audiomixer: rewrite the mixer functions

Let the mixer functions accumulate the intermediate results into a
larger size variable and then clamp to the final precission. This avoids
distortions because of intermediate clamping.

Although the access pattern of the reads are no longer sequential, the
writes are sequential and we don't need to read intermediate values.
Together with the avoided clamping this is probably faster overall.

Add a unit test for the various cases.
This commit is contained in:
Wim Taymans 2022-07-09 18:11:13 +02:00
parent 3ffb9f4b26
commit 371b5a1836
4 changed files with 376 additions and 48 deletions

View file

@ -0,0 +1,272 @@
/* Spa
*
* Copyright © 2022 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 "config.h"
#include <string.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <time.h>
#include <spa/debug/mem.h>
#include "test-helper.h"
#include "mix-ops.c"
static uint32_t cpu_flags;
#define N_SAMPLES 1024
static uint8_t samp_out[N_SAMPLES * 8];
static void compare_mem(int i, int j, const void *m1, const void *m2, size_t size)
{
int res = memcmp(m1, m2, size);
if (res != 0) {
fprintf(stderr, "%d %d %zd:\n", i, j, size);
spa_debug_mem(0, m1, size);
spa_debug_mem(0, m2, size);
}
spa_assert_se(res == 0);
}
static int run_test(const char *name, const void *src[], uint32_t n_src, const void *dst,
size_t dst_size, uint32_t n_samples, mix_func_t mix)
{
struct mix_ops ops;
ops.fmt = SPA_AUDIO_FORMAT_F32;
ops.n_channels = 1;
ops.cpu_flags = cpu_flags;
mix_ops_init(&ops);
fprintf(stderr, "%s\n", name);
mix(&ops, (void *)samp_out, src, n_src, n_samples);
compare_mem(0, 0, samp_out, dst, dst_size);
return 0;
}
static void test_s8(void)
{
int8_t out[] = { 0x00, 0x00, 0x00, 0x00 };
int8_t in_1[] = { 0x00, 0x00, 0x00, 0x00 };
int8_t in_2[] = { 0x7f, 0x80, 0x40, 0xc0 };
int8_t in_3[] = { 0x40, 0xc0, 0xc0, 0x40 };
int8_t in_4[] = { 0xc0, 0x40, 0x40, 0xc0 };
int8_t out_4[] = { 0x7f, 0x80, 0x40, 0xc0 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_s8_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_s8_c);
run_test("test_s8_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_s8_c);
run_test("test_s8_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_s8_c);
}
static void test_u8(void)
{
uint8_t out[] = { 0x80, 0x80, 0x80, 0x80 };
uint8_t in_1[] = { 0x80, 0x80, 0x80, 0x80 };
uint8_t in_2[] = { 0xff, 0x00, 0xc0, 0x40 };
uint8_t in_3[] = { 0xc0, 0x40, 0x40, 0xc0 };
uint8_t in_4[] = { 0x40, 0xc0, 0xc0, 0x40 };
uint8_t out_4[] = { 0xff, 0x00, 0xc0, 0x40 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_u8_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_u8_c);
run_test("test_u8_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_u8_c);
run_test("test_u8_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_u8_c);
}
static void test_s16(void)
{
int16_t out[] = { 0x0000, 0x0000, 0x0000, 0x0000 };
int16_t in_1[] = { 0x0000, 0x0000, 0x0000, 0x0000 };
int16_t in_2[] = { 0x7fff, 0x8000, 0x4000, 0xc000 };
int16_t in_3[] = { 0x4000, 0xc000, 0xc000, 0x4000 };
int16_t in_4[] = { 0xc000, 0x4000, 0x4000, 0xc000 };
int16_t out_4[] = { 0x7fff, 0x8000, 0x4000, 0xc000 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_s16_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_s16_c);
run_test("test_s16_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_s16_c);
run_test("test_s16_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_s16_c);
}
static void test_u16(void)
{
uint16_t out[] = { 0x8000, 0x8000, 0x8000, 0x8000 };
uint16_t in_1[] = { 0x8000, 0x8000, 0x8000 , 0x8000};
uint16_t in_2[] = { 0xffff, 0x0000, 0xc000, 0x4000 };
uint16_t in_3[] = { 0xc000, 0x4000, 0x4000, 0xc000 };
uint16_t in_4[] = { 0x4000, 0xc000, 0xc000, 0x4000 };
uint16_t out_4[] = { 0xffff, 0x0000, 0xc000, 0x4000 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_u16_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_u16_c);
run_test("test_u16_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_u16_c);
run_test("test_u16_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_u16_c);
}
static void test_s24(void)
{
int24_t out[] = { S32_TO_S24(0x000000), S32_TO_S24(0x000000), S32_TO_S24(0x000000) };
int24_t in_1[] = { S32_TO_S24(0x000000), S32_TO_S24(0x000000), S32_TO_S24(0x000000) };
int24_t in_2[] = { S32_TO_S24(0x7fffff), S32_TO_S24(0xff800000), S32_TO_S24(0x400000) };
int24_t in_3[] = { S32_TO_S24(0x400000), S32_TO_S24(0xffc00000), S32_TO_S24(0xffc00000) };
int24_t in_4[] = { S32_TO_S24(0xffc00000), S32_TO_S24(0x400000), S32_TO_S24(0x400000) };
int24_t out_4[] = { S32_TO_S24(0x7fffff), S32_TO_S24(0xff800000), S32_TO_S24(0x400000) };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_s24_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_s24_c);
run_test("test_s24_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_s24_c);
run_test("test_s24_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_s24_c);
}
static void test_u24(void)
{
uint24_t out[] = { U32_TO_U24(0x800000), U32_TO_U24(0x800000), U32_TO_U24(0x800000) };
uint24_t in_1[] = { U32_TO_U24(0x800000), U32_TO_U24(0x800000), U32_TO_U24(0x800000) };
uint24_t in_2[] = { U32_TO_U24(0xffffffff), U32_TO_U24(0x000000), U32_TO_U24(0xffc00000) };
uint24_t in_3[] = { U32_TO_U24(0xffc00000), U32_TO_U24(0x400000), U32_TO_U24(0x400000) };
uint24_t in_4[] = { U32_TO_U24(0x400000), U32_TO_U24(0xffc00000), U32_TO_U24(0xffc00000) };
uint24_t out_4[] = { U32_TO_U24(0xffffffff), U32_TO_U24(0x000000), U32_TO_U24(0xffc00000) };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_u24_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_u24_c);
run_test("test_u24_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_u24_c);
run_test("test_u24_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_u24_c);
}
static void test_s32(void)
{
int32_t out[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000 };
int32_t in_1[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000 };
int32_t in_2[] = { 0x7fffffff, 0x80000000, 0x40000000, 0xc0000000 };
int32_t in_3[] = { 0x40000000, 0xc0000000, 0xc0000000, 0x40000000 };
int32_t in_4[] = { 0xc0000000, 0x40000000, 0x40000000, 0xc0000000 };
int32_t out_4[] = { 0x7fffffff, 0x80000000, 0x40000000, 0xc0000000 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_s32_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_s32_c);
run_test("test_s32_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_s32_c);
run_test("test_s32_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_s32_c);
}
static void test_u32(void)
{
uint32_t out[] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 };
uint32_t in_1[] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 };
uint32_t in_2[] = { 0xffffffff, 0x00000000, 0xc0000000, 0x40000000 };
uint32_t in_3[] = { 0xc0000000, 0x40000000, 0x40000000, 0xc0000000 };
uint32_t in_4[] = { 0x40000000, 0xc0000000, 0xc0000000, 0x40000000 };
uint32_t out_4[] = { 0xffffffff, 0x00000000, 0xc0000000, 0x40000000 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_u32_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_u32_c);
run_test("test_u32_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_u32_c);
run_test("test_u32_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_u32_c);
}
static void test_s24_32(void)
{
int32_t out[] = { 0x000000, 0x000000, 0x000000, 0x000000 };
int32_t in_1[] = { 0x000000, 0x000000, 0x000000, 0x000000 };
int32_t in_2[] = { 0x7fffff, 0xff800000, 0x400000, 0xffc00000 };
int32_t in_3[] = { 0x400000, 0xffc00000, 0xffc00000, 0x400000 };
int32_t in_4[] = { 0xffc00000, 0x400000, 0x400000, 0xffc00000 };
int32_t out_4[] = { 0x7fffff, 0xff800000, 0x400000, 0xffc00000 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_s24_32_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_s24_32_c);
run_test("test_s24_32_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_s24_32_c);
run_test("test_s24_32_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_s24_32_c);
}
static void test_u24_32(void)
{
uint32_t out[] = { 0x800000, 0x800000, 0x800000, 0x800000 };
uint32_t in_1[] = { 0x800000, 0x800000, 0x800000, 0x800000 };
uint32_t in_2[] = { 0xffffff, 0x000000, 0xc00000, 0x400000 };
uint32_t in_3[] = { 0xc00000, 0x400000, 0x400000, 0xc00000 };
uint32_t in_4[] = { 0x400000, 0xc00000, 0xc00000, 0x400000 };
uint32_t out_4[] = { 0xffffff, 0x000000, 0xc00000, 0x400000 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_u24_32_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_u24_32_c);
run_test("test_u24_32_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_u24_32_c);
run_test("test_u24_32_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_u24_32_c);
}
static void test_f32(void)
{
float out[] = { 0.0f, 0.0f, 0.0f, 0.0f };
float in_1[] = { 0.0f, 0.0f, 0.0f, 0.0f };
float in_2[] = { 1.0f, -1.0f, 0.5f, -0.5f };
float in_3[] = { 0.5f, -0.5f, -0.5f, 0.5f };
float in_4[] = { -0.5f, 1.0f, 0.5f, -0.5f };
float out_4[] = { 1.0f, -0.5f, 0.5f, -0.5f };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_f32_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_f32_c);
run_test("test_f32_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_f32_c);
run_test("test_f32_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_f32_c);
}
static void test_f64(void)
{
double out[] = { 0.0, 0.0, 0.0, 0.0 };
double in_1[] = { 0.0, 0.0, 0.0, 0.0 };
double in_2[] = { 1.0, -1.0, 0.5, -0.5 };
double in_3[] = { 0.5, -0.5, -0.5, 0.5 };
double in_4[] = { -0.5, 1.0, 0.5, -0.5 };
double out_4[] = { 1.0, -0.5, 0.5, -0.5 };
const void *src[6] = { in_1, in_2, in_3, in_4 };
run_test("test_f64_0", NULL, 0, out, sizeof(out), SPA_N_ELEMENTS(out), mix_f64_c);
run_test("test_f64_1", src, 1, in_1, sizeof(in_1), SPA_N_ELEMENTS(in_1), mix_f64_c);
run_test("test_f64_4", src, 4, out_4, sizeof(out_4), SPA_N_ELEMENTS(out_4), mix_f64_c);
}
int main(int argc, char *argv[])
{
cpu_flags = get_cpu_flags();
printf("got get CPU flags %d\n", cpu_flags);
test_s8();
test_u8();
test_s16();
test_u16();
test_s24();
test_u24();
test_s32();
test_u32();
test_s24_32();
test_u24_32();
test_f32();
test_f64();
return 0;
}