Index: include/core/SkUtils.h |
diff --git a/include/core/SkUtils.h b/include/core/SkUtils.h |
index bca76ed54f831d4326997ba54bcd4941c6e158e6..89afdb68cee5b5e97e354e8330d2a2f1bc253980 100644 |
--- a/include/core/SkUtils.h |
+++ b/include/core/SkUtils.h |
@@ -10,56 +10,44 @@ |
#include "SkTypes.h" |
-/////////////////////////////////////////////////////////////////////////////// |
+namespace SkOpts { |
+ extern void (*memset16)(uint16_t[], uint16_t, int); |
djsollen
2015/07/31 17:25:23
does this mean that SkOpts.h should be in include/
mtklein_C
2015/07/31 17:36:25
Yeah, borderline case. Done.
|
+ extern void (*memset32)(uint32_t[], uint32_t, int); |
+} |
-// 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 |
-#else |
- // Platforms like Chrome on ARMv7 that don't typically compile with NEON globally. |
- #define SK_SMALL_MEMSET 10 |
-#endif |
+/////////////////////////////////////////////////////////////////////////////// |
+// The inlining heuristics below were determined using bench/MemsetBench.cpp |
+// on a x86 desktop, a Nexus 7 with and without NEON, and a Nexus 9. |
/** 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_large(uint16_t dst[], uint16_t value, int count); |
-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); |
- } |
+static inline void sk_memset16(uint16_t buffer[], uint16_t value, int count) { |
+#if defined(SK_CPU_ARM64) |
djsollen
2015/07/31 17:25:23
can you add some comments as to why we don't just
mtklein_C
2015/07/31 17:36:25
Done.
|
+ while (count --> 0) { *buffer++ = value; } return; |
+#elif defined(SK_CPU_ARM32) |
+ if (count <= 10) { while (count --> 0) { *buffer++ = value; } return; } |
+#endif |
+ SkOpts::memset16(buffer, 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_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); |
- } |
+static inline void sk_memset32(uint32_t buffer[], uint32_t value, int count) { |
+#if defined(SK_CPU_ARM64) |
djsollen
2015/07/31 17:25:23
further why not put this logic in the neon opts fi
mtklein_C
2015/07/31 17:36:25
Done.
|
+ while (count --> 0) { *buffer++ = value; } return; |
+#elif defined(SK_CPU_ARM32) |
+ if (count <= 10) { while (count --> 0) { *buffer++ = value; } return; } |
+#endif |
+ SkOpts::memset32(buffer, value, count); |
} |
-typedef void (*SkMemset32Proc)(uint32_t dst[], uint32_t value, int count); |
-SkMemset32Proc SkMemset32GetPlatformProc(); |
- |
-#undef SK_SMALL_MEMSET |
/////////////////////////////////////////////////////////////////////////////// |