1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-26 19:01:44 +02:00

avfilter/af_acrossover: move coefficients and state to simple arrays

This commit is contained in:
Paul B Mahol 2020-12-01 11:34:50 +01:00
parent aa3566a8ee
commit ba6e2a2d05

View File

@ -39,21 +39,17 @@
#define MAX_SPLITS 16
#define MAX_BANDS MAX_SPLITS + 1
#define B0 0
#define B1 1
#define B2 2
#define A1 3
#define A2 4
typedef struct BiquadCoeffs {
double b0, b1, b2;
double a1, a2;
double cd[5];
float cf[5];
} BiquadCoeffs;
typedef struct BiquadContext {
double z1, z2;
} BiquadContext;
typedef struct CrossoverChannel {
BiquadContext lp[MAX_BANDS][20];
BiquadContext hp[MAX_BANDS][20];
BiquadContext ap[MAX_BANDS][MAX_BANDS][20];
} CrossoverChannel;
typedef struct AudioCrossoverContext {
const AVClass *class;
@ -72,7 +68,7 @@ typedef struct AudioCrossoverContext {
BiquadCoeffs hp[MAX_BANDS][20];
BiquadCoeffs ap[MAX_BANDS][20];
CrossoverChannel *xover;
AVFrame *xover;
AVFrame *input_frame;
AVFrame *frames[MAX_BANDS];
@ -178,11 +174,17 @@ static void set_lp(BiquadCoeffs *b, double fc, double q, double sr)
double a1 = -2. * cosine;
double a2 = 1. - alpha;
b->b0 = b0 / a0;
b->b1 = b1 / a0;
b->b2 = b2 / a0;
b->a1 = -a1 / a0;
b->a2 = -a2 / a0;
b->cd[B0] = b0 / a0;
b->cd[B1] = b1 / a0;
b->cd[B2] = b2 / a0;
b->cd[A1] = -a1 / a0;
b->cd[A2] = -a2 / a0;
b->cf[B0] = b->cd[B0];
b->cf[B1] = b->cd[B1];
b->cf[B2] = b->cd[B2];
b->cf[A1] = b->cd[A1];
b->cf[A2] = b->cd[A2];
}
static void set_hp(BiquadCoeffs *b, double fc, double q, double sr)
@ -198,11 +200,17 @@ static void set_hp(BiquadCoeffs *b, double fc, double q, double sr)
double a1 = -2. * cosine;
double a2 = 1. - alpha;
b->b0 = b0 / a0;
b->b1 = b1 / a0;
b->b2 = b2 / a0;
b->a1 = -a1 / a0;
b->a2 = -a2 / a0;
b->cd[B0] = b0 / a0;
b->cd[B1] = b1 / a0;
b->cd[B2] = b2 / a0;
b->cd[A1] = -a1 / a0;
b->cd[A2] = -a2 / a0;
b->cf[B0] = b->cd[B0];
b->cf[B1] = b->cd[B1];
b->cf[B2] = b->cd[B2];
b->cf[A1] = b->cd[A1];
b->cf[A2] = b->cd[A2];
}
static void set_ap(BiquadCoeffs *b, double fc, double q, double sr)
@ -218,22 +226,34 @@ static void set_ap(BiquadCoeffs *b, double fc, double q, double sr)
double b1 = a1;
double b2 = a0;
b->b0 = b0 / a0;
b->b1 = b1 / a0;
b->b2 = b2 / a0;
b->a1 = -a1 / a0;
b->a2 = -a2 / a0;
b->cd[B0] = b0 / a0;
b->cd[B1] = b1 / a0;
b->cd[B2] = b2 / a0;
b->cd[A1] = -a1 / a0;
b->cd[A2] = -a2 / a0;
b->cf[B0] = b->cd[B0];
b->cf[B1] = b->cd[B1];
b->cf[B2] = b->cd[B2];
b->cf[A1] = b->cd[A1];
b->cf[A2] = b->cd[A2];
}
static void set_ap1(BiquadCoeffs *b, double fc, double sr)
{
double omega = 2. * M_PI * fc / sr;
b->a1 = exp(-omega);
b->a2 = 0.;
b->b0 = -b->a1;
b->b1 = 1.;
b->b2 = 0.;
b->cd[A1] = exp(-omega);
b->cd[A2] = 0.;
b->cd[B0] = -b->cd[A1];
b->cd[B1] = 1.;
b->cd[B2] = 0.;
b->cf[B0] = b->cd[B0];
b->cf[B1] = b->cd[B1];
b->cf[B2] = b->cd[B2];
b->cf[A1] = b->cd[A1];
b->cf[A2] = b->cd[A2];
}
static void calc_q_factors(int order, double *q)
@ -275,18 +295,18 @@ static int query_formats(AVFilterContext *ctx)
}
#define BIQUAD_PROCESS(name, type) \
static void biquad_process_## name(const BiquadCoeffs *const c,\
BiquadContext *b, \
static void biquad_process_## name(const type *const c, \
type *b, \
type *dst, const type *src, \
int nb_samples) \
{ \
const type b0 = c->b0; \
const type b1 = c->b1; \
const type b2 = c->b2; \
const type a1 = c->a1; \
const type a2 = c->a2; \
type z1 = b->z1; \
type z2 = b->z2; \
const type b0 = c[B0]; \
const type b1 = c[B1]; \
const type b2 = c[B2]; \
const type a1 = c[A1]; \
const type a2 = c[A2]; \
type z1 = b[0]; \
type z2 = b[1]; \
\
for (int n = 0; n + 1 < nb_samples; n++) { \
type in = src[n]; \
@ -316,8 +336,8 @@ static void biquad_process_## name(const BiquadCoeffs *const c,\
dst[n] = out; \
} \
\
b->z1 = z1; \
b->z2 = z2; \
b[0] = z1; \
b[1] = z2; \
}
BIQUAD_PROCESS(fltp, float)
@ -332,40 +352,41 @@ static int filter_channels_## name(AVFilterContext *ctx, void *arg, int jobnr, i
const int start = (in->channels * jobnr) / nb_jobs; \
const int end = (in->channels * (jobnr+1)) / nb_jobs; \
const int nb_samples = in->nb_samples; \
const int nb_outs = ctx->nb_outputs; \
\
for (int ch = start; ch < end; ch++) { \
const type *src = (const type *)in->extended_data[ch]; \
CrossoverChannel *xover = &s->xover[ch]; \
type *xover = (type *)s->xover->extended_data[ch]; \
\
s->fdsp->vector_## ff ##mul_scalar((type *)frames[0]->extended_data[ch], src, \
s->level_in, FFALIGN(nb_samples, sizeof(type))); \
\
for (int band = 0; band < ctx->nb_outputs; band++) { \
for (int f = 0; band + 1 < ctx->nb_outputs && f < s->filter_count; f++) { \
for (int band = 0; band < nb_outs; band++) { \
for (int f = 0; band + 1 < nb_outs && f < s->filter_count; f++) { \
const type *prv = (const type *)frames[band]->extended_data[ch]; \
type *dst = (type *)frames[band + 1]->extended_data[ch]; \
const type *hsrc = f == 0 ? prv : dst; \
BiquadContext *hp = &xover->hp[band][f]; \
BiquadCoeffs *hpc = &s->hp[band][f]; \
type *hp = xover + nb_outs * 20 + band * 20 + f * 2; \
const type *const hpc = (type *)&s->hp[band][f].c ## ff; \
\
biquad_process_## name(hpc, hp, dst, hsrc, nb_samples); \
} \
\
for (int f = 0; band + 1 < ctx->nb_outputs && f < s->filter_count; f++) { \
for (int f = 0; band + 1 < nb_outs && f < s->filter_count; f++) { \
type *dst = (type *)frames[band]->extended_data[ch]; \
const type *lsrc = dst; \
BiquadContext *lp = &xover->lp[band][f]; \
BiquadCoeffs *lpc = &s->lp[band][f]; \
type *lp = xover + band * 20 + f * 2; \
const type *const lpc = (type *)&s->lp[band][f].c ## ff; \
\
biquad_process_## name(lpc, lp, dst, lsrc, nb_samples); \
} \
\
for (int aband = band + 1; aband + 1 < ctx->nb_outputs; aband++) { \
for (int aband = band + 1; aband + 1 < nb_outs; aband++) { \
if (s->first_order) { \
const type *asrc = (const type *)frames[band]->extended_data[ch]; \
type *dst = (type *)frames[band]->extended_data[ch]; \
BiquadContext *ap = &xover->ap[band][aband][0]; \
BiquadCoeffs *apc = &s->ap[aband][0]; \
type *ap = xover + nb_outs * 40 + (aband * nb_outs + band) * 20; \
const type *const apc = (type *)&s->ap[aband][0].c ## ff; \
\
biquad_process_## name(apc, ap, dst, asrc, nb_samples); \
} \
@ -373,15 +394,15 @@ static int filter_channels_## name(AVFilterContext *ctx, void *arg, int jobnr, i
for (int f = s->first_order; f < s->ap_filter_count; f++) { \
const type *asrc = (const type *)frames[band]->extended_data[ch]; \
type *dst = (type *)frames[band]->extended_data[ch]; \
BiquadContext *ap = &xover->ap[band][aband][f]; \
BiquadCoeffs *apc = &s->ap[aband][f]; \
type *ap = xover + nb_outs * 40 + (aband * nb_outs + band) * 20 + f * 2;\
const type *const apc = (type *)&s->ap[aband][f].c ## ff; \
\
biquad_process_## name(apc, ap, dst, asrc, nb_samples); \
} \
} \
} \
\
for (int band = 0; band < ctx->nb_outputs && s->first_order; band++) { \
for (int band = 0; band < nb_outs && s->first_order; band++) { \
if (band & 1) { \
type *dst = (type *)frames[band]->extended_data[ch]; \
s->fdsp->vector_## ff ##mul_scalar(dst, dst, -one, \
@ -403,10 +424,6 @@ static int config_input(AVFilterLink *inlink)
int sample_rate = inlink->sample_rate;
double q[16];
s->xover = av_calloc(inlink->channels, sizeof(*s->xover));
if (!s->xover)
return AVERROR(ENOMEM);
s->order = (s->order_opt + 1) * 2;
s->filter_count = s->order / 2;
s->first_order = s->filter_count & 1;
@ -441,6 +458,11 @@ static int config_input(AVFilterLink *inlink)
case AV_SAMPLE_FMT_DBLP: s->filter_channels = filter_channels_dblp; break;
}
s->xover = ff_get_audio_buffer(inlink, 2 * (ctx->nb_outputs * 10 + ctx->nb_outputs * 10 +
ctx->nb_outputs * ctx->nb_outputs * 10));
if (!s->xover)
return AVERROR(ENOMEM);
return 0;
}
@ -492,7 +514,7 @@ static av_cold void uninit(AVFilterContext *ctx)
av_freep(&s->fdsp);
av_freep(&s->splits);
av_freep(&s->xover);
av_frame_free(&s->xover);
for (i = 0; i < ctx->nb_outputs; i++)
av_freep(&ctx->output_pads[i].name);