Properly enable S32_D16_filter_DX_SSE2 optimization.
authorcommit-bot@chromium.org <commit-bot@chromium.org@2bbb7eff-a529-9590-31e7-b0007b416f81>
Wed, 23 Apr 2014 19:26:15 +0000 (19:26 +0000)
committercommit-bot@chromium.org <commit-bot@chromium.org@2bbb7eff-a529-9590-31e7-b0007b416f81>
Wed, 23 Apr 2014 19:26:15 +0000 (19:26 +0000)
Currently, the S32_D16_filter_DX_SSE2 optimization is only used in
configurations where the maximum SSE level is SSE2.
This patch enables it for higher levels, as well.
Also, refactored the function a bit, to make future modifications
less error-prone.

Author: henrik.smiding@intel.com

Signed-off-by: Henrik Smiding <henrik.smiding@intel.com>
R=reed@google.com, mtklein@google.com, tomhudson@google.com, djsollen@google.com, joakim.landberg@intel.com

Author: henrik.smiding@intel.com

Review URL: https://codereview.chromium.org/239453010

git-svn-id: http://skia.googlecode.com/svn/trunk@14333 2bbb7eff-a529-9590-31e7-b0007b416f81

src/opts/opts_check_SSE2.cpp

index e7677220cb9326480e4fea631e2249398547e1a9..64a087bd61993e524f0c453f93d7eea29283eb4b 100644 (file)
@@ -123,46 +123,54 @@ void SkBitmapProcState::platformConvolutionProcs(SkConvolutionProcs* procs) {
 }
 
 void SkBitmapProcState::platformProcs() {
-    if (cachedHasSSSE3()) {
-        if (fSampleProc32 == S32_opaque_D32_filter_DX) {
+    /* Every optimization in the function requires at least SSE2 */
+    if (!cachedHasSSE2()) {
+        return;
+    }
+
+    /* Check fSampleProc32 */
+    if (fSampleProc32 == S32_opaque_D32_filter_DX) {
+        if (cachedHasSSSE3()) {
             fSampleProc32 = S32_opaque_D32_filter_DX_SSSE3;
-        } else if (fSampleProc32 == S32_alpha_D32_filter_DX) {
-            fSampleProc32 = S32_alpha_D32_filter_DX_SSSE3;
+        } else {
+            fSampleProc32 = S32_opaque_D32_filter_DX_SSE2;
         }
-
-        if (fSampleProc32 == S32_opaque_D32_filter_DXDY) {
+    } else if (fSampleProc32 == S32_opaque_D32_filter_DXDY) {
+        if (cachedHasSSSE3()) {
             fSampleProc32 = S32_opaque_D32_filter_DXDY_SSSE3;
-        } else if (fSampleProc32 == S32_alpha_D32_filter_DXDY) {
-            fSampleProc32 = S32_alpha_D32_filter_DXDY_SSSE3;
         }
-    } else if (cachedHasSSE2()) {
-        if (fSampleProc32 == S32_opaque_D32_filter_DX) {
-            fSampleProc32 = S32_opaque_D32_filter_DX_SSE2;
-        } else if (fSampleProc32 == S32_alpha_D32_filter_DX) {
+    } else if (fSampleProc32 == S32_alpha_D32_filter_DX) {
+        if (cachedHasSSSE3()) {
+            fSampleProc32 = S32_alpha_D32_filter_DX_SSSE3;
+        } else {
             fSampleProc32 = S32_alpha_D32_filter_DX_SSE2;
         }
-
-        if (fSampleProc16 == S32_D16_filter_DX) {
-            fSampleProc16 = S32_D16_filter_DX_SSE2;
+    } else if (fSampleProc32 == S32_alpha_D32_filter_DXDY) {
+        if (cachedHasSSSE3()) {
+            fSampleProc32 = S32_alpha_D32_filter_DXDY_SSSE3;
         }
     }
 
-    if (cachedHasSSSE3() || cachedHasSSE2()) {
-        if (fMatrixProc == ClampX_ClampY_filter_scale) {
-            fMatrixProc = ClampX_ClampY_filter_scale_SSE2;
-        } else if (fMatrixProc == ClampX_ClampY_nofilter_scale) {
-            fMatrixProc = ClampX_ClampY_nofilter_scale_SSE2;
-        }
+    /* Check fSampleProc16 */
+    if (fSampleProc16 == S32_D16_filter_DX) {
+        fSampleProc16 = S32_D16_filter_DX_SSE2;
+    }
 
-        if (fMatrixProc == ClampX_ClampY_filter_affine) {
-            fMatrixProc = ClampX_ClampY_filter_affine_SSE2;
-        } else if (fMatrixProc == ClampX_ClampY_nofilter_affine) {
-            fMatrixProc = ClampX_ClampY_nofilter_affine_SSE2;
-        }
-        if (c_hqfilter_sse) {
-            if (fShaderProc32 == highQualityFilter32) {
-                fShaderProc32 = highQualityFilter_SSE2;
-            }
+    /* Check fMatrixProc */
+    if (fMatrixProc == ClampX_ClampY_filter_scale) {
+        fMatrixProc = ClampX_ClampY_filter_scale_SSE2;
+    } else if (fMatrixProc == ClampX_ClampY_nofilter_scale) {
+        fMatrixProc = ClampX_ClampY_nofilter_scale_SSE2;
+    } else if (fMatrixProc == ClampX_ClampY_filter_affine) {
+        fMatrixProc = ClampX_ClampY_filter_affine_SSE2;
+    } else if (fMatrixProc == ClampX_ClampY_nofilter_affine) {
+        fMatrixProc = ClampX_ClampY_nofilter_affine_SSE2;
+    }
+
+    /* Check fShaderProc32 */
+    if (c_hqfilter_sse) {
+        if (fShaderProc32 == highQualityFilter32) {
+            fShaderProc32 = highQualityFilter_SSE2;
         }
     }
 }