Add MSA optimized rotate functions (used 16x16 transpose)

source/rotate.cc

 /*
  *  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 11 matching lines @@
 #endif
 
 LIBYUV_API
 void TransposePlane(const uint8* src,
                     int src_stride,
                     uint8* dst,
                     int dst_stride,
                     int width,
                     int height) {
   int i = height;
+#if defined(HAS_TRANSPOSEWX16_MSA)
+  void (*TransposeWx16)(const uint8* src, int src_stride, uint8* dst,
+                        int dst_stride, int width) = TransposeWx16_C;
+#else
   void (*TransposeWx8)(const uint8* src, int src_stride, uint8* dst,
                        int dst_stride, int width) = TransposeWx8_C;
+#endif
 #if defined(HAS_TRANSPOSEWX8_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     TransposeWx8 = TransposeWx8_NEON;
   }
 #endif
 #if defined(HAS_TRANSPOSEWX8_SSSE3)
   if (TestCpuFlag(kCpuHasSSSE3)) {
     TransposeWx8 = TransposeWx8_Any_SSSE3;
     if (IS_ALIGNED(width, 8)) {
       TransposeWx8 = TransposeWx8_SSSE3;
@@ skipping 11 matching lines @@
 #if defined(HAS_TRANSPOSEWX8_DSPR2)
   if (TestCpuFlag(kCpuHasDSPR2)) {
     if (IS_ALIGNED(width, 4) && IS_ALIGNED(src, 4) &&
         IS_ALIGNED(src_stride, 4)) {
       TransposeWx8 = TransposeWx8_Fast_DSPR2;
     } else {
       TransposeWx8 = TransposeWx8_DSPR2;
     }
   }
 #endif
-#if defined(HAS_TRANSPOSEWX8_MSA)
+#if defined(HAS_TRANSPOSEWX16_MSA)
   if (TestCpuFlag(kCpuHasMSA)) {
-    TransposeWx8 = TransposeWx8_Any_MSA;
+    TransposeWx16 = TransposeWx16_Any_MSA;
     if (IS_ALIGNED(width, 16)) {
-      TransposeWx8 = TransposeWx8_MSA;
+      TransposeWx16 = TransposeWx16_MSA;
     }
   }
 #endif
 
+#if defined(HAS_TRANSPOSEWX16_MSA)
+  // Work across the source in 16x16 tiles
+  while (i >= 16) {
+    TransposeWx16(src, src_stride, dst, dst_stride, width);
+    src += 16 * src_stride;  // Go down 16 rows.
+    dst += 16;               // Move over 16 columns.
+    i -= 16;
+  }
+#else
   // Work across the source in 8x8 tiles
   while (i >= 8) {
     TransposeWx8(src, src_stride, dst, dst_stride, width);
     src += 8 * src_stride;  // Go down 8 rows.
     dst += 8;               // Move over 8 columns.
     i -= 8;
   }
+#endif
 
   if (i > 0) {
     TransposeWxH_C(src, src_stride, dst, dst_stride, width, i);
   }
 }
 
 LIBYUV_API
 void RotatePlane90(const uint8* src,
                    int src_stride,
                    uint8* dst,
@@ skipping 120 matching lines @@
 LIBYUV_API
 void TransposeUV(const uint8* src,
                  int src_stride,
                  uint8* dst_a,
                  int dst_stride_a,
                  uint8* dst_b,
                  int dst_stride_b,
                  int width,
                  int height) {
   int i = height;
+#if defined(HAS_TRANSPOSEUVWX16_MSA)
+  void (*TransposeUVWx16)(const uint8* src, int src_stride, uint8* dst_a,
+                          int dst_stride_a, uint8* dst_b, int dst_stride_b,
+                          int width) = TransposeUVWx16_C;
+#else
   void (*TransposeUVWx8)(const uint8* src, int src_stride, uint8* dst_a,
                          int dst_stride_a, uint8* dst_b, int dst_stride_b,
                          int width) = TransposeUVWx8_C;
+#endif
 #if defined(HAS_TRANSPOSEUVWX8_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     TransposeUVWx8 = TransposeUVWx8_NEON;
   }
 #endif
 #if defined(HAS_TRANSPOSEUVWX8_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     TransposeUVWx8 = TransposeUVWx8_Any_SSE2;
     if (IS_ALIGNED(width, 8)) {
       TransposeUVWx8 = TransposeUVWx8_SSE2;
     }
   }
 #endif
 #if defined(HAS_TRANSPOSEUVWX8_DSPR2)
   if (TestCpuFlag(kCpuHasDSPR2) && IS_ALIGNED(width, 2) && IS_ALIGNED(src, 4) &&
       IS_ALIGNED(src_stride, 4)) {
     TransposeUVWx8 = TransposeUVWx8_DSPR2;
   }
 #endif
-#if defined(HAS_TRANSPOSEUVWX8_MSA)
+#if defined(HAS_TRANSPOSEUVWX16_MSA)
   if (TestCpuFlag(kCpuHasMSA)) {
-    TransposeUVWx8 = TransposeUVWx8_Any_MSA;
+    TransposeUVWx16 = TransposeUVWx16_Any_MSA;
     if (IS_ALIGNED(width, 8)) {
-      TransposeUVWx8 = TransposeUVWx8_MSA;
+      TransposeUVWx16 = TransposeUVWx16_MSA;
     }
   }
 #endif
 
+#if defined(HAS_TRANSPOSEUVWX16_MSA)
+  // Work through the source in 8x8 tiles.
+  while (i >= 16) {
+    TransposeUVWx16(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
+                    width);
+    src += 16 * src_stride;  // Go down 16 rows.
+    dst_a += 16;             // Move over 8 columns.
+    dst_b += 16;             // Move over 8 columns.
+    i -= 16;
+  }
+#else
   // Work through the source in 8x8 tiles.
   while (i >= 8) {
     TransposeUVWx8(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
                    width);
     src += 8 * src_stride;  // Go down 8 rows.
     dst_a += 8;             // Move over 8 columns.
     dst_b += 8;             // Move over 8 columns.
     i -= 8;
   }
+#endif
 
   if (i > 0) {
     TransposeUVWxH_C(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
                      width, i);
   }
 }
 
 LIBYUV_API
 void RotateUV90(const uint8* src,
                 int src_stride,
@@ skipping 231 matching lines @@
     default:
       break;
   }
   return -1;
 }
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif