x/ffmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2026-08-10 07:37:43 +00:00
Code Issues Packages Projects Releases Wiki Activity

avcodec/x86/ttadsp: Optimize updating dl[4..7]

Browse Source 
ttaencdsp old:
  filter_process_c:         15.0
  filter_process_ssse3:      9.0 ( 1.66x)
  filter_process_sse4:       6.9 ( 2.18x)

ttaencdsp new:
  filter_process_c:         14.9
  filter_process_ssse3:      8.3 ( 1.79x)
  filter_process_sse4:       6.4 ( 2.32x)

ttadsp old:
  filter_process_c:         14.7
  filter_process_ssse3:      8.7 ( 1.70x)
  filter_process_sse4:       6.6 ( 2.24x)

ttadsp new:
  filter_process_c:         14.6
  filter_process_ssse3:      8.1 ( 1.81x)
  filter_process_sse4:       6.3 ( 2.30x)

Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This commit is contained in:
Andreas Rheinhardt
2026-08-09 17:23:06 +02:00
parent a8d3ae2a71
commit ee83c175f8
1 changed files with 8 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
libavcodec/x86/ttadsp.asm
+8 -9
View File
@@ -107,6 +107,7 @@ cglobal tta%3_filter_process, 5,5,%2, qm, dx, dl, error, in, shift, round
movd m2, shiftm ;
movd m0, [inq] ;
psrad m4, m2 ;
psrldq m1, 4 ; dl5, dl6, dl7, 0
psubd m3, m0, m4 ;
movd [inq], m3 ; *in -= (sum >> filter->shift);
movd [errorq], m3 ; filter->error = *in;
@@ -116,19 +117,17 @@ cglobal tta%3_filter_process, 5,5,%2, qm, dx, dl, error, in, shift, round
movd m2, shiftm ; *in += (sum >> filter->shift);
psrad m4, m2 ;
psrldq m1, 4 ; dl5, dl6, dl7, 0
paddd m0, m4 ;
movd [inq], m0 ;
%endif
psrldq m1, 4 ;
pslldq m0, 12 ; filter->dl[4] = -filter->dl[5];
pshufd m0, m0, 0xf0 ; filter->dl[5] = -filter->dl[6];
psubd m0, m1 ; filter->dl[6] = *in - filter->dl[7];
psrldq m1, m0, 4 ; filter->dl[7] = *in;
pshufd m1, m1, 0xf4 ; filter->dl[5] += filter->dl[6];
paddd m0, m1 ; filter->dl[4] += filter->dl[5];
psrldq m1, 4 ;
paddd m0, m1 ;
pshufd m2, m1, q3321 ; dl6, dl7, 0, 0
pshufd m0, m0, 0
paddd m1, m2 ; dl5+dl6, dl6+dl7, dl7, 0
psrldq m2, 4 ; dl7
paddd m1, m2 ; dl5+dl6+dl7, dl5+dl6+dl7, dl6+dl7, dl7
psubd m0, m1
mova [dlq + 0x10], m0 ;
RET
%endmacro