Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Optimize NEON encoding for high dynamic range blocks #663

Closed
wants to merge 1 commit into from
Closed

Optimize NEON encoding for high dynamic range blocks #663

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions src/libFLAC/include/private/lpc.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -160,6 +160,7 @@ FLAC__bool FLAC__lpc_compute_residual_from_qlp_coefficients_limit_residual_33bit
# ifdef FLAC__CPU_ARM64
void FLAC__lpc_compute_residual_from_qlp_coefficients_intrin_neon(const FLAC__int32 *data, uint32_t data_len, const FLAC__int32 qlp_coeff[], uint32_t order, int lp_quantization, FLAC__int32 residual[]);
void FLAC__lpc_compute_residual_from_qlp_coefficients_wide_intrin_neon(const FLAC__int32 *data, uint32_t data_len, const FLAC__int32 qlp_coeff[], uint32_t order, int lp_quantization, FLAC__int32 residual[]);
FLAC__bool FLAC__lpc_compute_residual_from_qlp_coefficients_limit_residual_neon(const FLAC__int32 * data, uint32_t data_len, const FLAC__int32 qlp_coeff[], uint32_t order, int lp_quantization, FLAC__int32 residual[]);
# endif

# if (defined FLAC__CPU_IA32 || defined FLAC__CPU_X86_64) && FLAC__HAS_X86INTRIN
Expand Down
56 changes: 56 additions & 0 deletions src/libFLAC/lpc_intrin_neon.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1268,6 +1268,62 @@ void FLAC__lpc_compute_residual_from_qlp_coefficients_wide_intrin_neon(const FLA
return;
}



FLAC__bool FLAC__lpc_compute_residual_from_qlp_coefficients_limit_residual_neon(const FLAC__int32 * data, uint32_t data_len, const FLAC__int32 qlp_coeff[], uint32_t order, int lp_quantization, FLAC__int32 residual[])
{
uint32_t i, j, k, data_len_unroll;
const FLAC__int32 *pData;
FLAC__int32 ordered_qlp_coeff[FLAC__MAX_LPC_ORDER];
FLAC__int64 sum[4], residual_to_check[4];

FLAC__ASSERT(order > 0);
FLAC__ASSERT(order <= 32);

// To allow compiler's autovectorization of the multplications we inverse the order of the qlp coefficients -
// their indices increase same as the corresponding data indices.
for(j = 0; j < order; j++) {
ordered_qlp_coeff[j] = qlp_coeff[order - 1 - j];
}

data_len_unroll = data_len & (-4);
for(i = 0; i < data_len_unroll; i += 4) {
pData = data + i + 0 - (int)order;
sum[0] = sum[1] = sum[2] = sum[3] = 0;

for(j = 0; j < order; j++) {
for(k = 0; k < 4; k++) {
sum[k] += ordered_qlp_coeff[j] * (FLAC__int64)pData[j + k];
}
}
/* residual must not be INT32_MIN because abs(INT32_MIN) is undefined */
for(k = 0; k < 4; k++) {
residual_to_check[k] = data[i + k] - (sum[k] >> lp_quantization);
if(residual_to_check[k] <= INT32_MIN || residual_to_check[k] > INT32_MAX)
return false;
residual[i + k] = residual_to_check[k];
}
}

// Leftover of the data_len
for( ; i < data_len; i++) {
pData = data + i + 0 - (int)order;
sum[0] = 0;

for(j = 0; j < order; j++) {
sum[0] += ordered_qlp_coeff[j] * (FLAC__int64)pData[j];
}
/* residual must not be INT32_MIN because abs(INT32_MIN) is undefined */
residual_to_check[0] = data[i] - (sum[0] >> lp_quantization);
if(residual_to_check[0] <= INT32_MIN || residual_to_check[0] > INT32_MAX)
return false;
residual[i] = residual_to_check[0];
}

return true;
}


#endif /* FLAC__CPU_ARM64 && FLAC__HAS_ARCH64INTRIN */
#endif /* FLAC__NO_ASM */
#endif /* FLAC__INTEGER_ONLY_LIBRARY */
11 changes: 7 additions & 4 deletions src/libFLAC/stream_encoder.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -448,6 +448,7 @@ typedef struct FLAC__StreamEncoderPrivate {
void (*local_lpc_compute_residual_from_qlp_coefficients)(const FLAC__int32 *data, uint32_t data_len, const FLAC__int32 qlp_coeff[], uint32_t order, int lp_quantization, FLAC__int32 residual[]);
void (*local_lpc_compute_residual_from_qlp_coefficients_64bit)(const FLAC__int32 *data, uint32_t data_len, const FLAC__int32 qlp_coeff[], uint32_t order, int lp_quantization, FLAC__int32 residual[]);
void (*local_lpc_compute_residual_from_qlp_coefficients_16bit)(const FLAC__int32 *data, uint32_t data_len, const FLAC__int32 qlp_coeff[], uint32_t order, int lp_quantization, FLAC__int32 residual[]);
FLAC__bool (*local_lpc_compute_residual_from_qlp_coefficients_limit_residual)(const FLAC__int32 * flac_restrict data, uint32_t data_len, const FLAC__int32 * flac_restrict qlp_coeff, uint32_t order, int lp_quantization, FLAC__int32 * flac_restrict residual);
#endif
FLAC__bool disable_mmx;
FLAC__bool disable_sse2;
Expand Down Expand Up @@ -981,6 +982,7 @@ static FLAC__StreamEncoderInitStatus init_stream_internal_(
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients = FLAC__lpc_compute_residual_from_qlp_coefficients;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_64bit = FLAC__lpc_compute_residual_from_qlp_coefficients_wide;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_16bit = FLAC__lpc_compute_residual_from_qlp_coefficients;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_limit_residual = FLAC__lpc_compute_residual_from_qlp_coefficients_limit_residual;
#endif
/* now override with asm where appropriate */
#ifndef FLAC__INTEGER_ONLY_LIBRARY
Expand All @@ -996,9 +998,10 @@ static FLAC__StreamEncoderInitStatus init_stream_internal_(
else
encoder->private_->local_lpc_compute_autocorrelation = FLAC__lpc_compute_autocorrelation;
#endif
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_16bit = FLAC__lpc_compute_residual_from_qlp_coefficients_intrin_neon;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients = FLAC__lpc_compute_residual_from_qlp_coefficients_intrin_neon;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_64bit = FLAC__lpc_compute_residual_from_qlp_coefficients_wide_intrin_neon;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_16bit = FLAC__lpc_compute_residual_from_qlp_coefficients_intrin_neon;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients = FLAC__lpc_compute_residual_from_qlp_coefficients_intrin_neon;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_64bit = FLAC__lpc_compute_residual_from_qlp_coefficients_wide_intrin_neon;
encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_limit_residual = FLAC__lpc_compute_residual_from_qlp_coefficients_limit_residual_neon;
#endif /* defined FLAC__CPU_ARM64 && FLAC__HAS_NEONINTRIN */

if(encoder->private_->cpuinfo.use_asm) {
Expand Down Expand Up @@ -4511,7 +4514,7 @@ uint32_t evaluate_lpc_subframe_(

if(FLAC__lpc_max_residual_bps(subframe_bps, qlp_coeff, order, quantization) > 32) {
if(subframe_bps <= 32){
if(!FLAC__lpc_compute_residual_from_qlp_coefficients_limit_residual(((FLAC__int32 *)signal)+order, residual_samples, qlp_coeff, order, quantization, residual))
if(!encoder->private_->local_lpc_compute_residual_from_qlp_coefficients_limit_residual(((FLAC__int32 *)signal)+order, residual_samples, qlp_coeff, order, quantization, residual))
return 0;
}
else
Expand Down