diff options
| author | Christophe GISQUET <christophe.gisquet@gmail.com> | 2012-02-23 19:48:58 +0100 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2012-02-23 15:50:06 -0800 |
| commit | 34454c761f01275d4adaf40df6d70a59011c4a6c (patch) | |
| tree | a25a23c028ddee97c1195567f855ce064bdbe916 /libavcodec/sbrdsp.h | |
| parent | 2e74a5abc2fda6cfbc86589852d6194d502332cb (diff) | |
SBR DSP x86: implement SSE sbr_sum_square_sse
The 32bits targets have been compiled with -mfpmath=sse for proper reference.
sbr_sum_square C /32bits: 82c (unrolled)/102c
C /64bits: 69c (unrolled)/82c
SSE/32bits: 42c
SSE/64bits: 31c
Use of SSE4.1 dpps to perform the final sum is slower.
Not unrolling to perform 8 operations in a loop yields 10 more cycles.
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
Diffstat (limited to 'libavcodec/sbrdsp.h')
| -rw-r--r-- | libavcodec/sbrdsp.h | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/libavcodec/sbrdsp.h b/libavcodec/sbrdsp.h index ee5d5a0972..fe91957ce4 100644 --- a/libavcodec/sbrdsp.h +++ b/libavcodec/sbrdsp.h @@ -46,5 +46,6 @@ extern const float ff_sbr_noise_table[][2]; void ff_sbrdsp_init(SBRDSPContext *s); void ff_sbrdsp_init_arm(SBRDSPContext *s); +void ff_sbrdsp_init_x86(SBRDSPContext *s); #endif |