1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-26 19:01:44 +02:00
FFmpeg/libavcodec/dcadsp.c
James Almer 8ae7447941 x86/dcadec: add ff_lfe_fir0_float_{sse,sse2,avx,fma3}
Up to ~4 times faster on x86_64, ~8 times on x86_32 if compiling using x87 fp math.

Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
2016-02-06 01:36:55 -03:00

417 lines
13 KiB
C

/*
* Copyright (C) 2016 foo86
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/mem.h"
#include "dcadsp.h"
#include "dcamath.h"
static void decode_hf_c(int32_t **dst,
const int32_t *vq_index,
const int8_t hf_vq[1024][32],
int32_t scale_factors[32][2],
ptrdiff_t sb_start, ptrdiff_t sb_end,
ptrdiff_t ofs, ptrdiff_t len)
{
int i, j;
for (i = sb_start; i < sb_end; i++) {
const int8_t *coeff = hf_vq[vq_index[i]];
int32_t scale = scale_factors[i][0];
for (j = 0; j < len; j++)
dst[i][j + ofs] = clip23(coeff[j] * scale + (1 << 3) >> 4);
}
}
static void decode_joint_c(int32_t **dst, int32_t **src,
const int32_t *scale_factors,
ptrdiff_t sb_start, ptrdiff_t sb_end,
ptrdiff_t ofs, ptrdiff_t len)
{
int i, j;
for (i = sb_start; i < sb_end; i++) {
int32_t scale = scale_factors[i];
for (j = 0; j < len; j++)
dst[i][j + ofs] = clip23(mul17(src[i][j + ofs], scale));
}
}
static void lfe_fir_float_c(float *pcm_samples, int32_t *lfe_samples,
const float *filter_coeff, ptrdiff_t npcmblocks,
int dec_select)
{
// Select decimation factor
int factor = 64 << dec_select;
int ncoeffs = 8 >> dec_select;
int nlfesamples = npcmblocks >> (dec_select + 1);
int i, j, k;
for (i = 0; i < nlfesamples; i++) {
// One decimated sample generates 64 or 128 interpolated ones
for (j = 0; j < factor / 2; j++) {
float a = 0;
float b = 0;
for (k = 0; k < ncoeffs; k++) {
a += filter_coeff[ j * ncoeffs + k] * lfe_samples[-k];
b += filter_coeff[255 - j * ncoeffs - k] * lfe_samples[-k];
}
pcm_samples[ j] = a;
pcm_samples[factor / 2 + j] = b;
}
lfe_samples++;
pcm_samples += factor;
}
}
static void lfe_fir0_float_c(float *pcm_samples, int32_t *lfe_samples,
const float *filter_coeff, ptrdiff_t npcmblocks)
{
lfe_fir_float_c(pcm_samples, lfe_samples, filter_coeff, npcmblocks, 0);
}
static void lfe_fir1_float_c(float *pcm_samples, int32_t *lfe_samples,
const float *filter_coeff, ptrdiff_t npcmblocks)
{
lfe_fir_float_c(pcm_samples, lfe_samples, filter_coeff, npcmblocks, 1);
}
static void lfe_x96_float_c(float *dst, const float *src,
float *hist, ptrdiff_t len)
{
float prev = *hist;
int i;
for (i = 0; i < len; i++) {
float a = 0.25f * src[i] + 0.75f * prev;
float b = 0.75f * src[i] + 0.25f * prev;
prev = src[i];
*dst++ = a;
*dst++ = b;
}
*hist = prev;
}
static void sub_qmf32_float_c(SynthFilterContext *synth,
FFTContext *imdct,
float *pcm_samples,
int32_t **subband_samples_lo,
int32_t **subband_samples_hi,
float *hist1, int *offset, float *hist2,
const float *filter_coeff, ptrdiff_t npcmblocks,
float scale)
{
LOCAL_ALIGNED(32, float, input, [32]);
int i, j;
for (j = 0; j < npcmblocks; j++) {
// Load in one sample from each subband
for (i = 0; i < 32; i++) {
if ((i - 1) & 2)
input[i] = -subband_samples_lo[i][j];
else
input[i] = subband_samples_lo[i][j];
}
// One subband sample generates 32 interpolated ones
synth->synth_filter_float(imdct, hist1, offset,
hist2, filter_coeff,
pcm_samples, input, scale);
pcm_samples += 32;
}
}
static void sub_qmf64_float_c(SynthFilterContext *synth,
FFTContext *imdct,
float *pcm_samples,
int32_t **subband_samples_lo,
int32_t **subband_samples_hi,
float *hist1, int *offset, float *hist2,
const float *filter_coeff, ptrdiff_t npcmblocks,
float scale)
{
LOCAL_ALIGNED(32, float, input, [64]);
int i, j;
if (!subband_samples_hi)
memset(&input[32], 0, sizeof(input[0]) * 32);
for (j = 0; j < npcmblocks; j++) {
// Load in one sample from each subband
if (subband_samples_hi) {
// Full 64 subbands, first 32 are residual coded
for (i = 0; i < 32; i++) {
if ((i - 1) & 2)
input[i] = -subband_samples_lo[i][j] - subband_samples_hi[i][j];
else
input[i] = subband_samples_lo[i][j] + subband_samples_hi[i][j];
}
for (i = 32; i < 64; i++) {
if ((i - 1) & 2)
input[i] = -subband_samples_hi[i][j];
else
input[i] = subband_samples_hi[i][j];
}
} else {
// Only first 32 subbands
for (i = 0; i < 32; i++) {
if ((i - 1) & 2)
input[i] = -subband_samples_lo[i][j];
else
input[i] = subband_samples_lo[i][j];
}
}
// One subband sample generates 64 interpolated ones
synth->synth_filter_float_64(imdct, hist1, offset,
hist2, filter_coeff,
pcm_samples, input, scale);
pcm_samples += 64;
}
}
static void lfe_fir_fixed_c(int32_t *pcm_samples, int32_t *lfe_samples,
const int32_t *filter_coeff, ptrdiff_t npcmblocks)
{
// Select decimation factor
int nlfesamples = npcmblocks >> 1;
int i, j, k;
for (i = 0; i < nlfesamples; i++) {
// One decimated sample generates 64 interpolated ones
for (j = 0; j < 32; j++) {
int64_t a = 0;
int64_t b = 0;
for (k = 0; k < 8; k++) {
a += (int64_t)filter_coeff[ j * 8 + k] * lfe_samples[-k];
b += (int64_t)filter_coeff[255 - j * 8 - k] * lfe_samples[-k];
}
pcm_samples[ j] = clip23(norm23(a));
pcm_samples[32 + j] = clip23(norm23(b));
}
lfe_samples++;
pcm_samples += 64;
}
}
static void lfe_x96_fixed_c(int32_t *dst, const int32_t *src,
int32_t *hist, ptrdiff_t len)
{
int32_t prev = *hist;
int i;
for (i = 0; i < len; i++) {
int64_t a = INT64_C(2097471) * src[i] + INT64_C(6291137) * prev;
int64_t b = INT64_C(6291137) * src[i] + INT64_C(2097471) * prev;
prev = src[i];
*dst++ = clip23(norm23(a));
*dst++ = clip23(norm23(b));
}
*hist = prev;
}
static void sub_qmf32_fixed_c(SynthFilterContext *synth,
DCADCTContext *imdct,
int32_t *pcm_samples,
int32_t **subband_samples_lo,
int32_t **subband_samples_hi,
int32_t *hist1, int *offset, int32_t *hist2,
const int32_t *filter_coeff, ptrdiff_t npcmblocks)
{
LOCAL_ALIGNED(32, int32_t, input, [32]);
int i, j;
for (j = 0; j < npcmblocks; j++) {
// Load in one sample from each subband
for (i = 0; i < 32; i++)
input[i] = subband_samples_lo[i][j];
// One subband sample generates 32 interpolated ones
synth->synth_filter_fixed(imdct, hist1, offset,
hist2, filter_coeff,
pcm_samples, input);
pcm_samples += 32;
}
}
static void sub_qmf64_fixed_c(SynthFilterContext *synth,
DCADCTContext *imdct,
int32_t *pcm_samples,
int32_t **subband_samples_lo,
int32_t **subband_samples_hi,
int32_t *hist1, int *offset, int32_t *hist2,
const int32_t *filter_coeff, ptrdiff_t npcmblocks)
{
LOCAL_ALIGNED(32, int32_t, input, [64]);
int i, j;
if (!subband_samples_hi)
memset(&input[32], 0, sizeof(input[0]) * 32);
for (j = 0; j < npcmblocks; j++) {
// Load in one sample from each subband
if (subband_samples_hi) {
// Full 64 subbands, first 32 are residual coded
for (i = 0; i < 32; i++)
input[i] = subband_samples_lo[i][j] + subband_samples_hi[i][j];
for (i = 32; i < 64; i++)
input[i] = subband_samples_hi[i][j];
} else {
// Only first 32 subbands
for (i = 0; i < 32; i++)
input[i] = subband_samples_lo[i][j];
}
// One subband sample generates 64 interpolated ones
synth->synth_filter_fixed_64(imdct, hist1, offset,
hist2, filter_coeff,
pcm_samples, input);
pcm_samples += 64;
}
}
static void decor_c(int32_t *dst, const int32_t *src, int coeff, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++)
dst[i] += src[i] * coeff + (1 << 2) >> 3;
}
static void dmix_sub_xch_c(int32_t *dst1, int32_t *dst2,
const int32_t *src, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++) {
int32_t cs = mul23(src[i], 5931520 /* M_SQRT1_2 * (1 << 23) */);
dst1[i] -= cs;
dst2[i] -= cs;
}
}
static void dmix_sub_c(int32_t *dst, const int32_t *src, int coeff, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++)
dst[i] -= mul15(src[i], coeff);
}
static void dmix_add_c(int32_t *dst, const int32_t *src, int coeff, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++)
dst[i] += mul15(src[i], coeff);
}
static void dmix_scale_c(int32_t *dst, int scale, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++)
dst[i] = mul15(dst[i], scale);
}
static void dmix_scale_inv_c(int32_t *dst, int scale_inv, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++)
dst[i] = mul16(dst[i], scale_inv);
}
static void filter0(int32_t *dst, const int32_t *src, int32_t coeff, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++)
dst[i] -= mul22(src[i], coeff);
}
static void filter1(int32_t *dst, const int32_t *src, int32_t coeff, ptrdiff_t len)
{
int i;
for (i = 0; i < len; i++)
dst[i] -= mul23(src[i], coeff);
}
static void assemble_freq_bands_c(int32_t *dst, int32_t *src0, int32_t *src1,
const int32_t *coeff, ptrdiff_t len)
{
int i;
filter0(src0, src1, coeff[0], len);
filter0(src1, src0, coeff[1], len);
filter0(src0, src1, coeff[2], len);
filter0(src1, src0, coeff[3], len);
for (i = 0; i < 8; i++, src0--) {
filter1(src0, src1, coeff[i + 4], len);
filter1(src1, src0, coeff[i + 12], len);
filter1(src0, src1, coeff[i + 4], len);
}
for (i = 0; i < len; i++) {
*dst++ = *src1++;
*dst++ = *++src0;
}
}
av_cold void ff_dcadsp_init(DCADSPContext *s)
{
s->decode_hf = decode_hf_c;
s->decode_joint = decode_joint_c;
s->lfe_fir_float[0] = lfe_fir0_float_c;
s->lfe_fir_float[1] = lfe_fir1_float_c;
s->lfe_x96_float = lfe_x96_float_c;
s->sub_qmf_float[0] = sub_qmf32_float_c;
s->sub_qmf_float[1] = sub_qmf64_float_c;
s->lfe_fir_fixed = lfe_fir_fixed_c;
s->lfe_x96_fixed = lfe_x96_fixed_c;
s->sub_qmf_fixed[0] = sub_qmf32_fixed_c;
s->sub_qmf_fixed[1] = sub_qmf64_fixed_c;
s->decor = decor_c;
s->dmix_sub_xch = dmix_sub_xch_c;
s->dmix_sub = dmix_sub_c;
s->dmix_add = dmix_add_c;
s->dmix_scale = dmix_scale_c;
s->dmix_scale_inv = dmix_scale_inv_c;
s->assemble_freq_bands = assemble_freq_bands_c;
if (ARCH_X86)
ff_dcadsp_init_x86(s);
}