diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile index 6c2ce3001a..69ccd0896d 100644 --- a/libavcodec/riscv/Makefile +++ b/libavcodec/riscv/Makefile @@ -58,5 +58,7 @@ OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_init.o RVV-OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_rvv.o OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_init.o RVV-OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_rvv.o +OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9dsp_init.o +RVV-OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9_intra_rvv.o OBJS-$(CONFIG_VORBIS_DECODER) += riscv/vorbisdsp_init.o RVV-OBJS-$(CONFIG_VORBIS_DECODER) += riscv/vorbisdsp_rvv.o diff --git a/libavcodec/riscv/vp9_intra_rvv.S b/libavcodec/riscv/vp9_intra_rvv.S new file mode 100644 index 0000000000..db9774c263 --- /dev/null +++ b/libavcodec/riscv/vp9_intra_rvv.S @@ -0,0 +1,115 @@ +/* + * Copyright (c) 2024 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" + +.macro avgdc size + vwredsumu.vs v16, v8, v16 + vsetivli zero, 1, e16, m1, ta, ma + vmv.x.s t1, v16 + addi t1, t1, 1 << (\size - 1) + srai t1, t1, \size +.endm + +.macro getdc type size +.ifc \type,top + vmv.v.x v16, zero + vle8.v v8, (a3) + avgdc \size +.elseif \type == left + vmv.v.x v16, zero + vle8.v v8, (a2) + avgdc \size +.elseif \type == dc + vmv.v.x v16, zero + vle8.v v8, (a2) + vwredsumu.vs v16, v8, v16 + vle8.v v8, (a3) + avgdc \size +.else + li t1, \type +.endif +.endm + +.macro dc_e32 type size n restore +.ifc \size,32 + li t0, 32 + vsetvli zero, t0, e8, m2, ta, ma +.else + vsetivli zero, 16, e8, m1, ta, ma +.endif + getdc \type \n + +.if \restore == 1 && \size == 32 + vsetvli zero, t0, e8, m2, ta, ma +.elseif \restore == 1 && \size == 16 + vsetivli zero, 16, e8, m1, ta, ma +.endif + vmv.v.x v0, t1 + + .rept \size + vse8.v v0, (a0) + add a0, a0, a1 + .endr + + ret +.endm + +.macro dc_e64 type size n restore + vsetivli zero, 8, e8, mf2, ta, ma + getdc \type \n + + li t0, 64 + vsetvli zero, t0, e8, m4, ta, ma + vmv.v.x v0, t1 + vsetivli zero, 8, e8, mf2, ta, ma + vsse64.v v0, (a0), a1 + + ret +.endm + +.macro func_dc name size type n restore ext +func ff_\()\name\()_\()\size\()x\size\()_rvv, \ext +.ifc \size,8 + dc_e64 \type \size \n \restore +.else + dc_e32 \type \size \n \restore +.endif +endfunc +.endm + +func_dc dc_127 32 127 0 0 zve32x +func_dc dc_127 16 127 0 0 zve32x +func_dc dc_127 8 127 0 0 zve64x +func_dc dc_128 32 128 0 0 zve32x +func_dc dc_128 16 128 0 0 zve32x +func_dc dc_128 8 128 0 0 zve64x +func_dc dc_129 32 129 0 0 zve32x +func_dc dc_129 16 129 0 0 zve32x +func_dc dc_129 8 129 0 0 zve64x +func_dc dc 32 dc 6 1 zve32x +func_dc dc 16 dc 5 1 zve32x +func_dc dc 8 dc 4 0 zve64x +func_dc dc_left 32 left 5 1 zve32x +func_dc dc_left 16 left 4 1 zve32x +func_dc dc_left 8 left 3 0 zve64x +func_dc dc_top 32 top 5 1 zve32x +func_dc dc_top 16 top 4 1 zve32x +func_dc dc_top 8 top 3 0 zve64x diff --git a/libavcodec/riscv/vp9dsp.h b/libavcodec/riscv/vp9dsp.h new file mode 100644 index 0000000000..0540c7303d --- /dev/null +++ b/libavcodec/riscv/vp9dsp.h @@ -0,0 +1,171 @@ +/* + * Copyright (c) 2024 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 + */ + +#ifndef AVCODEC_RISCV_VP9DSP_RISCV_H +#define AVCODEC_RISCV_VP9DSP_RISCV_H + +#include +#include + +void ff_dc_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_top_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_top_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_top_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_left_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_left_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_left_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_127_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_127_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_127_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_128_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_128_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_128_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_129_32x32_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_129_16x16_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); +void ff_dc_129_8x8_rvv(uint8_t *dst, ptrdiff_t stride, const uint8_t *l, + const uint8_t *a); + +#define VP9_8TAP_RISCV_RVV_FUNC(SIZE, type, type_idx) \ +void ff_put_8tap_##type##_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_8tap_##type##_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_8tap_##type##_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_8tap_##type##_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_8tap_##type##_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_8tap_##type##_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, \ + ptrdiff_t srcstride, \ + int h, int mx, int my); + +#define VP9_BILINEAR_RISCV_RVV_FUNC(SIZE) \ +void ff_put_bilin_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_bilin_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_put_bilin_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_bilin_##SIZE##h_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_bilin_##SIZE##v_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg_bilin_##SIZE##hv_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); + +#define VP9_COPY_AVG_RISCV_RVV_FUNC(SIZE) \ +void ff_copy##SIZE##_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); \ + \ +void ff_avg##SIZE##_rvv(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); + +VP9_8TAP_RISCV_RVV_FUNC(64, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(32, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(16, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(8, regular, FILTER_8TAP_REGULAR); +VP9_8TAP_RISCV_RVV_FUNC(4, regular, FILTER_8TAP_REGULAR); + +VP9_8TAP_RISCV_RVV_FUNC(64, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(32, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(16, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(8, sharp, FILTER_8TAP_SHARP); +VP9_8TAP_RISCV_RVV_FUNC(4, sharp, FILTER_8TAP_SHARP); + +VP9_8TAP_RISCV_RVV_FUNC(64, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(32, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(16, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(8, smooth, FILTER_8TAP_SMOOTH); +VP9_8TAP_RISCV_RVV_FUNC(4, smooth, FILTER_8TAP_SMOOTH); + +VP9_BILINEAR_RISCV_RVV_FUNC(64); +VP9_BILINEAR_RISCV_RVV_FUNC(32); +VP9_BILINEAR_RISCV_RVV_FUNC(16); +VP9_BILINEAR_RISCV_RVV_FUNC(8); +VP9_BILINEAR_RISCV_RVV_FUNC(4); + +VP9_COPY_AVG_RISCV_RVV_FUNC(64); +VP9_COPY_AVG_RISCV_RVV_FUNC(32); +VP9_COPY_AVG_RISCV_RVV_FUNC(16); +VP9_COPY_AVG_RISCV_RVV_FUNC(8); +VP9_COPY_AVG_RISCV_RVV_FUNC(4); + +#define VP9_COPY_RISCV_RVI_FUNC(SIZE) \ +void ff_copy##SIZE##_rvi(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride, \ + int h, int mx, int my); + +VP9_COPY_RISCV_RVI_FUNC(8); +VP9_COPY_RISCV_RVI_FUNC(4); + +#undef VP9_8TAP_RISCV_RVV_FUNC +#undef VP9_BILINEAR_RISCV_RVV_FUNC +#undef VP9_COPY_AVG_RISCV_RVV_FUNC + +#endif // #ifndef AVCODEC_RISCV_VP9DSP_RISCV_H diff --git a/libavcodec/riscv/vp9dsp_init.c b/libavcodec/riscv/vp9dsp_init.c new file mode 100644 index 0000000000..69ab39004c --- /dev/null +++ b/libavcodec/riscv/vp9dsp_init.c @@ -0,0 +1,61 @@ +/* + * Copyright (c) 2024 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 Lervvr 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 + * Lervvr General Public License for more details. + * + * You should have received a copy of the GNU Lervvr 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/attributes.h" +#include "libavutil/cpu.h" +#include "libavutil/riscv/cpu.h" +#include "libavcodec/vp9dsp.h" +#include "vp9dsp.h" + +static av_cold void vp9dsp_intrapred_init_rvv(VP9DSPContext *dsp, int bpp) +{ + #if HAVE_RVV + int flags = av_get_cpu_flags(); + + if (bpp == 8 && flags & AV_CPU_FLAG_RVV_I64 && ff_get_rv_vlenb() >= 16) { + dsp->intra_pred[TX_8X8][DC_PRED] = ff_dc_8x8_rvv; + dsp->intra_pred[TX_8X8][LEFT_DC_PRED] = ff_dc_left_8x8_rvv; + dsp->intra_pred[TX_8X8][DC_127_PRED] = ff_dc_127_8x8_rvv; + dsp->intra_pred[TX_8X8][DC_128_PRED] = ff_dc_128_8x8_rvv; + dsp->intra_pred[TX_8X8][DC_129_PRED] = ff_dc_129_8x8_rvv; + dsp->intra_pred[TX_8X8][TOP_DC_PRED] = ff_dc_top_8x8_rvv; + } + + if (bpp == 8 && flags & AV_CPU_FLAG_RVV_I32 && ff_get_rv_vlenb() >= 16) { + dsp->intra_pred[TX_32X32][DC_PRED] = ff_dc_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_PRED] = ff_dc_16x16_rvv; + dsp->intra_pred[TX_32X32][LEFT_DC_PRED] = ff_dc_left_32x32_rvv; + dsp->intra_pred[TX_16X16][LEFT_DC_PRED] = ff_dc_left_16x16_rvv; + dsp->intra_pred[TX_32X32][DC_127_PRED] = ff_dc_127_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_127_PRED] = ff_dc_127_16x16_rvv; + dsp->intra_pred[TX_32X32][DC_128_PRED] = ff_dc_128_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_128_PRED] = ff_dc_128_16x16_rvv; + dsp->intra_pred[TX_32X32][DC_129_PRED] = ff_dc_129_32x32_rvv; + dsp->intra_pred[TX_16X16][DC_129_PRED] = ff_dc_129_16x16_rvv; + dsp->intra_pred[TX_32X32][TOP_DC_PRED] = ff_dc_top_32x32_rvv; + dsp->intra_pred[TX_16X16][TOP_DC_PRED] = ff_dc_top_16x16_rvv; + } + #endif +} + +av_cold void ff_vp9dsp_init_riscv(VP9DSPContext *dsp, int bpp, int bitexact) +{ + vp9dsp_intrapred_init_rvv(dsp, bpp); +} diff --git a/libavcodec/vp9dsp.c b/libavcodec/vp9dsp.c index d8ddf74d4f..967e6e1e1a 100644 --- a/libavcodec/vp9dsp.c +++ b/libavcodec/vp9dsp.c @@ -100,6 +100,8 @@ av_cold void ff_vp9dsp_init(VP9DSPContext *dsp, int bpp, int bitexact) ff_vp9dsp_init_aarch64(dsp, bpp); #elif ARCH_ARM ff_vp9dsp_init_arm(dsp, bpp); +#elif ARCH_RISCV + ff_vp9dsp_init_riscv(dsp, bpp, bitexact); #elif ARCH_X86 ff_vp9dsp_init_x86(dsp, bpp, bitexact); #elif ARCH_MIPS diff --git a/libavcodec/vp9dsp.h b/libavcodec/vp9dsp.h index be0ac0b181..772848e349 100644 --- a/libavcodec/vp9dsp.h +++ b/libavcodec/vp9dsp.h @@ -131,6 +131,7 @@ void ff_vp9dsp_init_12(VP9DSPContext *dsp); void ff_vp9dsp_init_aarch64(VP9DSPContext *dsp, int bpp); void ff_vp9dsp_init_arm(VP9DSPContext *dsp, int bpp); +void ff_vp9dsp_init_riscv(VP9DSPContext *dsp, int bpp, int bitexact); void ff_vp9dsp_init_x86(VP9DSPContext *dsp, int bpp, int bitexact); void ff_vp9dsp_init_mips(VP9DSPContext *dsp, int bpp); void ff_vp9dsp_init_loongarch(VP9DSPContext *dsp, int bpp);