mixer: Move floatmix to the audio mixer plugin

Move floatmix to the audiomixer plugin and change the name to
AUDIO_MIXER_DSP.
Add runtime selectable sse and sse2 optimizations.
Load the port mixer plugin dynamically based on the factory_name.
Add some more debug
This commit is contained in:
Wim Taymans 2019-10-03 16:20:12 +02:00
parent 27eabede35
commit 0ecbe4844e
13 changed files with 482 additions and 404 deletions

View file

@ -1,6 +1,6 @@
/* Spa
*
* Copyright © 2018 Wim Taymans
* Copyright © 2019 Wim Taymans
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
@ -22,262 +22,86 @@
* DEALINGS IN THE SOFTWARE.
*/
#include <string.h>
#include <stdio.h>
#include <math.h>
#include <spa/support/cpu.h>
#include <spa/utils/defs.h>
#include <spa/param/audio/format-utils.h>
#include "mix-ops.h"
static void
clear_s16(void *dst, int n_bytes)
{
memset(dst, 0, n_bytes);
}
typedef void (*mix_func_t) (struct mix_ops *ops, void * SPA_RESTRICT dst,
const void * SPA_RESTRICT src[], uint32_t n_src, uint32_t n_samples);
static void
clear_f32(void *dst, int n_bytes)
{
memset(dst, 0, n_bytes);
}
struct mix_info {
uint32_t fmt;
uint32_t n_channels;
uint32_t cpu_flags;
uint32_t stride;
mix_func_t process;
};
static void
copy_s16(void *dst, const void *src, int n_bytes)
static struct mix_info mix_table[] =
{
memcpy(dst, src, n_bytes);
}
/* f32 */
#if defined (HAVE_SSE)
{ SPA_AUDIO_FORMAT_F32, 1, SPA_CPU_FLAG_SSE, 4, mix_f32_sse },
{ SPA_AUDIO_FORMAT_F32P, 1, SPA_CPU_FLAG_SSE, 4, mix_f32_sse },
#endif
{ SPA_AUDIO_FORMAT_F32, 1, 0, 4, mix_f32_c },
{ SPA_AUDIO_FORMAT_F32P, 1, 0, 4, mix_f32_c },
static void
copy_f32(void *dst, const void *src, int n_bytes)
#if defined (HAVE_SSE2)
{ SPA_AUDIO_FORMAT_F64, 1, SPA_CPU_FLAG_SSE2, 8, mix_f64_sse2 },
{ SPA_AUDIO_FORMAT_F64P, 1, SPA_CPU_FLAG_SSE2, 8, mix_f64_sse2 },
#endif
{ SPA_AUDIO_FORMAT_F64, 1, 0, 8, mix_f64_c },
{ SPA_AUDIO_FORMAT_F64P, 1, 0, 8, mix_f64_c },
};
#define MATCH_CHAN(a,b) ((a) == 0 || (a) == (b))
#define MATCH_CPU_FLAGS(a,b) ((a) == 0 || ((a) & (b)) == a)
static const struct mix_info *find_mix_info(uint32_t fmt,
uint32_t n_channels, uint32_t cpu_flags)
{
memcpy(dst, src, n_bytes);
}
size_t i;
static void
add_s16(void *dst, const void *src, int n_bytes)
{
const int16_t *s = src;
int16_t *d = dst;
int32_t t;
n_bytes /= sizeof(int16_t);
while (n_bytes--) {
t = *d + *s;
*d = SPA_CLAMP(t, INT16_MIN, INT16_MAX);
d++;
s++;
for (i = 0; i < SPA_N_ELEMENTS(mix_table); i++) {
if (mix_table[i].fmt == fmt &&
MATCH_CHAN(mix_table[i].n_channels, n_channels) &&
MATCH_CPU_FLAGS(mix_table[i].cpu_flags, cpu_flags))
return &mix_table[i];
}
return NULL;
}
static void
add_f32(void *dst, const void *src, int n_bytes)
static void impl_mix_ops_clear(struct mix_ops *ops, void * SPA_RESTRICT dst, uint32_t n_samples)
{
const float *s = src;
float *d = dst;
n_bytes /= sizeof(float);
while (n_bytes--) {
*d += *s;
d++;
s++;
}
const struct mix_info *info = ops->priv;
memset(dst, 0, n_samples * info->stride);
}
static void
copy_scale_s16(void *dst, const void *src, const double scale, int n_bytes)
static void impl_mix_ops_free(struct mix_ops *ops)
{
const int16_t *s = src;
int16_t *d = dst;;
int32_t v = scale * (1 << 11), t;
n_bytes /= sizeof(int16_t);
while (n_bytes--) {
t = (*s * v) >> 11;
*d = SPA_CLAMP(t, INT16_MIN, INT16_MAX);
d++;
s++;
}
spa_zero(*ops);
}
static void
copy_scale_f32(void *dst, const void *src, const double scale, int n_bytes)
int mix_ops_init(struct mix_ops *ops)
{
const float *s = src;
float *d = dst;
float v = scale;
const struct mix_info *info;
n_bytes /= sizeof(float);
while (n_bytes--) {
*d = *s * v;
d++;
s++;
}
}
static void
add_scale_s16(void *dst, const void *src, const double scale, int n_bytes)
{
const int16_t *s = src;
int16_t *d = dst;
int32_t v = scale * (1 << 11), t;
n_bytes /= sizeof(int16_t);
while (n_bytes--) {
t = *d + ((*s * v) >> 11);
*d = SPA_CLAMP(t, INT16_MIN, INT16_MAX);
d++;
s++;
}
}
static void
add_scale_f32(void *dst, const void *src, const double scale, int n_bytes)
{
const float *s = src;
float *d = dst;
float v = scale;
n_bytes /= sizeof(float);
while (n_bytes--) {
*d += *s * v;
d++;
s++;
}
}
static void
copy_s16_i(void *dst, int dst_stride, const void *src, int src_stride, int n_bytes)
{
const int16_t *s = src;
int16_t *d = dst;
n_bytes /= sizeof(int16_t);
while (n_bytes--) {
*d = *s;
d += dst_stride;
s += src_stride;
}
}
static void
copy_f32_i(void *dst, int dst_stride, const void *src, int src_stride, int n_bytes)
{
const float *s = src;
float *d = dst;
n_bytes /= sizeof(float);
while (n_bytes--) {
*d = *s;
d += dst_stride;
s += src_stride;
}
}
static void
add_s16_i(void *dst, int dst_stride, const void *src, int src_stride, int n_bytes)
{
const int16_t *s = src;
int16_t *d = dst;
int32_t t;
n_bytes /= sizeof(int16_t);
while (n_bytes--) {
t = *d + *s;
*d = SPA_CLAMP(t, INT16_MIN, INT16_MAX);
d += dst_stride;
s += src_stride;
}
}
static void
add_f32_i(void *dst, int dst_stride, const void *src, int src_stride, int n_bytes)
{
const float *s = src;
float *d = dst;
n_bytes /= sizeof(float);
while (n_bytes--) {
*d += *s;
d += dst_stride;
s += src_stride;
}
}
static void
copy_scale_s16_i(void *dst, int dst_stride, const void *src, int src_stride, const double scale, int n_bytes)
{
const int16_t *s = src;
int16_t *d = dst;
int32_t v = scale * (1 << 11), t;
n_bytes /= sizeof(int16_t);
while (n_bytes--) {
t = (*s * v) >> 11;
*d = SPA_CLAMP(t, INT16_MIN, INT16_MAX);
d += dst_stride;
s += src_stride;
}
}
static void
copy_scale_f32_i(void *dst, int dst_stride, const void *src, int src_stride, const double scale, int n_bytes)
{
const float *s = src;
float *d = dst;
float v = scale;
n_bytes /= sizeof(float);
while (n_bytes--) {
*d = *s * v;
d += dst_stride;
s += src_stride;
}
}
static void
add_scale_s16_i(void *dst, int dst_stride, const void *src, int src_stride, const double scale, int n_bytes)
{
const int16_t *s = src;
int16_t *d = dst;
int32_t v = scale * (1 << 11), t;
n_bytes /= sizeof(int16_t);
while (n_bytes--) {
t = *d + ((*s * v) >> 11);
*d = SPA_CLAMP(t, INT16_MIN, INT16_MAX);
d += dst_stride;
s += src_stride;
}
}
static void
add_scale_f32_i(void *dst, int dst_stride, const void *src, int src_stride, const double scale, int n_bytes)
{
const float *s = src;
float *d = dst;
float v = scale;
n_bytes /= sizeof(float);
while (n_bytes--) {
*d += *s * v;
d += dst_stride;
s += src_stride;
}
}
void spa_audiomixer_get_ops(struct spa_audiomixer_ops *ops)
{
ops->clear[FMT_S16] = clear_s16;
ops->clear[FMT_F32] = clear_f32;
ops->copy[FMT_S16] = copy_s16;
ops->copy[FMT_F32] = copy_f32;
ops->add[FMT_S16] = add_s16;
ops->add[FMT_F32] = add_f32;
ops->copy_scale[FMT_S16] = copy_scale_s16;
ops->copy_scale[FMT_F32] = copy_scale_f32;
ops->add_scale[FMT_S16] = add_scale_s16;
ops->add_scale[FMT_F32] = add_scale_f32;
ops->copy_i[FMT_S16] = copy_s16_i;
ops->copy_i[FMT_F32] = copy_f32_i;
ops->add_i[FMT_S16] = add_s16_i;
ops->add_i[FMT_F32] = add_f32_i;
ops->copy_scale_i[FMT_S16] = copy_scale_s16_i;
ops->copy_scale_i[FMT_F32] = copy_scale_f32_i;
ops->add_scale_i[FMT_S16] = add_scale_s16_i;
ops->add_scale_i[FMT_F32] = add_scale_f32_i;
info = find_mix_info(ops->fmt, ops->n_channels, ops->cpu_flags);
if (info == NULL)
return -ENOTSUP;
ops->priv = info;
ops->cpu_flags = info->cpu_flags;
ops->clear = impl_mix_ops_clear;
ops->process = info->process;
ops->free = impl_mix_ops_free;
return 0;
}