Modify sample buffer size for larger displays.

src/core/SkBitmapProcShader.cpp

 
 /*
  * Copyright 2011 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 #include "SkColorPriv.h"
 #include "SkReadBuffer.h"
 #include "SkWriteBuffer.h"
@@ skipping 187 matching lines @@
 
     fFlags = flags;
 }
 
 SkBitmapProcShader::BitmapProcShaderContext::~BitmapProcShaderContext() {
     // The bitmap proc state has been created outside of the context on memory that will be freed
     // elsewhere. Only call the destructor but leave the freeing of the memory to the caller.
     fState->~SkBitmapProcState();
 }
 
-#define BUF_MAX     128
+/* Defines the buffer size for sample pixel indexes, used in the sample proc function calls.
+ * If the operation is bigger than the buffer, it's split into multiple calls. This split is bad
+ * for the performance of SIMD optimizations.
+ * A display in portrait mode, with a width of 720 pixels, requires a buffer size of at least 361
+ * to run uninterrupted in the most basic operations. (Formula: Width / 2 + 1)
+ */
+#ifdef ANDROID_LARGE_MEMORY_DEVICE

[djsollen, 2014/04/23 14:40:53]

my goal is to remove ANDROID_LARGE_MEMORY_DEVICE so I'm in agreement with Mike
that we should just go with the larger number.

[henrik.smiding, 2014/04/23 15:20:21]

Done.

+    #define BUF_MAX     541 // 1080p phone display
+#else
+    #define BUF_MAX     241 // 480p phone display
+#endif
 
 #define TEST_BUFFER_OVERRITEx
 
 #ifdef TEST_BUFFER_OVERRITE
     #define TEST_BUFFER_EXTRA   32
-    #define TEST_PATTERN    0x88888888
+    #define TEST_PATTERN        0x88888888
 #else
     #define TEST_BUFFER_EXTRA   0
 #endif
 
 void SkBitmapProcShader::BitmapProcShaderContext::shadeSpan(int x, int y, SkPMColor dstC[],
                                                             int count) {
     const SkBitmapProcState& state = *fState;
     if (state.getShaderProc32()) {
         state.getShaderProc32()(state, x, y, dstC, count);
         return;
     }
 
-    uint32_t buffer[BUF_MAX + TEST_BUFFER_EXTRA];
+    // Align buffer to 16 bytes to enable more efficient SIMD optimizations.
+    uint32_t SK_ALIGN(16) buffer[BUF_MAX + TEST_BUFFER_EXTRA];
+
     SkBitmapProcState::MatrixProc   mproc = state.getMatrixProc();
     SkBitmapProcState::SampleProc32 sproc = state.getSampleProc32();
     int max = state.maxCountForBufferSize(sizeof(buffer[0]) * BUF_MAX);
 
     SkASSERT(state.fBitmap->getPixels());
     SkASSERT(state.fBitmap->pixelRef() == NULL ||
              state.fBitmap->pixelRef()->isLocked());
 
     for (;;) {
         int n = count;
@@ skipping 248 matching lines @@
     GrEffectRef* effect = NULL;
     if (paintFilterLevel == SkPaint::kHigh_FilterLevel) {
         effect = GrBicubicEffect::Create(texture, matrix, tm);
     } else {
         effect = GrSimpleTextureEffect::Create(texture, matrix, params);
     }
     GrUnlockAndUnrefCachedBitmapTexture(texture);
     return effect;
 }
 #endif