Add SSSE3 acceleration for S32_D16_filter_DXDY

src/opts/SkBitmapProcState_opts_SSSE3.cpp

Index: src/opts/SkBitmapProcState_opts_SSSE3.cpp
diff --git a/src/opts/SkBitmapProcState_opts_SSSE3.cpp b/src/opts/SkBitmapProcState_opts_SSSE3.cpp
index 165f1f5ce068ffe560fa85ddbbaf80f93ad0ea68..746a264920d11c163f296c54b20a6c0f02243e4a 100644
--- a/src/opts/SkBitmapProcState_opts_SSSE3.cpp
+++ b/src/opts/SkBitmapProcState_opts_SSSE3.cpp
@@ -6,6 +6,7 @@
  */
 
 #include "SkBitmapProcState_opts_SSSE3.h"
+#include "SkColorPriv.h"
 #include "SkPaint.h"
 #include "SkUtils.h"
 
@@ -720,17 +721,27 @@ void S32_alpha_D32_filter_DX_SSSE3(const SkBitmapProcState& s,
 }
 
 void S32_opaque_D32_filter_DXDY_SSSE3(const SkBitmapProcState& s,
-                                    const uint32_t* xy,
-                                    int count, uint32_t* colors) {
+                                      const uint32_t* xy,
+                                      int count, uint32_t* colors) {
     S32_generic_D32_filter_DXDY_SSSE3<false>(s, xy, count, colors);
 }
 
 void S32_alpha_D32_filter_DXDY_SSSE3(const SkBitmapProcState& s,
-                                   const uint32_t* xy,
-                                   int count, uint32_t* colors) {
+                                     const uint32_t* xy,
+                                     int count, uint32_t* colors) {
     S32_generic_D32_filter_DXDY_SSSE3<true>(s, xy, count, colors);
 }
 
+void S32_D16_filter_DXDY_SSSE3(const SkBitmapProcState& s,
+                               const uint32_t* xy,
+                               int count, uint16_t* colors) {
+    SkAutoTMalloc<uint32_t> colors32(count);

[qiankun, 2014/12/02 10:51:41]

If "uint32_t colors32[128];" used here, it's more faster. But the array size
depends on count. In current code, count will not exceed 64. The data as
follows:
         bitmap_BGRA_8888_update_scale_rotate_bilerp     112us -> 72.3us       
0.64x
bitmap_BGRA_8888_update_volatile_scale_rotate_bilerp     115us ->   67us       
0.58x
                bitmap_BGRA_8888_scale_rotate_bilerp     115us -> 66.9us       
0.58x

[mtklein, 2014/12/02 15:19:05]

Might be worth trying SkAutoSTMalloc<64, uint32_t>, which should be both safe
and stay on the stack when count <= 64.

[qiankun1, 2014/12/02 15:58:43]

Done. Where is these performance improvement from? Just heap vs. stack?
Now the data are:
                bitmap_BGRA_8888_scale_rotate_bilerp     115us -> 70.5us       
0.61x
bitmap_BGRA_8888_update_volatile_scale_rotate_bilerp     115us -> 70.5us       
0.61x
         bitmap_BGRA_8888_update_scale_rotate_bilerp     112us ->   68us       
0.6x

[mtklein, 2014/12/02 16:52:22]

Right, just heap vs. stack.  There's no difference between heap memory and stack
memory, except
  - the stack usually has better cache locality, and
  - allocating on the stack is essentially free compared to calling malloc/free.

When we have no statically provable upper bound, but do have a practical upper
bound, we find it's generally a pretty big win to allocate that practical upper
bound on the stack and fall back to the heap only in exceptional cases.

+    S32_generic_D32_filter_DXDY_SSSE3<false>(s, xy, count, colors32);
+    for(int i = 0; i < count; i++) {
+        *colors++ = SkPixel32ToPixel16(colors32[i]);
+    }
+}
+
 #else // SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSSE3
 
 void S32_opaque_D32_filter_DX_SSSE3(const SkBitmapProcState& s,
@@ -746,14 +757,20 @@ void S32_alpha_D32_filter_DX_SSSE3(const SkBitmapProcState& s,
 }
 
 void S32_opaque_D32_filter_DXDY_SSSE3(const SkBitmapProcState& s,
-                                    const uint32_t* xy,
-                                    int count, uint32_t* colors) {
+                                      const uint32_t* xy,
+                                      int count, uint32_t* colors) {
     sk_throw();
 }
 
 void S32_alpha_D32_filter_DXDY_SSSE3(const SkBitmapProcState& s,
-                                   const uint32_t* xy,
-                                   int count, uint32_t* colors) {
+                                     const uint32_t* xy,
+                                     int count, uint32_t* colors) {
+    sk_throw();
+}
+
+void S32_D16_filter_DXDY_SSSE3(const SkBitmapProcState& s,
+                               const uint32_t* xy,
+                               int count, uint32_t* colors) {
     sk_throw();
 }