OrgACRN
/
sof
mirror of https://github.com/thesofproject/sof.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

Math: FIR: Simplify and optimize generic C FIR filter 

This change saves 18 MCPS as gcc build build for xtensa on TGL-H
from 134 MCPS to 115 MCPS in a stereo 48 kHz 64 tap filter case.

Signed-off-by: Seppo Ingalsuo <seppo.ingalsuo@linux.intel.com>
This commit is contained in:
Seppo Ingalsuo 2022-02-14 19:35:41 +02:00 committed by Liam Girdwood
parent da8ee993a7
commit fc30b4b870
1 changed files with 8 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
src/math/fir_generic.c
View File

@ -70,6 +70,9 @@ int32_t fir_32x16(struct fir_state_32x16 *fir, int32_t x)
int n1; int n1;
int n2; int n2;
int n; int n;
const int length = fir->length;
const int taps = fir->taps;
const int shift = 15 + fir->out_shift;
/* Bypass is set with length set to zero. */ /* Bypass is set with length set to zero. */
if (!fir->length) if (!fir->length)
@ -82,23 +85,11 @@ int32_t fir_32x16(struct fir_state_32x16 *fir, int32_t x)
* to process before circular wrap. * to process before circular wrap.
*/ */
n1 = ++fir->rwi; n1 = ++fir->rwi;
if (fir->rwi == fir->length) if (fir->rwi == length)
fir->rwi = 0; fir->rwi = 0;
/* Check if no need to un-wrap FIR data. */
if (n1 > fir->length) {
/* Data is Q1.31, coef is Q1.15, product is Q2.46 */
for (n = 0; n < fir->length; n++) {
y += (int64_t)(*coef) * (*data);
coef++;
data--;
}
/* Q2.46 -> Q2.31, saturate to Q1.31 */
return sat_int32(y >> (15 + fir->out_shift));
}
/* Part 1, loop n1 times */ /* Part 1, loop n1 times */
n1 = MIN(n1, taps);
for (n = 0; n < n1; n++) { for (n = 0; n < n1; n++) {
y += (int64_t)(*coef) * (*data); y += (int64_t)(*coef) * (*data);
coef++; coef++;
@ -106,8 +97,8 @@ int32_t fir_32x16(struct fir_state_32x16 *fir, int32_t x)
} }
/* Part 2, un-wrap data, continue n2 times */ /* Part 2, un-wrap data, continue n2 times */
n2 = fir->length - n1; n2 = taps - n1;
data = &fir->delay[fir->length - 1]; data = &fir->delay[length - 1];
for (n = 0; n < n2; n++) { for (n = 0; n < n2; n++) {
y += (int64_t)(*coef) * (*data); y += (int64_t)(*coef) * (*data);
coef++; coef++;
@ -115,7 +106,7 @@ int32_t fir_32x16(struct fir_state_32x16 *fir, int32_t x)
} }
/* Q2.46 -> Q2.31, saturate to Q1.31 */ /* Q2.46 -> Q2.31, saturate to Q1.31 */
return sat_int32(y >> (15 + fir->out_shift)); return sat_int32(y >> shift);
} }
#endif #endif