[WIP] Supporting arm_neon_optional flag for blink platform.

Source/platform/audio/VectorMath.cpp

Index: Source/platform/audio/VectorMath.cpp
diff --git a/Source/platform/audio/VectorMath.cpp b/Source/platform/audio/VectorMath.cpp
index 219ed5463977ac5cc172f1117fd0cca3f9008a9a..9bef4f41c6c66003e2a73cecc1e58739a3046a10 100644
--- a/Source/platform/audio/VectorMath.cpp
+++ b/Source/platform/audio/VectorMath.cpp
@@ -27,11 +27,17 @@
 #if ENABLE(WEB_AUDIO)
 
 #include "platform/audio/VectorMath.h"
+
 #include "wtf/Assertions.h"
 #include "wtf/CPU.h"
+#include <algorithm>
+#include <math.h>
 #include <stdint.h>
 
 #if OS(MACOSX)
+// On the Mac we use the highly optimized versions in Accelerate.framework
+// In 32-bit mode (__ppc__ or __i386__) <Accelerate/Accelerate.h> includes <vecLib/vDSP_translate.h> which defines macros of the same name as
+// our namespaced function names, so we must handle this case differently. Other architectures (64bit, ARM, etc.) do not include this header file.
 #include <Accelerate/Accelerate.h>
 #endif
 
@@ -39,46 +45,43 @@
 #include <emmintrin.h>
 #endif
 
-#if HAVE(ARM_NEON_INTRINSICS)
-#include <arm_neon.h>
-#endif
-
-#include <math.h>
-#include <algorithm>
 
 namespace blink {
 
 namespace VectorMath {
 
 #if OS(MACOSX)
-// On the Mac we use the highly optimized versions in Accelerate.framework
-// In 32-bit mode (__ppc__ or __i386__) <Accelerate/Accelerate.h> includes <vecLib/vDSP_translate.h> which defines macros of the same name as
-// our namespaced function names, so we must handle this case differently. Other architectures (64bit, ARM, etc.) do not include this header file.
 
 void vsmul(const float* sourceP, int sourceStride, const float* scale, float* destP, int destStride, size_t framesToProcess)
 {
-#if CPU(X86)
-    ::vsmul(sourceP, sourceStride, scale, destP, destStride, framesToProcess);
-#else
+#if OS(MACOSX) && !CPU(X86)
     vDSP_vsmul(sourceP, sourceStride, scale, destP, destStride, framesToProcess);
+#elif HAVE(ARM_NEON_INTRINSICS)
+    WTF_CPU_ARM_NEON_WRAP(vsmul)(sourceP, sourceStride, scale, destP, destStride, framesToProcess);
+#else
+    ::vsmul(sourceP, sourceStride, scale, destP, destStride, framesToProcess);
 #endif
 }
 
 void vadd(const float* source1P, int sourceStride1, const float* source2P, int sourceStride2, float* destP, int destStride, size_t framesToProcess)
 {
-#if CPU(X86)
-    ::vadd(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
-#else
+#if OS(MACOSX) && !CPU(X86)
     vDSP_vadd(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
+#elif HAVE(ARM_NEON_INTRINSICS)
+    WTF_CPU_ARM_NEON_WRAP(vadd)(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
+#else
+    ::vadd(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
 #endif
 }
 
 void vmul(const float* source1P, int sourceStride1, const float* source2P, int sourceStride2, float* destP, int destStride, size_t framesToProcess)
 {
-#if CPU(X86)
-    ::vmul(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
-#else
+#if OS(MACOSX) && !CPU(X86)
     vDSP_vmul(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
+#elif HAVE(ARM_NEON_INTRINSICS)
+    WTF_CPU_ARM_NEON_WRAP(vmul)(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
+#else
+    ::vmul(source1P, sourceStride1, source2P, sourceStride2, destP, destStride, framesToProcess);
 #endif
 }