Message ID | 20220913115824.60792-3-hum@semihalf.com |
---|---|
State | New |
Headers | show |
Series | Provide arm64 neon implementations | expand |
Context | Check | Description |
---|---|---|
yinshiyou/make_loongarch64 | success | Make finished |
yinshiyou/make_fate_loongarch64 | success | Make fate finished |
andriy/make_x86 | success | Make finished |
andriy/make_fate_x86 | success | Make fate finished |
On Tue, 13 Sep 2022, Hubert Mazur wrote: > Provide optimized implementation for pix_median_abs16 function. You've forgot to update this part of the commit message. > Performance comparison tests are shown below. > - vsad_5_c: 94.7 > - vsad_5_neon: 20.7 > > Benchmarks and tests run with checkasm tool on AWS Graviton 3. > > Signed-off-by: Hubert Mazur <hum@semihalf.com> > --- > libavcodec/aarch64/me_cmp_init_aarch64.c | 3 ++ > libavcodec/aarch64/me_cmp_neon.S | 42 ++++++++++++++++++++++++ > 2 files changed, 45 insertions(+) > > diff --git a/libavcodec/aarch64/me_cmp_init_aarch64.c b/libavcodec/aarch64/me_cmp_init_aarch64.c > index fb51a833be..d3fa047a86 100644 > --- a/libavcodec/aarch64/me_cmp_init_aarch64.c > +++ b/libavcodec/aarch64/me_cmp_init_aarch64.c > @@ -45,6 +45,8 @@ int vsad16_neon(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, > ptrdiff_t stride, int h); > int vsad_intra16_neon(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, > ptrdiff_t stride, int h) ; > +int vsad_intra8_neon(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, > + ptrdiff_t stride, int h) ; > int vsse16_neon(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, > ptrdiff_t stride, int h); > int vsse_intra16_neon(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, > @@ -75,6 +77,7 @@ av_cold void ff_me_cmp_init_aarch64(MECmpContext *c, AVCodecContext *avctx) > > c->vsad[0] = vsad16_neon; > c->vsad[4] = vsad_intra16_neon; > + c->vsad[5] = vsad_intra8_neon; > > c->vsse[0] = vsse16_neon; > c->vsse[4] = vsse_intra16_neon; > diff --git a/libavcodec/aarch64/me_cmp_neon.S b/libavcodec/aarch64/me_cmp_neon.S > index a4a4344f42..73701bd353 100644 > --- a/libavcodec/aarch64/me_cmp_neon.S > +++ b/libavcodec/aarch64/me_cmp_neon.S > @@ -1050,3 +1050,45 @@ function pix_median_abs16_neon, export=1 > ret > > endfunc > + > +function vsad_intra8_neon, export=1 > + // x0 unused > + // x1 uint8_t *pix1 > + // x2 uint8_t *dummy > + // x3 ptrdiff_t stride > + // w4 int h > + > + ld1 {v0.8b}, [x1], x3 > + sub w4, w4, #1 // we need to make h-1 iterations > + cmp w4, #3 > + movi v16.8h, #0 > + b.lt 2f > + > +1: > + // v = abs( pix1[0] - pix1[0 + stride] ) > + // score = sum(v) > + ld1 {v1.8b}, [x1], x3 > + ld1 {v2.8b}, [x1], x3 > + uabal v16.8h, v0.8b, v1.8b > + ld1 {v3.8b}, [x1], x3 > + sub w4, w4, #3 Instinctively, I'd prefer to move the sub instruction up to between the first two ld1 instructions here. However I don't see any change in benchmarks on Cortex A53 due to that, so it's not strictly necessary, but I'd prefer it that way. Other than that, this looks very reasonable and straightforward. // Martin
diff --git a/libavcodec/aarch64/me_cmp_init_aarch64.c b/libavcodec/aarch64/me_cmp_init_aarch64.c index fb51a833be..d3fa047a86 100644 --- a/libavcodec/aarch64/me_cmp_init_aarch64.c +++ b/libavcodec/aarch64/me_cmp_init_aarch64.c @@ -45,6 +45,8 @@ int vsad16_neon(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride, int h); int vsad_intra16_neon(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, ptrdiff_t stride, int h) ; +int vsad_intra8_neon(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, + ptrdiff_t stride, int h) ; int vsse16_neon(MpegEncContext *c, const uint8_t *s1, const uint8_t *s2, ptrdiff_t stride, int h); int vsse_intra16_neon(MpegEncContext *c, const uint8_t *s, const uint8_t *dummy, @@ -75,6 +77,7 @@ av_cold void ff_me_cmp_init_aarch64(MECmpContext *c, AVCodecContext *avctx) c->vsad[0] = vsad16_neon; c->vsad[4] = vsad_intra16_neon; + c->vsad[5] = vsad_intra8_neon; c->vsse[0] = vsse16_neon; c->vsse[4] = vsse_intra16_neon; diff --git a/libavcodec/aarch64/me_cmp_neon.S b/libavcodec/aarch64/me_cmp_neon.S index a4a4344f42..73701bd353 100644 --- a/libavcodec/aarch64/me_cmp_neon.S +++ b/libavcodec/aarch64/me_cmp_neon.S @@ -1050,3 +1050,45 @@ function pix_median_abs16_neon, export=1 ret endfunc + +function vsad_intra8_neon, export=1 + // x0 unused + // x1 uint8_t *pix1 + // x2 uint8_t *dummy + // x3 ptrdiff_t stride + // w4 int h + + ld1 {v0.8b}, [x1], x3 + sub w4, w4, #1 // we need to make h-1 iterations + cmp w4, #3 + movi v16.8h, #0 + b.lt 2f + +1: + // v = abs( pix1[0] - pix1[0 + stride] ) + // score = sum(v) + ld1 {v1.8b}, [x1], x3 + ld1 {v2.8b}, [x1], x3 + uabal v16.8h, v0.8b, v1.8b + ld1 {v3.8b}, [x1], x3 + sub w4, w4, #3 + uabal v16.8h, v1.8b, v2.8b + cmp w4, #3 + mov v0.8b, v3.8b + uabal v16.8h, v2.8b, v3.8b + b.ge 1b + cbz w4, 3f + +2: + ld1 {v1.8b}, [x1], x3 + subs w4, w4, #1 + uabal v16.8h, v0.8b, v1.8b + mov v0.8b, v1.8b + cbnz w4, 2b + +3: + uaddlv s17, v16.8h + fmov w0, s17 + + ret +endfunc
Provide optimized implementation for pix_median_abs16 function. Performance comparison tests are shown below. - vsad_5_c: 94.7 - vsad_5_neon: 20.7 Benchmarks and tests run with checkasm tool on AWS Graviton 3. Signed-off-by: Hubert Mazur <hum@semihalf.com> --- libavcodec/aarch64/me_cmp_init_aarch64.c | 3 ++ libavcodec/aarch64/me_cmp_neon.S | 42 ++++++++++++++++++++++++ 2 files changed, 45 insertions(+)