blender/intern/cycles/util/util_simd.h
Sv. Lockal ab32a1807d Cycles: SSE optimization for Voronoi cells texture
Gives 5-6% speedup for Caterpillar_PatazStudio.blend.

Reviewed By: brecht, dingto

Differential Revision: https://developer.blender.org/D419
2014-04-03 23:35:10 +04:00

277 lines
8.3 KiB
C++

/*
* Copyright 2011-2013 Blender Foundation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License
*/
#ifndef __UTIL_SIMD_H__
#define __UTIL_SIMD_H__
CCL_NAMESPACE_BEGIN
#ifdef __KERNEL_SSE2__
/* SSE shuffle utility functions */
#ifdef __KERNEL_SSSE3__
/* faster version for SSSE3 */
typedef __m128i shuffle_swap_t;
ccl_device_inline const shuffle_swap_t shuffle_swap_identity(void)
{
return _mm_set_epi8(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
}
ccl_device_inline const shuffle_swap_t shuffle_swap_swap(void)
{
return _mm_set_epi8(7, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8);
}
ccl_device_inline const __m128 shuffle_swap(const __m128& a, const shuffle_swap_t& shuf)
{
return _mm_castsi128_ps(_mm_shuffle_epi8(_mm_castps_si128(a), shuf));
}
#else
/* somewhat slower version for SSE2 */
typedef int shuffle_swap_t;
ccl_device_inline const shuffle_swap_t shuffle_swap_identity(void)
{
return 0;
}
ccl_device_inline const shuffle_swap_t shuffle_swap_swap(void)
{
return 1;
}
ccl_device_inline const __m128 shuffle_swap(const __m128& a, shuffle_swap_t shuf)
{
/* shuffle value must be a constant, so we need to branch */
if(shuf)
return _mm_shuffle_ps(a, a, _MM_SHUFFLE(1, 0, 3, 2));
else
return _mm_shuffle_ps(a, a, _MM_SHUFFLE(3, 2, 1, 0));
}
#endif
#ifdef __KERNEL_SSE41__
ccl_device_inline void gen_idirsplat_swap(const __m128 &pn, const shuffle_swap_t &shuf_identity, const shuffle_swap_t &shuf_swap,
const float3& idir, __m128 idirsplat[3], shuffle_swap_t shufflexyz[3])
{
const __m128 idirsplat_raw[] = { _mm_set_ps1(idir.x), _mm_set_ps1(idir.y), _mm_set_ps1(idir.z) };
idirsplat[0] = _mm_xor_ps(idirsplat_raw[0], pn);
idirsplat[1] = _mm_xor_ps(idirsplat_raw[1], pn);
idirsplat[2] = _mm_xor_ps(idirsplat_raw[2], pn);
const __m128 signmask = _mm_castsi128_ps(_mm_set1_epi32(0x80000000));
const __m128 shuf_identity_f = _mm_castsi128_ps(shuf_identity);
const __m128 shuf_swap_f = _mm_castsi128_ps(shuf_swap);
shufflexyz[0] = _mm_castps_si128(_mm_blendv_ps(shuf_identity_f, shuf_swap_f, _mm_and_ps(idirsplat_raw[0], signmask)));
shufflexyz[1] = _mm_castps_si128(_mm_blendv_ps(shuf_identity_f, shuf_swap_f, _mm_and_ps(idirsplat_raw[1], signmask)));
shufflexyz[2] = _mm_castps_si128(_mm_blendv_ps(shuf_identity_f, shuf_swap_f, _mm_and_ps(idirsplat_raw[2], signmask)));
}
#else
ccl_device_inline void gen_idirsplat_swap(const __m128 &pn, const shuffle_swap_t &shuf_identity, const shuffle_swap_t &shuf_swap,
const float3& idir, __m128 idirsplat[3], shuffle_swap_t shufflexyz[3])
{
idirsplat[0] = _mm_xor_ps(_mm_set_ps1(idir.x), pn);
idirsplat[1] = _mm_xor_ps(_mm_set_ps1(idir.y), pn);
idirsplat[2] = _mm_xor_ps(_mm_set_ps1(idir.z), pn);
shufflexyz[0] = (idir.x >= 0)? shuf_identity: shuf_swap;
shufflexyz[1] = (idir.y >= 0)? shuf_identity: shuf_swap;
shufflexyz[2] = (idir.z >= 0)? shuf_identity: shuf_swap;
}
#endif
template<size_t i0, size_t i1, size_t i2, size_t i3> ccl_device_inline const __m128 shuffle(const __m128& a, const __m128& b)
{
return _mm_shuffle_ps(a, b, _MM_SHUFFLE(i3, i2, i1, i0));
}
template<size_t i0, size_t i1, size_t i2, size_t i3> ccl_device_inline const __m128 shuffle(const __m128& a)
{
return _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(a), _MM_SHUFFLE(i3, i2, i1, i0)));
}
template<> __forceinline const __m128 shuffle<0, 1, 0, 1>(const __m128& a)
{
return _mm_movelh_ps(a, a);
}
template<> __forceinline const __m128 shuffle<2, 3, 2, 3>(const __m128& a)
{
return _mm_movehl_ps(a, a);
}
template<size_t i0, size_t i1, size_t i2, size_t i3> ccl_device_inline const __m128i shuffle(const __m128i& a)
{
return _mm_shuffle_epi32(a, _MM_SHUFFLE(i3, i2, i1, i0));
}
template<size_t i0, size_t i1, size_t i2, size_t i3> ccl_device_inline const __m128i shuffle(const __m128i& a, const __m128i& b)
{
return _mm_castps_si128(_mm_shuffle_ps(_mm_castsi128_ps(a), _mm_castsi128_ps(b), _MM_SHUFFLE(i3, i2, i1, i0)));
}
/* Blend 2 vectors based on mask: (a[i] & mask[i]) | (b[i] & ~mask[i]) */
#ifdef __KERNEL_SSE41__
ccl_device_inline const __m128 blend(const __m128& mask, const __m128& a, const __m128& b)
{
return _mm_blendv_ps(b, a, mask);
}
#else
ccl_device_inline const __m128 blend(const __m128& mask, const __m128& a, const __m128& b)
{
return _mm_or_ps(_mm_and_ps(mask, a), _mm_andnot_ps(mask, b));
}
#endif
/* calculate a*b+c (replacement for fused multiply-add on SSE CPUs) */
ccl_device_inline const __m128 fma(const __m128& a, const __m128& b, const __m128& c)
{
return _mm_add_ps(_mm_mul_ps(a, b), c);
}
/* calculate a*b-c (replacement for fused multiply-subtract on SSE CPUs) */
ccl_device_inline const __m128 fms(const __m128& a, const __m128& b, const __m128& c)
{
return _mm_sub_ps(_mm_mul_ps(a, b), c);
}
/* calculate -a*b+c (replacement for fused negated-multiply-subtract on SSE CPUs) */
ccl_device_inline const __m128 fnma(const __m128& a, const __m128& b, const __m128& c)
{
return _mm_sub_ps(c, _mm_mul_ps(a, b));
}
template<size_t N> ccl_device_inline const __m128 broadcast(const __m128& a)
{
return _mm_castsi128_ps(_mm_shuffle_epi32(_mm_castps_si128(a), _MM_SHUFFLE(N, N, N, N)));
}
template<size_t N> ccl_device_inline const __m128i broadcast(const __m128i& a)
{
return _mm_shuffle_epi32(a, _MM_SHUFFLE(N, N, N, N));
}
ccl_device_inline const __m128 uint32_to_float(const __m128i &in)
{
__m128i a = _mm_srli_epi32(in, 16);
__m128i b = _mm_and_si128(in, _mm_set1_epi32(0x0000ffff));
__m128i c = _mm_or_si128(a, _mm_set1_epi32(0x53000000));
__m128 d = _mm_cvtepi32_ps(b);
__m128 e = _mm_sub_ps(_mm_castsi128_ps(c), _mm_castsi128_ps(_mm_set1_epi32(0x53000000)));
return _mm_add_ps(e, d);
}
template<size_t S1, size_t S2, size_t S3, size_t S4>
ccl_device_inline const __m128 set_sign_bit(const __m128 &a)
{
return _mm_xor_ps(a, _mm_castsi128_ps(_mm_setr_epi32(S1 << 31, S2 << 31, S3 << 31, S4 << 31)));
}
#ifdef __KERNEL_WITH_SSE_ALIGN__
ccl_device_inline const __m128 load_m128(const float4 &vec)
{
return _mm_load_ps(&vec.x);
}
ccl_device_inline const __m128 load_m128(const float3 &vec)
{
return _mm_load_ps(&vec.x);
}
#else
ccl_device_inline const __m128 load_m128(const float4 &vec)
{
return _mm_loadu_ps(&vec.x);
}
ccl_device_inline const __m128 load_m128(const float3 &vec)
{
return _mm_loadu_ps(&vec.x);
}
#endif /* __KERNEL_WITH_SSE_ALIGN__ */
ccl_device_inline const __m128 dot3_splat(const __m128& a, const __m128& b)
{
#ifdef __KERNEL_SSE41__
return _mm_dp_ps(a, b, 0x7f);
#else
__m128 t = _mm_mul_ps(a, b);
return _mm_set1_ps(((float*)&t)[0] + ((float*)&t)[1] + ((float*)&t)[2]);
#endif
}
/* squared length taking only specified axes into account */
template<size_t X, size_t Y, size_t Z, size_t W>
ccl_device_inline float len_squared(const __m128& a)
{
#ifndef __KERNEL_SSE41__
float4& t = (float4 &)a;
return (X ? t.x * t.x : 0.0f) + (Y ? t.y * t.y : 0.0f) + (Z ? t.z * t.z : 0.0f) + (W ? t.w * t.w : 0.0f);
#else
return _mm_cvtss_f32(_mm_dp_ps(a, a, (X << 4) | (Y << 5) | (Z << 6) | (W << 7) | 0xf));
#endif
}
ccl_device_inline float dot3(const __m128& a, const __m128& b)
{
#ifdef __KERNEL_SSE41__
return _mm_cvtss_f32(_mm_dp_ps(a, b, 0x7f));
#else
__m128 t = _mm_mul_ps(a, b);
return ((float*)&t)[0] + ((float*)&t)[1] + ((float*)&t)[2];
#endif
}
ccl_device_inline const __m128 len3_squared_splat(const __m128& a)
{
return dot3_splat(a, a);
}
ccl_device_inline float len3_squared(const __m128& a)
{
return dot3(a, a);
}
ccl_device_inline float len3(const __m128& a)
{
return _mm_cvtss_f32(_mm_sqrt_ss(dot3_splat(a, a)));
}
/* calculate shuffled cross product, useful when order of components does not matter */
ccl_device_inline const __m128 cross_zxy(const __m128& a, const __m128& b)
{
return fms(a, shuffle<1, 2, 0, 3>(b), _mm_mul_ps(b, shuffle<1, 2, 0, 3>(a)));
}
ccl_device_inline const __m128 cross(const __m128& a, const __m128& b)
{
return shuffle<1, 2, 0, 3>(cross_zxy(a, b));
}
#endif /* __KERNEL_SSE2__ */
CCL_NAMESPACE_END
#endif /* __UTIL_SIMD_H__ */