diff options
Diffstat (limited to 'src/jumper')
-rw-r--r-- | src/jumper/SkJumper_stages_lowp.cpp | 54 |
1 files changed, 30 insertions, 24 deletions
diff --git a/src/jumper/SkJumper_stages_lowp.cpp b/src/jumper/SkJumper_stages_lowp.cpp index 21b257f9c1..8c0274f911 100644 --- a/src/jumper/SkJumper_stages_lowp.cpp +++ b/src/jumper/SkJumper_stages_lowp.cpp @@ -183,14 +183,28 @@ SI D join(S lo, S hi) { memcpy((char*)&v + 1*sizeof(S), &hi, sizeof(S)); return v; } -template <typename V, typename H> -SI V map(V v, H (*fn)(H)) { - H lo,hi; - split(v, &lo,&hi); - lo = fn(lo); - hi = fn(hi); - return join<V>(lo,hi); -} + +#if defined(__AVX2__) + template <typename V, typename Fn> + SI auto map(V v, Fn&& fn) -> decltype( + __builtin_shufflevector(fn(__builtin_shufflevector(v,v, 0,1, 2, 3, 4, 5, 6, 7)), + fn(__builtin_shufflevector(v,v, 8,9,10,11,12,13,14,15)), + 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15)) { + return __builtin_shufflevector(fn(__builtin_shufflevector(v,v, 0,1, 2, 3, 4, 5, 6, 7)), + fn(__builtin_shufflevector(v,v, 8,9,10,11,12,13,14,15)), + 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15); + } +#else + template <typename V, typename Fn> + SI auto map(V v, Fn&& fn) -> decltype( + __builtin_shufflevector(fn(__builtin_shufflevector(v,v, 0,1,2,3)), + fn(__builtin_shufflevector(v,v, 4,5,6,7)), + 0,1,2,3,4,5,6,7)) { + return __builtin_shufflevector(fn(__builtin_shufflevector(v,v, 0,1,2,3)), + fn(__builtin_shufflevector(v,v, 4,5,6,7)), + 0,1,2,3,4,5,6,7); + } +#endif // TODO: do we need platform-specific intrinsics for any of these? SI F if_then_else(I32 c, F t, F e) { @@ -208,7 +222,7 @@ SI F rcp(F x) { #elif defined(__SSE__) return map(x, _mm_rcp_ps); #elif defined(__ARM_NEON) - return map(x, +[](float32x4_t v) { + return map(x, [](float32x4_t v) { auto est = vrecpeq_f32(v); return vrecpsq_f32(v,est)*est; }); @@ -224,7 +238,7 @@ SI F sqrt_(F x) { #elif defined(__aarch64__) return map(x, vsqrtq_f32); #elif defined(__ARM_NEON) - return map(x, +[](float32x4_t v) { + return map(x, [](float32x4_t v) { auto est = vrsqrteq_f32(v); // Estimate and two refinement steps for est = rsqrt(v). est *= vrsqrtsq_f32(v,est*est); est *= vrsqrtsq_f32(v,est*est); @@ -242,9 +256,9 @@ SI F floor_(F x) { #if defined(__aarch64__) return map(x, vrndmq_f32); #elif defined(__AVX2__) - return map(x, +[](__m256 v){ return _mm256_floor_ps(v); }); // _mm256_floor_ps is a macro... + return map(x, [](__m256 v){ return _mm256_floor_ps(v); }); // _mm256_floor_ps is a macro... #elif defined(__SSE4_1__) - return map(x, +[](__m128 v){ return _mm_floor_ps(v); }); // _mm_floor_ps() is a macro too. + return map(x, [](__m128 v){ return _mm_floor_ps(v); }); // _mm_floor_ps() is a macro too. #else F roundtrip = cast<F>(cast<I32>(x)); return roundtrip - if_then_else(roundtrip > x, F(1), F(0)); @@ -480,21 +494,13 @@ SI void store(T* ptr, size_t tail, V v) { } template<> - F gather(const float* p, U32 ix) { - __m256i lo, hi; - split(ix, &lo, &hi); - - return join<F>(_mm256_i32gather_ps(p, lo, 4), - _mm256_i32gather_ps(p, hi, 4)); + F gather(const float* ptr, U32 ix) { + return map(ix, [&](__m256i ix) { return _mm256_i32gather_ps(ptr, ix, 4); }); } template<> - U32 gather(const uint32_t* p, U32 ix) { - __m256i lo, hi; - split(ix, &lo, &hi); - - return join<U32>(_mm256_i32gather_epi32(p, lo, 4), - _mm256_i32gather_epi32(p, hi, 4)); + U32 gather(const uint32_t* ptr, U32 ix) { + return map(ix, [&](__m256i ix) { return _mm256_i32gather_epi32(ptr, ix, 4); }); } #else template <typename V, typename T> |