Commit ac05f903 authored by Justin Ruggles's avatar Justin Ruggles

Convert floating-point MDCT coefficients to 24-bit fixed-point all at once

instead of doing it separately in 2 different functions.
This makes float AC-3 encoding approx. 3-7% faster overall.
Also, the coefficient conversion can now be easily SIMD-optimized.

Originally committed as revision 26232 to svn://svn.ffmpeg.org/ffmpeg/trunk
parent 87c1410d
......@@ -62,6 +62,7 @@
typedef struct AC3Block {
uint8_t **bap; ///< bit allocation pointers (bap)
CoefType **mdct_coef; ///< MDCT coefficients
int32_t **fixed_coef; ///< fixed-point MDCT coefficients
uint8_t **exp; ///< original exponents
uint8_t **grouped_exp; ///< grouped exponents
int16_t **psd; ///< psd per frequency bin
......@@ -128,6 +129,7 @@ typedef struct AC3EncodeContext {
uint8_t *bap_buffer;
uint8_t *bap1_buffer;
CoefType *mdct_coef_buffer;
int32_t *fixed_coef_buffer;
uint8_t *exp_buffer;
uint8_t *grouped_exp_buffer;
int16_t *psd_buffer;
......@@ -153,6 +155,8 @@ static void apply_window(SampleType *output, const SampleType *input,
static int normalize_samples(AC3EncodeContext *s);
static void scale_coefficients(AC3EncodeContext *s);
/**
* LUT for number of exponent groups.
......@@ -286,11 +290,11 @@ static void extract_exponents(AC3EncodeContext *s)
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) {
AC3Block *block = &s->blocks[blk];
uint8_t *exp = block->exp[ch];
CoefType *coef = block->mdct_coef[ch];
int32_t *coef = block->fixed_coef[ch];
int exp_shift = block->exp_shift[ch];
for (i = 0; i < AC3_MAX_COEFS; i++) {
int e;
int v = abs(SCALE_COEF(coef[i]));
int v = abs(coef[i]);
if (v == 0)
e = 24;
else {
......@@ -1017,7 +1021,7 @@ static inline int asym_quant(int c, int e, int qbits)
/**
* Quantize a set of mantissas for a single channel in a single block.
*/
static void quantize_mantissas_blk_ch(AC3EncodeContext *s, CoefType *mdct_coef,
static void quantize_mantissas_blk_ch(AC3EncodeContext *s, int32_t *fixed_coef,
int8_t exp_shift, uint8_t *exp,
uint8_t *bap, uint16_t *qmant, int n)
{
......@@ -1025,7 +1029,7 @@ static void quantize_mantissas_blk_ch(AC3EncodeContext *s, CoefType *mdct_coef,
for (i = 0; i < n; i++) {
int v;
int c = SCALE_COEF(mdct_coef[i]);
int c = fixed_coef[i];
int e = exp[i] - exp_shift;
int b = bap[i];
switch (b) {
......@@ -1122,7 +1126,7 @@ static void quantize_mantissas(AC3EncodeContext *s)
s->qmant1_ptr = s->qmant2_ptr = s->qmant4_ptr = NULL;
for (ch = 0; ch < s->channels; ch++) {
quantize_mantissas_blk_ch(s, block->mdct_coef[ch], block->exp_shift[ch],
quantize_mantissas_blk_ch(s, block->fixed_coef[ch], block->exp_shift[ch],
block->exp[ch], block->bap[ch],
block->qmant[ch], s->nb_coefs[ch]);
}
......@@ -1390,6 +1394,8 @@ static int ac3_encode_frame(AVCodecContext *avctx, unsigned char *frame,
apply_mdct(s);
scale_coefficients(s);
process_exponents(s);
ret = compute_bit_allocation(s);
......@@ -1420,6 +1426,7 @@ static av_cold int ac3_encode_close(AVCodecContext *avctx)
av_freep(&s->bap_buffer);
av_freep(&s->bap1_buffer);
av_freep(&s->mdct_coef_buffer);
av_freep(&s->fixed_coef_buffer);
av_freep(&s->exp_buffer);
av_freep(&s->grouped_exp_buffer);
av_freep(&s->psd_buffer);
......@@ -1430,6 +1437,7 @@ static av_cold int ac3_encode_close(AVCodecContext *avctx)
AC3Block *block = &s->blocks[blk];
av_freep(&block->bap);
av_freep(&block->mdct_coef);
av_freep(&block->fixed_coef);
av_freep(&block->exp);
av_freep(&block->grouped_exp);
av_freep(&block->psd);
......@@ -1639,6 +1647,26 @@ static av_cold int allocate_buffers(AVCodecContext *avctx)
}
}
if (CONFIG_AC3ENC_FLOAT) {
FF_ALLOC_OR_GOTO(avctx, s->fixed_coef_buffer, AC3_MAX_BLOCKS * s->channels *
AC3_MAX_COEFS * sizeof(*s->fixed_coef_buffer), alloc_fail);
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) {
AC3Block *block = &s->blocks[blk];
FF_ALLOCZ_OR_GOTO(avctx, block->fixed_coef, s->channels *
sizeof(*block->fixed_coef), alloc_fail);
for (ch = 0; ch < s->channels; ch++)
block->fixed_coef[ch] = &s->fixed_coef_buffer[AC3_MAX_COEFS * (blk * s->channels + ch)];
}
} else {
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) {
AC3Block *block = &s->blocks[blk];
FF_ALLOCZ_OR_GOTO(avctx, block->fixed_coef, s->channels *
sizeof(*block->fixed_coef), alloc_fail);
for (ch = 0; ch < s->channels; ch++)
block->fixed_coef[ch] = (int32_t *)block->mdct_coef[ch];
}
}
return 0;
alloc_fail:
return AVERROR(ENOMEM);
......
......@@ -319,6 +319,17 @@ static int normalize_samples(AC3EncodeContext *s)
}
/**
* Scale MDCT coefficients from float to fixed-point.
*/
static void scale_coefficients(AC3EncodeContext *s)
{
/* scaling/conversion is obviously not needed for the fixed-point encoder
since the coefficients are already fixed-point. */
return;
}
#ifdef TEST
/*************************************************************************/
/* TEST */
......
......@@ -35,8 +35,6 @@
typedef int16_t SampleType;
typedef int32_t CoefType;
#define SCALE_COEF(a) (a)
/**
* Compex number.
......
......@@ -102,6 +102,17 @@ static int normalize_samples(AC3EncodeContext *s)
}
/**
* Scale MDCT coefficients from float to 24-bit fixed-point.
*/
static void scale_coefficients(AC3EncodeContext *s)
{
int i;
for (i = 0; i < AC3_MAX_COEFS * AC3_MAX_BLOCKS * s->channels; i++)
s->fixed_coef_buffer[i] = SCALE_FLOAT(s->mdct_coef_buffer[i], 24);
}
AVCodec ac3_encoder = {
"ac3",
AVMEDIA_TYPE_AUDIO,
......
......@@ -35,8 +35,6 @@
typedef float SampleType;
typedef float CoefType;
#define SCALE_COEF(a) SCALE_FLOAT((a), 24)
typedef struct AC3MDCTContext {
const float *window; ///< MDCT window function
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment