Unified · Split
| @@ -16,6 +16,7 @@ RVV-OBJS-$(CONFIG_BLOCKDSP) += riscv/blockdsp_rvv.o · 15 unmodified lines | |||
| 16 | OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_init.o | 16 | OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_init.o |
| 17 | RV-OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_rvb.o | 17 | RV-OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_rvb.o |
| 18 | RVV-OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_rvv.o | 18 | RVV-OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_rvv.o |
| 19 | RVVB-OBJS-$(CONFIG_BSWAPDSP) += riscv/bswapdsp_rvvb.o | ||
| 19 | OBJS-$(CONFIG_EXR_DECODER) += riscv/exrdsp_init.o | 20 | OBJS-$(CONFIG_EXR_DECODER) += riscv/exrdsp_init.o |
| 20 | RVV-OBJS-$(CONFIG_EXR_DECODER) += riscv/exrdsp_rvv.o | 21 | RVV-OBJS-$(CONFIG_EXR_DECODER) += riscv/exrdsp_rvv.o |
| 21 | OBJS-$(CONFIG_FLAC_DECODER) += riscv/flacdsp_init.o | 22 | OBJS-$(CONFIG_FLAC_DECODER) += riscv/flacdsp_init.o |
| @@ -27,6 +27,8 @@ · 26 unmodified lines | |||
| 27 | 27 | ||
| 28 | void ff_bswap32_buf_rvb(uint32_t *dst, const uint32_t *src, int len); | 28 | void ff_bswap32_buf_rvb(uint32_t *dst, const uint32_t *src, int len); |
| 29 | void ff_bswap16_buf_rvv(uint16_t *dst, const uint16_t *src, int len); | 29 | void ff_bswap16_buf_rvv(uint16_t *dst, const uint16_t *src, int len); |
| 30 | void ff_bswap32_buf_rvvb(uint32_t *dst, const uint32_t *src, int len); | ||
| 31 | void ff_bswap16_buf_rvvb(uint16_t *dst, const uint16_t *src, int len); | ||
| 30 | 32 | ||
| 31 | av_cold void ff_bswapdsp_init_riscv(BswapDSPContext *c) | 33 | av_cold void ff_bswapdsp_init_riscv(BswapDSPContext *c) |
| 32 | { | 34 | { |
| @@ -38,9 +40,14 @@ av_cold void ff_bswapdsp_init_riscv(BswapDSPContext *c) · 5 unmodified lines | |||
| 38 | c->bswap_buf = ff_bswap32_buf_rvb; | 40 | c->bswap_buf = ff_bswap32_buf_rvb; |
| 39 | #endif | 41 | #endif |
| 40 | #if HAVE_RVV | 42 | #if HAVE_RVV |
| 41 | if (flags & AV_CPU_FLAG_RVB_ADDR) { | 43 | if ((flags & AV_CPU_FLAG_RVV_I32) && (flags & AV_CPU_FLAG_RVB)) { |
| 42 | if (flags & AV_CPU_FLAG_RVV_I32) | 44 | c->bswap16_buf = ff_bswap16_buf_rvv; |
| 43 | c->bswap16_buf = ff_bswap16_buf_rvv; | 45 | #if HAVE_RV_ZVBB |
| 46 | if (flags & AV_CPU_FLAG_RV_ZVBB) { | ||
| 47 | c->bswap_buf = ff_bswap32_buf_rvvb; | ||
| 48 | c->bswap16_buf = ff_bswap16_buf_rvvb; | ||
| 49 | } | ||
| 50 | #endif | ||
| 44 | } | 51 | } |
| 45 | #endif | 52 | #endif |
| 46 | #endif | 53 | #endif |
| @@ -0,0 +1,52 @@ | |||
| 1 | /* | ||
| 2 | * Copyright © 2024 Rémi Denis-Courmont. | ||
| 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 "config.h" | ||
| 22 | #include "libavutil/riscv/asm.S" | ||
| 23 | |||
| 24 | func ff_bswap32_buf_rvvb, zve32x, zvbb, zba | ||
| 25 | lpad 0 | ||
| 26 | 1: | ||
| 27 | vsetvli t0, a2, e32, m8, ta, ma | ||
| 28 | vle32.v v8, (a1) | ||
| 29 | sub a2, a2, t0 | ||
| 30 | vrev8.v v8, v8 | ||
| 31 | sh2add a1, t0, a1 | ||
| 32 | vse32.v v8, (a0) | ||
| 33 | sh2add a0, t0, a0 | ||
| 34 | bnez a2, 1b | ||
| 35 | |||
| 36 | ret | ||
| 37 | endfunc | ||
| 38 | |||
| 39 | func ff_bswap16_buf_rvvb, zve32x, zvbb, zba | ||
| 40 | lpad 0 | ||
| 41 | 1: | ||
| 42 | vsetvli t0, a2, e16, m8, ta, ma | ||
| 43 | vle16.v v8, (a1) | ||
| 44 | sub a2, a2, t0 | ||
| 45 | vrev8.v v8, v8 | ||
| 46 | sh1add a1, t0, a1 | ||
| 47 | vse16.v v8, (a0) | ||
| 48 | sh1add a0, t0, a0 | ||
| 49 | bnez a2, 1b | ||
| 50 | |||
| 51 | ret | ||
| 52 | endfunc | ||