Merge third_party/libvpx from https://chromium.googlesource.com/chromium/deps/libvpx.git at efe9712d52c2d216fb3d1ceb508b8148847a7e4b

This commit was generated by merge_from_chromium.py.

Change-Id: I001fde9953b66e6dbdb5e0c751ef66aecc5b3f08

2 files changed, 44 insertions, 49 deletions

diff --git a/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c b/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c
index 0bec7fb..e103476 100644
--- a/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c
+++ b/source/libvpx/vp8/common/arm/neon/loopfilter_neon.c
@@ -251,9 +251,38 @@ void vp8_loop_filter_horizontal_edge_uv_neon(
     return;
 }
 
+#if (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))
+#warning Using GCC 4.6 is not recommended
+// Some versions of gcc4.6 do not correctly process vst4_lane_u8. When built
+// with any gcc4.6, use the C code.
+extern void vp8_loop_filter_vertical_edge_c(unsigned char *s, int p,
+                                            const unsigned char *blimit,
+                                            const unsigned char *limit,
+                                            const unsigned char *thresh,
+                                            int count);
+
+void vp8_loop_filter_vertical_edge_y_neon(
+        unsigned char *src,
+        int pitch,
+        unsigned char blimit,
+        unsigned char limit,
+        unsigned char thresh) {
+  vp8_loop_filter_vertical_edge_c(src, pitch, &blimit, &limit, &thresh, 2);
+}
+
+void vp8_loop_filter_vertical_edge_uv_neon(
+        unsigned char *u,
+        int pitch,
+        unsigned char blimit,
+        unsigned char limit,
+        unsigned char thresh,
+        unsigned char *v) {
+  vp8_loop_filter_vertical_edge_c(u, pitch, &blimit, &limit, &thresh, 1);
+  vp8_loop_filter_vertical_edge_c(v, pitch, &blimit, &limit, &thresh, 1);
+}
+#else
 static INLINE void write_4x8(unsigned char *dst, int pitch,
                              const uint8x8x4_t result) {
-#if (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7))
     vst4_lane_u8(dst, result, 0);
     dst += pitch;
     vst4_lane_u8(dst, result, 1);
@@ -269,54 +298,6 @@ static INLINE void write_4x8(unsigned char *dst, int pitch,
     vst4_lane_u8(dst, result, 6);
     dst += pitch;
     vst4_lane_u8(dst, result, 7);
-#else
-    /*
-     * uint8x8x4_t result
-    00 01 02 03 | 04 05 06 07
-    10 11 12 13 | 14 15 16 17
-    20 21 22 23 | 24 25 26 27
-    30 31 32 33 | 34 35 36 37
-    ---
-    * after vtrn_u16
-    00 01 20 21 | 04 05 24 25
-    02 03 22 23 | 06 07 26 27
-    10 11 30 31 | 14 15 34 35
-    12 13 32 33 | 16 17 36 37
-    ---
-    * after vtrn_u8
-    00 10 20 30 | 04 14 24 34
-    01 11 21 31 | 05 15 25 35
-    02 12 22 32 | 06 16 26 36
-    03 13 23 33 | 07 17 27 37
-    */
-    const uint16x4x2_t r02_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[0]),
-                                          vreinterpret_u16_u8(result.val[2]));
-    const uint16x4x2_t r13_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[1]),
-                                          vreinterpret_u16_u8(result.val[3]));
-    const uint8x8x2_t r01_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[0]),
-                                       vreinterpret_u8_u16(r13_u16.val[0]));
-    const uint8x8x2_t r23_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[1]),
-                                       vreinterpret_u8_u16(r13_u16.val[1]));
-    const uint32x2_t x_0_4 = vreinterpret_u32_u8(r01_u8.val[0]);
-    const uint32x2_t x_1_5 = vreinterpret_u32_u8(r01_u8.val[1]);
-    const uint32x2_t x_2_6 = vreinterpret_u32_u8(r23_u8.val[0]);
-    const uint32x2_t x_3_7 = vreinterpret_u32_u8(r23_u8.val[1]);
-    vst1_lane_u32((uint32_t *)dst, x_0_4, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_1_5, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_2_6, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_3_7, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_0_4, 1);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_1_5, 1);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_2_6, 1);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_3_7, 1);
-#endif
 }
 
 void vp8_loop_filter_vertical_edge_y_neon(
@@ -547,3 +528,4 @@ void vp8_loop_filter_vertical_edge_uv_neon(
     vd = v - 2;
     write_4x8(vd, pitch, q4ResultH);
 }
+#endif  // (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))
diff --git a/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c b/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c
index d6b67f8..ffa3d91 100644
--- a/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c
+++ b/source/libvpx/vp8/encoder/arm/neon/vp8_shortwalsh4x4_neon.c
@@ -10,6 +10,18 @@
 
 #include <arm_neon.h>
 
+#if (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))
+#warning Using GCC 4.6 is not recommended
+// Some versions of gcc4.6 do not correctly process this function. When built
+// with any gcc4.6, use the C code.
+#include "./vp8_rtcd.h"
+void vp8_short_walsh4x4_neon(
+        int16_t *input,
+        int16_t *output,
+        int pitch) {
+  vp8_short_walsh4x4_c(input, output, pitch);
+}
+#else
 void vp8_short_walsh4x4_neon(
         int16_t *input,
         int16_t *output,
@@ -116,3 +128,4 @@ void vp8_short_walsh4x4_neon(
     vst1q_s16(output + 8, q1s16);
     return;
 }
+#endif  // (__GNUC__ == 4 && (__GNUC_MINOR__ == 6))