Rewrite memset benches, then use results to add a small-N optimization.

include/core/SkUtils.h

Index: include/core/SkUtils.h
diff --git a/include/core/SkUtils.h b/include/core/SkUtils.h
index d522ae0dea139555f6e8790ac06d8287e5771fd7..bb7214c7d9e6ac50ebd8472c3de3652cb31e75b0 100644
--- a/include/core/SkUtils.h
+++ b/include/core/SkUtils.h
@@ -12,12 +12,31 @@
 
 ///////////////////////////////////////////////////////////////////////////////
 
+// Determined empirically using bench/MemsetBench.cpp on a Nexus 7, Nexus 9, and desktop.
+#if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2 || defined(SK_ARM_HAS_NEON)
+    // Platforms where we can assume an autovectorizer will give us a good inline memset.
+    #define SK_SMALL_MEMSET 1000

[reed1, 2015/04/09 20:30:24]

I know I suggested this be a #define. Since this is a public header, and its
#defines will be visible to clients code, I wonder if we think this name should
be longer/more-obscure, and/or a const int, or just leave it as is.

[mtklein, 2015/04/09 20:36:42]

Hmm.  Actually, I can just undef it when we're done with it so it doesn't leak
out.

+#else
+    // Platforms like Chrome on ARMv7 that don't typically compile with NEON globally.
+    #define SK_SMALL_MEMSET 10
+#endif
+
+
 /** Similar to memset(), but it assigns a 16bit value into the buffer.
     @param buffer   The memory to have value copied into it
     @param value    The 16bit value to be copied into buffer
     @param count    The number of times value should be copied into the buffer.
 */
-void sk_memset16(uint16_t dst[], uint16_t value, int count);
+void sk_memset16_large(uint16_t dst[], uint16_t value, int count);

[reed1, 2015/04/09 20:30:24]

These two "large" functions. I know they have to be declared for the inliner,
but its a shame we have to expose them to our clients. (ugh if a client started
explicitly using sk_memset16_large).

Maybe the real fix is to move all of these mem functions into a new private
header, since only skia itself should get to use our goodies.

[mtklein, 2015/04/09 20:36:42]

Moving this code around sounds good to me.  I hadn't realized anyone was
actually using SkUtils.h outside Skia itself.

+inline void sk_memset16(uint16_t dst[], uint16_t value, int count) {
+    if (count <= SK_SMALL_MEMSET) {
+        for (int i = 0; i < count; i++) {
+            dst[i] = value;
+        }
+    } else {
+        sk_memset16_large(dst, value, count);
+    }
+}
 typedef void (*SkMemset16Proc)(uint16_t dst[], uint16_t value, int count);
 SkMemset16Proc SkMemset16GetPlatformProc();
 
@@ -26,7 +45,17 @@ SkMemset16Proc SkMemset16GetPlatformProc();
     @param value    The 32bit value to be copied into buffer
     @param count    The number of times value should be copied into the buffer.
 */
-void sk_memset32(uint32_t dst[], uint32_t value, int count);
+void sk_memset32_large(uint32_t dst[], uint32_t value, int count);
+inline void sk_memset32(uint32_t dst[], uint32_t value, int count) {
+    if (count <= SK_SMALL_MEMSET) {
+        for (int i = 0; i < count; i++) {
+            dst[i] = value;
+        }
+    } else {
+        sk_memset32_large(dst, value, count);
+    }
+}
+
 typedef void (*SkMemset32Proc)(uint32_t dst[], uint32_t value, int count);
 SkMemset32Proc SkMemset32GetPlatformProc();