm1)                            \
     "lea        " MEMLEA(0x4, [u_buf]) ",%[u_buf]               \n"            \
     "punpcklbw  %%xmm1,%%xmm0                                   \n"            \
-    "punpcklwd  %%xmm0,%%xmm0                                   \n"
+    "punpcklwd  %%xmm0,%%xmm0                                   \n"            \
+    "movq       " MEMACCESS([y_buf]) ",%%xmm4                   \n"            \
+    "punpcklbw  %%xmm4,%%xmm4                                   \n"            \
+    "lea        " MEMLEA(0x8, [y_buf]) ",%[y_buf]               \n"
 
 // Read 2 UV from 411, upsample to 8 UV
 #define READYUV411                                                             \
     "movd       " MEMACCESS([u_buf]) ",%%xmm0                   \n"            \
     MEMOPREG(movd, 0x00, [u_buf], [v_buf], 1, xmm1)                            \
     "lea        " MEMLEA(0x2, [u_buf]) ",%[u_buf]               \n"            \
     "punpcklbw  %%xmm1,%%xmm0                                   \n"            \
     "punpcklwd  %%xmm0,%%xmm0                                   \n"            \
-    "punpckldq  %%xmm0,%%xmm0                                   \n"
+    "punpckldq  %%xmm0,%%xmm0                                   \n"            \
+    "movq       " MEMACCESS([y_buf]) ",%%xmm4                   \n"            \
+    "punpcklbw  %%xmm4,%%xmm4                                   \n"            \
+    "lea        " MEMLEA(0x8, [y_buf]) ",%[y_buf]               \n"
 
 // Read 4 UV from NV12, upsample to 8 UV
 #define READNV12                                                               \
     "movq       " MEMACCESS([uv_buf]) ",%%xmm0                  \n"            \
     "lea        " MEMLEA(0x8, [uv_buf]) ",%[uv_buf]             \n"            \
-    "punpcklwd  %%xmm0,%%xmm0                                   \n"
+    "punpcklwd  %%xmm0,%%xmm0                                   \n"            \
+    "movq       " MEMACCESS([y_buf]) ",%%xmm4                   \n"            \
+    "punpcklbw  %%xmm4,%%xmm4                                   \n"            \
+    "lea        " MEMLEA(0x8, [y_buf]) ",%[y_buf]               \n"
+
+// YUY2 shuf 8 Y to 16 Y.
+static const vec8 kShuffleYUY2Y = {
+  0, 0, 2, 2, 4, 4, 6, 6, 8, 8, 10, 10, 12, 12, 14, 14
+};
+
+// YUY2 shuf 4 UV to 8 UV.
+static const vec8 kShuffleYUY2UV = {
+  1, 3, 1, 3, 5, 7, 5, 7, 9, 11, 9, 11, 13, 15, 13, 15
+};
+
+// Read 4 YUY2 with 8 Y and update 4 UV to 8 UV.
+#define READYUY2                                                               \
+    "movdqu     " MEMACCESS([yuy2_buf]) ",%%xmm4                \n"            \
+    "pshufb     %[kShuffleYUY2Y], %%xmm4                        \n"            \
+    "movdqu     " MEMACCESS([yuy2_buf]) ",%%xmm0                \n"            \
+    "pshufb     %[kShuffleYUY2UV], %%xmm0                       \n"            \
+    "lea        " MEMLEA(0x10, [yuy2_buf]) ",%[yuy2_buf]        \n"
+
+// UYVY shuf 8 Y to 16 Y.
+static const vec8 kShuffleUYVYY = {
+  1, 1, 3, 3, 5, 5, 7, 7, 9, 9, 11, 11, 13, 13, 15, 15
+};
+
+// UYVY shuf 4 UV to 8 UV.
+static const vec8 kShuffleUYVYUV = {
+  0, 2, 0, 2, 4, 6, 4, 6, 8, 10, 8, 10, 12, 14, 12, 14
+};
+
+// Read 4 UYVY with 8 Y and update 4 UV to 8 UV.
+#define READUYVY                                                               \
+    "movdqu     " MEMACCESS([uyvy_buf]) ",%%xmm4                \n"            \
+    "pshufb     %[kShuffleUYVYY], %%xmm4                        \n"            \
+    "movdqu     " MEMACCESS([uyvy_buf]) ",%%xmm0                \n"            \
+    "pshufb     %[kShuffleUYVYUV], %%xmm0                       \n"            \
+    "lea        " MEMLEA(0x10, [uyvy_buf]) ",%[uyvy_buf]        \n"
 
 // Convert 8 pixels: 8 UV and 8 Y
 #define YUVTORGB(yuvconstants)                                                 \
     "movdqa     %%xmm0,%%xmm1                                   \n"            \
     "movdqa     %%xmm0,%%xmm2                                   \n"            \
     "movdqa     %%xmm0,%%xmm3                                   \n"            \
     "movdqa     " MEMACCESS2(96, [yuvconstants]) ",%%xmm0       \n"            \
     "pmaddubsw  " MEMACCESS([yuvconstants]) ",%%xmm1            \n"            \
     "psubw      %%xmm1,%%xmm0                                   \n"            \
     "movdqa     " MEMACCESS2(128, [yuvconstants]) ",%%xmm1      \n"            \
     "pmaddubsw  " MEMACCESS2(32, [yuvconstants]) ",%%xmm2       \n"            \
     "psubw      %%xmm2,%%xmm1                                   \n"            \
     "movdqa     " MEMACCESS2(160, [yuvconstants]) ",%%xmm2      \n"            \
     "pmaddubsw  " MEMACCESS2(64, [yuvconstants]) ",%%xmm3       \n"            \
     "psubw      %%xmm3,%%xmm2                                   \n"            \
-    "movq       " MEMACCESS([y_buf]) ",%%xmm3                   \n"            \
-    "lea        " MEMLEA(0x8, [y_buf]) ",%[y_buf]               \n"            \
-    "punpcklbw  %%xmm3,%%xmm3                                   \n"            \
-    "pmulhuw    " MEMACCESS2(192, [yuvconstants]) ",%%xmm3      \n"            \
-    "paddsw     %%xmm3,%%xmm0                                   \n"            \
-    "paddsw     %%xmm3,%%xmm1                                   \n"            \
-    "paddsw     %%xmm3,%%xmm2                                   \n"            \
+    "pmulhuw    " MEMACCESS2(192, [yuvconstants]) ",%%xmm4      \n"            \
+    "paddsw     %%xmm4,%%xmm0                                   \n"            \
+    "paddsw     %%xmm4,%%xmm1                                   \n"            \
+    "paddsw     %%xmm4,%%xmm2                                   \n"            \
     "psraw      $0x6,%%xmm0                                     \n"            \
     "psraw      $0x6,%%xmm1                                     \n"            \
     "psraw      $0x6,%%xmm2                                     \n"            \
     "packuswb   %%xmm0,%%xmm0                                   \n"            \
     "packuswb   %%xmm1,%%xmm1                                   \n"            \
     "packuswb   %%xmm2,%%xmm2                                   \n"
 
 // Store 8 ARGB values. Assumes XMM5 is zero.
 #define STOREARGB                                                              \
     "punpcklbw  %%xmm1,%%xmm0                                    \n"           \
@@ skipping 56 matching lines @@
     STOREARGB
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_argb]"+r"(dst_argb),  // %[dst_argb]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
 void OMITFP I444ToABGRRow_SSSE3(const uint8* y_buf,
                                 const uint8* u_buf,
                                 const uint8* v_buf,
                                 uint8* dst_abgr,
                                 struct YuvConstants* yuvconstants,
                                 int width) {
   asm volatile (
     "sub       %[u_buf],%[v_buf]               \n"
     "pcmpeqb   %%xmm5,%%xmm5                   \n"
     LABELALIGN
   "1:                                          \n"
     READYUV444
     YUVTORGB(yuvconstants)
     STOREABGR
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_abgr]"+r"(dst_abgr),  // %[dst_abgr]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
 // TODO(fbarchard): Consider putting masks into constants.
 void OMITFP I422ToRGB24Row_SSSE3(const uint8* y_buf,
                                  const uint8* u_buf,
                                  const uint8* v_buf,
                                  uint8* dst_rgb24,
                                  struct YuvConstants* yuvconstants,
                                  int width) {
@@ skipping 25 matching lines @@
 // TODO(fbarchard): Make width a register for 32 bit.
 #if defined(__i386__) && defined(__pic__)
     [width]"+m"(width)     // %[width]
 #else
     [width]"+rm"(width)    // %[width]
 #endif
   : [yuvconstants]"r"(yuvconstants),  // %[yuvconstants]
     [kShuffleMaskARGBToRGB24_0]"m"(kShuffleMaskARGBToRGB24_0),
     [kShuffleMaskARGBToRGB24]"m"(kShuffleMaskARGBToRGB24)
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5", "xmm6"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6"
   );
 }
 
 void OMITFP I422ToRAWRow_SSSE3(const uint8* y_buf,
                                const uint8* u_buf,
                                const uint8* v_buf,
                                uint8* dst_raw,
                                struct YuvConstants* yuvconstants,
                                int width) {
   asm volatile (
@@ skipping 24 matching lines @@
 // TODO(fbarchard): Make width a register for 32 bit.
 #if defined(__i386__) && defined(__pic__)
     [width]"+m"(width)    // %[width]
 #else
     [width]"+rm"(width)    // %[width]
 #endif
   : [yuvconstants]"r"(yuvconstants),  // %[yuvconstants]
     [kShuffleMaskARGBToRAW_0]"m"(kShuffleMaskARGBToRAW_0),
     [kShuffleMaskARGBToRAW]"m"(kShuffleMaskARGBToRAW)
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5", "xmm6"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6"
   );
 }
 
 void OMITFP I422ToARGBRow_SSSE3(const uint8* y_buf,
                                 const uint8* u_buf,
                                 const uint8* v_buf,
                                 uint8* dst_argb,
                                 struct YuvConstants* yuvconstants,
                                 int width) {
   asm volatile (
     "sub       %[u_buf],%[v_buf]               \n"
     "pcmpeqb   %%xmm5,%%xmm5                   \n"
     LABELALIGN
   "1:                                          \n"
     READYUV422
     YUVTORGB(yuvconstants)
     STOREARGB
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_argb]"+r"(dst_argb),  // %[dst_argb]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
 void OMITFP I411ToARGBRow_SSSE3(const uint8* y_buf,
                                 const uint8* u_buf,
                                 const uint8* v_buf,
                                 uint8* dst_argb,
                                 struct YuvConstants* yuvconstants,
                                 int width) {
   asm volatile (
     "sub       %[u_buf],%[v_buf]               \n"
     "pcmpeqb   %%xmm5,%%xmm5                   \n"
     LABELALIGN
   "1:                                          \n"
     READYUV411
     YUVTORGB(yuvconstants)
     STOREARGB
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_argb]"+r"(dst_argb),  // %[dst_argb]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
 void OMITFP NV12ToARGBRow_SSSE3(const uint8* y_buf,
                                 const uint8* uv_buf,
                                 uint8* dst_argb,
                                 struct YuvConstants* yuvconstants,
                                 int width) {
   asm volatile (
     "pcmpeqb   %%xmm5,%%xmm5                   \n"
     LABELALIGN
   "1:                                          \n"
     READNV12
     YUVTORGB(yuvconstants)
     STOREARGB
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [uv_buf]"+r"(uv_buf),    // %[uv_buf]
     [dst_argb]"+r"(dst_argb),  // %[dst_argb]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   // Does not use r14.
-  : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+  : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
+void OMITFP YUY2ToARGBRow_SSSE3(const uint8* yuy2_buf,
+                                uint8* dst_argb,
+                                struct YuvConstants* yuvconstants,
+                                int width) {
+  asm volatile (
+    "pcmpeqb   %%xmm5,%%xmm5                   \n"
+    LABELALIGN
+  "1:                                          \n"
+    READYUY2
+    YUVTORGB(yuvconstants)
+    STOREARGB
+    "sub       $0x8,%[width]                   \n"
+    "jg        1b                              \n"
+  : [yuy2_buf]"+r"(yuy2_buf),    // %[yuy2_buf]
+    [dst_argb]"+r"(dst_argb),  // %[dst_argb]
+    [width]"+rm"(width)    // %[width]
+  : [yuvconstants]"r"(yuvconstants), // %[yuvconstants]
+    [kShuffleYUY2Y]"m"(kShuffleYUY2Y),
+    [kShuffleYUY2UV]"m"(kShuffleYUY2UV)
+  // Does not use r14.
+  : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
+  );
+}
+
+void OMITFP UYVYToARGBRow_SSSE3(const uint8* uyvy_buf,
+                                uint8* dst_argb,
+                                struct YuvConstants* yuvconstants,
+                                int width) {
+  asm volatile (
+    "pcmpeqb   %%xmm5,%%xmm5                   \n"
+    LABELALIGN
+  "1:                                          \n"
+    READUYVY
+    YUVTORGB(yuvconstants)
+    STOREARGB
+    "sub       $0x8,%[width]                   \n"
+    "jg        1b                              \n"
+  : [uyvy_buf]"+r"(uyvy_buf),    // %[uyvy_buf]
+    [dst_argb]"+r"(dst_argb),  // %[dst_argb]
+    [width]"+rm"(width)    // %[width]
+  : [yuvconstants]"r"(yuvconstants), // %[yuvconstants]
+    [kShuffleUYVYY]"m"(kShuffleUYVYY),
+    [kShuffleUYVYUV]"m"(kShuffleUYVYUV)
+  // Does not use r14.
+  : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
+  );
+}
+
 void OMITFP I422ToBGRARow_SSSE3(const uint8* y_buf,
                                 const uint8* u_buf,
                                 const uint8* v_buf,
                                 uint8* dst_bgra,
                                 struct YuvConstants* yuvconstants,
                                 int width) {
   asm volatile (
     "sub       %[u_buf],%[v_buf]               \n"
     "pcmpeqb   %%xmm5,%%xmm5                   \n"
     LABELALIGN
   "1:                                          \n"
     READYUV422
     YUVTORGB(yuvconstants)
     STOREBGRA
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_bgra]"+r"(dst_bgra),  // %[dst_bgra]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
 void OMITFP I422ToABGRRow_SSSE3(const uint8* y_buf,
                                 const uint8* u_buf,
                                 const uint8* v_buf,
                                 uint8* dst_abgr,
                                 struct YuvConstants* yuvconstants,
                                 int width) {
   asm volatile (
     "sub       %[u_buf],%[v_buf]               \n"
     "pcmpeqb   %%xmm5,%%xmm5                   \n"
     LABELALIGN
   "1:                                          \n"
     READYUV422
     YUVTORGB(yuvconstants)
     STOREABGR
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_abgr]"+r"(dst_abgr),  // %[dst_abgr]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
 void OMITFP I422ToRGBARow_SSSE3(const uint8* y_buf,
                                 const uint8* u_buf,
                                 const uint8* v_buf,
                                 uint8* dst_rgba,
                                 struct YuvConstants* yuvconstants,
                                 int width) {
   asm volatile (
     "sub       %[u_buf],%[v_buf]               \n"
     "pcmpeqb   %%xmm5,%%xmm5                   \n"
     LABELALIGN
   "1:                                          \n"
     READYUV422
     YUVTORGB(yuvconstants)
     STORERGBA
     "sub       $0x8,%[width]                   \n"
     "jg        1b                              \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_rgba]"+r"(dst_rgba),  // %[dst_rgba]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 
 #endif  // HAS_I422TOARGBROW_SSSE3
 
 // Read 8 UV from 422, upsample to 16 UV.
 #define READYUV422_AVX2                                                        \
     "vmovq       " MEMACCESS([u_buf]) ",%%xmm0                      \n"        \
     MEMOPREG(vmovq, 0x00, [u_buf], [v_buf], 1, xmm1)                           \
     "lea        " MEMLEA(0x8, [u_buf]) ",%[u_buf]                   \n"        \
@@ skipping 58 matching lines @@
     "sub       $0x10,%[width]                  \n"
     "jg        1b                              \n"
     "vzeroupper                                \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_bgra]"+r"(dst_bgra),  // %[dst_bgra]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 #endif  // HAS_I422TOBGRAROW_AVX2
 
 #if defined(HAS_I422TOARGBROW_AVX2)
 // 16 pixels
 // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ARGB (64 bytes).
 void OMITFP I422ToARGBRow_AVX2(const uint8* y_buf,
                                const uint8* u_buf,
                                const uint8* v_buf,
@@ skipping 22 matching lines @@
     "sub       $0x10,%[width]                  \n"
     "jg        1b                              \n"
     "vzeroupper                                \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_argb]"+r"(dst_argb),  // %[dst_argb]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 #endif  // HAS_I422TOARGBROW_AVX2
 
 #if defined(HAS_I422TOABGRROW_AVX2)
 // 16 pixels
 // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 ABGR (64 bytes).
 void OMITFP I422ToABGRRow_AVX2(const uint8* y_buf,
                                const uint8* u_buf,
                                const uint8* v_buf,
@@ skipping 21 matching lines @@
     "sub       $0x10,%[width]                  \n"
     "jg        1b                              \n"
     "vzeroupper                                \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_argb]"+r"(dst_argb),  // %[dst_argb]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 #endif  // HAS_I422TOABGRROW_AVX2
 
 #if defined(HAS_I422TORGBAROW_AVX2)
 // 16 pixels
 // 8 UV values upsampled to 16 UV, mixed with 16 Y producing 16 RGBA (64 bytes).
 void OMITFP I422ToRGBARow_AVX2(const uint8* y_buf,
                                const uint8* u_buf,
                                const uint8* v_buf,
@@ skipping 21 matching lines @@
     "sub       $0x10,%[width]                  \n"
     "jg        1b                              \n"
     "vzeroupper                                \n"
   : [y_buf]"+r"(y_buf),    // %[y_buf]
     [u_buf]"+r"(u_buf),    // %[u_buf]
     [v_buf]"+r"(v_buf),    // %[v_buf]
     [dst_argb]"+r"(dst_argb),  // %[dst_argb]
     [width]"+rm"(width)    // %[width]
   : [yuvconstants]"r"(yuvconstants)  // %[yuvconstants]
   : "memory", "cc", NACL_R14
-    "xmm0", "xmm1", "xmm2", "xmm3", "xmm5"
+    "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5"
   );
 }
 #endif  // HAS_I422TORGBAROW_AVX2
 
 #ifdef HAS_I400TOARGBROW_SSE2
 void I400ToARGBRow_SSE2(const uint8* y_buf, uint8* dst_argb, int width) {
   asm volatile (
     "mov       $0x4a354a35,%%eax               \n"  // 4a35 = 18997 = 1.164
     "movd      %%eax,%%xmm2                    \n"
     "pshufd    $0x0,%%xmm2,%%xmm2              \n"
@@ skipping 3258 matching lines @@
   );
 }
 #endif  // HAS_ARGBLUMACOLORTABLEROW_SSSE3
 
 #endif  // defined(__x86_64__) || defined(__i386__)
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif