[FFmpeg-devel] [PATCH 5/7] lavc/me_cmp: R-V V vsse vsad

[flow gg]

.macro vabsaddu dst src tmp
        - vneg.v          \tmp, \src
        - vmax.vv         \tmp, \src, \tmp
        + vfabs.v             \tmp, \src
        vwaddu.wv       \dst, \dst, \tmp
.endm

After making this change, the tests did not pass. I'm not quite clear on
how to understand the differences..

checkasm: 4 of 21 tests have failed
benchmarking with native FFmpeg timers
nop: 7.0
func: vsad_0, x=46 y=13 h=4, error: asm=720974 c=6162
func: vsad_1, x=16 y=14 h=10, error: asm=1146753 c=9353
func: vsad_4, x=13 y=32 h=12, error: asm=2654565 c=14573
func: vsad_5, x=32 y=1 h=8, error: asm=917745 c=3865

Rémi Denis-Courmont <remi at remlab.net> 于2024年2月22日周四 02:07写道：

> Le tiistaina 6. helmikuuta 2024, 17.56.32 EET flow gg a écrit :
> >
>
> Did you try to compute integral absolute values with the ad-hoc (floating
> point) instruction instead of vneg/vmax? It should work since the sign is
> in
> the same place, though I don't know if it will be faster.
>
> --
> レミ・デニ-クールモン
> http://www.remlab.net/
>
>
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>