diff options
author | mtklein <mtklein@chromium.org> | 2016-01-20 11:55:51 -0800 |
---|---|---|
committer | Commit bot <commit-bot@chromium.org> | 2016-01-20 11:55:51 -0800 |
commit | 550e9b0ef1c0fba42e2e902a467af322ad2413da (patch) | |
tree | 6e032834e21c115c024c539e7b60ad4c7ad209d9 /src/core/SkMipMap.cpp | |
parent | 88651aeb551fb02003a7600679e8e7df8a589e7f (diff) |
SkNx miplevel building
All sizes approximately twice as fast.
Before:
micros bench
1649.35 mipmap_build_512x512 nonrendering
1824.42 mipmap_build_511x512 nonrendering
2100.66 ? mipmap_build_512x511 nonrendering
2375.94 mipmap_build_511x511 nonrendering
After:
micros bench
730.32 ! mipmap_build_512x512 nonrendering
922.12 mipmap_build_511x512 nonrendering
999.07 mipmap_build_512x511 nonrendering
1342.93 ! mipmap_build_511x511 nonrendering
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1606013003
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot
Committed: https://skia.googlesource.com/skia/+/3bd5aba2a0e165997f683cf3aa306661e71464f6
Review URL: https://codereview.chromium.org/1606013003
Diffstat (limited to 'src/core/SkMipMap.cpp')
-rw-r--r-- | src/core/SkMipMap.cpp | 18 |
1 files changed, 15 insertions, 3 deletions
diff --git a/src/core/SkMipMap.cpp b/src/core/SkMipMap.cpp index d468507ac7..14e87a33a3 100644 --- a/src/core/SkMipMap.cpp +++ b/src/core/SkMipMap.cpp @@ -8,6 +8,7 @@ #include "SkMipMap.h" #include "SkBitmap.h" #include "SkColorPriv.h" +#include "SkNx.h" // // ColorTypeFilter is the "Type" we pass to some downsample template functions. @@ -18,12 +19,23 @@ struct ColorTypeFilter_8888 { typedef uint32_t Type; +#if defined(SKNX_IS_FAST) + static Sk4h Expand(uint32_t x) { + return SkNx_cast<uint16_t>(Sk4b::Load((const uint8_t*)&x)); + } + static uint32_t Compact(const Sk4h& x) { + uint32_t r; + SkNx_cast<uint8_t>(x).store((uint8_t*)&r); + return r; + } +#else static uint64_t Expand(uint32_t x) { return (x & 0xFF00FF) | ((uint64_t)(x & 0xFF00FF00) << 24); } static uint32_t Compact(uint64_t x) { return (uint32_t)((x & 0xFF00FF) | ((x >> 24) & 0xFF00FF00)); } +#endif }; struct ColorTypeFilter_565 { @@ -56,7 +68,7 @@ struct ColorTypeFilter_8 { } }; -template <typename T> T add_121(T a, T b, T c) { +template <typename T> T add_121(const T& a, const T& b, const T& c) { return a + b + b + c; } @@ -93,7 +105,7 @@ template <typename F> void downsample_3_2(void* dst, const void* src, size_t src auto p0 = static_cast<const typename F::Type*>(src); auto p1 = (const typename F::Type*)((const char*)p0 + srcRB); auto d = static_cast<typename F::Type*>(dst); - + auto c02 = F::Expand(p0[0]); auto c12 = F::Expand(p1[0]); for (int i = 0; i < count; ++i) { @@ -116,7 +128,7 @@ template <typename F> void downsample_2_3(void* dst, const void* src, size_t src auto p1 = (const typename F::Type*)((const char*)p0 + srcRB); auto p2 = (const typename F::Type*)((const char*)p1 + srcRB); auto d = static_cast<typename F::Type*>(dst); - + for (int i = 0; i < count; ++i) { auto c00 = F::Expand(p0[0]); auto c01 = F::Expand(p0[1]); |