| 1 | /* |
| 2 | * Copyright (c) 2009 Mans Rullgard <mans@mansr.com> |
| 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 | #ifndef AVUTIL_AVR32_INTREADWRITE_H |
| 22 | #define AVUTIL_AVR32_INTREADWRITE_H |
| 23 | |
| 24 | #include <stdint.h> |
| 25 | #include "config.h" |
| 26 | #include "libavutil/bswap.h" |
| 27 | |
| 28 | /* |
| 29 | * AVR32 does not support unaligned memory accesses, except for the AP |
| 30 | * series which suppports unaligned 32-bit loads and stores. 16-bit |
| 31 | * and 64-bit accesses must be aligned to 16 and 32 bits, respectively. |
| 32 | * This means we cannot use the byte-swapping load/store instructions |
| 33 | * here. |
| 34 | * |
| 35 | * For 16-bit, 24-bit, and (on UC series) 32-bit loads, we instead use |
| 36 | * the LDINS.B instruction, which gcc fails to utilise with the |
| 37 | * generic code. GCC also fails to use plain LD.W and ST.W even for |
| 38 | * AP processors, so we override the generic code. The 64-bit |
| 39 | * versions are improved by using our optimised 32-bit functions. |
| 40 | */ |
| 41 | |
| 42 | #define AV_RL16 AV_RL16 |
| 43 | static av_always_inline uint16_t AV_RL16(const void *p) |
| 44 | { |
| 45 | uint16_t v; |
| 46 | __asm__ ("ld.ub %0, %1 \n\t" |
| 47 | "ldins.b %0:l, %2 \n\t" |
| 48 | : "=&r"(v) |
| 49 | : "m"(*(const uint8_t*)p), "RKs12"(*((const uint8_t*)p+1))); |
| 50 | return v; |
| 51 | } |
| 52 | |
| 53 | #define AV_RB16 AV_RB16 |
| 54 | static av_always_inline uint16_t AV_RB16(const void *p) |
| 55 | { |
| 56 | uint16_t v; |
| 57 | __asm__ ("ld.ub %0, %2 \n\t" |
| 58 | "ldins.b %0:l, %1 \n\t" |
| 59 | : "=&r"(v) |
| 60 | : "RKs12"(*(const uint8_t*)p), "m"(*((const uint8_t*)p+1))); |
| 61 | return v; |
| 62 | } |
| 63 | |
| 64 | #define AV_RB24 AV_RB24 |
| 65 | static av_always_inline uint32_t AV_RB24(const void *p) |
| 66 | { |
| 67 | uint32_t v; |
| 68 | __asm__ ("ld.ub %0, %3 \n\t" |
| 69 | "ldins.b %0:l, %2 \n\t" |
| 70 | "ldins.b %0:u, %1 \n\t" |
| 71 | : "=&r"(v) |
| 72 | : "RKs12"(* (const uint8_t*)p), |
| 73 | "RKs12"(*((const uint8_t*)p+1)), |
| 74 | "m" (*((const uint8_t*)p+2))); |
| 75 | return v; |
| 76 | } |
| 77 | |
| 78 | #define AV_RL24 AV_RL24 |
| 79 | static av_always_inline uint32_t AV_RL24(const void *p) |
| 80 | { |
| 81 | uint32_t v; |
| 82 | __asm__ ("ld.ub %0, %1 \n\t" |
| 83 | "ldins.b %0:l, %2 \n\t" |
| 84 | "ldins.b %0:u, %3 \n\t" |
| 85 | : "=&r"(v) |
| 86 | : "m" (* (const uint8_t*)p), |
| 87 | "RKs12"(*((const uint8_t*)p+1)), |
| 88 | "RKs12"(*((const uint8_t*)p+2))); |
| 89 | return v; |
| 90 | } |
| 91 | |
| 92 | #if ARCH_AVR32_AP |
| 93 | |
| 94 | #define AV_RB32 AV_RB32 |
| 95 | static av_always_inline uint32_t AV_RB32(const void *p) |
| 96 | { |
| 97 | uint32_t v; |
| 98 | __asm__ ("ld.w %0, %1" : "=r"(v) : "m"(*(const uint32_t*)p)); |
| 99 | return v; |
| 100 | } |
| 101 | |
| 102 | #define AV_WB32 AV_WB32 |
| 103 | static av_always_inline void AV_WB32(void *p, uint32_t v) |
| 104 | { |
| 105 | __asm__ ("st.w %0, %1" : "=m"(*(uint32_t*)p) : "r"(v)); |
| 106 | } |
| 107 | |
| 108 | /* These two would be defined by generic code, but we need them sooner. */ |
| 109 | #define AV_RL32(p) av_bswap32(AV_RB32(p)) |
| 110 | #define AV_WL32(p, v) AV_WB32(p, av_bswap32(v)) |
| 111 | |
| 112 | #define AV_WB64 AV_WB64 |
| 113 | static av_always_inline void AV_WB64(void *p, uint64_t v) |
| 114 | { |
| 115 | union { uint64_t v; uint32_t hl[2]; } vv = { v }; |
| 116 | AV_WB32(p, vv.hl[0]); |
| 117 | AV_WB32((uint32_t*)p+1, vv.hl[1]); |
| 118 | } |
| 119 | |
| 120 | #define AV_WL64 AV_WL64 |
| 121 | static av_always_inline void AV_WL64(void *p, uint64_t v) |
| 122 | { |
| 123 | union { uint64_t v; uint32_t hl[2]; } vv = { v }; |
| 124 | AV_WL32(p, vv.hl[1]); |
| 125 | AV_WL32((uint32_t*)p+1, vv.hl[0]); |
| 126 | } |
| 127 | |
| 128 | #else /* ARCH_AVR32_AP */ |
| 129 | |
| 130 | #define AV_RB32 AV_RB32 |
| 131 | static av_always_inline uint32_t AV_RB32(const void *p) |
| 132 | { |
| 133 | uint32_t v; |
| 134 | __asm__ ("ld.ub %0, %4 \n\t" |
| 135 | "ldins.b %0:l, %3 \n\t" |
| 136 | "ldins.b %0:u, %2 \n\t" |
| 137 | "ldins.b %0:t, %1 \n\t" |
| 138 | : "=&r"(v) |
| 139 | : "RKs12"(* (const uint8_t*)p), |
| 140 | "RKs12"(*((const uint8_t*)p+1)), |
| 141 | "RKs12"(*((const uint8_t*)p+2)), |
| 142 | "m" (*((const uint8_t*)p+3))); |
| 143 | return v; |
| 144 | } |
| 145 | |
| 146 | #define AV_RL32 AV_RL32 |
| 147 | static av_always_inline uint32_t AV_RL32(const void *p) |
| 148 | { |
| 149 | uint32_t v; |
| 150 | __asm__ ("ld.ub %0, %1 \n\t" |
| 151 | "ldins.b %0:l, %2 \n\t" |
| 152 | "ldins.b %0:u, %3 \n\t" |
| 153 | "ldins.b %0:t, %4 \n\t" |
| 154 | : "=&r"(v) |
| 155 | : "m" (* (const uint8_t*)p), |
| 156 | "RKs12"(*((const uint8_t*)p+1)), |
| 157 | "RKs12"(*((const uint8_t*)p+2)), |
| 158 | "RKs12"(*((const uint8_t*)p+3))); |
| 159 | return v; |
| 160 | } |
| 161 | |
| 162 | #endif /* ARCH_AVR32_AP */ |
| 163 | |
| 164 | #define AV_RB64 AV_RB64 |
| 165 | static av_always_inline uint64_t AV_RB64(const void *p) |
| 166 | { |
| 167 | union { uint64_t v; uint32_t hl[2]; } v; |
| 168 | v.hl[0] = AV_RB32(p); |
| 169 | v.hl[1] = AV_RB32((const uint32_t*)p+1); |
| 170 | return v.v; |
| 171 | } |
| 172 | |
| 173 | #define AV_RL64 AV_RL64 |
| 174 | static av_always_inline uint64_t AV_RL64(const void *p) |
| 175 | { |
| 176 | union { uint64_t v; uint32_t hl[2]; } v; |
| 177 | v.hl[1] = AV_RL32(p); |
| 178 | v.hl[0] = AV_RL32((const uint32_t*)p+1); |
| 179 | return v.v; |
| 180 | } |
| 181 | |
| 182 | #endif /* AVUTIL_AVR32_INTREADWRITE_H */ |