mirrors/pipewire
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/pipewire/pipewire.git synced 2025-11-02 09:01:50 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

fmt-ops: add avx2 optimized version

Browse source 
Only one optimized version but the sse2 version are compiled with
the avx2 flags so that they get optimized better.
This commit is contained in:
Wim Taymans 2020-03-16 16:11:29 +01:00
parent 6eca935e61
commit 3a911dfe3b
7 changed files with 821 additions and 28 deletions
Download patch file Download diff file Expand all files Collapse all files

34
spa/plugins/audioconvert/fmt-ops-sse2.c
View file

@ -31,8 +31,7 @@ conv_s16_to_f32d_1s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
uint32_t n_channels, uint32_t n_samples)
{
const int16_t *s = src;
float **d = (float **) dst;
float *d0 = d[0];
float *d0 = dst[0];
uint32_t n, unrolled;
__m128i in;
__m128 out, factor = _mm_set1_ps(1.0f / S16_SCALE);
@ -77,8 +76,7 @@ conv_s16_to_f32d_2_sse2(struct convert *conv, void * SPA_RESTRICT dst[], const v
uint32_t n_samples)
{
const int16_t *s = src[0];
float **d = (float **) dst;
float *d0 = d[0], *d1 = d[1];
float *d0 = dst[0], *d1 = dst[1];
uint32_t n, unrolled;
__m128i in[2], t[4];
__m128 out[4], factor = _mm_set1_ps(1.0f / S16_SCALE);
@ -135,8 +133,7 @@ conv_s24_to_f32d_1s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
uint32_t n_channels, uint32_t n_samples)
{
const uint8_t *s = src;
float **d = (float **) dst;
float *d0 = d[0];
float *d0 = dst[0];
uint32_t n, unrolled;
__m128i in;
__m128 out, factor = _mm_set1_ps(1.0f / S24_SCALE);
@ -175,8 +172,7 @@ conv_s24_to_f32d_2s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
uint32_t n_channels, uint32_t n_samples)
{
const uint8_t *s = src;
float **d = (float **) dst;
float *d0 = d[0], *d1 = d[1];
float *d0 = dst[0], *d1 = dst[1];
uint32_t n, unrolled;
__m128i in[2];
__m128 out[2], factor = _mm_set1_ps(1.0f / S24_SCALE);
@ -235,8 +231,7 @@ conv_s24_to_f32d_4s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
uint32_t n_channels, uint32_t n_samples)
{
const uint8_t *s = src;
float **d = (float **) dst;
float *d0 = d[0], *d1 = d[1], *d2 = d[2], *d3 = d[3];
float *d0 = dst[0], *d1 = dst[1], *d2 = dst[2], *d3 = dst[3];
uint32_t n, unrolled;
__m128i in[4];
__m128 out[4], factor = _mm_set1_ps(1.0f / S24_SCALE);
@ -340,8 +335,7 @@ conv_s32_to_f32d_1s_sse2(void *data, void * SPA_RESTRICT dst[], const void * SPA
uint32_t n_channels, uint32_t n_samples)
{
const int32_t *s = src;
float **d = (float **) dst;
float *d0 = d[0];
float *d0 = dst[0];
uint32_t n, unrolled;
__m128i in;
__m128 out, factor = _mm_set1_ps(1.0f / S24_SCALE);
@ -385,8 +379,7 @@ static void
conv_f32d_to_s32_1s_sse2(void *data, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
uint32_t n_channels, uint32_t n_samples)
{
const float **s = (const float **) src;
const float *s0 = s[0];
const float *s0 = src[0];
int32_t *d = dst;
uint32_t n, unrolled;
__m128 in[1];
@ -426,8 +419,7 @@ static void
conv_f32d_to_s32_2s_sse2(void *data, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
uint32_t n_channels, uint32_t n_samples)
{
const float **s = (const float **) src;
const float *s0 = s[0], *s1 = s[1];
const float *s0 = src[0], *s1 = src[1];
int32_t *d = dst;
uint32_t n, unrolled;
__m128 in[2];
@ -478,8 +470,7 @@ static void
conv_f32d_to_s32_4s_sse2(void *data, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
uint32_t n_channels, uint32_t n_samples)
{
const float **s = (const float **) src;
const float *s0 = s[0], *s1 = s[1], *s2 = s[2], *s3 = s[3];
const float *s0 = src[0], *s1 = src[1], *s2 = src[2], *s3 = src[3];
int32_t *d = dst;
uint32_t n, unrolled;
__m128 in[4];
@ -556,8 +547,7 @@ static void
conv_f32d_to_s16_1s_sse2(void *data, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
uint32_t n_channels, uint32_t n_samples)
{
const float **s = (const float **) src;
const float *s0 = s[0];
const float *s0 = src[0];
int16_t *d = dst;
uint32_t n, unrolled;
__m128 in[2];
@ -599,8 +589,7 @@ static void
conv_f32d_to_s16_2s_sse2(void *data, void * SPA_RESTRICT dst, const void * SPA_RESTRICT src[],
uint32_t n_channels, uint32_t n_samples)
{
const float **s = (const float **) src;
const float *s0 = s[0], *s1 = s[1];
const float *s0 = src[0], *s1 = src[1];
int16_t *d = dst;
uint32_t n, unrolled;
__m128 in[2];
@ -724,7 +713,6 @@ conv_f32d_to_s16_sse2(struct convert *conv, void * SPA_RESTRICT dst[], const voi
conv_f32d_to_s16_1s_sse2(conv, &d[i], &src[i], n_channels, n_samples);
}
void
conv_f32d_to_s16_2_sse2(struct convert *conv, void * SPA_RESTRICT dst[], const void * SPA_RESTRICT src[],
uint32_t n_samples)