[FFmpeg-devel] [PATCH] avcodec/vvc: Add aarch64 neon optimization for ALF

Nuo Mi nuomi2021 at gmail.com
Mon Jul 22 16:13:46 EEST 2024


On Tue, Jul 16, 2024 at 8:39 PM Nuo Mi <nuomi2021 at gmail.com> wrote:

> Hi Zhili,
> Good job. Appreciate it.
> With this patch, we're very close to smooth 4K at 30 playback on my M2.
>
Applied.
Thank you, Zhili, for our first Neon optimizations.


> On Tue, Jul 16, 2024 at 12:19 AM Zhao Zhili <quinkblack at foxmail.com>
> wrote:
>
>> From: Zhao Zhili <zhilizhao at tencent.com>
>>
>> vvc_alf_filter_chroma_4x4_8_c: 3.0
>> vvc_alf_filter_chroma_4x4_8_neon: 1.0
>> vvc_alf_filter_chroma_4x4_10_c: 2.7
>> vvc_alf_filter_chroma_4x4_10_neon: 1.0
>> vvc_alf_filter_chroma_4x4_12_c: 2.7
>> vvc_alf_filter_chroma_4x4_12_neon: 1.0
>> vvc_alf_filter_chroma_8x8_8_c: 10.2
>> vvc_alf_filter_chroma_8x8_8_neon: 3.0
>> vvc_alf_filter_chroma_8x8_10_c: 10.0
>> vvc_alf_filter_chroma_8x8_10_neon: 2.5
>> vvc_alf_filter_chroma_8x8_12_c: 10.0
>> vvc_alf_filter_chroma_8x8_12_neon: 2.5
>> vvc_alf_filter_chroma_16x16_8_c: 41.7
>> vvc_alf_filter_chroma_16x16_8_neon: 11.2
>> vvc_alf_filter_chroma_16x16_10_c: 39.0
>> vvc_alf_filter_chroma_16x16_10_neon: 10.0
>> vvc_alf_filter_chroma_16x16_12_c: 40.2
>> vvc_alf_filter_chroma_16x16_12_neon: 10.2
>> vvc_alf_filter_chroma_32x32_8_c: 162.0
>> vvc_alf_filter_chroma_32x32_8_neon: 45.0
>> vvc_alf_filter_chroma_32x32_10_c: 155.5
>> vvc_alf_filter_chroma_32x32_10_neon: 39.5
>> vvc_alf_filter_chroma_32x32_12_c: 155.5
>> vvc_alf_filter_chroma_32x32_12_neon: 40.0
>> vvc_alf_filter_chroma_64x64_8_c: 646.0
>> vvc_alf_filter_chroma_64x64_8_neon: 175.5
>> vvc_alf_filter_chroma_64x64_10_c: 708.2
>> vvc_alf_filter_chroma_64x64_10_neon: 166.7
>> vvc_alf_filter_chroma_64x64_12_c: 619.2
>> vvc_alf_filter_chroma_64x64_12_neon: 157.2
>> vvc_alf_filter_chroma_128x128_8_c: 2611.5
>> vvc_alf_filter_chroma_128x128_8_neon: 698.2
>> vvc_alf_filter_chroma_128x128_10_c: 2470.0
>> vvc_alf_filter_chroma_128x128_10_neon: 616.0
>> vvc_alf_filter_chroma_128x128_12_c: 2531.5
>> vvc_alf_filter_chroma_128x128_12_neon: 620.2
>> vvc_alf_filter_luma_8x8_8_c: 25.2
>> vvc_alf_filter_luma_8x8_8_neon: 4.2
>> vvc_alf_filter_luma_8x8_10_c: 18.5
>> vvc_alf_filter_luma_8x8_10_neon: 4.0
>> vvc_alf_filter_luma_8x8_12_c: 19.0
>> vvc_alf_filter_luma_8x8_12_neon: 4.0
>> vvc_alf_filter_luma_16x16_8_c: 106.5
>> vvc_alf_filter_luma_16x16_8_neon: 16.2
>> vvc_alf_filter_luma_16x16_10_c: 75.2
>> vvc_alf_filter_luma_16x16_10_neon: 14.7
>> vvc_alf_filter_luma_16x16_12_c: 79.7
>> vvc_alf_filter_luma_16x16_12_neon: 14.7
>> vvc_alf_filter_luma_32x32_8_c: 400.5
>> vvc_alf_filter_luma_32x32_8_neon: 63.2
>> vvc_alf_filter_luma_32x32_10_c: 299.2
>> vvc_alf_filter_luma_32x32_10_neon: 57.7
>> vvc_alf_filter_luma_32x32_12_c: 299.2
>> vvc_alf_filter_luma_32x32_12_neon: 57.7
>> vvc_alf_filter_luma_64x64_8_c: 1602.5
>> vvc_alf_filter_luma_64x64_8_neon: 251.7
>> vvc_alf_filter_luma_64x64_10_c: 1197.0
>> vvc_alf_filter_luma_64x64_10_neon: 235.5
>> vvc_alf_filter_luma_64x64_12_c: 1220.2
>> vvc_alf_filter_luma_64x64_12_neon: 235.7
>> vvc_alf_filter_luma_128x128_8_c: 6570.2
>> vvc_alf_filter_luma_128x128_8_neon: 1007.7
>> vvc_alf_filter_luma_128x128_10_c: 4822.7
>> vvc_alf_filter_luma_128x128_10_neon: 936.2
>> vvc_alf_filter_luma_128x128_12_c: 4791.2
>> vvc_alf_filter_luma_128x128_12_neon: 938.5
>>
>> Signed-off-by: Zhao Zhili <zhilizhao at tencent.com>
>> ---
>>  libavcodec/aarch64/vvc/Makefile       |   5 +
>>  libavcodec/aarch64/vvc/alf.S          | 293 ++++++++++++++++++++++++++
>>  libavcodec/aarch64/vvc/alf_template.c | 157 ++++++++++++++
>>  libavcodec/aarch64/vvc/dsp_init.c     |  57 +++++
>>  libavcodec/vvc/dsp.c                  |   4 +-
>>  libavcodec/vvc/dsp.h                  |   1 +
>>  6 files changed, 516 insertions(+), 1 deletion(-)
>>  create mode 100644 libavcodec/aarch64/vvc/Makefile
>>  create mode 100644 libavcodec/aarch64/vvc/alf.S
>>  create mode 100644 libavcodec/aarch64/vvc/alf_template.c
>>  create mode 100644 libavcodec/aarch64/vvc/dsp_init.c
>>
>> diff --git a/libavcodec/aarch64/vvc/Makefile
>> b/libavcodec/aarch64/vvc/Makefile
>> new file mode 100644
>> index 0000000000..58398d6e3d
>> --- /dev/null
>> +++ b/libavcodec/aarch64/vvc/Makefile
>> @@ -0,0 +1,5 @@
>> +clean::
>> +       $(RM) $(CLEANSUFFIXES:%=libavcodec/aarch64/vvc/%)
>> +
>> +OBJS-$(CONFIG_VVC_DECODER)                             +=
>> aarch64/vvc/dsp_init.o
>> +NEON-OBJS-$(CONFIG_VVC_DECODER)                        +=
>> aarch64/vvc/alf.o
>> diff --git a/libavcodec/aarch64/vvc/alf.S b/libavcodec/aarch64/vvc/alf.S
>> new file mode 100644
>> index 0000000000..beb36ac66b
>> --- /dev/null
>> +++ b/libavcodec/aarch64/vvc/alf.S
>> @@ -0,0 +1,293 @@
>> +/*
>> + * Copyright (c) 2024 Zhao Zhili <quinkblack at foxmail.com>
>> + *
>> + * 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/aarch64/asm.S"
>> +
>> +.macro alf_luma_filter_pixel index, pix_size, addr1, addr2, offset1,
>> offset2
>> +    .if \pix_size == 1
>> +        ldur            d3, [\addr1, #\offset1]
>> +        ldur            d4, [\addr2, #\offset2]
>> +        uxtl            v6.8h, v3.8b
>> +        uxtl            v7.8h, v4.8b
>> +    .else
>> +        ldur            q6, [\addr1, #(2*\offset1)]
>> +        ldur            q7, [\addr2, #(2*\offset2)]
>> +    .endif
>> +    .if \index < 8
>> +        dup             v17.4h, v0.h[\index]        // clip
>> +        dup             v18.4h, v16.h[\index]       // -clip
>> +        dup             v19.4h, v1.h[\index]        // filter
>> +
>> +        dup             v26.4h, v22.h[\index]       // clip
>> +        dup             v27.4h, v23.h[\index]       // -clip
>> +        dup             v28.4h, v24.h[\index]       // filter
>> +    .else
>> +        dup             v17.4h, v0.h[\index - 8]    // clip
>> +        dup             v18.4h, v16.h[\index - 8]   // -clip
>> +        dup             v19.4h, v1.h[\index - 8]    // filter
>> +
>> +        dup             v26.4h, v22.h[\index - 8]   // clip
>> +        dup             v27.4h, v23.h[\index - 8]   // -clip
>> +        dup             v28.4h, v24.h[\index - 8]   // filter
>> +    .endif
>> +        ins             v17.d[1], v26.d[0]
>> +        ins             v18.d[1], v27.d[0]
>> +        ins             v19.d[1], v28.d[0]
>> +
>> +        sub             v6.8h, v6.8h, v5.8h
>> +        sub             v7.8h, v7.8h, v5.8h
>> +        smin            v6.8h, v6.8h, v17.8h
>> +        smin            v7.8h, v7.8h, v17.8h
>> +        smax            v6.8h, v6.8h, v18.8h
>> +        smax            v7.8h, v7.8h, v18.8h
>> +        add             v6.8h, v6.8h, v7.8h
>> +        smlal           v20.4s, v19.4h, v6.4h       // v20: sum
>> +        smlal2          v21.4s, v19.8h, v6.8h       // v21: sum
>> +.endm
>> +
>> +/* x0: dst
>> + * x1: pp
>> + * x2: filter
>> + * x3: clip
>> + * w4: is_near_vb
>> + * w5: pix_max
>> + */
>> +.macro alf_filter_luma_kernel, pix_size
>> +        dst             .req x0
>> +        pp              .req x1
>> +        filter          .req x2
>> +        clip            .req x3
>> +        is_near_vb      .req w4
>> +        pix_max         .req w5
>> +    .if \pix_size > 1
>> +        dup             v25.8h, pix_max             // pix_max
>> +    .endif
>> +        ldr             q0, [clip]                  // clip
>> +        ldr             q1, [filter]                // filter
>> +        ldr             q22, [clip, #24]            // clip
>> +        ldr             q24, [filter, #24]          // filter
>> +
>> +        ldr             x5, [pp]                    // x5: p0
>> +        ldr             x6, [pp, #(5*8)]            // x6: p5
>> +        ldr             x7, [pp, #(6*8)]            // x7: p6
>> +        neg             v16.8h, v0.8h               // -clip
>> +        neg             v23.8h, v22.8h              // -clip
>> +
>> +    .if \pix_size == 1
>> +        ldr             d2, [x5]                    // curr
>> +    .else
>> +        ldr             q5, [x5]                    // curr
>> +    .endif
>> +        movi            v20.4s, 64
>> +        cbz             is_near_vb, 1f
>> +        shl             v20.4s, v20.4s, #3
>> +1:
>> +    .if \pix_size == 1
>> +        uxtl            v5.8h, v2.8b
>> +    .endif
>> +        mov             v21.16b, v20.16b
>> +        ldr             x8, [pp, #(3*8)]            // p3
>> +        ldr             x9, [pp, #(4*8)]            // p4
>> +        alf_luma_filter_pixel 0, \pix_size, x6, x7, 0, 0
>> +
>> +        ldr             x6, [pp, #(1*8)]            // p1
>> +        ldr             x7, [pp, #(2*8)]            // p2
>> +        alf_luma_filter_pixel 1, \pix_size, x8, x9, 1, -1
>> +        alf_luma_filter_pixel 2, \pix_size, x8, x9, 0, 0
>> +        alf_luma_filter_pixel 3, \pix_size, x8, x9, -1, 1
>> +
>> +        alf_luma_filter_pixel 4, \pix_size, x6, x7, 2, -2
>> +        alf_luma_filter_pixel 5, \pix_size, x6, x7, 1, -1
>> +        alf_luma_filter_pixel 6, \pix_size, x6, x7, 0, 0
>> +        alf_luma_filter_pixel 7, \pix_size, x6, x7, -1, 1
>> +
>> +        ldr             d0, [clip, #16]             // clip
>> +        ldr             d1, [filter, #16]           // filter
>> +        neg             v16.4h, v0.4h               // -clip
>> +
>> +        ldr             d22, [clip, #40]            // clip
>> +        ldr             d24, [filter, #40]          // filter
>> +        neg             v23.4h, v22.4h              // -clip
>> +        alf_luma_filter_pixel 8, \pix_size, x6, x7, -2, 2
>> +        alf_luma_filter_pixel 9, \pix_size, x5, x5, 3, -3
>> +        alf_luma_filter_pixel 10, \pix_size, x5, x5, 2, -2
>> +        alf_luma_filter_pixel 11, \pix_size, x5, x5, 1, -1
>> +
>> +        cbz             is_near_vb, 2f
>> +        sshr            v20.4s, v20.4s, #10
>> +        sshr            v21.4s, v21.4s, #10
>> +        b               3f
>> +2:
>> +        sshr            v20.4s, v20.4s, #7
>> +        sshr            v21.4s, v21.4s, #7
>> +3:
>> +        uxtl            v22.4s, v5.4h
>> +        uxtl2           v23.4s, v5.8h
>> +        add             v20.4s, v20.4s, v22.4s
>> +        add             v21.4s, v21.4s, v23.4s
>> +        sqxtun          v20.4h, v20.4s
>> +        sqxtun2         v20.8h, v21.4s
>> +    .if \pix_size == 1
>> +        sqxtun          v20.8b, v20.8h
>> +        str             d20, [dst]
>> +    .else
>> +        smin            v20.8h, v20.8h, v25.8h
>> +        str             q20, [dst]
>> +    .endif
>> +        ret
>> +
>> +        .unreq          dst
>> +        .unreq          pp
>> +        .unreq          filter
>> +        .unreq          clip
>> +        .unreq          is_near_vb
>> +        .unreq          pix_max
>> +.endm
>> +
>> +.macro alf_chroma_filter_pixel index, pix_size, addr1, addr2, offset1,
>> offset2
>> +    .if \pix_size == 1
>> +        ldur            s3, [\addr1, #\offset1]
>> +        ldur            s4, [\addr2, #\offset2]
>> +        uxtl            v6.8h, v3.8b
>> +        uxtl            v7.8h, v4.8b
>> +    .else
>> +        ldur            d6, [\addr1, #(2*\offset1)]
>> +        ldur            d7, [\addr2, #(2*\offset2)]
>> +    .endif
>> +    .if \index < 8
>> +        dup             v17.4h, v0.h[\index]        // v17: clip[0]
>> +        dup             v18.4h, v16.h[\index]       // v18: -clip[0]
>> +        dup             v19.4h, v1.h[\index]        // v19: filter[0]
>> +    .else
>> +        dup             v17.4h, v0.h[\index - 8]    // v17: clip[0]
>> +        dup             v18.4h, v16.h[\index - 8]   // v18: -clip[0]
>> +        dup             v19.4h, v1.h[\index - 8]    // v19: filter[0]
>> +    .endif
>> +
>> +        sub             v6.4h, v6.4h, v5.4h
>> +        sub             v7.4h, v7.4h, v5.4h
>> +        smin            v6.4h, v6.4h, v17.4h
>> +        smin            v7.4h, v7.4h, v17.4h
>> +        smax            v6.4h, v6.4h, v18.4h
>> +        smax            v7.4h, v7.4h, v18.4h
>> +        add             v6.4h, v6.4h, v7.4h
>> +        smlal           v20.4s, v19.4h, v6.4h       // v20: sum
>> +.endm
>> +
>> +/* x0: dst
>> + * x1: pp
>> + * x2: filter
>> + * x3: clip
>> + * w4: is_near_vb
>> + * w5: pix_max
>> + */
>> +.macro alf_filter_chroma_kernel, pix_size
>> +        dst             .req x0
>> +        pp              .req x1
>> +        filter          .req x2
>> +        clip            .req x3
>> +        is_near_vb      .req w4
>> +        pix_max         .req w5
>> +    .if \pix_size > 1
>> +        dup             v25.4h, pix_max             // pix_max
>> +    .endif
>> +        ldr             q0, [clip]                  // clip
>> +        ldr             q1, [filter]                // filter
>> +        ldr             x5, [pp]                    // p0
>> +        ldr             x6, [pp, #(3*8)]            // p3
>> +        ldr             x7, [pp, #(4*8)]            // p4
>> +        neg             v16.8h, v0.8h               // -clip
>> +
>> +    .if \pix_size == 1
>> +        ldr             s2, [x5]                    // curr
>> +    .else
>> +        ldr             d5, [x5]                    // curr
>> +    .endif
>> +        movi            v20.4s, 64
>> +        cbz             is_near_vb, 1f
>> +        shl             v20.4s, v20.4s, #3
>> +1:
>> +    .if \pix_size == 1
>> +        uxtl            v5.8h, v2.8b
>> +    .endif
>> +        ldr             x8, [pp, #(1*8)]            // p1
>> +        ldr             x9, [pp, #(2*8)]            // p2
>> +        alf_chroma_filter_pixel 0, \pix_size, x6, x7, 0, 0
>> +        alf_chroma_filter_pixel 1, \pix_size, x8, x9, 1, -1
>> +        alf_chroma_filter_pixel 2, \pix_size, x8, x9, 0, 0
>> +        alf_chroma_filter_pixel 3, \pix_size, x8, x9, -1, 1
>> +        alf_chroma_filter_pixel 4, \pix_size, x5, x5, 2, -2
>> +        alf_chroma_filter_pixel 5, \pix_size, x5, x5, 1, -1
>> +
>> +        uxtl            v22.4s, v5.4h
>> +        cbz             is_near_vb, 2f
>> +        sshr            v20.4s, v20.4s, #10
>> +        b               3f
>> +2:
>> +        sshr            v20.4s, v20.4s, #7
>> +3:
>> +        add             v20.4s, v20.4s, v22.4s
>> +        sqxtun          v20.4h, v20.4s
>> +    .if \pix_size == 1
>> +        sqxtun          v20.8b, v20.8h
>> +        str             s20, [dst]
>> +    .else
>> +        smin            v20.4h, v20.4h, v25.4h
>> +        str             d20, [dst]
>> +    .endif
>> +        ret
>> +
>> +        .unreq          dst
>> +        .unreq          pp
>> +        .unreq          filter
>> +        .unreq          clip
>> +        .unreq          is_near_vb
>> +        .unreq          pix_max
>> +.endm
>> +
>> +function ff_alf_filter_luma_kernel_8_neon, export=1
>> +        alf_filter_luma_kernel  1
>> +endfunc
>> +
>> +function ff_alf_filter_luma_kernel_12_neon, export=1
>> +        mov             w5, 4095
>> +        b               1f
>> +endfunc
>> +
>> +function ff_alf_filter_luma_kernel_10_neon, export=1
>> +        mov             w5, 1023
>> +1:
>> +        alf_filter_luma_kernel  2
>> +endfunc
>> +
>> +function ff_alf_filter_chroma_kernel_8_neon, export=1
>> +        alf_filter_chroma_kernel  1
>> +endfunc
>> +
>> +function ff_alf_filter_chroma_kernel_12_neon, export=1
>> +        mov             w5, 4095
>> +        b               1f
>> +endfunc
>> +
>> +function ff_alf_filter_chroma_kernel_10_neon, export=1
>> +        mov             w5, 1023
>> +1:
>> +        alf_filter_chroma_kernel  2
>> +endfunc
>> diff --git a/libavcodec/aarch64/vvc/alf_template.c
>> b/libavcodec/aarch64/vvc/alf_template.c
>> new file mode 100644
>> index 0000000000..b1927966d2
>> --- /dev/null
>> +++ b/libavcodec/aarch64/vvc/alf_template.c
>> @@ -0,0 +1,157 @@
>> +/*
>> + * VVC filters DSP
>> + *
>> + * Copyright (C) 2024 Zhao Zhili
>> + *
>> + * 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 "libavcodec/bit_depth_template.c"
>> +
>> +void FUNC2(ff_alf_filter_luma_kernel, BIT_DEPTH, _neon)(pixel *dst,
>> +                                            const pixel **p,
>> +                                            const int16_t *filter,
>> +                                            const int16_t *clip,
>> +                                            int is_near_vb);
>> +
>> +void FUNC2(ff_alf_filter_chroma_kernel, BIT_DEPTH, _neon)(pixel *dst,
>> +                                            const pixel **p,
>> +                                            const int16_t *filter,
>> +                                            const int16_t *clip,
>> +                                            int is_near_vb);
>> +
>> +static void FUNC2(alf_filter_luma, BIT_DEPTH, _neon)(uint8_t *_dst,
>> +                                            ptrdiff_t dst_stride,
>> +                                            const uint8_t *_src,
>> +                                            ptrdiff_t src_stride,
>> +                                            const int width, const int
>> height,
>> +                                            const int16_t *filter,
>> +                                            const int16_t *clip,
>> +                                            const int vb_pos)
>> +{
>> +    const pixel *src    = (pixel *)_src;
>> +
>> +    dst_stride /= sizeof(pixel);
>> +    src_stride /= sizeof(pixel);
>> +
>> +    for (int y = 0; y < height; y += ALF_BLOCK_SIZE) {
>> +        int far = (y + 3 < vb_pos - 3) || (y > vb_pos + 2);
>> +
>> +        for (int x = 0; x < width; x += 2 * ALF_BLOCK_SIZE) {
>> +            const pixel *s0 = src + y * src_stride + x;
>> +            const pixel *s1 = s0 + src_stride;
>> +            const pixel *s2 = s0 - src_stride;
>> +            const pixel *s3 = s1 + src_stride;
>> +            const pixel *s4 = s2 - src_stride;
>> +            const pixel *s5 = s3 + src_stride;
>> +            const pixel *s6 = s4 - src_stride;
>> +
>> +            for (int i = 0; i < ALF_BLOCK_SIZE; i++) {
>> +                pixel *dst = (pixel *) _dst + (y + i) * dst_stride + x;
>> +
>> +                const pixel *p0 = s0 + i * src_stride;
>> +                const pixel *p1 = s1 + i * src_stride;
>> +                const pixel *p2 = s2 + i * src_stride;
>> +                const pixel *p3 = s3 + i * src_stride;
>> +                const pixel *p4 = s4 + i * src_stride;
>> +                const pixel *p5 = s5 + i * src_stride;
>> +                const pixel *p6 = s6 + i * src_stride;
>> +                int is_near_vb = 0;
>> +
>> +                if (!far) {
>> +                    is_near_vb = (y + i == vb_pos - 1) || (y + i ==
>> vb_pos);
>> +                    if (is_near_vb) {
>> +                        p1 = p0;
>> +                        p2 = p0;
>> +                    }
>> +                    if (y + i >= vb_pos - 2 && y + i <= vb_pos + 1) {
>> +                        p3 = p1;
>> +                        p4 = p2;
>> +                    }
>> +                    if (y + i >= vb_pos - 3 && y + i <= vb_pos + 2) {
>> +                        p5 = p3;
>> +                        p6 = p4;
>> +                    }
>> +                }
>> +                FUNC2(ff_alf_filter_luma_kernel, BIT_DEPTH, _neon)(dst,
>> +                           (const pixel *[]) { p0, p1, p2, p3, p4, p5,
>> p6},
>> +                           filter,
>> +                           clip,
>> +                           is_near_vb);
>> +            }
>> +            filter += 2 * ALF_NUM_COEFF_LUMA;
>> +            clip += 2 * ALF_NUM_COEFF_LUMA;
>> +        }
>> +    }
>> +}
>> +
>> +static void FUNC2(alf_filter_chroma, BIT_DEPTH, _neon)(uint8_t *_dst,
>> +                                                         ptrdiff_t
>> dst_stride,
>> +                                                         const uint8_t
>> *_src,
>> +                                                         ptrdiff_t
>> src_stride,
>> +                                                         const int width,
>> +                                                         const int
>> height,
>> +                                                         const int16_t
>> *filter,
>> +                                                         const int16_t
>> *clip,
>> +                                                         const int
>> vb_pos)
>> +{
>> +    const pixel *src = (pixel *)_src;
>> +
>> +    dst_stride /= sizeof(pixel);
>> +    src_stride /= sizeof(pixel);
>> +
>> +    for (int y = 0; y < height; y += ALF_BLOCK_SIZE) {
>> +        int far = (y + 3 < vb_pos - 2) || (y > vb_pos + 1);
>> +
>> +        for (int x = 0; x < width; x += ALF_BLOCK_SIZE) {
>> +            const pixel *s0 = src + y * src_stride + x;
>> +            const pixel *s1 = s0 + src_stride;
>> +            const pixel *s2 = s0 - src_stride;
>> +            const pixel *s3 = s1 + src_stride;
>> +            const pixel *s4 = s2 - src_stride;
>> +
>> +            for (int i = 0; i < ALF_BLOCK_SIZE; i++) {
>> +                pixel *dst = (pixel *)_dst + (y + i) * dst_stride + x;
>> +
>> +                const pixel *p0 = s0 + i * src_stride;
>> +                const pixel *p1 = s1 + i * src_stride;
>> +                const pixel *p2 = s2 + i * src_stride;
>> +                const pixel *p3 = s3 + i * src_stride;
>> +                const pixel *p4 = s4 + i * src_stride;
>> +                int is_near_vb = 0;
>> +
>> +                if (!far) {
>> +                    is_near_vb = (y + i == vb_pos - 1) || (y + i ==
>> vb_pos);
>> +                    if (is_near_vb) {
>> +                        p1 = p0;
>> +                        p2 = p0;
>> +                    }
>> +
>> +                    if (y + i >= vb_pos - 2 && y + i <= vb_pos + 1) {
>> +                        p3 = p1;
>> +                        p4 = p2;
>> +                    }
>> +                }
>> +
>> +                FUNC2(ff_alf_filter_chroma_kernel, BIT_DEPTH, _neon)(dst,
>> +                             (const pixel *[]){p0, p1, p2, p3, p4},
>> +                             filter, clip,
>> +                             is_near_vb);
>> +            }
>> +        }
>> +    }
>> +}
>> \ No newline at end of file
>> diff --git a/libavcodec/aarch64/vvc/dsp_init.c
>> b/libavcodec/aarch64/vvc/dsp_init.c
>> new file mode 100644
>> index 0000000000..926e16f70a
>> --- /dev/null
>> +++ b/libavcodec/aarch64/vvc/dsp_init.c
>> @@ -0,0 +1,57 @@
>> +/*
>> + * VVC filters DSP
>> + *
>> + * Copyright (C) 2024 Zhao Zhili
>> + *
>> + * 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/cpu.h"
>> +#include "libavutil/aarch64/cpu.h"
>> +#include "libavcodec/vvc/dsp.h"
>> +#include "libavcodec/vvc/dec.h"
>> +#include "libavcodec/vvc/ctu.h"
>> +
>> +#define BIT_DEPTH 8
>> +#include "alf_template.c"
>> +#undef BIT_DEPTH
>> +
>> +#define BIT_DEPTH 10
>> +#include "alf_template.c"
>> +#undef BIT_DEPTH
>> +
>> +#define BIT_DEPTH 12
>> +#include "alf_template.c"
>> +#undef BIT_DEPTH
>> +
>> +void ff_vvc_dsp_init_aarch64(VVCDSPContext *const c, const int bd)
>> +{
>> +    int cpu_flags = av_get_cpu_flags();
>> +    if (!have_neon(cpu_flags))
>> +        return;
>> +
>> +    if (bd == 8) {
>> +        c->alf.filter[LUMA] = alf_filter_luma_8_neon;
>> +        c->alf.filter[CHROMA] = alf_filter_chroma_8_neon;
>> +    } else if (bd == 10) {
>> +        c->alf.filter[LUMA] = alf_filter_luma_10_neon;
>> +        c->alf.filter[CHROMA] = alf_filter_chroma_10_neon;
>> +    } else if (bd == 12) {
>> +        c->alf.filter[LUMA] = alf_filter_luma_12_neon;
>> +        c->alf.filter[CHROMA] = alf_filter_chroma_12_neon;
>> +    }
>> +}
>> \ No newline at end of file
>> diff --git a/libavcodec/vvc/dsp.c b/libavcodec/vvc/dsp.c
>> index 41e830a98a..648d54ebb2 100644
>> --- a/libavcodec/vvc/dsp.c
>> +++ b/libavcodec/vvc/dsp.c
>> @@ -121,7 +121,9 @@ void ff_vvc_dsp_init(VVCDSPContext *vvcdsp, int
>> bit_depth)
>>          break;
>>      }
>>
>> -#if ARCH_X86
>> +#if ARCH_AARCH64
>> +    ff_vvc_dsp_init_aarch64(vvcdsp, bit_depth);
>> +#elif ARCH_X86
>>      ff_vvc_dsp_init_x86(vvcdsp, bit_depth);
>>  #endif
>>  }
>> diff --git a/libavcodec/vvc/dsp.h b/libavcodec/vvc/dsp.h
>> index 1f14096c41..0b49b97021 100644
>> --- a/libavcodec/vvc/dsp.h
>> +++ b/libavcodec/vvc/dsp.h
>> @@ -180,6 +180,7 @@ typedef struct VVCDSPContext {
>>
>>  void ff_vvc_dsp_init(VVCDSPContext *hpc, int bit_depth);
>>
>> +void ff_vvc_dsp_init_aarch64(VVCDSPContext *hpc, const int bit_depth);
>>  void ff_vvc_dsp_init_x86(VVCDSPContext *hpc, const int bit_depth);
>>
>>  #endif /* AVCODEC_VVC_DSP_H */
>> --
>> 2.42.0
>>
>> _______________________________________________
>> 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