Enable ARGBToYMatrixRow_RVV/RGBAToYMatrixRow_RVV/RGBToYMatrixRow_RVV

Run on SiFive internal FPGA:

ARGBToJ400_Opt (~6x vs scalar)
RGBAToJ400_Opt (~6x vs scalar)
RGB24ToJ400_Opt (~5.5x vs scalar)

LIBYUV_WIDTH=1280 LIBYUV_HEIGHT=720 LIBYUV_REPEAT=10

Change-Id: Ia3ce8cea7962fbd8618cc23e850a7913c9cabf4f
Signed-off-by: Bruce Lai <bruce.lai@sifive.com>
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4521783
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>

1 files changed, 86 insertions, 1 deletions

diff --git a/source/convert_from_argb.cc b/source/convert_from_argb.cc
index 7e6d8647..897b2c11 100644
--- a/source/convert_from_argb.cc
+++ b/source/convert_from_argb.cc
@@ -132,6 +132,11 @@ int ARGBToI444(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     ARGBToUV444Row(src_argb, dst_u, dst_v, width);
@@ -256,6 +261,11 @@ int ARGBToI422(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     ARGBToUVRow(src_argb, 0, dst_u, dst_v, width);
@@ -373,6 +383,11 @@ int ARGBToNV12(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
 #if defined(HAS_MERGEUVROW_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     MergeUVRow_ = MergeUVRow_Any_SSE2;
@@ -551,6 +566,11 @@ int ARGBToNV21(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
 #if defined(HAS_MERGEUVROW_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     MergeUVRow_ = MergeUVRow_Any_SSE2;
@@ -726,6 +746,11 @@ int ABGRToNV12(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYRow = ABGRToYRow_RVV;
+  }
+#endif
 #if defined(HAS_MERGEUVROW_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     MergeUVRow_ = MergeUVRow_Any_SSE2;
@@ -902,6 +927,11 @@ int ABGRToNV21(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYRow = ABGRToYRow_RVV;
+  }
+#endif
 #if defined(HAS_MERGEUVROW_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     MergeUVRow_ = MergeUVRow_Any_SSE2;
@@ -1085,6 +1115,11 @@ int ARGBToYUY2(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
 #if defined(HAS_I422TOYUY2ROW_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     I422ToYUY2Row = I422ToYUY2Row_Any_SSE2;
@@ -1264,6 +1299,11 @@ int ARGBToUYVY(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
 #if defined(HAS_I422TOUYVYROW_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     I422ToUYVYRow = I422ToUYVYRow_Any_SSE2;
@@ -1405,6 +1445,11 @@ int ARGBToI400(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ARGBToYRow = ARGBToYRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     ARGBToYRow(src_argb, dst_y, width);
@@ -2104,6 +2149,11 @@ int ARGBToJ420(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+     ARGBToYJRow = ARGBToYJRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height - 1; y += 2) {
     ARGBToUVJRow(src_argb, src_stride_argb, dst_uj, dst_vj, width);
@@ -2235,6 +2285,11 @@ int ARGBToJ422(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+     ARGBToYJRow = ARGBToYJRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     ARGBToUVJRow(src_argb, 0, dst_uj, dst_vj, width);
@@ -2304,6 +2359,11 @@ int ARGBToJ400(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+     ARGBToYJRow = ARGBToYJRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     ARGBToYJRow(src_argb, dst_yj, width);
@@ -2386,6 +2446,11 @@ int RGBAToJ400(const uint8_t* src_rgba,
     }
   }
 #endif
+#if defined(HAS_RGBATOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RGBAToYJRow = RGBAToYJRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     RGBAToYJRow(src_rgba, dst_yj, width);
@@ -2496,6 +2561,11 @@ int ABGRToJ420(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYJRow = ABGRToYJRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height - 1; y += 2) {
     ABGRToUVJRow(src_abgr, src_stride_abgr, dst_uj, dst_vj, width);
@@ -2623,6 +2693,11 @@ int ABGRToJ422(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYJRow = ABGRToYJRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     ABGRToUVJRow(src_abgr, 0, dst_uj, dst_vj, width);
@@ -2708,6 +2783,11 @@ int ABGRToJ400(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    ABGRToYJRow = ABGRToYJRow_RVV;
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     ABGRToYJRow(src_abgr, dst_yj, width);
@@ -2846,7 +2926,7 @@ int ARGBToAB64(const uint8_t* src_argb,
 }
 
 // Enabled if 1 pass is available
-#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA)
+#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || defined(HAS_RAWTOYJROW_RVV)
 #define HAS_RAWTOYJROW
 #endif
 
@@ -2928,6 +3008,11 @@ int RAWToJNV21(const uint8_t* src_raw,
     }
   }
 #endif
+#if defined(HAS_RAWTOYJROW_RVV)
+  if (TestCpuFlag(kCpuHasRVV)) {
+    RAWToYJRow = RAWToYJRow_RVV;
+  }
+#endif
 
 // Other platforms do intermediate conversion from RAW to ARGB.
 #else  // HAS_RAWTOYJROW