diff options
author | Sven Gothel <[email protected]> | 2023-05-03 16:17:49 +0200 |
---|---|---|
committer | Sven Gothel <[email protected]> | 2023-05-03 16:17:49 +0200 |
commit | ec167fd05661a5b02dd406c87081f84a0f8dd77d (patch) | |
tree | 9c4669e471c9969bda59265381b18d2d416db060 /core/mixer/mixer_sse2.cpp | |
parent | 0d14d30808cfe7b9e3413353e3eef8a0f201399a (diff) | |
parent | d3875f333fb6abe2f39d82caca329414871ae53b (diff) |
Merge branch 'v1.23.1'
Resolved Conflicts:
CMakeLists.txt
Diffstat (limited to 'core/mixer/mixer_sse2.cpp')
-rw-r--r-- | core/mixer/mixer_sse2.cpp | 90 |
1 files changed, 90 insertions, 0 deletions
diff --git a/core/mixer/mixer_sse2.cpp b/core/mixer/mixer_sse2.cpp new file mode 100644 index 00000000..edaaf7a1 --- /dev/null +++ b/core/mixer/mixer_sse2.cpp @@ -0,0 +1,90 @@ +/** + * OpenAL cross platform audio library + * Copyright (C) 2014 by Timothy Arceri <[email protected]>. + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Library General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Library General Public License for more details. + * + * You should have received a copy of the GNU Library General Public + * License along with this library; if not, write to the + * Free Software Foundation, Inc., + * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + * Or go to http://www.gnu.org/copyleft/lgpl.html + */ + +#include "config.h" + +#include <xmmintrin.h> +#include <emmintrin.h> + +#include "alnumeric.h" +#include "defs.h" + +struct SSE2Tag; +struct LerpTag; + + +#if defined(__GNUC__) && !defined(__clang__) && !defined(__SSE2__) +#pragma GCC target("sse2") +#endif + +template<> +void Resample_<LerpTag,SSE2Tag>(const InterpState*, const float *RESTRICT src, uint frac, + const uint increment, const al::span<float> dst) +{ + ASSUME(frac < MixerFracOne); + + const __m128i increment4{_mm_set1_epi32(static_cast<int>(increment*4))}; + const __m128 fracOne4{_mm_set1_ps(1.0f/MixerFracOne)}; + const __m128i fracMask4{_mm_set1_epi32(MixerFracMask)}; + + alignas(16) uint pos_[4], frac_[4]; + InitPosArrays(frac, increment, frac_, pos_); + __m128i frac4{_mm_setr_epi32(static_cast<int>(frac_[0]), static_cast<int>(frac_[1]), + static_cast<int>(frac_[2]), static_cast<int>(frac_[3]))}; + __m128i pos4{_mm_setr_epi32(static_cast<int>(pos_[0]), static_cast<int>(pos_[1]), + static_cast<int>(pos_[2]), static_cast<int>(pos_[3]))}; + + auto dst_iter = dst.begin(); + for(size_t todo{dst.size()>>2};todo;--todo) + { + const int pos0{_mm_cvtsi128_si32(pos4)}; + const int pos1{_mm_cvtsi128_si32(_mm_srli_si128(pos4, 4))}; + const int pos2{_mm_cvtsi128_si32(_mm_srli_si128(pos4, 8))}; + const int pos3{_mm_cvtsi128_si32(_mm_srli_si128(pos4, 12))}; + const __m128 val1{_mm_setr_ps(src[pos0 ], src[pos1 ], src[pos2 ], src[pos3 ])}; + const __m128 val2{_mm_setr_ps(src[pos0+1], src[pos1+1], src[pos2+1], src[pos3+1])}; + + /* val1 + (val2-val1)*mu */ + const __m128 r0{_mm_sub_ps(val2, val1)}; + const __m128 mu{_mm_mul_ps(_mm_cvtepi32_ps(frac4), fracOne4)}; + const __m128 out{_mm_add_ps(val1, _mm_mul_ps(mu, r0))}; + + _mm_store_ps(dst_iter, out); + dst_iter += 4; + + frac4 = _mm_add_epi32(frac4, increment4); + pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, MixerFracBits)); + frac4 = _mm_and_si128(frac4, fracMask4); + } + + if(size_t todo{dst.size()&3}) + { + src += static_cast<uint>(_mm_cvtsi128_si32(pos4)); + frac = static_cast<uint>(_mm_cvtsi128_si32(frac4)); + + do { + *(dst_iter++) = lerpf(src[0], src[1], static_cast<float>(frac) * (1.0f/MixerFracOne)); + + frac += increment; + src += frac>>MixerFracBits; + frac &= MixerFracMask; + } while(--todo); + } +} |