in stream encoder, only allocate and calculate real signal if max lpc order > 0
This commit is contained in:
parent
2de1124e36
commit
c549f0f17e
@ -92,6 +92,7 @@
|
||||
<LI>
|
||||
General:
|
||||
<UL>
|
||||
<LI>Sped up encoding when not using LPC (i.e. when using <TT>flac</TT> options <TT>-0</TT>, <TT>-1</TT>, <TT>-2</TT>, or <TT>-l 0</TT>).</LI>
|
||||
</UL>
|
||||
</LI>
|
||||
<LI>
|
||||
@ -129,7 +130,8 @@
|
||||
<LI>
|
||||
libraries:
|
||||
<UL>
|
||||
<LI>libFLAC, libOggFLAC: Can now be compiled to use only integer instructions, including encoding. The decoder is almost completely integer anyway but there were a couple places that needed a fixed-point replacement. There is no fixed-point version of LPC analysis yet, so if libFLAC is compiled integer-only, it will behave as if the max LPC order is 0 (i.e. used fixed predictors only).</LI>
|
||||
<LI>libFLAC: Sped up encoding when not using LPC (i.e. <TT>max_lpc_order == 0</TT>).</LI>
|
||||
<LI>libFLAC, libOggFLAC: Can now be compiled to use only integer instructions, including encoding. The decoder is almost completely integer anyway but there were a couple places that needed a fixed-point replacement. There is no fixed-point version of LPC analysis yet, so if libFLAC is compiled integer-only, the encoder will behave as if the max LPC order is 0 (i.e. used fixed predictors only). LPC decoding is supported in all cases as it always was integer-only.</LI>
|
||||
</UL>
|
||||
</LI>
|
||||
<LI>
|
||||
|
@ -1374,59 +1374,121 @@ FLAC_API FLAC__bool FLAC__stream_encoder_process(FLAC__StreamEncoder *encoder, c
|
||||
FLAC__ASSERT(encoder->protected_->state == FLAC__STREAM_ENCODER_OK);
|
||||
|
||||
j = 0;
|
||||
if(encoder->protected_->do_mid_side_stereo && channels == 2) {
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
/*
|
||||
* we have several flavors of the same basic loop, optimized for
|
||||
* different conditions:
|
||||
*/
|
||||
if(encoder->protected_->max_lpc_order > 0) {
|
||||
if(encoder->protected_->do_mid_side_stereo && channels == 2) {
|
||||
/*
|
||||
* stereo coding: unroll channel loop
|
||||
* with LPC: calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
x = mid = side = buffer[0][j];
|
||||
encoder->private_->integer_signal[0][i] = x;
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
x = mid = side = buffer[0][j];
|
||||
encoder->private_->integer_signal[0][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[0][i] = (FLAC__real)x;
|
||||
encoder->private_->real_signal[0][i] = (FLAC__real)x;
|
||||
#endif
|
||||
x = buffer[1][j];
|
||||
encoder->private_->integer_signal[1][i] = x;
|
||||
x = buffer[1][j];
|
||||
encoder->private_->integer_signal[1][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[1][i] = (FLAC__real)x;
|
||||
encoder->private_->real_signal[1][i] = (FLAC__real)x;
|
||||
#endif
|
||||
mid += x;
|
||||
side -= x;
|
||||
mid >>= 1; /* NOTE: not the same as 'mid = (buffer[0][j] + buffer[1][j]) / 2' ! */
|
||||
encoder->private_->integer_signal_mid_side[1][i] = side;
|
||||
encoder->private_->integer_signal_mid_side[0][i] = mid;
|
||||
mid += x;
|
||||
side -= x;
|
||||
mid >>= 1; /* NOTE: not the same as 'mid = (buffer[0][j] + buffer[1][j]) / 2' ! */
|
||||
encoder->private_->integer_signal_mid_side[1][i] = side;
|
||||
encoder->private_->integer_signal_mid_side[0][i] = mid;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal_mid_side[1][i] = (FLAC__real)side;
|
||||
encoder->private_->real_signal_mid_side[0][i] = (FLAC__real)mid;
|
||||
encoder->private_->real_signal_mid_side[1][i] = (FLAC__real)side;
|
||||
encoder->private_->real_signal_mid_side[0][i] = (FLAC__real)mid;
|
||||
#endif
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
else {
|
||||
/*
|
||||
* independent channel coding: buffer each channel in inner loop
|
||||
* with LPC: calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
for(channel = 0; channel < channels; channel++) {
|
||||
x = buffer[channel][j];
|
||||
encoder->private_->integer_signal[channel][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[channel][i] = (FLAC__real)x;
|
||||
#endif
|
||||
}
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
}
|
||||
else {
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
if(encoder->protected_->do_mid_side_stereo && channels == 2) {
|
||||
/*
|
||||
* stereo coding: unroll channel loop
|
||||
* without LPC: no need to calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
for(channel = 0; channel < channels; channel++) {
|
||||
x = buffer[channel][j];
|
||||
encoder->private_->integer_signal[channel][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[channel][i] = (FLAC__real)x;
|
||||
#endif
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
encoder->private_->integer_signal[0][i] = mid = side = buffer[0][j];
|
||||
x = buffer[1][j];
|
||||
encoder->private_->integer_signal[1][i] = x;
|
||||
mid += x;
|
||||
side -= x;
|
||||
mid >>= 1; /* NOTE: not the same as 'mid = (buffer[0][j] + buffer[1][j]) / 2' ! */
|
||||
encoder->private_->integer_signal_mid_side[1][i] = side;
|
||||
encoder->private_->integer_signal_mid_side[0][i] = mid;
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
else {
|
||||
/*
|
||||
* independent channel coding: buffer each channel in inner loop
|
||||
* without LPC: no need to calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
for(channel = 0; channel < channels; channel++)
|
||||
encoder->private_->integer_signal[channel][i] = buffer[channel][j];
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
}
|
||||
|
||||
return true;
|
||||
@ -1442,59 +1504,121 @@ FLAC_API FLAC__bool FLAC__stream_encoder_process_interleaved(FLAC__StreamEncoder
|
||||
FLAC__ASSERT(encoder->protected_->state == FLAC__STREAM_ENCODER_OK);
|
||||
|
||||
j = k = 0;
|
||||
if(encoder->protected_->do_mid_side_stereo && channels == 2) {
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_interleaved_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
/*
|
||||
* we have several flavors of the same basic loop, optimized for
|
||||
* different conditions:
|
||||
*/
|
||||
if(encoder->protected_->max_lpc_order > 0) {
|
||||
if(encoder->protected_->do_mid_side_stereo && channels == 2) {
|
||||
/*
|
||||
* stereo coding: unroll channel loop
|
||||
* with LPC: calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_interleaved_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
x = mid = side = buffer[k++];
|
||||
encoder->private_->integer_signal[0][i] = x;
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
x = mid = side = buffer[k++];
|
||||
encoder->private_->integer_signal[0][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[0][i] = (FLAC__real)x;
|
||||
encoder->private_->real_signal[0][i] = (FLAC__real)x;
|
||||
#endif
|
||||
x = buffer[k++];
|
||||
encoder->private_->integer_signal[1][i] = x;
|
||||
x = buffer[k++];
|
||||
encoder->private_->integer_signal[1][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[1][i] = (FLAC__real)x;
|
||||
encoder->private_->real_signal[1][i] = (FLAC__real)x;
|
||||
#endif
|
||||
mid += x;
|
||||
side -= x;
|
||||
mid >>= 1; /* NOTE: not the same as 'mid = (left + right) / 2' ! */
|
||||
encoder->private_->integer_signal_mid_side[1][i] = side;
|
||||
encoder->private_->integer_signal_mid_side[0][i] = mid;
|
||||
mid += x;
|
||||
side -= x;
|
||||
mid >>= 1; /* NOTE: not the same as 'mid = (left + right) / 2' ! */
|
||||
encoder->private_->integer_signal_mid_side[1][i] = side;
|
||||
encoder->private_->integer_signal_mid_side[0][i] = mid;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal_mid_side[1][i] = (FLAC__real)side;
|
||||
encoder->private_->real_signal_mid_side[0][i] = (FLAC__real)mid;
|
||||
encoder->private_->real_signal_mid_side[1][i] = (FLAC__real)side;
|
||||
encoder->private_->real_signal_mid_side[0][i] = (FLAC__real)mid;
|
||||
#endif
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
else {
|
||||
/*
|
||||
* independent channel coding: buffer each channel in inner loop
|
||||
* with LPC: calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_interleaved_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
for(channel = 0; channel < channels; channel++) {
|
||||
x = buffer[k++];
|
||||
encoder->private_->integer_signal[channel][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[channel][i] = (FLAC__real)x;
|
||||
#endif
|
||||
}
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
}
|
||||
else {
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_interleaved_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
if(encoder->protected_->do_mid_side_stereo && channels == 2) {
|
||||
/*
|
||||
* stereo coding: unroll channel loop
|
||||
* without LPC: no need to calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_interleaved_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
for(channel = 0; channel < channels; channel++) {
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
encoder->private_->integer_signal[0][i] = mid = side = buffer[k++];
|
||||
x = buffer[k++];
|
||||
encoder->private_->integer_signal[channel][i] = x;
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
encoder->private_->real_signal[channel][i] = (FLAC__real)x;
|
||||
#endif
|
||||
encoder->private_->integer_signal[1][i] = x;
|
||||
mid += x;
|
||||
side -= x;
|
||||
mid >>= 1; /* NOTE: not the same as 'mid = (left + right) / 2' ! */
|
||||
encoder->private_->integer_signal_mid_side[1][i] = side;
|
||||
encoder->private_->integer_signal_mid_side[0][i] = mid;
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
else {
|
||||
/*
|
||||
* independent channel coding: buffer each channel in inner loop
|
||||
* without LPC: no need to calculate floating point version of signal
|
||||
*/
|
||||
do {
|
||||
if(encoder->protected_->verify)
|
||||
append_to_verify_fifo_interleaved_(&encoder->private_->verify.input_fifo, buffer, j, channels, min(blocksize-encoder->private_->current_sample_number, samples-j));
|
||||
|
||||
for(i = encoder->private_->current_sample_number; i < blocksize && j < samples; i++, j++) {
|
||||
for(channel = 0; channel < channels; channel++)
|
||||
encoder->private_->integer_signal[channel][i] = buffer[k++];
|
||||
encoder->private_->current_sample_number++;
|
||||
}
|
||||
if(i == blocksize) {
|
||||
if(!process_frame_(encoder, false)) /* false => not last frame */
|
||||
return false;
|
||||
}
|
||||
} while(j < samples);
|
||||
}
|
||||
}
|
||||
|
||||
return true;
|
||||
@ -1630,7 +1754,8 @@ FLAC__bool resize_buffers_(FLAC__StreamEncoder *encoder, unsigned new_size)
|
||||
for(i = 0; ok && i < encoder->protected_->channels; i++) {
|
||||
ok = ok && FLAC__memory_alloc_aligned_int32_array(new_size+4, &encoder->private_->integer_signal_unaligned[i], &encoder->private_->integer_signal[i]);
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
ok = ok && FLAC__memory_alloc_aligned_real_array(new_size, &encoder->private_->real_signal_unaligned[i], &encoder->private_->real_signal[i]);
|
||||
if(encoder->protected_->max_lpc_order > 0)
|
||||
ok = ok && FLAC__memory_alloc_aligned_real_array(new_size, &encoder->private_->real_signal_unaligned[i], &encoder->private_->real_signal[i]);
|
||||
#endif
|
||||
memset(encoder->private_->integer_signal[i], 0, sizeof(FLAC__int32)*4);
|
||||
encoder->private_->integer_signal[i] += 4;
|
||||
@ -1638,7 +1763,8 @@ FLAC__bool resize_buffers_(FLAC__StreamEncoder *encoder, unsigned new_size)
|
||||
for(i = 0; ok && i < 2; i++) {
|
||||
ok = ok && FLAC__memory_alloc_aligned_int32_array(new_size+4, &encoder->private_->integer_signal_mid_side_unaligned[i], &encoder->private_->integer_signal_mid_side[i]);
|
||||
#ifndef FLAC__INTEGER_ONLY_LIBRARY
|
||||
ok = ok && FLAC__memory_alloc_aligned_real_array(new_size, &encoder->private_->real_signal_mid_side_unaligned[i], &encoder->private_->real_signal_mid_side[i]);
|
||||
if(encoder->protected_->max_lpc_order > 0)
|
||||
ok = ok && FLAC__memory_alloc_aligned_real_array(new_size, &encoder->private_->real_signal_mid_side_unaligned[i], &encoder->private_->real_signal_mid_side[i]);
|
||||
#endif
|
||||
memset(encoder->private_->integer_signal_mid_side[i], 0, sizeof(FLAC__int32)*4);
|
||||
encoder->private_->integer_signal_mid_side[i] += 4;
|
||||
|
Loading…
Reference in New Issue
Block a user