C908: vc1dsp.vc1_inv_trans_4x4_dc_c: 125.7 vc1dsp.vc1_inv_trans_4x4_dc_rvv_i32: 53.5 vc1dsp.vc1_inv_trans_4x8_dc_c: 230.7 vc1dsp.vc1_inv_trans_4x8_dc_rvv_i32: 65.5 vc1dsp.vc1_inv_trans_8x4_dc_c: 228.7 vc1dsp.vc1_inv_trans_8x4_dc_rvv_i64: 64.5 vc1dsp.vc1_inv_trans_8x8_dc_c: 476.5 vc1dsp.vc1_inv_trans_8x8_dc_rvv_i64: 80.2
Unified · Split
| @@ -40,5 +40,7 @@ OBJS-$(CONFIG_PIXBLOCKDSP) += riscv/pixblockdsp_init.o \ · 39 unmodified lines | ||
| 40 | 40 | RVV-OBJS-$(CONFIG_PIXBLOCKDSP) += riscv/pixblockdsp_rvv.o |
| 41 | 41 | OBJS-$(CONFIG_UTVIDEO_DECODER) += riscv/utvideodsp_init.o |
| 42 | 42 | RVV-OBJS-$(CONFIG_UTVIDEO_DECODER) += riscv/utvideodsp_rvv.o |
| 43 | +OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_init.o | |
| 44 | +RVV-OBJS-$(CONFIG_VC1DSP) += riscv/vc1dsp_rvv.o | |
| 43 | 45 | OBJS-$(CONFIG_VORBIS_DECODER) += riscv/vorbisdsp_init.o |
| 44 | 46 | RVV-OBJS-$(CONFIG_VORBIS_DECODER) += riscv/vorbisdsp_rvv.o |
| @@ -0,0 +1,49 @@ | ||
| 1 | +/* | |
| 2 | + * Copyright (c) 2023 Institue of Software Chinese Academy of Sciences (ISCAS). | |
| 3 | + * | |
| 4 | + * This file is part of FFmpeg. | |
| 5 | + * | |
| 6 | + * FFmpeg is free software; you can redistribute it and/or | |
| 7 | + * modify it under the terms of the GNU Lesser General Public | |
| 8 | + * License as published by the Free Software Foundation; either | |
| 9 | + * version 2.1 of the License, or (at your option) any later version. | |
| 10 | + * | |
| 11 | + * FFmpeg is distributed in the hope that it will be useful, | |
| 12 | + * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 13 | + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 14 | + * Lesser General Public License for more details. | |
| 15 | + * | |
| 16 | + * You should have received a copy of the GNU Lesser General Public | |
| 17 | + * License along with FFmpeg; if not, write to the Free Software | |
| 18 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
| 19 | + */ | |
| 20 | + | |
| 21 | +#include <stdint.h> | |
| 22 | + | |
| 23 | +#include "libavutil/attributes.h" | |
| 24 | +#include "libavutil/cpu.h" | |
| 25 | +#include "libavutil/riscv/cpu.h" | |
| 26 | +#include "libavcodec/vc1.h" | |
| 27 | + | |
| 28 | +void ff_vc1_inv_trans_8x8_dc_rvv(uint8_t *dest, ptrdiff_t stride, int16_t *block); | |
| 29 | +void ff_vc1_inv_trans_4x8_dc_rvv(uint8_t *dest, ptrdiff_t stride, int16_t *block); | |
| 30 | +void ff_vc1_inv_trans_8x4_dc_rvv(uint8_t *dest, ptrdiff_t stride, int16_t *block); | |
| 31 | +void ff_vc1_inv_trans_4x4_dc_rvv(uint8_t *dest, ptrdiff_t stride, int16_t *block); | |
| 32 | + | |
| 33 | +av_cold void ff_vc1dsp_init_riscv(VC1DSPContext *dsp) | |
| 34 | +{ | |
| 35 | +#if HAVE_RVV | |
| 36 | + int flags = av_get_cpu_flags(); | |
| 37 | + | |
| 38 | + if (ff_get_rv_vlenb() >= 16) { | |
| 39 | + if (flags & AV_CPU_FLAG_RVV_I64) { | |
| 40 | + dsp->vc1_inv_trans_8x8_dc = ff_vc1_inv_trans_8x8_dc_rvv; | |
| 41 | + dsp->vc1_inv_trans_8x4_dc = ff_vc1_inv_trans_8x4_dc_rvv; | |
| 42 | + } | |
| 43 | + if (flags & AV_CPU_FLAG_RVV_I32) { | |
| 44 | + dsp->vc1_inv_trans_4x8_dc = ff_vc1_inv_trans_4x8_dc_rvv; | |
| 45 | + dsp->vc1_inv_trans_4x4_dc = ff_vc1_inv_trans_4x4_dc_rvv; | |
| 46 | + } | |
| 47 | + } | |
| 48 | +#endif | |
| 49 | +} | |
| @@ -0,0 +1,113 @@ | ||
| 1 | +/* | |
| 2 | + * Copyright (c) 2023 Institue of Software Chinese Academy of Sciences (ISCAS). | |
| 3 | + * | |
| 4 | + * This file is part of FFmpeg. | |
| 5 | + * | |
| 6 | + * FFmpeg is free software; you can redistribute it and/or | |
| 7 | + * modify it under the terms of the GNU Lesser General Public | |
| 8 | + * License as published by the Free Software Foundation; either | |
| 9 | + * version 2.1 of the License, or (at your option) any later version. | |
| 10 | + * | |
| 11 | + * FFmpeg is distributed in the hope that it will be useful, | |
| 12 | + * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 13 | + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 14 | + * Lesser General Public License for more details. | |
| 15 | + * | |
| 16 | + * You should have received a copy of the GNU Lesser General Public | |
| 17 | + * License along with FFmpeg; if not, write to the Free Software | |
| 18 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
| 19 | + */ | |
| 20 | + | |
| 21 | +#include "libavutil/riscv/asm.S" | |
| 22 | + | |
| 23 | +func ff_vc1_inv_trans_8x8_dc_rvv, zve64x | |
| 24 | + lh t2, (a2) | |
| 25 | + vsetivli zero, 8, e8, mf2, ta, ma | |
| 26 | + vlse64.v v0, (a0), a1 | |
| 27 | + sh1add t2, t2, t2 | |
| 28 | + addi t2, t2, 1 | |
| 29 | + srai t2, t2, 1 | |
| 30 | + sh1add t2, t2, t2 | |
| 31 | + addi t2, t2, 16 | |
| 32 | + srai t2, t2, 5 | |
| 33 | + li t0, 8*8 | |
| 34 | + vsetvli zero, t0, e16, m8, ta, ma | |
| 35 | + vzext.vf2 v8, v0 | |
| 36 | + vadd.vx v8, v8, t2 | |
| 37 | + vmax.vx v8, v8, zero | |
| 38 | + vsetvli zero, zero, e8, m4, ta, ma | |
| 39 | + vnclipu.wi v0, v8, 0 | |
| 40 | + vsetivli zero, 8, e8, mf2, ta, ma | |
| 41 | + vsse64.v v0, (a0), a1 | |
| 42 | + ret | |
| 43 | +endfunc | |
| 44 | + | |
| 45 | +func ff_vc1_inv_trans_4x8_dc_rvv, zve32x | |
| 46 | + lh t2, (a2) | |
| 47 | + vsetivli zero, 8, e8, mf2, ta, ma | |
| 48 | + vlse32.v v0, (a0), a1 | |
| 49 | + slli t1, t2, 4 | |
| 50 | + add t2, t2, t1 | |
| 51 | + addi t2, t2, 4 | |
| 52 | + srai t2, t2, 3 | |
| 53 | + sh1add t2, t2, t2 | |
| 54 | + slli t2, t2, 2 | |
| 55 | + addi t2, t2, 64 | |
| 56 | + srai t2, t2, 7 | |
| 57 | + li t0, 4*8 | |
| 58 | + vsetvli zero, t0, e16, m4, ta, ma | |
| 59 | + vzext.vf2 v4, v0 | |
| 60 | + vadd.vx v4, v4, t2 | |
| 61 | + vmax.vx v4, v4, zero | |
| 62 | + vsetvli zero, zero, e8, m2, ta, ma | |
| 63 | + vnclipu.wi v0, v4, 0 | |
| 64 | + vsetivli zero, 8, e8, mf2, ta, ma | |
| 65 | + vsse32.v v0, (a0), a1 | |
| 66 | + ret | |
| 67 | +endfunc | |
| 68 | + | |
| 69 | +func ff_vc1_inv_trans_8x4_dc_rvv, zve64x | |
| 70 | + lh t2, (a2) | |
| 71 | + vsetivli zero, 8, e8, mf2, ta, ma | |
| 72 | + vlse64.v v0, (a0), a1 | |
| 73 | + sh1add t2, t2, t2 | |
| 74 | + addi t2, t2, 1 | |
| 75 | + srai t2, t2, 1 | |
| 76 | + slli t1, t2, 4 | |
| 77 | + add t2, t2, t1 | |
| 78 | + addi t2, t2, 64 | |
| 79 | + srai t2, t2, 7 | |
| 80 | + li t0, 8*4 | |
| 81 | + vsetvli zero, t0, e16, m4, ta, ma | |
| 82 | + vzext.vf2 v4, v0 | |
| 83 | + vadd.vx v4, v4, t2 | |
| 84 | + vmax.vx v4, v4, zero | |
| 85 | + vsetvli zero, zero, e8, m2, ta, ma | |
| 86 | + vnclipu.wi v0, v4, 0 | |
| 87 | + vsetivli zero, 8, e8, mf2, ta, ma | |
| 88 | + vsse64.v v0, (a0), a1 | |
| 89 | + ret | |
| 90 | +endfunc | |
| 91 | + | |
| 92 | +func ff_vc1_inv_trans_4x4_dc_rvv, zve32x | |
| 93 | + lh t2, (a2) | |
| 94 | + vsetivli zero, 4, e8, mf2, ta, ma | |
| 95 | + vlse32.v v0, (a0), a1 | |
| 96 | + slli t1, t2, 4 | |
| 97 | + add t2, t2, t1 | |
| 98 | + addi t2, t2, 4 | |
| 99 | + srai t2, t2, 3 | |
| 100 | + slli t1, t2, 4 | |
| 101 | + add t2, t2, t1 | |
| 102 | + addi t2, t2, 64 | |
| 103 | + srai t2, t2, 7 | |
| 104 | + vsetivli zero, 4*4, e16, m2, ta, ma | |
| 105 | + vzext.vf2 v2, v0 | |
| 106 | + vadd.vx v2, v2, t2 | |
| 107 | + vmax.vx v2, v2, zero | |
| 108 | + vsetvli zero, zero, e8, m1, ta, ma | |
| 109 | + vnclipu.wi v0, v2, 0 | |
| 110 | + vsetivli zero, 4, e8, mf2, ta, ma | |
| 111 | + vsse32.v v0, (a0), a1 | |
| 112 | + ret | |
| 113 | +endfunc | |
| @@ -1039,6 +1039,8 @@ av_cold void ff_vc1dsp_init(VC1DSPContext *dsp) · 1038 unmodified lines | ||
| 1039 | 1039 | ff_vc1dsp_init_arm(dsp); |
| 1040 | 1040 | #elif ARCH_PPC |
| 1041 | 1041 | ff_vc1dsp_init_ppc(dsp); |
| 1042 | +#elif ARCH_RISCV | |
| 1043 | + ff_vc1dsp_init_riscv(dsp); | |
| 1042 | 1044 | #elif ARCH_X86 |
| 1043 | 1045 | ff_vc1dsp_init_x86(dsp); |
| 1044 | 1046 | #elif ARCH_MIPS |
| @@ -89,6 +89,7 @@ void ff_vc1dsp_init(VC1DSPContext* c); · 88 unmodified lines | ||
| 89 | 89 | void ff_vc1dsp_init_aarch64(VC1DSPContext* dsp); |
| 90 | 90 | void ff_vc1dsp_init_arm(VC1DSPContext* dsp); |
| 91 | 91 | void ff_vc1dsp_init_ppc(VC1DSPContext *c); |
| 92 | +void ff_vc1dsp_init_riscv(VC1DSPContext *c); | |
| 92 | 93 | void ff_vc1dsp_init_x86(VC1DSPContext* dsp); |
| 93 | 94 | void ff_vc1dsp_init_mips(VC1DSPContext* dsp); |
| 94 | 95 | void ff_vc1dsp_init_loongarch(VC1DSPContext* dsp); |