[FFmpeg-devel] Subject: [PATCH 3/3] lavc/dnxhdenc: R-V V get_pixels_8x4_sym

[flow gg]

> I expect that it would be faster to make one large load, and then 4 small
> stores, but that might work only for exactly 128-bit vectors?

This seems to require vle128, so I didn't modify it.

> That's not needed. You can use immediate values.
> You can reorder to avoid immediate data dependencies on the addresses.
> In any case, you need to check the vector length in init.

Okay, I've updated it in the reply.

Rémi Denis-Courmont <remi at remlab.net> 于2024年1月29日周一 23:41写道：

> Hi,
>
> +/*
> + * Copyright (c) 2023 Institue of Software Chinese Academy of Sciences
> (ISCAS).
> + *
> + * This file is part of FFmpeg.
> + *
> + * FFmpeg is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU Lesser General Public
> + * License as published by the Free Software Foundation; either
> + * version 2.1 of the License, or (at your option) any later version.
> + *
> + * FFmpeg is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> + * Lesser General Public License for more details.
> + *
> + * You should have received a copy of the GNU Lesser General Public
> + * License along with FFmpeg; if not, write to the Free Software
> + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
> 02110-1301
> USA
> + */
> +
> +#include "libavutil/riscv/asm.S"
> +
> +func ff_get_pixels_8x4_sym_rvv, zve64x
> +        vsetivli    zero, 8, e8, mf2, ta, ma
> +        vlse64.v    v16, (a1), a2
> +        li          t0, 8 * 8
> +        vsetvli     zero, t0, e16, m4, ta, ma
> +        vzext.vf2   v8, v16
> +        vse16.v     v8, (a0)
> +        li          a2, 8*2
>
> That's not needed. You can use immediate values.
>
> +        vsetivli    zero, 2, e8, mf8, ta, ma
> +        addi        a1, a0, 48
> +        addi        a0, a0, 32*2
> +        vle64.v     v0, (a1)
> +        vse64.v     v0, (a0)
> +        sub         a1, a1, a2
> +        vle64.v     v0, (a1)
> +        add         a0, a0, a2
> +        vse64.v     v0, (a0)
> +        sub         a1, a1, a2
> +        vle64.v     v0, (a1)
> +        add         a0, a0, a2
> +        vse64.v     v0, (a0)
> +        sub         a1, a1, a2
> +        vle64.v     v0, (a1)
> +        add         a0, a0, a2
> +        vse64.v     v0, (a0)
>
> You can reorder to avoid immediate data dependencies on the addresses.
>
> I expect that it would be faster to make one large load, and then 4 small
> stores, but that might work only for exactly 128-bit vectors?
>
> In any case, you need to check the vector length in init.
>
> +
> +        ret
> +endfunc
>
> --
> 雷米‧德尼-库尔蒙
> 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".
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0003-lavc-dnxhdenc-R-V-V-get_pixels_8x4_sym.patch
Type: text/x-patch
Size: 5748 bytes
Desc: not available
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20240130/64d62483/attachment.bin>