| 1 | /* |
| 2 | * ARM generic helpers for various arithmetical operations. |
| 3 | * |
| 4 | * This code is licensed under the GNU GPL v2 or later. |
| 5 | * |
| 6 | * SPDX-License-Identifier: GPL-2.0-or-later |
| 7 | */ |
| 8 | #include "qemu/osdep.h" |
| 9 | #include "qemu/bswap.h" |
| 10 | #include "qemu/crc32c.h" |
| 11 | #include "helper.h" |
| 12 | #include <zlib.h> /* for crc32 */ |
| 13 | |
| 14 | /* |
| 15 | * Note that signed overflow is undefined in C. The following routines are |
| 16 | * careful to use unsigned types where modulo arithmetic is required. |
| 17 | * Failure to do so _will_ break on newer gcc. |
| 18 | */ |
| 19 | |
| 20 | /* Signed saturating arithmetic. */ |
| 21 | |
| 22 | /* Perform 16-bit signed saturating addition. */ |
| 23 | static inline uint16_t add16_sat(uint16_t a, uint16_t b) |
| 24 | { |
| 25 | uint16_t res; |
| 26 | |
| 27 | res = a + b; |
| 28 | if (((res ^ a) & 0x8000) && !((a ^ b) & 0x8000)) { |
| 29 | if (a & 0x8000) { |
| 30 | res = 0x8000; |
| 31 | } else { |
| 32 | res = 0x7fff; |
| 33 | } |
| 34 | } |
| 35 | return res; |
| 36 | } |
| 37 | |
| 38 | /* Perform 8-bit signed saturating addition. */ |
| 39 | static inline uint8_t add8_sat(uint8_t a, uint8_t b) |
| 40 | { |
| 41 | uint8_t res; |
| 42 | |
| 43 | res = a + b; |
| 44 | if (((res ^ a) & 0x80) && !((a ^ b) & 0x80)) { |
| 45 | if (a & 0x80) { |
| 46 | res = 0x80; |
| 47 | } else { |
| 48 | res = 0x7f; |
| 49 | } |
| 50 | } |
| 51 | return res; |
| 52 | } |
| 53 | |
| 54 | /* Perform 16-bit signed saturating subtraction. */ |
| 55 | static inline uint16_t sub16_sat(uint16_t a, uint16_t b) |
| 56 | { |
| 57 | uint16_t res; |
| 58 | |
| 59 | res = a - b; |
| 60 | if (((res ^ a) & 0x8000) && ((a ^ b) & 0x8000)) { |
| 61 | if (a & 0x8000) { |
| 62 | res = 0x8000; |
| 63 | } else { |
| 64 | res = 0x7fff; |
| 65 | } |
| 66 | } |
| 67 | return res; |
| 68 | } |
| 69 | |
| 70 | /* Perform 8-bit signed saturating subtraction. */ |
| 71 | static inline uint8_t sub8_sat(uint8_t a, uint8_t b) |
| 72 | { |
| 73 | uint8_t res; |
| 74 | |
| 75 | res = a - b; |
| 76 | if (((res ^ a) & 0x80) && ((a ^ b) & 0x80)) { |
| 77 | if (a & 0x80) { |
| 78 | res = 0x80; |
| 79 | } else { |
| 80 | res = 0x7f; |
| 81 | } |
| 82 | } |
| 83 | return res; |
| 84 | } |
| 85 | |
| 86 | #define ADD16(a, b, n) RESULT(add16_sat(a, b), n, 16); |
| 87 | #define SUB16(a, b, n) RESULT(sub16_sat(a, b), n, 16); |
| 88 | #define ADD8(a, b, n) RESULT(add8_sat(a, b), n, 8); |
| 89 | #define SUB8(a, b, n) RESULT(sub8_sat(a, b), n, 8); |
| 90 | #define PFX q |
| 91 | |
| 92 | #include "op_addsub.c.inc" |
| 93 | |
| 94 | /* Unsigned saturating arithmetic. */ |
| 95 | static inline uint16_t add16_usat(uint16_t a, uint16_t b) |
| 96 | { |
| 97 | uint16_t res; |
| 98 | res = a + b; |
| 99 | if (res < a) { |
| 100 | res = 0xffff; |
| 101 | } |
| 102 | return res; |
| 103 | } |
| 104 | |
| 105 | static inline uint16_t sub16_usat(uint16_t a, uint16_t b) |
| 106 | { |
| 107 | if (a > b) { |
| 108 | return a - b; |
| 109 | } else { |
| 110 | return 0; |
| 111 | } |
| 112 | } |
| 113 | |
| 114 | static inline uint8_t add8_usat(uint8_t a, uint8_t b) |
| 115 | { |
| 116 | uint8_t res; |
| 117 | res = a + b; |
| 118 | if (res < a) { |
| 119 | res = 0xff; |
| 120 | } |
| 121 | return res; |
| 122 | } |
| 123 | |
| 124 | static inline uint8_t sub8_usat(uint8_t a, uint8_t b) |
| 125 | { |
| 126 | if (a > b) { |
| 127 | return a - b; |
| 128 | } else { |
| 129 | return 0; |
| 130 | } |
| 131 | } |
| 132 | |
| 133 | #define ADD16(a, b, n) RESULT(add16_usat(a, b), n, 16); |
| 134 | #define SUB16(a, b, n) RESULT(sub16_usat(a, b), n, 16); |
| 135 | #define ADD8(a, b, n) RESULT(add8_usat(a, b), n, 8); |
| 136 | #define SUB8(a, b, n) RESULT(sub8_usat(a, b), n, 8); |
| 137 | #define PFX uq |
| 138 | |
| 139 | #include "op_addsub.c.inc" |
| 140 | |
| 141 | /* Signed modulo arithmetic. */ |
| 142 | #define SARITH16(a, b, n, op) do { \ |
| 143 | int32_t sum; \ |
| 144 | sum = (int32_t)(int16_t)(a) op (int32_t)(int16_t)(b); \ |
| 145 | RESULT(sum, n, 16); \ |
| 146 | if (sum >= 0) \ |
| 147 | ge |= 3 << (n * 2); \ |
| 148 | } while (0) |
| 149 | |
| 150 | #define SARITH8(a, b, n, op) do { \ |
| 151 | int32_t sum; \ |
| 152 | sum = (int32_t)(int8_t)(a) op (int32_t)(int8_t)(b); \ |
| 153 | RESULT(sum, n, 8); \ |
| 154 | if (sum >= 0) \ |
| 155 | ge |= 1 << n; \ |
| 156 | } while (0) |
| 157 | |
| 158 | |
| 159 | #define ADD16(a, b, n) SARITH16(a, b, n, +) |
| 160 | #define SUB16(a, b, n) SARITH16(a, b, n, -) |
| 161 | #define ADD8(a, b, n) SARITH8(a, b, n, +) |
| 162 | #define SUB8(a, b, n) SARITH8(a, b, n, -) |
| 163 | #define PFX s |
| 164 | #define ARITH_GE |
| 165 | |
| 166 | #include "op_addsub.c.inc" |
| 167 | |
| 168 | /* Unsigned modulo arithmetic. */ |
| 169 | #define ADD16(a, b, n) do { \ |
| 170 | uint32_t sum; \ |
| 171 | sum = (uint32_t)(uint16_t)(a) + (uint32_t)(uint16_t)(b); \ |
| 172 | RESULT(sum, n, 16); \ |
| 173 | if ((sum >> 16) == 1) \ |
| 174 | ge |= 3 << (n * 2); \ |
| 175 | } while (0) |
| 176 | |
| 177 | #define ADD8(a, b, n) do { \ |
| 178 | uint32_t sum; \ |
| 179 | sum = (uint32_t)(uint8_t)(a) + (uint32_t)(uint8_t)(b); \ |
| 180 | RESULT(sum, n, 8); \ |
| 181 | if ((sum >> 8) == 1) \ |
| 182 | ge |= 1 << n; \ |
| 183 | } while (0) |
| 184 | |
| 185 | #define SUB16(a, b, n) do { \ |
| 186 | uint32_t sum; \ |
| 187 | sum = (uint32_t)(uint16_t)(a) - (uint32_t)(uint16_t)(b); \ |
| 188 | RESULT(sum, n, 16); \ |
| 189 | if ((sum >> 16) == 0) \ |
| 190 | ge |= 3 << (n * 2); \ |
| 191 | } while (0) |
| 192 | |
| 193 | #define SUB8(a, b, n) do { \ |
| 194 | uint32_t sum; \ |
| 195 | sum = (uint32_t)(uint8_t)(a) - (uint32_t)(uint8_t)(b); \ |
| 196 | RESULT(sum, n, 8); \ |
| 197 | if ((sum >> 8) == 0) \ |
| 198 | ge |= 1 << n; \ |
| 199 | } while (0) |
| 200 | |
| 201 | #define PFX u |
| 202 | #define ARITH_GE |
| 203 | |
| 204 | #include "op_addsub.c.inc" |
| 205 | |
| 206 | /* Halved signed arithmetic. */ |
| 207 | #define ADD16(a, b, n) \ |
| 208 | RESULT(((int32_t)(int16_t)(a) + (int32_t)(int16_t)(b)) >> 1, n, 16) |
| 209 | #define SUB16(a, b, n) \ |
| 210 | RESULT(((int32_t)(int16_t)(a) - (int32_t)(int16_t)(b)) >> 1, n, 16) |
| 211 | #define ADD8(a, b, n) \ |
| 212 | RESULT(((int32_t)(int8_t)(a) + (int32_t)(int8_t)(b)) >> 1, n, 8) |
| 213 | #define SUB8(a, b, n) \ |
| 214 | RESULT(((int32_t)(int8_t)(a) - (int32_t)(int8_t)(b)) >> 1, n, 8) |
| 215 | #define PFX sh |
| 216 | |
| 217 | #include "op_addsub.c.inc" |
| 218 | |
| 219 | /* Halved unsigned arithmetic. */ |
| 220 | #define ADD16(a, b, n) \ |
| 221 | RESULT(((uint32_t)(uint16_t)(a) + (uint32_t)(uint16_t)(b)) >> 1, n, 16) |
| 222 | #define SUB16(a, b, n) \ |
| 223 | RESULT(((uint32_t)(uint16_t)(a) - (uint32_t)(uint16_t)(b)) >> 1, n, 16) |
| 224 | #define ADD8(a, b, n) \ |
| 225 | RESULT(((uint32_t)(uint8_t)(a) + (uint32_t)(uint8_t)(b)) >> 1, n, 8) |
| 226 | #define SUB8(a, b, n) \ |
| 227 | RESULT(((uint32_t)(uint8_t)(a) - (uint32_t)(uint8_t)(b)) >> 1, n, 8) |
| 228 | #define PFX uh |
| 229 | |
| 230 | #include "op_addsub.c.inc" |
| 231 | |
| 232 | static inline uint8_t do_usad(uint8_t a, uint8_t b) |
| 233 | { |
| 234 | if (a > b) { |
| 235 | return a - b; |
| 236 | } else { |
| 237 | return b - a; |
| 238 | } |
| 239 | } |
| 240 | |
| 241 | /* Unsigned sum of absolute byte differences. */ |
| 242 | uint32_t HELPER(usad8)(uint32_t a, uint32_t b) |
| 243 | { |
| 244 | uint32_t sum; |
| 245 | sum = do_usad(a, b); |
| 246 | sum += do_usad(a >> 8, b >> 8); |
| 247 | sum += do_usad(a >> 16, b >> 16); |
| 248 | sum += do_usad(a >> 24, b >> 24); |
| 249 | return sum; |
| 250 | } |
| 251 | |
| 252 | /* For ARMv6 SEL instruction. */ |
| 253 | uint32_t HELPER(sel_flags)(uint32_t flags, uint32_t a, uint32_t b) |
| 254 | { |
| 255 | uint32_t mask; |
| 256 | |
| 257 | mask = 0; |
| 258 | if (flags & 1) { |
| 259 | mask |= 0xff; |
| 260 | } |
| 261 | if (flags & 2) { |
| 262 | mask |= 0xff00; |
| 263 | } |
| 264 | if (flags & 4) { |
| 265 | mask |= 0xff0000; |
| 266 | } |
| 267 | if (flags & 8) { |
| 268 | mask |= 0xff000000; |
| 269 | } |
| 270 | return (a & mask) | (b & ~mask); |
| 271 | } |
| 272 | |
| 273 | /* |
| 274 | * CRC helpers. |
| 275 | * The upper bytes of val (above the number specified by 'bytes') must have |
| 276 | * been zeroed out by the caller. |
| 277 | */ |
| 278 | uint32_t HELPER(crc32)(uint32_t acc, uint32_t val, uint32_t bytes) |
| 279 | { |
| 280 | uint8_t buf[4]; |
| 281 | |
| 282 | stl_le_p(buf, val); |
| 283 | |
| 284 | /* zlib crc32 converts the accumulator and output to one's complement. */ |
| 285 | return crc32(acc ^ 0xffffffff, buf, bytes) ^ 0xffffffff; |
| 286 | } |
| 287 | |
| 288 | uint32_t HELPER(crc32c)(uint32_t acc, uint32_t val, uint32_t bytes) |
| 289 | { |
| 290 | uint8_t buf[4]; |
| 291 | |
| 292 | stl_le_p(buf, val); |
| 293 | |
| 294 | /* Linux crc32c converts the output to one's complement. */ |
| 295 | return crc32c(acc, buf, bytes) ^ 0xffffffff; |
| 296 | } |