[FFmpeg-devel] [PATCH] avcodec/hevcdsp: Offset ff_hevc_.pel_filters to simplify addressing
Nuo Mi
nuomi2021 at gmail.com
Sun Feb 11 13:36:59 EET 2024
On Sun, Feb 11, 2024 at 4:21 PM Andreas Rheinhardt <
andreas.rheinhardt at outlook.com> wrote:
> Besides simplifying address computations (it saves 432B of .text
> in hevcdsp.o alone here) it also fixes undefined behaviour that
> occurs if mx or my are 0 (happens when the filters are unused)
> because they lead to an array index of -1 in the old code.
> This happens in the checkasm-hevc_pel FATE-test.
>
> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at outlook.com>
> ---
> The loongarch and mips parts of this are untested. Luckily we have a
> loongarch patchwork runner...
>
> libavcodec/hevcdsp.c | 6 +-
> libavcodec/hevcdsp.h | 5 +-
> libavcodec/hevcdsp_template.c | 38 ++--
> libavcodec/loongarch/hevc_mc.S | 224 +++++-------------------
> libavcodec/loongarch/hevc_mc_bi_lsx.c | 6 +-
> libavcodec/loongarch/hevc_mc_uni_lsx.c | 6 +-
> libavcodec/loongarch/hevc_mc_uniw_lsx.c | 4 +-
> libavcodec/loongarch/hevcdsp_lsx.c | 6 +-
> libavcodec/mips/hevc_mc_bi_msa.c | 6 +-
> libavcodec/mips/hevc_mc_biw_msa.c | 6 +-
> libavcodec/mips/hevc_mc_uni_msa.c | 6 +-
> libavcodec/mips/hevc_mc_uniw_msa.c | 6 +-
> libavcodec/mips/hevcdsp_mmi.c | 20 +--
> libavcodec/mips/hevcdsp_msa.c | 6 +-
> libavcodec/x86/hevcdsp_init.c | 4 +-
> 15 files changed, 112 insertions(+), 237 deletions(-)
>
> diff --git a/libavcodec/hevcdsp.c b/libavcodec/hevcdsp.c
> index 2ca551df1d..630fdc012e 100644
> --- a/libavcodec/hevcdsp.c
> +++ b/libavcodec/hevcdsp.c
> @@ -91,7 +91,8 @@ static const int8_t transform[32][32] = {
> 90, -90, 88, -85, 82, -78, 73, -67, 61, -54, 46, -38, 31,
> -22, 13, -4 },
> };
>
> -DECLARE_ALIGNED(16, const int8_t, ff_hevc_epel_filters)[7][4] = {
> +DECLARE_ALIGNED(16, const int8_t, ff_hevc_epel_filters)[8][4] = {
> + { 0 },
> { -2, 58, 10, -2},
> { -4, 54, 16, -2},
> { -6, 46, 28, -4},
> @@ -101,7 +102,8 @@ DECLARE_ALIGNED(16, const int8_t,
> ff_hevc_epel_filters)[7][4] = {
> { -2, 10, 58, -2},
> };
>
> -DECLARE_ALIGNED(16, const int8_t, ff_hevc_qpel_filters)[3][16] = {
> +DECLARE_ALIGNED(16, const int8_t, ff_hevc_qpel_filters)[4][16] = {
>
Do you know why this is [4][16]? [4][8] should suffice.
If some architecture requires 16, we might need to update
VVC_INTER_LUMA_TAPS to 16 in the future.
Thank you
> + { 0 },
> { -1, 4,-10, 58, 17, -5, 1, 0, -1, 4,-10, 58, 17, -5, 1, 0},
> { -1, 4,-11, 40, 40,-11, 4, -1, -1, 4,-11, 40, 40,-11, 4, -1},
> { 0, 1, -5, 17, 58,-10, 4, -1, 0, 1, -5, 17, 58,-10, 4, -1}
> diff --git a/libavcodec/hevcdsp.h b/libavcodec/hevcdsp.h
> index 1b9c5bb6bc..a5933dcac4 100644
>
> --
> 2.34.1
>
> _______________________________________________
> 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".
>
More information about the ffmpeg-devel
mailing list