har0ke/FFmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

avcodec/x86/lossless_audiodsp: support len %16 == 8 in scalarproduct_and_madd_int16()

Browse Source 
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2014-11-22 20:38:30 +01:00
parent 2f6550bb9a
commit 4327088da3
2 changed files with 7 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
libavcodec/lossless_audiodsp.h
View File

@ -29,7 +29,7 @@ typedef struct LLAudDSPContext {
/** /**
* Calculate scalar product of v1 and v2, * Calculate scalar product of v1 and v2,
* and v1[i] += v3[i] * mul * and v1[i] += v3[i] * mul
* @param len length of vectors, should be multiple of 16 * @param len length of vectors, should be multiple of 8
*/ */
int32_t (*scalarproduct_and_madd_int16)(int16_t *v1 /* align 16 */, int32_t (*scalarproduct_and_madd_int16)(int16_t *v1 /* align 16 */,
const int16_t *v2, const int16_t *v2,

6
libavcodec/x86/lossless_audiodsp.asm
View File

@ -26,6 +26,10 @@ SECTION_TEXT
; int ff_scalarproduct_and_madd_int16(int16_t *v1, int16_t *v2, int16_t *v3, ; int ff_scalarproduct_and_madd_int16(int16_t *v1, int16_t *v2, int16_t *v3,
; int order, int mul) ; int order, int mul)
cglobal scalarproduct_and_madd_int16, 4,4,8, v1, v2, v3, order, mul cglobal scalarproduct_and_madd_int16, 4,4,8, v1, v2, v3, order, mul
%if mmsize == 16
test orderq, 8
jnz scalarproduct_and_madd_int16_mmxext;
%endif
shl orderq, 1 shl orderq, 1
movd m7, mulm movd m7, mulm
%if mmsize == 16 %if mmsize == 16
@ -117,6 +121,8 @@ align 16
; int order, int mul) ; int order, int mul)
INIT_XMM ssse3 INIT_XMM ssse3
cglobal scalarproduct_and_madd_int16, 4,5,10, v1, v2, v3, order, mul cglobal scalarproduct_and_madd_int16, 4,5,10, v1, v2, v3, order, mul
test orderq, 8
jnz scalarproduct_and_madd_int16_mmxext;
shl orderq, 1 shl orderq, 1
movd m7, mulm movd m7, mulm
pshuflw m7, m7, 0 pshuflw m7, m7, 0