mirror of
https://gitlab.freedesktop.org/pipewire/pipewire.git
synced 2025-11-04 13:30:12 -05:00
resample: add SSE peaks resampler
This commit is contained in:
parent
a1e567b985
commit
0a4ef3d2c0
4 changed files with 116 additions and 10 deletions
83
spa/plugins/audioconvert/resample-peaks-sse.h
Normal file
83
spa/plugins/audioconvert/resample-peaks-sse.h
Normal file
|
|
@ -0,0 +1,83 @@
|
||||||
|
/* Spa
|
||||||
|
*
|
||||||
|
* Copyright © 2018 Wim Taymans
|
||||||
|
*
|
||||||
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
||||||
|
* copy of this software and associated documentation files (the "Software"),
|
||||||
|
* to deal in the Software without restriction, including without limitation
|
||||||
|
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
||||||
|
* and/or sell copies of the Software, and to permit persons to whom the
|
||||||
|
* Software is furnished to do so, subject to the following conditions:
|
||||||
|
*
|
||||||
|
* The above copyright notice and this permission notice (including the next
|
||||||
|
* paragraph) shall be included in all copies or substantial portions of the
|
||||||
|
* Software.
|
||||||
|
*
|
||||||
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||||
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||||
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
||||||
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||||
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
||||||
|
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
|
||||||
|
* DEALINGS IN THE SOFTWARE.
|
||||||
|
*/
|
||||||
|
|
||||||
|
#include <math.h>
|
||||||
|
|
||||||
|
#include <xmmintrin.h>
|
||||||
|
|
||||||
|
static inline float hmax_ps(__m128 val)
|
||||||
|
{
|
||||||
|
__m128 t = _mm_movehl_ps(val, val);
|
||||||
|
t = _mm_max_ps(t, val);
|
||||||
|
val = _mm_shuffle_ps(val, t, 0x55);
|
||||||
|
val = _mm_max_ss(t, val);
|
||||||
|
return _mm_cvtss_f32(val);
|
||||||
|
}
|
||||||
|
|
||||||
|
static void impl_peaks_process_sse(struct resample *r, int channel,
|
||||||
|
void *src, uint32_t *in_len, void *dst, uint32_t *out_len)
|
||||||
|
{
|
||||||
|
struct peaks_data *pd = r->data;
|
||||||
|
float *s = src, *d = dst, m;
|
||||||
|
int i, o, end, chunk, unrolled;
|
||||||
|
__m128 in, max, mask = _mm_set_ps1(-0.0f);
|
||||||
|
|
||||||
|
o = i = 0;
|
||||||
|
|
||||||
|
m = pd->max_f[channel];
|
||||||
|
max = _mm_set1_ps(m);
|
||||||
|
|
||||||
|
while (i < *in_len && o < *out_len) {
|
||||||
|
end = ((uint64_t) (pd->o_count + 1) * r->i_rate) / r->o_rate;
|
||||||
|
end = end > pd->i_count ? end - pd->i_count : 0;
|
||||||
|
chunk = SPA_MIN(end, *in_len);
|
||||||
|
|
||||||
|
unrolled = chunk - (chunk & 3);
|
||||||
|
|
||||||
|
for (; i < unrolled; i+=4) {
|
||||||
|
in = _mm_loadu_ps(&s[i]);
|
||||||
|
in = _mm_andnot_ps(mask, in);
|
||||||
|
max = _mm_max_ps(max, in);
|
||||||
|
}
|
||||||
|
for (; i < chunk; i++)
|
||||||
|
m = SPA_MAX(fabsf(s[i]), m);
|
||||||
|
|
||||||
|
if (i == end) {
|
||||||
|
d[o++] = SPA_MAX(hmax_ps(max), m);
|
||||||
|
m = 0.0f;
|
||||||
|
max = _mm_set1_ps(m);
|
||||||
|
pd->o_count++;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
pd->max_f[channel] = SPA_MAX(hmax_ps(max), m);
|
||||||
|
|
||||||
|
*out_len = o;
|
||||||
|
*in_len = i;
|
||||||
|
pd->i_count += i;
|
||||||
|
|
||||||
|
while (pd->i_count >= r->i_rate) {
|
||||||
|
pd->i_count -= r->i_rate;
|
||||||
|
pd->o_count -= r->o_rate;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
@ -30,6 +30,10 @@ struct peaks_data {
|
||||||
float max_f[0];
|
float max_f[0];
|
||||||
};
|
};
|
||||||
|
|
||||||
|
#if defined (__SSE__)
|
||||||
|
#include "resample-peaks-sse.h"
|
||||||
|
#endif
|
||||||
|
|
||||||
static void impl_peaks_free(struct resample *r)
|
static void impl_peaks_free(struct resample *r)
|
||||||
{
|
{
|
||||||
if (r->data)
|
if (r->data)
|
||||||
|
|
@ -45,28 +49,28 @@ static void impl_peaks_process(struct resample *r, int channel,
|
||||||
void *src, uint32_t *in_len, void *dst, uint32_t *out_len)
|
void *src, uint32_t *in_len, void *dst, uint32_t *out_len)
|
||||||
{
|
{
|
||||||
struct peaks_data *pd = r->data;
|
struct peaks_data *pd = r->data;
|
||||||
float *s = src, *d = dst;
|
float *s = src, *d = dst, m;
|
||||||
int i, o, end, chunk;
|
int i, o, end, chunk;
|
||||||
|
|
||||||
o = i = 0;
|
o = i = 0;
|
||||||
|
m = pd->max_f[channel];
|
||||||
|
|
||||||
while (i < *in_len && o < *out_len) {
|
while (i < *in_len && o < *out_len) {
|
||||||
end = ((uint64_t) (pd->o_count + 1) * r->i_rate) / r->o_rate;
|
end = ((uint64_t) (pd->o_count + 1) * r->i_rate) / r->o_rate;
|
||||||
end = end > pd->i_count ? end - pd->i_count : 0;
|
end = end > pd->i_count ? end - pd->i_count : 0;
|
||||||
chunk = SPA_MIN(end, *in_len);
|
chunk = SPA_MIN(end, *in_len);
|
||||||
|
|
||||||
for (; i < chunk; i++) {
|
for (; i < chunk; i++)
|
||||||
float n = fabsf(s[i]);
|
m = SPA_MAX(fabsf(s[i]), m);
|
||||||
if (n > pd->max_f[channel])
|
|
||||||
pd->max_f[channel] = n;
|
|
||||||
}
|
|
||||||
|
|
||||||
if (i == end) {
|
if (i == end) {
|
||||||
d[o++] = pd->max_f[channel];
|
d[o++] = m;
|
||||||
pd->max_f[channel] = 0.0f;
|
m = 0.0f;
|
||||||
pd->o_count++;
|
pd->o_count++;
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
pd->max_f[channel] = m;
|
||||||
|
|
||||||
*out_len = o;
|
*out_len = o;
|
||||||
*in_len = i;
|
*in_len = i;
|
||||||
pd->i_count += i;
|
pd->i_count += i;
|
||||||
|
|
@ -89,7 +93,13 @@ static int impl_peaks_init(struct resample *r)
|
||||||
|
|
||||||
r->free = impl_peaks_free;
|
r->free = impl_peaks_free;
|
||||||
r->update_rate = impl_peaks_update_rate;
|
r->update_rate = impl_peaks_update_rate;
|
||||||
|
#if defined (__SSE__)
|
||||||
|
if (r->cpu_flags & SPA_CPU_FLAG_SSE)
|
||||||
|
r->process = impl_peaks_process_sse;
|
||||||
|
else
|
||||||
|
#endif
|
||||||
r->process = impl_peaks_process;
|
r->process = impl_peaks_process;
|
||||||
|
|
||||||
r->reset = impl_peaks_reset;
|
r->reset = impl_peaks_reset;
|
||||||
d = r->data = calloc(1, sizeof(struct peaks_data) * sizeof(float) * r->channels);
|
d = r->data = calloc(1, sizeof(struct peaks_data) * sizeof(float) * r->channels);
|
||||||
if (r->data == NULL)
|
if (r->data == NULL)
|
||||||
|
|
|
||||||
|
|
@ -92,6 +92,7 @@ struct impl {
|
||||||
struct spa_node node;
|
struct spa_node node;
|
||||||
|
|
||||||
struct spa_log *log;
|
struct spa_log *log;
|
||||||
|
struct spa_cpu *cpu;
|
||||||
|
|
||||||
struct props props;
|
struct props props;
|
||||||
|
|
||||||
|
|
@ -925,9 +926,18 @@ impl_init(const struct spa_handle_factory *factory,
|
||||||
this = (struct impl *) handle;
|
this = (struct impl *) handle;
|
||||||
|
|
||||||
for (i = 0; i < n_support; i++) {
|
for (i = 0; i < n_support; i++) {
|
||||||
if (support[i].type == SPA_TYPE_INTERFACE_Log)
|
switch (support[i].type) {
|
||||||
|
case SPA_TYPE_INTERFACE_Log:
|
||||||
this->log = support[i].data;
|
this->log = support[i].data;
|
||||||
|
break;
|
||||||
|
case SPA_TYPE_INTERFACE_CPU:
|
||||||
|
this->cpu = support[i].data;
|
||||||
|
break;
|
||||||
}
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
if (this->cpu)
|
||||||
|
this->resample.cpu_flags = spa_cpu_get_flags(this->cpu);
|
||||||
|
|
||||||
if (info != NULL && (str = spa_dict_lookup(info, "resample.peaks")) != NULL)
|
if (info != NULL && (str = spa_dict_lookup(info, "resample.peaks")) != NULL)
|
||||||
this->monitor = atoi(str);
|
this->monitor = atoi(str);
|
||||||
|
|
|
||||||
|
|
@ -22,7 +22,10 @@
|
||||||
* DEALINGS IN THE SOFTWARE.
|
* DEALINGS IN THE SOFTWARE.
|
||||||
*/
|
*/
|
||||||
|
|
||||||
|
#include <spa/support/cpu.h>
|
||||||
|
|
||||||
struct resample {
|
struct resample {
|
||||||
|
uint32_t cpu_flags;
|
||||||
uint32_t channels;
|
uint32_t channels;
|
||||||
uint32_t i_rate;
|
uint32_t i_rate;
|
||||||
uint32_t o_rate;
|
uint32_t o_rate;
|
||||||
|
|
|
||||||
Loading…
Add table
Add a link
Reference in a new issue