Use shuffle+cvt to extract SIMD values instead of storing to memory
[openal-soft.git] / Alc / mixer / mixer_sse2.c
blobe01980224b7d0e9c1d4e8d403ff3e9bf194a8396
1 /**
2 * OpenAL cross platform audio library
3 * Copyright (C) 2014 by Timothy Arceri <t_arceri@yahoo.com.au>.
4 * This library is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Library General Public
6 * License as published by the Free Software Foundation; either
7 * version 2 of the License, or (at your option) any later version.
9 * This library is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Library General Public License for more details.
14 * You should have received a copy of the GNU Library General Public
15 * License along with this library; if not, write to the
16 * Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
18 * Or go to http://www.gnu.org/copyleft/lgpl.html
21 #include "config.h"
23 #include <xmmintrin.h>
24 #include <emmintrin.h>
26 #include "alu.h"
27 #include "defs.h"
30 const ALfloat *Resample_lerp_SSE2(const InterpState* UNUSED(state),
31 const ALfloat *restrict src, ALsizei frac, ALint increment,
32 ALfloat *restrict dst, ALsizei numsamples)
34 const __m128i increment4 = _mm_set1_epi32(increment*4);
35 const __m128 fracOne4 = _mm_set1_ps(1.0f/FRACTIONONE);
36 const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
37 union { alignas(16) ALint i[4]; float f[4]; } pos_;
38 union { alignas(16) ALsizei i[4]; float f[4]; } frac_;
39 __m128i frac4, pos4;
40 ALint pos;
41 ALsizei i;
43 ASSUME(numsamples > 0);
45 InitiatePositionArrays(frac, increment, frac_.i, pos_.i, 4);
47 frac4 = _mm_castps_si128(_mm_load_ps(frac_.f));
48 pos4 = _mm_castps_si128(_mm_load_ps(pos_.f));
50 for(i = 0;numsamples-i > 3;i += 4)
52 const __m128 val1 = _mm_setr_ps(src[pos_.i[0] ], src[pos_.i[1] ], src[pos_.i[2] ], src[pos_.i[3] ]);
53 const __m128 val2 = _mm_setr_ps(src[pos_.i[0]+1], src[pos_.i[1]+1], src[pos_.i[2]+1], src[pos_.i[3]+1]);
55 /* val1 + (val2-val1)*mu */
56 const __m128 r0 = _mm_sub_ps(val2, val1);
57 const __m128 mu = _mm_mul_ps(_mm_cvtepi32_ps(frac4), fracOne4);
58 const __m128 out = _mm_add_ps(val1, _mm_mul_ps(mu, r0));
60 _mm_store_ps(&dst[i], out);
62 frac4 = _mm_add_epi32(frac4, increment4);
63 pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
64 frac4 = _mm_and_si128(frac4, fracMask4);
66 pos_.i[0] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(0, 0, 0, 0)));
67 pos_.i[1] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(1, 1, 1, 1)));
68 pos_.i[2] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(2, 2, 2, 2)));
69 pos_.i[3] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(3, 3, 3, 3)));
72 /* NOTE: These four elements represent the position *after* the last four
73 * samples, so the lowest element is the next position to resample.
75 pos = pos_.i[0];
76 frac = _mm_cvtsi128_si32(frac4);
78 for(;i < numsamples;i++)
80 dst[i] = lerp(src[pos], src[pos+1], frac * (1.0f/FRACTIONONE));
82 frac += increment;
83 pos += frac>>FRACTIONBITS;
84 frac &= FRACTIONMASK;
86 return dst;