Re: [FFmpeg-devel] [PATCH v4] lavc/vvc_mc: R-V V avg w_avg

2024-06-26 Thread Rémi Denis-Courmont
Le tiistaina 11. kesäkuuta 2024, 21.38.10 EEST u...@foxmail.com a écrit : > +#include "libavutil/riscv/asm.S" > + > +.macro vsetvlstatic8 w, vlen, is_w > +.if \w <= 2 > +vsetivlizero, \w, e8, mf8, ta, ma > +.elseif \w <= 4 && \vlen == 128 > +v

Re: [FFmpeg-devel] [PATCH v4] lavc/vvc_mc: R-V V avg w_avg

2024-06-11 Thread flow gg
> Nit: for overall code base consistency, I'd use csrwi here. Reason being that > for other rounding modes, csrwi is the better option. > > Probably faster to swap the two above, to avoid stalling on LD. > > If you check more than one length, better to get ff_get_rv_vlenb() into a local > variable.

[FFmpeg-devel] [PATCH v4] lavc/vvc_mc: R-V V avg w_avg

2024-06-11 Thread uk7b
From: sunyuechi C908 X60 avg_8_2x2_c:1.21.0 avg_8_2x2_rvv_i32 :1.01.0 avg_8_2x4_c:2.02.0 avg_8_2x4_rvv_i