Revert "libyuv: update to 2f101fdb"

Compile failures on linux platform.  

BUG=webm:1253

This reverts commit aa81375d73ee33d382e7f717c519db6159e497ee.

Change-Id: Ibab2c4827bc21518dc03c6e9716b5015cff56fc7

1 files changed, 150 insertions, 21 deletions

diff --git a/third_party/libyuv/source/row_mips.cc b/third_party/libyuv/source/row_mips.cc
index 285f0b5ad..cfc9ffe03 100644
--- a/third_party/libyuv/source/row_mips.cc
+++ b/third_party/libyuv/source/row_mips.cc
@@ -375,13 +375,13 @@ void CopyRow_MIPS(const uint8* src, uint8* dst, int count) {
 }
 #endif  // HAS_COPYROW_MIPS
 
-// DSPR2 functions
+// MIPS DSPR2 functions
 #if !defined(LIBYUV_DISABLE_MIPS) && defined(__mips_dsp) && \
     (__mips_dsp_rev >= 2) && \
     (_MIPS_SIM == _MIPS_SIM_ABI32) && (__mips_isa_rev < 6)
 
-void SplitUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
-                      int width) {
+void SplitUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
+                           int width) {
   __asm__ __volatile__ (
     ".set push                                     \n"
     ".set noreorder                                \n"
@@ -389,6 +389,7 @@ void SplitUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
     "blez            $t4, 2f                       \n"
     " andi           %[width], %[width], 0xf       \n"  // residual
 
+    ".p2align        2                             \n"
   "1:                                              \n"
     "addiu           $t4, $t4, -1                  \n"
     "lw              $t0, 0(%[src_uv])             \n"  // V1 | U1 | V0 | U0
@@ -446,7 +447,7 @@ void SplitUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
   );
 }
 
-void MirrorRow_DSPR2(const uint8* src, uint8* dst, int width) {
+void MirrorRow_MIPS_DSPR2(const uint8* src, uint8* dst, int width) {
   __asm__ __volatile__ (
     ".set push                             \n"
     ".set noreorder                        \n"
@@ -456,6 +457,7 @@ void MirrorRow_DSPR2(const uint8* src, uint8* dst, int width) {
     "blez      $t4, 2f                     \n"
     " addu     %[src], %[src], %[width]    \n"  // src += width
 
+    ".p2align  2                           \n"
    "1:                                     \n"
     "lw        $t0, -16(%[src])            \n"  // |3|2|1|0|
     "lw        $t1, -12(%[src])            \n"  // |7|6|5|4|
@@ -496,10 +498,10 @@ void MirrorRow_DSPR2(const uint8* src, uint8* dst, int width) {
   );
 }
 
-void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
-                       int width) {
-  int x;
-  int y;
+void MirrorUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
+                            int width) {
+  int x = 0;
+  int y = 0;
   __asm__ __volatile__ (
     ".set push                                    \n"
     ".set noreorder                               \n"
@@ -510,6 +512,7 @@ void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
     "blez            %[x], 2f                     \n"
     " addu           %[src_uv], %[src_uv], $t4    \n"
 
+    ".p2align        2                            \n"
    "1:                                            \n"
     "lw              $t0, -32(%[src_uv])          \n"  // |3|2|1|0|
     "lw              $t1, -28(%[src_uv])          \n"  // |7|6|5|4|
@@ -579,7 +582,7 @@ void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
         [dst_u] "+r" (dst_u),
         [dst_v] "+r" (dst_v),
         [x] "=&r" (x),
-        [y] "=&r" (y)
+        [y] "+r" (y)
       : [width] "r" (width)
       : "t0", "t1", "t2", "t3", "t4",
       "t5", "t7", "t8", "t9"
@@ -593,7 +596,7 @@ void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
 // t8 = | 0 | G1 | 0 | g1 |
 // t2 = | 0 | R0 | 0 | r0 |
 // t1 = | 0 | R1 | 0 | r1 |
-#define YUVTORGB                                                               \
+#define I422ToTransientMipsRGB                                                 \
       "lw                $t0, 0(%[y_buf])       \n"                            \
       "lhu               $t1, 0(%[u_buf])       \n"                            \
       "lhu               $t2, 0(%[v_buf])       \n"                            \
@@ -652,13 +655,11 @@ void MirrorUVRow_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
       "addu.ph           $t2, $t2, $s5          \n"                            \
       "addu.ph           $t1, $t1, $s5          \n"
 
-// TODO(fbarchard): accept yuv conversion constants.
-void I422ToARGBRow_DSPR2(const uint8* y_buf,
-                         const uint8* u_buf,
-                         const uint8* v_buf,
-                         uint8* rgb_buf,
-                         const struct YuvConstants* yuvconstants,
-                         int width) {
+void I422ToARGBRow_MIPS_DSPR2(const uint8* y_buf,
+                              const uint8* u_buf,
+                              const uint8* v_buf,
+                              uint8* rgb_buf,
+                              int width) {
   __asm__ __volatile__ (
     ".set push                                \n"
     ".set noreorder                           \n"
@@ -672,8 +673,9 @@ void I422ToARGBRow_DSPR2(const uint8* y_buf,
     "lui               $s6, 0xff00            \n"
     "ori               $s6, 0xff00            \n"  // |ff|00|ff|00|ff|
 
+    ".p2align          2                      \n"
    "1:                                        \n"
-      YUVTORGB
+      I422ToTransientMipsRGB
 // Arranging into argb format
     "precr.qb.ph       $t4, $t8, $t4          \n"  // |G1|g1|B1|b1|
     "precr.qb.ph       $t5, $t9, $t5          \n"  // |G0|g0|B0|b0|
@@ -715,10 +717,136 @@ void I422ToARGBRow_DSPR2(const uint8* y_buf,
   );
 }
 
+void I422ToABGRRow_MIPS_DSPR2(const uint8* y_buf,
+                              const uint8* u_buf,
+                              const uint8* v_buf,
+                              uint8* rgb_buf,
+                              int width) {
+  __asm__ __volatile__ (
+    ".set push                                \n"
+    ".set noreorder                           \n"
+    "beqz              %[width], 2f           \n"
+    " repl.ph          $s0, 74                \n"  // |YG|YG| = |74|74|
+    "repl.ph           $s1, -25               \n"  // |UG|UG| = |-25|-25|
+    "repl.ph           $s2, -52               \n"  // |VG|VG| = |-52|-52|
+    "repl.ph           $s3, 102               \n"  // |VR|VR| = |102|102|
+    "repl.ph           $s4, 16                \n"  // |0|16|0|16|
+    "repl.ph           $s5, 128               \n"  // |128|128|
+    "lui               $s6, 0xff00            \n"
+    "ori               $s6, 0xff00            \n"  // |ff|00|ff|00|
+
+    ".p2align          2                       \n"
+   "1:                                         \n"
+      I422ToTransientMipsRGB
+// Arranging into abgr format
+    "precr.qb.ph      $t0, $t8, $t1           \n"  // |G1|g1|R1|r1|
+    "precr.qb.ph      $t3, $t9, $t2           \n"  // |G0|g0|R0|r0|
+    "precrq.qb.ph     $t8, $t0, $t3           \n"  // |G1|R1|G0|R0|
+    "precr.qb.ph      $t9, $t0, $t3           \n"  // |g1|r1|g0|r0|
+
+    "precr.qb.ph       $t2, $t4, $t5          \n"  // |B1|b1|B0|b0|
+    "addiu             %[width], -4           \n"
+    "addiu             %[y_buf], 4            \n"
+    "preceu.ph.qbla    $t1, $t2               \n"  // |0 |B1|0 |B0|
+    "preceu.ph.qbra    $t2, $t2               \n"  // |0 |b1|0 |b0|
+    "or                $t1, $t1, $s6          \n"  // |ff|B1|ff|B0|
+    "or                $t2, $t2, $s6          \n"  // |ff|b1|ff|b0|
+    "precrq.ph.w       $t0, $t2, $t9          \n"  // |ff|b1|g1|r1|
+    "precrq.ph.w       $t3, $t1, $t8          \n"  // |ff|B1|G1|R1|
+    "sll               $t9, $t9, 16           \n"
+    "sll               $t8, $t8, 16           \n"
+    "packrl.ph         $t2, $t2, $t9          \n"  // |ff|b0|g0|r0|
+    "packrl.ph         $t1, $t1, $t8          \n"  // |ff|B0|G0|R0|
+// Store results.
+    "sw                $t2, 0(%[rgb_buf])     \n"
+    "sw                $t0, 4(%[rgb_buf])     \n"
+    "sw                $t1, 8(%[rgb_buf])     \n"
+    "sw                $t3, 12(%[rgb_buf])    \n"
+    "bnez              %[width], 1b           \n"
+    " addiu            %[rgb_buf], 16         \n"
+   "2:                                        \n"
+    ".set pop                                 \n"
+      :[y_buf] "+r" (y_buf),
+       [u_buf] "+r" (u_buf),
+       [v_buf] "+r" (v_buf),
+       [width] "+r" (width),
+       [rgb_buf] "+r" (rgb_buf)
+      :
+      : "t0", "t1",  "t2", "t3",  "t4", "t5",
+      "t6", "t7", "t8", "t9",
+      "s0", "s1", "s2", "s3",
+      "s4", "s5", "s6"
+  );
+}
+
+void I422ToBGRARow_MIPS_DSPR2(const uint8* y_buf,
+                              const uint8* u_buf,
+                              const uint8* v_buf,
+                              uint8* rgb_buf,
+                              int width) {
+  __asm__ __volatile__ (
+    ".set push                                \n"
+    ".set noreorder                           \n"
+    "beqz              %[width], 2f           \n"
+    " repl.ph          $s0, 74                \n"  // |YG|YG| = |74 |74 |
+    "repl.ph           $s1, -25               \n"  // |UG|UG| = |-25|-25|
+    "repl.ph           $s2, -52               \n"  // |VG|VG| = |-52|-52|
+    "repl.ph           $s3, 102               \n"  // |VR|VR| = |102|102|
+    "repl.ph           $s4, 16                \n"  // |0|16|0|16|
+    "repl.ph           $s5, 128               \n"  // |128|128|
+    "lui               $s6, 0xff              \n"
+    "ori               $s6, 0xff              \n"  // |00|ff|00|ff|
+
+    ".p2align          2                      \n"
+   "1:                                        \n"
+      I422ToTransientMipsRGB
+      // Arranging into bgra format
+    "precr.qb.ph       $t4, $t4, $t8          \n"  // |B1|b1|G1|g1|
+    "precr.qb.ph       $t5, $t5, $t9          \n"  // |B0|b0|G0|g0|
+    "precrq.qb.ph      $t8, $t4, $t5          \n"  // |B1|G1|B0|G0|
+    "precr.qb.ph       $t9, $t4, $t5          \n"  // |b1|g1|b0|g0|
+
+    "precr.qb.ph       $t2, $t1, $t2          \n"  // |R1|r1|R0|r0|
+    "addiu             %[width], -4           \n"
+    "addiu             %[y_buf], 4            \n"
+    "preceu.ph.qbla    $t1, $t2               \n"  // |0 |R1|0 |R0|
+    "preceu.ph.qbra    $t2, $t2               \n"  // |0 |r1|0 |r0|
+    "sll               $t1, $t1, 8            \n"  // |R1|0 |R0|0 |
+    "sll               $t2, $t2, 8            \n"  // |r1|0 |r0|0 |
+    "or                $t1, $t1, $s6          \n"  // |R1|ff|R0|ff|
+    "or                $t2, $t2, $s6          \n"  // |r1|ff|r0|ff|
+    "precrq.ph.w       $t0, $t9, $t2          \n"  // |b1|g1|r1|ff|
+    "precrq.ph.w       $t3, $t8, $t1          \n"  // |B1|G1|R1|ff|
+    "sll               $t1, $t1, 16           \n"
+    "sll               $t2, $t2, 16           \n"
+    "packrl.ph         $t2, $t9, $t2          \n"  // |b0|g0|r0|ff|
+    "packrl.ph         $t1, $t8, $t1          \n"  // |B0|G0|R0|ff|
+// Store results.
+    "sw                $t2, 0(%[rgb_buf])     \n"
+    "sw                $t0, 4(%[rgb_buf])     \n"
+    "sw                $t1, 8(%[rgb_buf])     \n"
+    "sw                $t3, 12(%[rgb_buf])    \n"
+    "bnez              %[width], 1b           \n"
+    " addiu            %[rgb_buf], 16         \n"
+   "2:                                        \n"
+    ".set pop                                 \n"
+      :[y_buf] "+r" (y_buf),
+       [u_buf] "+r" (u_buf),
+       [v_buf] "+r" (v_buf),
+       [width] "+r" (width),
+       [rgb_buf] "+r" (rgb_buf)
+      :
+      : "t0", "t1",  "t2", "t3",  "t4", "t5",
+      "t6", "t7", "t8", "t9",
+      "s0", "s1", "s2", "s3",
+      "s4", "s5", "s6"
+  );
+}
+
 // Bilinear filter 8x2 -> 8x1
-void InterpolateRow_DSPR2(uint8* dst_ptr, const uint8* src_ptr,
-                          ptrdiff_t src_stride, int dst_width,
-                          int source_y_fraction) {
+void InterpolateRow_MIPS_DSPR2(uint8* dst_ptr, const uint8* src_ptr,
+                               ptrdiff_t src_stride, int dst_width,
+                               int source_y_fraction) {
     int y0_fraction = 256 - source_y_fraction;
     const uint8* src_ptr1 = src_ptr + src_stride;
 
@@ -729,6 +857,7 @@ void InterpolateRow_DSPR2(uint8* dst_ptr, const uint8* src_ptr,
      "replv.ph          $t0, %[y0_fraction]               \n"
      "replv.ph          $t1, %[source_y_fraction]         \n"
 
+    ".p2align           2                                 \n"
    "1:                                                    \n"
      "lw                $t2, 0(%[src_ptr])                \n"
      "lw                $t3, 0(%[src_ptr1])               \n"