diff options
| author | gxw <guxiwei-hf@loongson.cn> | 2019-02-25 18:13:51 +0800 |
|---|---|---|
| committer | Michael Niedermayer <michael@niedermayer.cc> | 2019-02-27 01:51:40 +0100 |
| commit | 4571c7c05d8488cbc2ae0b337751f132f9670fec (patch) | |
| tree | 674088dd8a00aaeceef98a0b54e56cd36623cfa3 /libavutil | |
| parent | 97f47fd601d0e380109f501e4aabff71ffbad052 (diff) | |
| download | ffmpeg-4571c7c05d8488cbc2ae0b337751f132f9670fec.tar.gz | |
avcodec/mips: [loongson] mmi optimizations for VP9 put and avg functions
VP9 decoding speed improved about 60.5%(from 38fps to 61fps, tested on loongson 3A3000).
Reviewed-by: Shiyou Yin <yinshiyou-hf@loongson.cn>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Diffstat (limited to 'libavutil')
| -rw-r--r-- | libavutil/mips/mmiutils.h | 15 |
1 files changed, 15 insertions, 0 deletions
diff --git a/libavutil/mips/mmiutils.h b/libavutil/mips/mmiutils.h index 5fc1908ee3..05f6b31155 100644 --- a/libavutil/mips/mmiutils.h +++ b/libavutil/mips/mmiutils.h @@ -345,5 +345,20 @@ PSRAH_4_MMI(fp1, fp2, fp3, fp4, shift) \ PSRAH_4_MMI(fp5, fp6, fp7, fp8, shift) +/** + * brief: (((value) + (1 << ((n) - 1))) >> (n)) + * fr_i0: src & dst + * fr_i1: Operand number + * fr_t0, fr_t1: temporary FPR + * gr_t0: temporary GPR + */ +#define ROUND_POWER_OF_TWO_MMI(fr_i0, fr_i1, fr_t0, fr_t1, gr_t0) \ + "li "#gr_t0", 0x01 \n\t" \ + "dmtc1 "#gr_t0", "#fr_t0" \n\t" \ + "punpcklwd "#fr_t0", "#fr_t0", "#fr_t0" \n\t" \ + "psubw "#fr_t1", "#fr_i1", "#fr_t0" \n\t" \ + "psllw "#fr_t1", "#fr_t0", "#fr_t1" \n\t" \ + "paddw "#fr_i0", "#fr_i0", "#fr_t1" \n\t" \ + "psraw "#fr_i0", "#fr_i0", "#fr_i1" \n\t" #endif /* AVUTILS_MIPS_MMIUTILS_H */ |