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

include/core/SkUtils.h

 /*
  * Copyright 2006 The Android Open Source Project
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #ifndef SkUtils_DEFINED
 #define SkUtils_DEFINED
 
 #include "SkTypes.h"
 
 ///////////////////////////////////////////////////////////////////////////////
 
+// 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();
 
 /** Similar to memset(), but it assigns a 32bit value into the buffer.
     @param buffer   The memory to have value copied into it
     @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();
 
 /** Similar to memcpy(), but it copies count 32bit values from src to dst.
     @param dst      The memory to have value copied into it
     @param src      The memory to have value copied from it
     @param count    The number of values should be copied.
 */
 void sk_memcpy32(uint32_t dst[], const uint32_t src[], int count);
 typedef void (*SkMemcpy32Proc)(uint32_t dst[], const uint32_t src[], int count);
@@ skipping 69 matching lines @@
     }
     ~SkAutoTrace() {
         SkDebugf("--- trace: %s Leave\n", fLabel);
     }
 private:
     const char* fLabel;
 };
 #define SkAutoTrace(...) SK_REQUIRE_LOCAL_VAR(SkAutoTrace)
 
 #endif