Message ID | tencent_872A8605CA104F71ED1C05D671D81C4FB206@qq.com |
---|---|
State | New |
Headers | show |
Series | [FFmpeg-devel,01/10] lavc/vp8dsp: R-V put_vp8_pixels | expand |
Context | Check | Description |
---|---|---|
andriy/make_x86 | success | Make finished |
andriy/make_fate_x86 | success | Make fate finished |
Made these changes according to the previous review: moved func into macro, added macro vset to reduce if else, used rvi, supplemented __riscv_xlen <uk7b@foxmail.com> 于2024年5月6日周一 00:45写道: > From: sunyuechi <sunyuechi@iscas.ac.cn> > > C908: > vp8_put_pixels4_c: 78.0 > vp8_put_pixels4_rvi: 33.7 > vp8_put_pixels8_c: 278.0 > vp8_put_pixels8_rvi: 55.0 > vp8_put_pixels16_c: 999.0 > vp8_put_pixels16_rvi: 86.7 > --- > libavcodec/riscv/Makefile | 1 + > libavcodec/riscv/vp8dsp.h | 75 ++++++++++++++++++++++++++++++++++ > libavcodec/riscv/vp8dsp_init.c | 22 ++++++++++ > libavcodec/riscv/vp8dsp_rvi.S | 61 +++++++++++++++++++++++++++ > libavcodec/vp8dsp.c | 2 + > libavcodec/vp8dsp.h | 1 + > 6 files changed, 162 insertions(+) > create mode 100644 libavcodec/riscv/vp8dsp.h > create mode 100644 libavcodec/riscv/vp8dsp_rvi.S > > diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile > index 050c08ee61..526cb5c97c 100644 > --- a/libavcodec/riscv/Makefile > +++ b/libavcodec/riscv/Makefile > @@ -61,6 +61,7 @@ RVV-OBJS-$(CONFIG_UTVIDEO_DECODER) += > riscv/utvideodsp_rvv.o > OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_init.o > RVV-OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_rvv.o > OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_init.o > +RV-OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_rvi.o > RVV-OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_rvv.o > OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9dsp_init.o > RVV-OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9_intra_rvv.o > diff --git a/libavcodec/riscv/vp8dsp.h b/libavcodec/riscv/vp8dsp.h > new file mode 100644 > index 0000000000..971c5c0a96 > --- /dev/null > +++ b/libavcodec/riscv/vp8dsp.h > @@ -0,0 +1,75 @@ > +/* > + * 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_VP8DSP_H > +#define AVCODEC_RISCV_VP8DSP_H > + > +#include "libavcodec/vp8dsp.h" > + > +#define VP8_LF_Y(hv, inner, opt) > \ > + void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst, > \ > + ptrdiff_t stride, > \ > + int flim_E, int > flim_I, \ > + int hev_thresh) > + > +#define VP8_LF_UV(hv, inner, opt) > \ > + void ff_vp8_##hv##_loop_filter8uv##inner##_##opt(uint8_t *dstU, > \ > + uint8_t *dstV, > \ > + ptrdiff_t stride, > \ > + int flim_E, int > flim_I, \ > + int hev_thresh) > + > +#define VP8_LF_SIMPLE(hv, opt) \ > + void ff_vp8_##hv##_loop_filter16_simple_##opt(uint8_t *dst, \ > + ptrdiff_t stride, \ > + int flim) > + > +#define VP8_LF_HV(inner, opt) \ > + VP8_LF_Y(h, inner, opt); \ > + VP8_LF_Y(v, inner, opt); \ > + VP8_LF_UV(h, inner, opt); \ > + VP8_LF_UV(v, inner, opt) > + > +#define VP8_LF(opt) \ > + VP8_LF_HV(, opt); \ > + VP8_LF_HV(_inner, opt); \ > + VP8_LF_SIMPLE(h, opt); \ > + VP8_LF_SIMPLE(v, opt) > + > +#define VP8_MC(n, opt) \ > + void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride, \ > + const uint8_t *src, ptrdiff_t srcstride,\ > + int h, int x, int y) > + > +#define VP8_EPEL(w, opt) \ > + VP8_MC(pixels ## w, opt); \ > + VP8_MC(epel ## w ## _h4, opt); \ > + VP8_MC(epel ## w ## _h6, opt); \ > + VP8_MC(epel ## w ## _v4, opt); \ > + VP8_MC(epel ## w ## _h4v4, opt); \ > + VP8_MC(epel ## w ## _h6v4, opt); \ > + VP8_MC(epel ## w ## _v6, opt); \ > + VP8_MC(epel ## w ## _h4v6, opt); \ > + VP8_MC(epel ## w ## _h6v6, opt) > + > +#define VP8_BILIN(w, opt) \ > + VP8_MC(bilin ## w ## _h, opt); \ > + VP8_MC(bilin ## w ## _v, opt); \ > + VP8_MC(bilin ## w ## _hv, opt) > + > +#endif /* AVCODEC_RISCV_VP8DSP_H */ > diff --git a/libavcodec/riscv/vp8dsp_init.c > b/libavcodec/riscv/vp8dsp_init.c > index af57aabb71..fa3feeacf7 100644 > --- a/libavcodec/riscv/vp8dsp_init.c > +++ b/libavcodec/riscv/vp8dsp_init.c > @@ -24,11 +24,33 @@ > #include "libavutil/cpu.h" > #include "libavutil/riscv/cpu.h" > #include "libavcodec/vp8dsp.h" > +#include "vp8dsp.h" > > void ff_vp8_idct_dc_add_rvv(uint8_t *dst, int16_t block[16], ptrdiff_t > stride); > void ff_vp8_idct_dc_add4y_rvv(uint8_t *dst, int16_t block[4][16], > ptrdiff_t stride); > void ff_vp8_idct_dc_add4uv_rvv(uint8_t *dst, int16_t block[4][16], > ptrdiff_t stride); > > +VP8_EPEL(16, rvi); > +VP8_EPEL(8, rvi); > +VP8_EPEL(4, rvi); > + > +av_cold void ff_vp78dsp_init_riscv(VP8DSPContext *c) > +{ > +#if HAVE_RV > + int flags = av_get_cpu_flags(); > + if (flags & AV_CPU_FLAG_RVI) { > +#if __riscv_xlen >= 64 > + c->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_rvi; > + c->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_rvi; > + c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_rvi; > + c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_rvi; > +#endif > + c->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_rvi; > + c->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_rvi; > + } > +#endif > +} > + > av_cold void ff_vp8dsp_init_riscv(VP8DSPContext *c) > { > #if HAVE_RVV > diff --git a/libavcodec/riscv/vp8dsp_rvi.S b/libavcodec/riscv/vp8dsp_rvi.S > new file mode 100644 > index 0000000000..50ba4f293f > --- /dev/null > +++ b/libavcodec/riscv/vp8dsp_rvi.S > @@ -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 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" > + > +#if __riscv_xlen >= 64 > +func ff_put_vp8_pixels16_rvi > +1: > + addi a4, a4, -1 > + ld t0, (a2) > + ld t1, 8(a2) > + sd t0, (a0) > + sd t1, 8(a0) > + add a2, a2, a3 > + add a0, a0, a1 > + bnez a4, 1b > + > + ret > +endfunc > + > +func ff_put_vp8_pixels8_rvi > +1: > + addi a4, a4, -1 > + ld t0, (a2) > + sd t0, (a0) > + add a2, a2, a3 > + add a0, a0, a1 > + bnez a4, 1b > + > + ret > +endfunc > +#endif > + > +func ff_put_vp8_pixels4_rvi > +1: > + addi a4, a4, -1 > + lw t0, (a2) > + sw t0, (a0) > + add a2, a2, a3 > + add a0, a0, a1 > + bnez a4, 1b > + > + ret > +endfunc > diff --git a/libavcodec/vp8dsp.c b/libavcodec/vp8dsp.c > index df7bd12424..f7c9c9899c 100644 > --- a/libavcodec/vp8dsp.c > +++ b/libavcodec/vp8dsp.c > @@ -1402,6 +1402,8 @@ dsp->put_vp8_epel_pixels_tab[2][2][2] = > put_vp8_epel4_h6v6_c; > ff_vp78dsp_init_arm(dsp); > #elif ARCH_PPC > ff_vp78dsp_init_ppc(dsp); > +#elif ARCH_RISCV > + ff_vp78dsp_init_riscv(dsp); > #elif ARCH_X86 > ff_vp78dsp_init_x86(dsp); > #endif > diff --git a/libavcodec/vp8dsp.h b/libavcodec/vp8dsp.h > index 30dc2c6cc1..3bf12b6b45 100644 > --- a/libavcodec/vp8dsp.h > +++ b/libavcodec/vp8dsp.h > @@ -87,6 +87,7 @@ void ff_vp78dsp_init(VP8DSPContext *c); > void ff_vp78dsp_init_aarch64(VP8DSPContext *c); > void ff_vp78dsp_init_arm(VP8DSPContext *c); > void ff_vp78dsp_init_ppc(VP8DSPContext *c); > +void ff_vp78dsp_init_riscv(VP8DSPContext *c); > void ff_vp78dsp_init_x86(VP8DSPContext *c); > > void ff_vp8dsp_init(VP8DSPContext *c); > -- > 2.45.0 > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe". >
diff --git a/libavcodec/riscv/Makefile b/libavcodec/riscv/Makefile index 050c08ee61..526cb5c97c 100644 --- a/libavcodec/riscv/Makefile +++ b/libavcodec/riscv/Makefile @@ -61,6 +61,7 @@ RVV-OBJS-$(CONFIG_UTVIDEO_DECODER) += riscv/utvideodsp_rvv.o OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_init.o RVV-OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_rvv.o OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_init.o +RV-OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_rvi.o RVV-OBJS-$(CONFIG_VP8DSP) += riscv/vp8dsp_rvv.o OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9dsp_init.o RVV-OBJS-$(CONFIG_VP9_DECODER) += riscv/vp9_intra_rvv.o diff --git a/libavcodec/riscv/vp8dsp.h b/libavcodec/riscv/vp8dsp.h new file mode 100644 index 0000000000..971c5c0a96 --- /dev/null +++ b/libavcodec/riscv/vp8dsp.h @@ -0,0 +1,75 @@ +/* + * 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_VP8DSP_H +#define AVCODEC_RISCV_VP8DSP_H + +#include "libavcodec/vp8dsp.h" + +#define VP8_LF_Y(hv, inner, opt) \ + void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst, \ + ptrdiff_t stride, \ + int flim_E, int flim_I, \ + int hev_thresh) + +#define VP8_LF_UV(hv, inner, opt) \ + void ff_vp8_##hv##_loop_filter8uv##inner##_##opt(uint8_t *dstU, \ + uint8_t *dstV, \ + ptrdiff_t stride, \ + int flim_E, int flim_I, \ + int hev_thresh) + +#define VP8_LF_SIMPLE(hv, opt) \ + void ff_vp8_##hv##_loop_filter16_simple_##opt(uint8_t *dst, \ + ptrdiff_t stride, \ + int flim) + +#define VP8_LF_HV(inner, opt) \ + VP8_LF_Y(h, inner, opt); \ + VP8_LF_Y(v, inner, opt); \ + VP8_LF_UV(h, inner, opt); \ + VP8_LF_UV(v, inner, opt) + +#define VP8_LF(opt) \ + VP8_LF_HV(, opt); \ + VP8_LF_HV(_inner, opt); \ + VP8_LF_SIMPLE(h, opt); \ + VP8_LF_SIMPLE(v, opt) + +#define VP8_MC(n, opt) \ + void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride, \ + const uint8_t *src, ptrdiff_t srcstride,\ + int h, int x, int y) + +#define VP8_EPEL(w, opt) \ + VP8_MC(pixels ## w, opt); \ + VP8_MC(epel ## w ## _h4, opt); \ + VP8_MC(epel ## w ## _h6, opt); \ + VP8_MC(epel ## w ## _v4, opt); \ + VP8_MC(epel ## w ## _h4v4, opt); \ + VP8_MC(epel ## w ## _h6v4, opt); \ + VP8_MC(epel ## w ## _v6, opt); \ + VP8_MC(epel ## w ## _h4v6, opt); \ + VP8_MC(epel ## w ## _h6v6, opt) + +#define VP8_BILIN(w, opt) \ + VP8_MC(bilin ## w ## _h, opt); \ + VP8_MC(bilin ## w ## _v, opt); \ + VP8_MC(bilin ## w ## _hv, opt) + +#endif /* AVCODEC_RISCV_VP8DSP_H */ diff --git a/libavcodec/riscv/vp8dsp_init.c b/libavcodec/riscv/vp8dsp_init.c index af57aabb71..fa3feeacf7 100644 --- a/libavcodec/riscv/vp8dsp_init.c +++ b/libavcodec/riscv/vp8dsp_init.c @@ -24,11 +24,33 @@ #include "libavutil/cpu.h" #include "libavutil/riscv/cpu.h" #include "libavcodec/vp8dsp.h" +#include "vp8dsp.h" void ff_vp8_idct_dc_add_rvv(uint8_t *dst, int16_t block[16], ptrdiff_t stride); void ff_vp8_idct_dc_add4y_rvv(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride); void ff_vp8_idct_dc_add4uv_rvv(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride); +VP8_EPEL(16, rvi); +VP8_EPEL(8, rvi); +VP8_EPEL(4, rvi); + +av_cold void ff_vp78dsp_init_riscv(VP8DSPContext *c) +{ +#if HAVE_RV + int flags = av_get_cpu_flags(); + if (flags & AV_CPU_FLAG_RVI) { +#if __riscv_xlen >= 64 + c->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_rvi; + c->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_rvi; + c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_rvi; + c->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_rvi; +#endif + c->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_rvi; + c->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_rvi; + } +#endif +} + av_cold void ff_vp8dsp_init_riscv(VP8DSPContext *c) { #if HAVE_RVV diff --git a/libavcodec/riscv/vp8dsp_rvi.S b/libavcodec/riscv/vp8dsp_rvi.S new file mode 100644 index 0000000000..50ba4f293f --- /dev/null +++ b/libavcodec/riscv/vp8dsp_rvi.S @@ -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 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" + +#if __riscv_xlen >= 64 +func ff_put_vp8_pixels16_rvi +1: + addi a4, a4, -1 + ld t0, (a2) + ld t1, 8(a2) + sd t0, (a0) + sd t1, 8(a0) + add a2, a2, a3 + add a0, a0, a1 + bnez a4, 1b + + ret +endfunc + +func ff_put_vp8_pixels8_rvi +1: + addi a4, a4, -1 + ld t0, (a2) + sd t0, (a0) + add a2, a2, a3 + add a0, a0, a1 + bnez a4, 1b + + ret +endfunc +#endif + +func ff_put_vp8_pixels4_rvi +1: + addi a4, a4, -1 + lw t0, (a2) + sw t0, (a0) + add a2, a2, a3 + add a0, a0, a1 + bnez a4, 1b + + ret +endfunc diff --git a/libavcodec/vp8dsp.c b/libavcodec/vp8dsp.c index df7bd12424..f7c9c9899c 100644 --- a/libavcodec/vp8dsp.c +++ b/libavcodec/vp8dsp.c @@ -1402,6 +1402,8 @@ dsp->put_vp8_epel_pixels_tab[2][2][2] = put_vp8_epel4_h6v6_c; ff_vp78dsp_init_arm(dsp); #elif ARCH_PPC ff_vp78dsp_init_ppc(dsp); +#elif ARCH_RISCV + ff_vp78dsp_init_riscv(dsp); #elif ARCH_X86 ff_vp78dsp_init_x86(dsp); #endif diff --git a/libavcodec/vp8dsp.h b/libavcodec/vp8dsp.h index 30dc2c6cc1..3bf12b6b45 100644 --- a/libavcodec/vp8dsp.h +++ b/libavcodec/vp8dsp.h @@ -87,6 +87,7 @@ void ff_vp78dsp_init(VP8DSPContext *c); void ff_vp78dsp_init_aarch64(VP8DSPContext *c); void ff_vp78dsp_init_arm(VP8DSPContext *c); void ff_vp78dsp_init_ppc(VP8DSPContext *c); +void ff_vp78dsp_init_riscv(VP8DSPContext *c); void ff_vp78dsp_init_x86(VP8DSPContext *c); void ff_vp8dsp_init(VP8DSPContext *c);
From: sunyuechi <sunyuechi@iscas.ac.cn> C908: vp8_put_pixels4_c: 78.0 vp8_put_pixels4_rvi: 33.7 vp8_put_pixels8_c: 278.0 vp8_put_pixels8_rvi: 55.0 vp8_put_pixels16_c: 999.0 vp8_put_pixels16_rvi: 86.7 --- libavcodec/riscv/Makefile | 1 + libavcodec/riscv/vp8dsp.h | 75 ++++++++++++++++++++++++++++++++++ libavcodec/riscv/vp8dsp_init.c | 22 ++++++++++ libavcodec/riscv/vp8dsp_rvi.S | 61 +++++++++++++++++++++++++++ libavcodec/vp8dsp.c | 2 + libavcodec/vp8dsp.h | 1 + 6 files changed, 162 insertions(+) create mode 100644 libavcodec/riscv/vp8dsp.h create mode 100644 libavcodec/riscv/vp8dsp_rvi.S