Add MSA optimized ARGBToRGB24Row_MSA and ARGBToRAWRow_MSA functions

source/row_msa.cc

 /*
  *  Copyright 2016 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 211 matching lines @@
     YUVTORGB(src0, src1, src2, vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg,
              vec_br, vec_yg, vec0, vec1, vec2);
     STOREARGB(vec0, vec1, vec2, const_255, rgb_buf);
     src_y += 8;
     src_u += 4;
     src_v += 4;
     rgb_buf += 32;
   }
 }
 
-void YUVTORGBARow_MSA(const uint8* src_y,
-                      const uint8* src_u,
-                      const uint8* src_v,
-                      uint8* rgb_buf,
-                      const struct YuvConstants* yuvconstants,
-                      int width) {
+void I422ToRGBARow_MSA(const uint8* src_y,
+                       const uint8* src_u,
+                       const uint8* src_v,
+                       uint8* rgb_buf,
+                       const struct YuvConstants* yuvconstants,
+                       int width) {
   int x;
   v16u8 src0, src1, src2;
   v8i16 vec0, vec1, vec2;
   v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;
   v16u8 const_255 = (v16u8)__msa_ldi_b(255);
 
   YUVTORGB_SETUP(yuvconstants, vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg,
                  vec_br, vec_yg);
 
   for (x = 0; x < width; x += 8) {
@@ skipping 34 matching lines @@
     src3 = (v16u8)__msa_ilvr_b((v16i8)src3, (v16i8)src3);
     STOREARGB(vec0, vec1, vec2, src3, rgb_buf);
     src_y += 8;
     src_u += 4;
     src_v += 4;
     src_a += 8;
     rgb_buf += 32;
   }
 }
 
-void YUVTORGB24Row_MSA(const uint8* src_y,
-                       const uint8* src_u,
-                       const uint8* src_v,
-                       uint8* rgb_buf,
-                       const struct YuvConstants* yuvconstants,
-                       int32 width) {
+void I422ToRGB24Row_MSA(const uint8* src_y,
+                        const uint8* src_u,
+                        const uint8* src_v,
+                        uint8* rgb_buf,
+                        const struct YuvConstants* yuvconstants,
+                        int32 width) {
   int x;
   int64 data_u, data_v;
   v16u8 src0, src1, src2, src3, src4, src5, dst0, dst1, dst2;
   v8i16 vec0, vec1, vec2, vec3, vec4, vec5;
   v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;
   v16u8 reg0, reg1, reg2, reg3;
   v2i64 zero = {0};
   v16i8 shuffler0 = {0, 1, 16, 2, 3, 17, 4, 5, 18, 6, 7, 19, 8, 9, 20, 10};
   v16i8 shuffler1 = {0, 21, 1, 2, 22, 3, 4, 23, 5, 6, 24, 7, 8, 25, 9, 10};
   v16i8 shuffler2 = {26, 6,  7,  27, 8,  9,  28, 10,
@@ skipping 25 matching lines @@
     ST_UB2(dst0, dst1, rgb_buf, 16);
     ST_UB(dst2, (rgb_buf + 32));
     src_y += 16;
     src_u += 8;
     src_v += 8;
     rgb_buf += 48;
   }
 }
 
 // TODO(fbarchard): Consider AND instead of shift to isolate 5 upper bits of R.
-void YUVTORGB565Row_MSA(const uint8* src_y,
-                        const uint8* src_u,
-                        const uint8* src_v,
-                        uint8* dst_rgb565,
-                        const struct YuvConstants* yuvconstants,
-                        int width) {
+void I422ToRGB565Row_MSA(const uint8* src_y,
+                         const uint8* src_u,
+                         const uint8* src_v,
+                         uint8* dst_rgb565,
+                         const struct YuvConstants* yuvconstants,
+                         int width) {
   int x;
   v16u8 src0, src1, src2, dst0;
   v8i16 vec0, vec1, vec2;
   v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;
 
   YUVTORGB_SETUP(yuvconstants, vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg,
                  vec_br, vec_yg);
 
   for (x = 0; x < width; x += 8) {
     READYUV422(src_y, src_u, src_v, src0, src1, src2);
@@ skipping 374 matching lines @@
     dst1 = (v16u8)__msa_pckev_b((v16i8)reg5, (v16i8)reg4);
     ST_UB(dst0, dst_u);
     ST_UB(dst1, dst_v);
     src_argb0 += 128;
     src_argb0_next += 128;
     dst_u += 16;
     dst_v += 16;
   }
 }
 
+void ARGBToRGB24Row_MSA(const uint8* src_argb, uint8* dst_rgb, int width) {
+  int x;
+  v16u8 src0, src1, src2, src3, dst0, dst1, dst2;
+  v16i8 shuffler0 = {0, 1, 2, 4, 5, 6, 8, 9, 10, 12, 13, 14, 16, 17, 18, 20};
+  v16i8 shuffler1 = { 5,  6,  8,  9, 10, 12, 13, 14,
+                     16, 17, 18, 20, 21, 22, 24, 25};
+  v16i8 shuffler2 = {10, 12, 13, 14, 16, 17, 18, 20,
+                     21, 22, 24, 25, 26, 28, 29, 30};
+
+  for (x = 0; x < width; x += 16) {
+    src0 = (v16u8)__msa_ld_b((v16i8*)src_argb, 0);
+    src1 = (v16u8)__msa_ld_b((v16i8*)src_argb, 16);
+    src2 = (v16u8)__msa_ld_b((v16i8*)src_argb, 32);
+    src3 = (v16u8)__msa_ld_b((v16i8*)src_argb, 48);
+    dst0 = (v16u8)__msa_vshf_b(shuffler0, (v16i8)src1, (v16i8)src0);
+    dst1 = (v16u8)__msa_vshf_b(shuffler1, (v16i8)src2, (v16i8)src1);
+    dst2 = (v16u8)__msa_vshf_b(shuffler2, (v16i8)src3, (v16i8)src2);
+    ST_UB2(dst0, dst1, dst_rgb, 16);
+    ST_UB(dst2, (dst_rgb + 32));
+    src_argb += 64;
+    dst_rgb += 48;
+  }
+}
+
+void ARGBToRAWRow_MSA(const uint8* src_argb, uint8* dst_rgb, int width) {
+  int x;
+  v16u8 src0, src1, src2, src3, dst0, dst1, dst2;
+  v16i8 shuffler0 = {2, 1, 0, 6, 5, 4, 10, 9, 8, 14, 13, 12, 18, 17, 16, 22};
+  v16i8 shuffler1 = { 5,  4, 10,  9,  8, 14, 13, 12,
+                     18, 17, 16, 22, 21, 20, 26, 25};
+  v16i8 shuffler2 = { 8, 14, 13, 12, 18, 17, 16, 22,
+                     21, 20, 26, 25, 24, 30, 29, 28};
+
+  for (x = 0; x < width; x += 16) {
+    src0 = (v16u8)__msa_ld_b((v16i8*)src_argb, 0);
+    src1 = (v16u8)__msa_ld_b((v16i8*)src_argb, 16);
+    src2 = (v16u8)__msa_ld_b((v16i8*)src_argb, 32);
+    src3 = (v16u8)__msa_ld_b((v16i8*)src_argb, 48);
+    dst0 = (v16u8)__msa_vshf_b(shuffler0, (v16i8)src1, (v16i8)src0);
+    dst1 = (v16u8)__msa_vshf_b(shuffler1, (v16i8)src2, (v16i8)src1);
+    dst2 = (v16u8)__msa_vshf_b(shuffler2, (v16i8)src3, (v16i8)src2);
+    ST_UB2(dst0, dst1, dst_rgb, 16);
+    ST_UB(dst2, (dst_rgb + 32));
+    src_argb += 64;
+    dst_rgb += 48;
+  }
+}
+
 void ARGB4444ToARGBRow_MSA(const uint8* src_argb4444,
                            uint8* dst_argb,
                            int width) {
   int x;
   v16u8 src0, src1;
   v8u16 vec0, vec1, vec2, vec3;
   v16u8 dst0, dst1, dst2, dst3;
 
   for (x = 0; x < width; x += 16) {
     src0 = (v16u8)__msa_ld_b((v16u8*)src_argb4444, 0);
@@ skipping 15 matching lines @@
     dst_argb += 64;
   }
 }
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif
 
 #endif  // !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)