mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2026-04-14 08:21:47 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

spa: use the right AVX2 flags

Browse source 
Our AVX optimizations are really AVX2 so rename the files and functions
and use the right HAVE_AVX2 and cpu flags to compile and select the
right functions.

Fixes #5072
This commit is contained in:
Wim Taymans 2026-01-13 12:03:09 +01:00
parent 5871f88b81
commit 8eb60b132c
15 changed files with 66 additions and 66 deletions
Download patch file Download diff file Expand all files Collapse all files

68
spa/plugins/audiomixer/mix-ops-avx.c
View file

@ -1,68 +0,0 @@
/* Spa */
/* SPDX-FileCopyrightText: Copyright © 2019 Wim Taymans */
/* SPDX-License-Identifier: MIT */
#include <string.h>
#include <stdio.h>
#include <math.h>
#include <spa/utils/defs.h>
#include "mix-ops.h"
#include <immintrin.h>
void
mix_f32_avx(struct mix_ops *ops, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
uint32_t n_src, uint32_t n_samples)
{
n_samples *= ops->n_channels;
if (n_src == 0)
memset(dst, 0, n_samples * ops->n_channels * sizeof(float));
else if (n_src == 1) {
if (dst != src[0])
spa_memcpy(dst, src[0], n_samples * sizeof(float));
} else {
uint32_t i, n, unrolled;
const float **s = (const float **)src;
float *d = dst;
if (SPA_LIKELY(SPA_IS_ALIGNED(dst, 32))) {
unrolled = n_samples & ~31;
for (i = 0; i < n_src; i++) {
if (SPA_UNLIKELY(!SPA_IS_ALIGNED(src[i], 32))) {
unrolled = 0;
break;
}
}
} else
unrolled = 0;
for (n = 0; n < unrolled; n += 32) {
__m256 in[4];
in[0] = _mm256_load_ps(&s[0][n + 0]);
in[1] = _mm256_load_ps(&s[0][n + 8]);
in[2] = _mm256_load_ps(&s[0][n + 16]);
in[3] = _mm256_load_ps(&s[0][n + 24]);
for (i = 1; i < n_src; i++) {
in[0] = _mm256_add_ps(in[0], _mm256_load_ps(&s[i][n + 0]));
in[1] = _mm256_add_ps(in[1], _mm256_load_ps(&s[i][n + 8]));
in[2] = _mm256_add_ps(in[2], _mm256_load_ps(&s[i][n + 16]));
in[3] = _mm256_add_ps(in[3], _mm256_load_ps(&s[i][n + 24]));
}
_mm256_store_ps(&d[n + 0], in[0]);
_mm256_store_ps(&d[n + 8], in[1]);
_mm256_store_ps(&d[n + 16], in[2]);
_mm256_store_ps(&d[n + 24], in[3]);
}
for (; n < n_samples; n++) {
__m128 in[1];
in[0] = _mm_load_ss(&s[0][n]);
for (i = 1; i < n_src; i++)
in[0] = _mm_add_ss(in[0], _mm_load_ss(&s[i][n]));
_mm_store_ss(&d[n], in[0]);
}
}
}