From 228ca70d485e2660c2e381d7112cbcca65c156a0 Mon Sep 17 00:00:00 2001 From: diego Date: Wed, 2 Jun 2004 22:59:04 +0000 Subject: update to the 2.0 release of faad, patch by adland git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@12528 b3059339-0415-0410-9bf9-f77b7e298cf2 --- libfaad2/output.c | 515 +++++++++++++++++++++++++++++++++++++++--------------- 1 file changed, 375 insertions(+), 140 deletions(-) (limited to 'libfaad2/output.c') diff --git a/libfaad2/output.c b/libfaad2/output.c index 9baf62bd6c..71dac845b4 100644 --- a/libfaad2/output.c +++ b/libfaad2/output.c @@ -33,189 +33,366 @@ #ifndef FIXED_POINT -#include "dither.h" - - -#define ftol(A,B) {tmp = *(int32_t*) & A - 0x4B7F8000; \ - B = (int16_t)((tmp==(int16_t)tmp) ? tmp : (tmp>>31)^0x7FFF);} - -#define ROUND(x) ((x >= 0) ? (int32_t)floor((x) + 0.5) : (int32_t)ceil((x) + 0.5)) - -#define ROUND32(x) ROUND(x) - -#define ROUND64(x) (doubletmp = (x) + Dither.Add + (int64_t)0x001FFFFD80000000L, *(int64_t*)(&doubletmp) - (int64_t)0x433FFFFD80000000L) #define FLOAT_SCALE (1.0f/(1<<15)) -dither_t Dither; -double doubletmp; +#define DM_MUL REAL_CONST(0.3203772410170407) // 1/(1+sqrt(2) + 1/sqrt(2)) +#define RSQRT2 REAL_CONST(0.7071067811865475244) // 1/sqrt(2) -#define DM_MUL ((real_t)1.0/((real_t)1.0+(real_t)sqrt(2.0))) static INLINE real_t get_sample(real_t **input, uint8_t channel, uint16_t sample, - uint8_t downMatrix, uint8_t *internal_channel) + uint8_t down_matrix, uint8_t *internal_channel) { - if (downMatrix) + if (!down_matrix) + return input[internal_channel[channel]][sample]; + + if (channel == 0) { - if (channel == 0) - { - return DM_MUL * (input[internal_channel[1]][sample] + - input[internal_channel[0]][sample]/(real_t)sqrt(2.) + - input[internal_channel[3]][sample]/(real_t)sqrt(2.)); - } else { - return DM_MUL * (input[internal_channel[2]][sample] + - input[internal_channel[0]][sample]/(real_t)sqrt(2.) + - input[internal_channel[4]][sample]/(real_t)sqrt(2.)); - } + return DM_MUL * (input[internal_channel[1]][sample] + + input[internal_channel[0]][sample] * RSQRT2 + + input[internal_channel[3]][sample] * RSQRT2); } else { - return input[internal_channel[channel]][sample]; + return DM_MUL * (input[internal_channel[2]][sample] + + input[internal_channel[0]][sample] * RSQRT2 + + input[internal_channel[4]][sample] * RSQRT2); } } -void* output_to_PCM(faacDecHandle hDecoder, - real_t **input, void *sample_buffer, uint8_t channels, - uint16_t frame_len, uint8_t format) -{ - uint8_t ch; - uint16_t i, j = 0; - uint8_t internal_channel; +#ifndef HAS_LRINTF +#define CLIP(sample, max, min) \ +if (sample >= 0.0f) \ +{ \ + sample += 0.5f; \ + if (sample >= max) \ + sample = max; \ +} else { \ + sample += -0.5f; \ + if (sample <= min) \ + sample = min; \ +} +#else +#define CLIP(sample, max, min) \ +if (sample >= 0.0f) \ +{ \ + if (sample >= max) \ + sample = max; \ +} else { \ + if (sample <= min) \ + sample = min; \ +} +#endif - int16_t *short_sample_buffer = (int16_t*)sample_buffer; - int32_t *int_sample_buffer = (int32_t*)sample_buffer; - float32_t *float_sample_buffer = (float32_t*)sample_buffer; - double *double_sample_buffer = (double*)sample_buffer; +#define CONV(a,b) ((a<<1)|(b&0x1)) - /* Copy output to a standard PCM buffer */ - for (ch = 0; ch < channels; ch++) +static void to_PCM_16bit(faacDecHandle hDecoder, real_t **input, + uint8_t channels, uint16_t frame_len, + int16_t **sample_buffer) +{ + uint8_t ch, ch1; + uint16_t i; + + switch (CONV(channels,hDecoder->downMatrix)) { - internal_channel = hDecoder->internal_channel[ch]; + case CONV(1,0): + case CONV(1,1): + for(i = 0; i < frame_len; i++) + { + real_t inp = input[hDecoder->internal_channel[0]][i]; - switch (format) + CLIP(inp, 32767.0f, -32768.0f); + + (*sample_buffer)[i] = (int16_t)lrintf(inp); + } + break; + case CONV(2,0): + ch = hDecoder->internal_channel[0]; + ch1 = hDecoder->internal_channel[1]; + for(i = 0; i < frame_len; i++) + { + real_t inp0 = input[ch ][i]; + real_t inp1 = input[ch1][i]; + + CLIP(inp0, 32767.0f, -32768.0f); + CLIP(inp1, 32767.0f, -32768.0f); + + (*sample_buffer)[(i*2)+0] = (int16_t)lrintf(inp0); + (*sample_buffer)[(i*2)+1] = (int16_t)lrintf(inp1); + } + break; + default: + for (ch = 0; ch < channels; ch++) { - case FAAD_FMT_16BIT: for(i = 0; i < frame_len; i++) { - int32_t tmp; - real_t ftemp; - //real_t inp = input[internal_channel][i]; real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); - ftemp = inp + 0xff8000; - ftol(ftemp, short_sample_buffer[(i*channels)+ch]); - } - break; - case FAAD_FMT_16BIT_DITHER: - for(i = 0; i < frame_len; i++, j++) - { - //real_t inp = input[internal_channel][i]; - real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); - double Sum = inp * 65535.f; - int64_t val; - if(j > 31) - j = 0; - val = dither_output(1, 0, j, Sum, ch) / 65536; - if (val > (1<<15)-1) - val = (1<<15)-1; - else if (val < -(1<<15)) - val = -(1<<15); - short_sample_buffer[(i*channels)+ch] = (int16_t)val; - } - break; - case FAAD_FMT_16BIT_L_SHAPE: - case FAAD_FMT_16BIT_M_SHAPE: - case FAAD_FMT_16BIT_H_SHAPE: - for(i = 0; i < frame_len; i++, j++) - { - //real_t inp = input[internal_channel][i]; - real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); - double Sum = inp * 65535.f; - int64_t val; - if(j > 31) - j = 0; - val = dither_output(1, 1, j, Sum, ch) / 65536; - if (val > (1<<15)-1) - val = (1<<15)-1; - else if (val < -(1<<15)) - val = -(1<<15); - short_sample_buffer[(i*channels)+ch] = (int16_t)val; + CLIP(inp, 32767.0f, -32768.0f); + + (*sample_buffer)[(i*channels)+ch] = (int16_t)lrintf(inp); } - break; - case FAAD_FMT_24BIT: + } + break; + } +} + +static void to_PCM_24bit(faacDecHandle hDecoder, real_t **input, + uint8_t channels, uint16_t frame_len, + int32_t **sample_buffer) +{ + uint8_t ch, ch1; + uint16_t i; + + switch (CONV(channels,hDecoder->downMatrix)) + { + case CONV(1,0): + case CONV(1,1): + for(i = 0; i < frame_len; i++) + { + real_t inp = input[hDecoder->internal_channel[0]][i]; + + inp *= 256.0f; + CLIP(inp, 8388607.0f, -8388608.0f); + + (*sample_buffer)[i] = (int32_t)lrintf(inp); + } + break; + case CONV(2,0): + ch = hDecoder->internal_channel[0]; + ch1 = hDecoder->internal_channel[1]; + for(i = 0; i < frame_len; i++) + { + real_t inp0 = input[ch ][i]; + real_t inp1 = input[ch1][i]; + + inp0 *= 256.0f; + inp1 *= 256.0f; + CLIP(inp0, 8388607.0f, -8388608.0f); + CLIP(inp1, 8388607.0f, -8388608.0f); + + (*sample_buffer)[(i*2)+0] = (int32_t)lrintf(inp0); + (*sample_buffer)[(i*2)+1] = (int32_t)lrintf(inp1); + } + break; + default: + for (ch = 0; ch < channels; ch++) + { for(i = 0; i < frame_len; i++) { - //real_t inp = input[internal_channel][i]; real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); - if (inp > (1<<15)-1) - inp = (1<<15)-1; - else if (inp < -(1<<15)) - inp = -(1<<15); - int_sample_buffer[(i*channels)+ch] = ROUND(inp*(1<<8)); + + inp *= 256.0f; + CLIP(inp, 8388607.0f, -8388608.0f); + + (*sample_buffer)[(i*channels)+ch] = (int32_t)lrintf(inp); } - break; - case FAAD_FMT_32BIT: + } + break; + } +} + +static void to_PCM_32bit(faacDecHandle hDecoder, real_t **input, + uint8_t channels, uint16_t frame_len, + int32_t **sample_buffer) +{ + uint8_t ch, ch1; + uint16_t i; + + switch (CONV(channels,hDecoder->downMatrix)) + { + case CONV(1,0): + case CONV(1,1): + for(i = 0; i < frame_len; i++) + { + real_t inp = input[hDecoder->internal_channel[0]][i]; + + inp *= 65536.0f; + CLIP(inp, 2147483647.0f, -2147483648.0f); + + (*sample_buffer)[i] = (int32_t)lrintf(inp); + } + break; + case CONV(2,0): + ch = hDecoder->internal_channel[0]; + ch1 = hDecoder->internal_channel[1]; + for(i = 0; i < frame_len; i++) + { + real_t inp0 = input[ch ][i]; + real_t inp1 = input[ch1][i]; + + inp0 *= 65536.0f; + inp1 *= 65536.0f; + CLIP(inp0, 2147483647.0f, -2147483648.0f); + CLIP(inp1, 2147483647.0f, -2147483648.0f); + + (*sample_buffer)[(i*2)+0] = (int32_t)lrintf(inp0); + (*sample_buffer)[(i*2)+1] = (int32_t)lrintf(inp1); + } + break; + default: + for (ch = 0; ch < channels; ch++) + { for(i = 0; i < frame_len; i++) { - //real_t inp = input[internal_channel][i]; real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); - if (inp > (1<<15)-1) - inp = (1<<15)-1; - else if (inp < -(1<<15)) - inp = -(1<<15); - int_sample_buffer[(i*channels)+ch] = ROUND32(inp*(1<<16)); + + inp *= 65536.0f; + CLIP(inp, 2147483647.0f, -2147483648.0f); + + (*sample_buffer)[(i*channels)+ch] = (int32_t)lrintf(inp); } - break; - case FAAD_FMT_FLOAT: + } + break; + } +} + +static void to_PCM_float(faacDecHandle hDecoder, real_t **input, + uint8_t channels, uint16_t frame_len, + float32_t **sample_buffer) +{ + uint8_t ch, ch1; + uint16_t i; + + switch (CONV(channels,hDecoder->downMatrix)) + { + case CONV(1,0): + case CONV(1,1): + for(i = 0; i < frame_len; i++) + { + real_t inp = input[hDecoder->internal_channel[0]][i]; + (*sample_buffer)[i] = inp*FLOAT_SCALE; + } + break; + case CONV(2,0): + ch = hDecoder->internal_channel[0]; + ch1 = hDecoder->internal_channel[1]; + for(i = 0; i < frame_len; i++) + { + real_t inp0 = input[ch ][i]; + real_t inp1 = input[ch1][i]; + (*sample_buffer)[(i*2)+0] = inp0*FLOAT_SCALE; + (*sample_buffer)[(i*2)+1] = inp1*FLOAT_SCALE; + } + break; + default: + for (ch = 0; ch < channels; ch++) + { for(i = 0; i < frame_len; i++) { - //real_t inp = input[internal_channel][i]; real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); - float_sample_buffer[(i*channels)+ch] = inp*FLOAT_SCALE; + (*sample_buffer)[(i*channels)+ch] = inp*FLOAT_SCALE; } - break; - case FAAD_FMT_DOUBLE: + } + break; + } +} + +static void to_PCM_double(faacDecHandle hDecoder, real_t **input, + uint8_t channels, uint16_t frame_len, + double **sample_buffer) +{ + uint8_t ch, ch1; + uint16_t i; + + switch (CONV(channels,hDecoder->downMatrix)) + { + case CONV(1,0): + case CONV(1,1): + for(i = 0; i < frame_len; i++) + { + real_t inp = input[hDecoder->internal_channel[0]][i]; + (*sample_buffer)[i] = (double)inp*FLOAT_SCALE; + } + break; + case CONV(2,0): + ch = hDecoder->internal_channel[0]; + ch1 = hDecoder->internal_channel[1]; + for(i = 0; i < frame_len; i++) + { + real_t inp0 = input[ch ][i]; + real_t inp1 = input[ch1][i]; + (*sample_buffer)[(i*2)+0] = (double)inp0*FLOAT_SCALE; + (*sample_buffer)[(i*2)+1] = (double)inp1*FLOAT_SCALE; + } + break; + default: + for (ch = 0; ch < channels; ch++) + { for(i = 0; i < frame_len; i++) { - //real_t inp = input[internal_channel][i]; real_t inp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); - double_sample_buffer[(i*channels)+ch] = (double)inp*FLOAT_SCALE; + (*sample_buffer)[(i*channels)+ch] = (double)inp*FLOAT_SCALE; } - break; } + break; } +} + +void *output_to_PCM(faacDecHandle hDecoder, + real_t **input, void *sample_buffer, uint8_t channels, + uint16_t frame_len, uint8_t format) +{ + int16_t *short_sample_buffer = (int16_t*)sample_buffer; + int32_t *int_sample_buffer = (int32_t*)sample_buffer; + float32_t *float_sample_buffer = (float32_t*)sample_buffer; + double *double_sample_buffer = (double*)sample_buffer; + +#ifdef PROFILE + int64_t count = faad_get_ts(); +#endif + + /* Copy output to a standard PCM buffer */ + switch (format) + { + case FAAD_FMT_16BIT: + to_PCM_16bit(hDecoder, input, channels, frame_len, &short_sample_buffer); + break; + case FAAD_FMT_24BIT: + to_PCM_24bit(hDecoder, input, channels, frame_len, &int_sample_buffer); + break; + case FAAD_FMT_32BIT: + to_PCM_32bit(hDecoder, input, channels, frame_len, &int_sample_buffer); + break; + case FAAD_FMT_FLOAT: + to_PCM_float(hDecoder, input, channels, frame_len, &float_sample_buffer); + break; + case FAAD_FMT_DOUBLE: + to_PCM_double(hDecoder, input, channels, frame_len, &double_sample_buffer); + break; + } + +#ifdef PROFILE + count = faad_get_ts() - count; + hDecoder->output_cycles += count; +#endif return sample_buffer; } +#else + +#define DM_MUL FRAC_CONST(0.3203772410170407) // 1/(1+sqrt(2) + 1/sqrt(2)) +#define RSQRT2 FRAC_CONST(0.7071067811865475244) // 1/sqrt(2) -/* Dither output */ -static int64_t dither_output(uint8_t dithering, uint8_t shapingtype, uint16_t i, double Sum, uint8_t k) +static INLINE real_t get_sample(real_t **input, uint8_t channel, uint16_t sample, + uint8_t down_matrix, uint8_t *internal_channel) { - double Sum2; - int64_t val; - if(dithering) + if (!down_matrix) + return input[internal_channel[channel]][sample]; + + if (channel == 0) { - if(!shapingtype) - { - double tmp = Random_Equi(Dither.Dither); - Sum2 = tmp - (double)Dither.LastRandomNumber[k]; - Dither.LastRandomNumber[k] = (int32_t)tmp; - Sum2 = Sum += Sum2; - val = ROUND64(Sum2)&Dither.Mask; - } else { - Sum2 = Random_Triangular(Dither.Dither) - scalar16(Dither.DitherHistory[k], Dither.FilterCoeff + i); - Sum += Dither.DitherHistory[k][(-1-i)&15] = (float32_t)Sum2; - Sum2 = Sum + scalar16(Dither.ErrorHistory[k], Dither.FilterCoeff + i ); - val = ROUND64(Sum2)&Dither.Mask; - Dither.ErrorHistory[k][(-1-i)&15] = (float)(Sum - val); - } - return val; + real_t C = MUL_F(input[internal_channel[0]][sample], RSQRT2); + real_t L_S = MUL_F(input[internal_channel[3]][sample], RSQRT2); + real_t cum = input[internal_channel[1]][sample] + C + L_S; + return MUL_F(cum, DM_MUL); + } else { + real_t C = MUL_F(input[internal_channel[0]][sample], RSQRT2); + real_t R_S = MUL_F(input[internal_channel[4]][sample], RSQRT2); + real_t cum = input[internal_channel[2]][sample] + C + R_S; + return MUL_F(cum, DM_MUL); } - else - return ROUND64 (Sum); } -#else - void* output_to_PCM(faacDecHandle hDecoder, real_t **input, void *sample_buffer, uint8_t channels, uint16_t frame_len, uint8_t format) @@ -223,18 +400,76 @@ void* output_to_PCM(faacDecHandle hDecoder, uint8_t ch; uint16_t i; int16_t *short_sample_buffer = (int16_t*)sample_buffer; + int32_t *int_sample_buffer = (int32_t*)sample_buffer; /* Copy output to a standard PCM buffer */ for (ch = 0; ch < channels; ch++) { - for(i = 0; i < frame_len; i++) + switch (format) { - int32_t tmp = input[ch][i]; - tmp += (1 << (REAL_BITS-1)); - tmp >>= REAL_BITS; - if (tmp > 0x7fff) tmp = 0x7fff; - else if (tmp <= -32768) tmp = -32768; - short_sample_buffer[(i*channels)+ch] = (int16_t)tmp; + case FAAD_FMT_16BIT: + for(i = 0; i < frame_len; i++) + { + //int32_t tmp = input[ch][i]; + int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); + if (tmp >= 0) + { + tmp += (1 << (REAL_BITS-1)); + if (tmp >= REAL_CONST(32767)) + { + tmp = REAL_CONST(32767); + } + } else { + tmp += -(1 << (REAL_BITS-1)); + if (tmp <= REAL_CONST(-32768)) + { + tmp = REAL_CONST(-32768); + } + } + tmp >>= REAL_BITS; + short_sample_buffer[(i*channels)+ch] = (int16_t)tmp; + } + break; + case FAAD_FMT_24BIT: + for(i = 0; i < frame_len; i++) + { + //int32_t tmp = input[ch][i]; + int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); + if (tmp >= 0) + { + tmp += (1 << (REAL_BITS-9)); + tmp >>= (REAL_BITS-8); + if (tmp >= 8388607) + { + tmp = 8388607; + } + } else { + tmp += -(1 << (REAL_BITS-9)); + tmp >>= (REAL_BITS-8); + if (tmp <= -8388608) + { + tmp = -8388608; + } + } + int_sample_buffer[(i*channels)+ch] = (int32_t)tmp; + } + break; + case FAAD_FMT_32BIT: + for(i = 0; i < frame_len; i++) + { + //int32_t tmp = input[ch][i]; + int32_t tmp = get_sample(input, ch, i, hDecoder->downMatrix, hDecoder->internal_channel); + if (tmp >= 0) + { + tmp += (1 << (16-REAL_BITS-1)); + tmp <<= (16-REAL_BITS); + } else { + tmp += -(1 << (16-REAL_BITS-1)); + tmp <<= (16-REAL_BITS); + } + int_sample_buffer[(i*channels)+ch] = (int32_t)tmp; + } + break; } } -- cgit v1.2.3