mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-11-03 09:01:54 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

audio-dsp: add some SSE2

Browse source
This commit is contained in:
Wim Taymans 2018-07-17 09:52:27 +02:00
parent fba00fb791
commit 2b1b356455
1 changed files with 61 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

62
src/modules/module-media-session/audio-dsp.c
View file

@ -23,6 +23,13 @@
#include <math.h>
#include <time.h>
#if defined (__SSE2__) && !defined (__sun__)
#include <emmintrin.h>
#ifdef __SSE4_1__
#include <smmintrin.h>
#endif
#endif
#include "config.h"
#include <spa/node/node.h>
@ -275,16 +282,69 @@ static void fill_s16(void *dst, int index, int n_samples, int stride)
_v * S32_SCALE; \
})
#define SAMPLE_24BIT_MAX_F 8388607.0f
static void conv_f32_s32(void *dst, void *src, int index, int n_samples, int stride)
{
int32_t *d = dst;
float *s = src;
int i;
d += index;
#if defined (__SSE2__) && !defined (__sun__)
__m128 int_max = _mm_set1_ps(SAMPLE_24BIT_MAX_F);
__m128 int_min = _mm_sub_ps(_mm_setzero_ps(), int_max);
__m128 factor = int_max;
unsigned long unrolled = n_samples / 4;
n_samples = n_samples & 3;
while (unrolled--) {
__m128 in = _mm_load_ps(s);
__m128 scaled = _mm_mul_ps(in, factor);
__m128 clipped = _mm_min_ps(int_max, _mm_max_ps(scaled, int_min));
__m128i y = _mm_cvttps_epi32(clipped);
__m128i shifted = _mm_slli_epi32(y, 8);
#ifdef __SSE4_1__
*d = _mm_extract_epi32(shifted, 0);
*(d+stride) = _mm_extract_epi32(shifted, 1);
*(d+2*stride) = _mm_extract_epi32(shifted, 2);
*(d+3*stride) = _mm_extract_epi32(shifted, 3);
#else
__m128i shuffled1 = _mm_shuffle_epi32(shifted, _MM_SHUFFLE(0, 3, 2, 1));
__m128i shuffled2 = _mm_shuffle_epi32(shifted, _MM_SHUFFLE(1, 0, 3, 2));
__m128i shuffled3 = _mm_shuffle_epi32(shifted, _MM_SHUFFLE(2, 1, 0, 3));
_mm_store_ss((float*)d, (__m128)shifted);
_mm_store_ss((float*)(d+stride), (__m128)shuffled1);
_mm_store_ss((float*)(d+2*stride), (__m128)shuffled2);
_mm_store_ss((float*)(d+3*stride), (__m128)shuffled3);
#endif
d += 4 * stride;
s += 4;
}
while (n_samples--) {
__m128 in = _mm_load_ss(s);
__m128 scaled = _mm_mul_ss(in, factor);
__m128 clipped = _mm_min_ss(int_max, _mm_max_ss(scaled, int_min));
*d = _mm_cvttss_si32(clipped) << 8;
d += stride;
s++;
}
#else
int i;
for (i = 0; i < n_samples; i++) {
*d = F32_TO_S32(s[i]);
d += stride;
}
#endif
}
#define S32_TO_F32(v) ((v) * (1.0f / S32_SCALE))