Thank you, Jianhua.
This patch mixes many things.
Could you help split it into smaller, more atomic patches?
For example, one for moving code blocks and another for fixing
--disable-ssse3.
On Fri, Mar 1, 2024 at 1:42 AM <toq...@outlook.com> wrote:
> From: Wu Jianhua <toq...@outlook.com>
>
> Signed-off-by: Wu Jianhua <toq...@outlook.com>
> ---
> libavcodec/x86/vvc/vvcdsp_init.c | 115 ++++++++++++++++++++++---------
> 1 file changed, 82 insertions(+), 33 deletions(-)
>
> diff --git a/libavcodec/x86/vvc/vvcdsp_init.c
> b/libavcodec/x86/vvc/vvcdsp_init.c
> index 0d2c683f0f..9ae84bda48 100644
> --- a/libavcodec/x86/vvc/vvcdsp_init.c
> +++ b/libavcodec/x86/vvc/vvcdsp_init.c
> @@ -31,9 +31,67 @@
> #include "libavcodec/vvc/vvcdsp.h"
> #include "libavcodec/x86/h26x/h2656dsp.h"
>
> +#define PUT_PROTOTYPE(name, depth, opt) \
> +void ff_vvc_put_ ## name ## _ ## depth ## _##opt(int16_t *dst, const
> uint8_t *src, ptrdiff_t srcstride, int height, const int8_t *hf, const
> int8_t *vf, int width);
> +
> +#define PUT_PROTOTYPES(name, bitd, opt) \
> + PUT_PROTOTYPE(name##2, bitd, opt) \
> + PUT_PROTOTYPE(name##4, bitd, opt) \
> + PUT_PROTOTYPE(name##8, bitd, opt) \
> + PUT_PROTOTYPE(name##12, bitd, opt) \
> + PUT_PROTOTYPE(name##16, bitd, opt) \
> + PUT_PROTOTYPE(name##24, bitd, opt) \
> + PUT_PROTOTYPE(name##32, bitd, opt) \
> + PUT_PROTOTYPE(name##48, bitd, opt) \
> + PUT_PROTOTYPE(name##64, bitd, opt) \
> + PUT_PROTOTYPE(name##128, bitd, opt)
> +
> +#define PUT_BPC_PROTOTYPES(name, opt) \
> + PUT_PROTOTYPES(name, 8, opt) \
> + PUT_PROTOTYPES(name, 10, opt) \
> + PUT_PROTOTYPES(name, 12, opt)
> +
> +#define PUT_TAP_PROTOTYPES(n, opt) \
> + PUT_BPC_PROTOTYPES(n##tap_h, opt) \
> + PUT_BPC_PROTOTYPES(n##tap_v, opt) \
> + PUT_BPC_PROTOTYPES(n##tap_hv, opt)
> +
> +PUT_BPC_PROTOTYPES(pixels, sse4)
> +PUT_BPC_PROTOTYPES(pixels, avx2)
> +
> +PUT_TAP_PROTOTYPES(4, sse4)
> +PUT_TAP_PROTOTYPES(8, sse4)
> +PUT_TAP_PROTOTYPES(4, avx2)
> +PUT_TAP_PROTOTYPES(8, avx2)
> +
> +#define bf(fn, bd, opt) fn##_##bd##_##opt
> +#define BF(fn, bpc, opt) fn##_##bpc##bpc_##opt
> +
> +#define AVG_BPC_PROTOTYPES(bpc, opt)
> \
> +void BF(ff_vvc_avg, bpc, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> + const int16_t *src0, const int16_t *src1, intptr_t width, intptr_t
> height, intptr_t pixel_max); \
> +void BF(ff_vvc_w_avg, bpc, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> + const int16_t *src0, const int16_t *src1, intptr_t width, intptr_t
> height, \
> + intptr_t denom, intptr_t w0, intptr_t w1, intptr_t o0, intptr_t o1,
> intptr_t pixel_max);
> +
> +#define AVG_PROTOTYPES(bd, opt)
> \
> +void bf(ff_vvc_avg, bd, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> + const int16_t *src0, const int16_t *src1, int width, int height);
> \
> +void bf(ff_vvc_w_avg, bd, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> + const int16_t *src0, const int16_t *src1, int width, int height,
> \
> + int denom, int w0, int w1, int o0, int o1);
> +
> +AVG_BPC_PROTOTYPES( 8, avx2)
> +AVG_BPC_PROTOTYPES(16, avx2)
> +
> +AVG_PROTOTYPES( 8, avx2)
> +AVG_PROTOTYPES(10, avx2)
> +AVG_PROTOTYPES(12, avx2)
> +
> #if ARCH_X86_64
> +#if HAVE_SSE4_EXTERNAL
> #define FW_PUT(name, depth, opt) \
> -static void ff_vvc_put_ ## name ## _ ## depth ## _##opt(int16_t *dst,
> const uint8_t *src, ptrdiff_t srcstride, \
> +void ff_vvc_put_ ## name ## _ ## depth ## _##opt(int16_t *dst, const
> uint8_t *src, ptrdiff_t srcstride, \
> int height, const int8_t
> *hf, const int8_t *vf, int width) \
> {
> \
> ff_h2656_put_## name ## _ ## depth ## _##opt(dst, 2 * MAX_PB_SIZE,
> src, srcstride, height, hf, vf, width); \
> @@ -69,7 +127,9 @@ static void ff_vvc_put_ ## name ## _ ## depth ##
> _##opt(int16_t *dst, const uint
> FW_PUT_SSE4( 8)
> FW_PUT_SSE4(10)
> FW_PUT_SSE4(12)
> +#endif
>
> +#if HAVE_AVX2_EXTERNAL
> #define FW_PUT_TAP_AVX2(n, bitd) \
> FW_PUT(n ## tap_h32, bitd, avx2) \
> FW_PUT(n ## tap_h64, bitd, avx2) \
> @@ -105,6 +165,25 @@ FW_PUT_AVX2(12)
> FW_PUT_16BPC_AVX2(10)
> FW_PUT_16BPC_AVX2(12)
>
> +#define AVG_FUNCS(bpc, bd, opt)
> \
> +void bf(ff_vvc_avg, bd, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> + const int16_t *src0, const int16_t *src1, int width, int height)
> \
> +{
> \
> + BF(ff_vvc_avg, bpc, opt)(dst, dst_stride, src0, src1, width, height,
> (1 << bd) - 1); \
> +}
> \
> +void bf(ff_vvc_w_avg, bd, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> + const int16_t *src0, const int16_t *src1, int width, int height,
> \
> + int denom, int w0, int w1, int o0, int o1)
> \
> +{
> \
> + BF(ff_vvc_w_avg, bpc, opt)(dst, dst_stride, src0, src1, width,
> height, \
> + denom, w0, w1, o0, o1, (1 << bd) - 1);
> \
> +}
> +
> +AVG_FUNCS(8, 8, avx2)
> +AVG_FUNCS(16, 10, avx2)
> +AVG_FUNCS(16, 12, avx2)
> +#endif
> +
> #define PEL_LINK(dst, C, W, idx1, idx2, name, D, opt)
> \
> dst[C][W][idx1][idx2] = ff_vvc_put_## name ## _ ## D ## _##opt;
> \
> dst ## _uni[C][W][idx1][idx2] = ff_h2656_put_uni_ ## name ## _ ## D
> ## _##opt; \
> @@ -170,40 +249,10 @@ FW_PUT_16BPC_AVX2(12)
> MC_TAP_LINKS_16BPC_AVX2(LUMA, 8, bd); \
> MC_TAP_LINKS_16BPC_AVX2(CHROMA, 4, bd);
>
> -#define bf(fn, bd, opt) fn##_##bd##_##opt
> -#define BF(fn, bpc, opt) fn##_##bpc##bpc_##opt
> -
> -#define AVG_BPC_FUNC(bpc, opt)
> \
> -void BF(ff_vvc_avg, bpc, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> - const int16_t *src0, const int16_t *src1, intptr_t width, intptr_t
> height, intptr_t pixel_max); \
> -void BF(ff_vvc_w_avg, bpc, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> - const int16_t *src0, const int16_t *src1, intptr_t width, intptr_t
> height, \
> - intptr_t denom, intptr_t w0, intptr_t w1, intptr_t o0, intptr_t o1,
> intptr_t pixel_max);
> -
> -#define AVG_FUNCS(bpc, bd, opt)
> \
> -static void bf(avg, bd, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> - const int16_t *src0, const int16_t *src1, int width, int height)
> \
> -{
> \
> - BF(ff_vvc_avg, bpc, opt)(dst, dst_stride, src0, src1, width, height,
> (1 << bd) - 1); \
> -}
> \
> -static void bf(w_avg, bd, opt)(uint8_t *dst, ptrdiff_t dst_stride,
> \
> - const int16_t *src0, const int16_t *src1, int width, int height,
> \
> - int denom, int w0, int w1, int o0, int o1)
> \
> -{
> \
> - BF(ff_vvc_w_avg, bpc, opt)(dst, dst_stride, src0, src1, width,
> height, \
> - denom, w0, w1, o0, o1, (1 << bd) - 1);
> \
> -}
> -
> -AVG_BPC_FUNC(8, avx2)
> -AVG_BPC_FUNC(16, avx2)
> -
> -AVG_FUNCS(8, 8, avx2)
> -AVG_FUNCS(16, 10, avx2)
> -AVG_FUNCS(16, 12, avx2)
>
> #define AVG_INIT(bd, opt) do { \
> - c->inter.avg = bf(avg, bd, opt); \
> - c->inter.w_avg = bf(w_avg, bd, opt); \
> + c->inter.avg = bf(ff_vvc_avg, bd, opt); \
> + c->inter.w_avg = bf(ff_vvc_w_avg, bd, opt); \
>
Why change the function scope to fix a compilation issue?
> } while (0)
> #endif
>
> --
> 2.33.0.windows.2
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
