diff options
| author | James Almer <jamrial@gmail.com> | 2017-02-02 17:51:21 -0300 |
|---|---|---|
| committer | James Almer <jamrial@gmail.com> | 2017-02-02 17:51:21 -0300 |
| commit | c8467abbadab424757ea23f71a1036abfb7f14b4 (patch) | |
| tree | 41a4db9cdb0453253f65df734df858cf0e08ca7a /libavcodec/x86/rv34dsp.asm | |
| parent | ab5c4d006da0b3e7aea6fddcebdf5d3d9ad7fd30 (diff) | |
| download | ffmpeg-c8467abbadab424757ea23f71a1036abfb7f14b4.tar.gz | |
x86/rv34dsp: add ff_rv34_idct_dc_add_sse2
Also disable ff_rv34_idct_dc_add_mmx on x86_64 as the presence of sse2
is guaranteed in such builds.
Signed-off-by: James Almer <jamrial@gmail.com>
Diffstat (limited to 'libavcodec/x86/rv34dsp.asm')
| -rw-r--r-- | libavcodec/x86/rv34dsp.asm | 19 |
1 files changed, 18 insertions, 1 deletions
diff --git a/libavcodec/x86/rv34dsp.asm b/libavcodec/x86/rv34dsp.asm index 7732d65b2a..692b4acfcd 100644 --- a/libavcodec/x86/rv34dsp.asm +++ b/libavcodec/x86/rv34dsp.asm @@ -64,6 +64,7 @@ rv34_idct dc rv34_idct dc_noround ; ff_rv34_idct_dc_add_mmx(uint8_t *dst, int stride, int dc); +%if ARCH_X86_32 INIT_MMX mmx cglobal rv34_idct_dc_add, 3, 3 ; calculate DC @@ -97,6 +98,7 @@ cglobal rv34_idct_dc_add, 3, 3 movh [r2], m4 movh [r2+r1], m5 RET +%endif ; Load coeffs and perform row transform ; Output: coeffs in mm[0467], rounder in mm5 @@ -167,7 +169,7 @@ cglobal rv34_idct_add, 3,3,0, d, s, b ret ; ff_rv34_idct_dc_add_sse4(uint8_t *dst, int stride, int dc); -INIT_XMM sse4 +%macro RV34_IDCT_DC_ADD 0 cglobal rv34_idct_dc_add, 3, 3, 6 ; load data IDCT_DC_ROUND r2 @@ -190,7 +192,22 @@ cglobal rv34_idct_dc_add, 3, 3, 6 paddw m4, m0 packuswb m2, m4 movd [r0], m2 +%if cpuflag(sse4) pextrd [r0+r1], m2, 1 pextrd [r2], m2, 2 pextrd [r2+r1], m2, 3 +%else + psrldq m2, 4 + movd [r0+r1], m2 + psrldq m2, 4 + movd [r2], m2 + psrldq m2, 4 + movd [r2+r1], m2 +%endif RET +%endmacro + +INIT_XMM sse2 +RV34_IDCT_DC_ADD +INIT_XMM sse4 +RV34_IDCT_DC_ADD |