| 1 | /* |
| 2 | * QEMU TCG support -- s390x vector integer instruction support |
| 3 | * |
| 4 | * Copyright (C) 2019 Red Hat Inc |
| 5 | * |
| 6 | * Authors: |
| 7 | * David Hildenbrand <david@redhat.com> |
| 8 | * |
| 9 | * This work is licensed under the terms of the GNU GPL, version 2 or later. |
| 10 | * See the COPYING file in the top-level directory. |
| 11 | */ |
| 12 | #include "qemu/osdep.h" |
| 13 | #include "cpu.h" |
| 14 | #include "vec.h" |
| 15 | #include "exec/helper-proto.h" |
| 16 | #include "tcg/tcg-gvec-desc.h" |
| 17 | #include "crypto/clmul.h" |
| 18 | |
| 19 | static bool s390_vec_is_zero(const S390Vector *v) |
| 20 | { |
| 21 | return !v->doubleword[0] && !v->doubleword[1]; |
| 22 | } |
| 23 | |
| 24 | static void s390_vec_and(S390Vector *res, const S390Vector *a, |
| 25 | const S390Vector *b) |
| 26 | { |
| 27 | res->doubleword[0] = a->doubleword[0] & b->doubleword[0]; |
| 28 | res->doubleword[1] = a->doubleword[1] & b->doubleword[1]; |
| 29 | } |
| 30 | |
| 31 | static bool s390_vec_equal(const S390Vector *a, const S390Vector *b) |
| 32 | { |
| 33 | return a->doubleword[0] == b->doubleword[0] && |
| 34 | a->doubleword[1] == b->doubleword[1]; |
| 35 | } |
| 36 | |
| 37 | static void s390_vec_shl(S390Vector *d, const S390Vector *a, uint64_t count) |
| 38 | { |
| 39 | uint64_t tmp; |
| 40 | |
| 41 | g_assert(count < 128); |
| 42 | if (count == 0) { |
| 43 | d->doubleword[0] = a->doubleword[0]; |
| 44 | d->doubleword[1] = a->doubleword[1]; |
| 45 | } else if (count == 64) { |
| 46 | d->doubleword[0] = a->doubleword[1]; |
| 47 | d->doubleword[1] = 0; |
| 48 | } else if (count < 64) { |
| 49 | tmp = extract64(a->doubleword[1], 64 - count, count); |
| 50 | d->doubleword[1] = a->doubleword[1] << count; |
| 51 | d->doubleword[0] = (a->doubleword[0] << count) | tmp; |
| 52 | } else { |
| 53 | d->doubleword[0] = a->doubleword[1] << (count - 64); |
| 54 | d->doubleword[1] = 0; |
| 55 | } |
| 56 | } |
| 57 | |
| 58 | static void s390_vec_sar(S390Vector *d, const S390Vector *a, uint64_t count) |
| 59 | { |
| 60 | uint64_t tmp; |
| 61 | |
| 62 | if (count == 0) { |
| 63 | d->doubleword[0] = a->doubleword[0]; |
| 64 | d->doubleword[1] = a->doubleword[1]; |
| 65 | } else if (count == 64) { |
| 66 | tmp = (int64_t)a->doubleword[0] >> 63; |
| 67 | d->doubleword[1] = a->doubleword[0]; |
| 68 | d->doubleword[0] = tmp; |
| 69 | } else if (count < 64) { |
| 70 | tmp = a->doubleword[1] >> count; |
| 71 | d->doubleword[1] = deposit64(tmp, 64 - count, count, a->doubleword[0]); |
| 72 | d->doubleword[0] = (int64_t)a->doubleword[0] >> count; |
| 73 | } else { |
| 74 | tmp = (int64_t)a->doubleword[0] >> 63; |
| 75 | d->doubleword[1] = (int64_t)a->doubleword[0] >> (count - 64); |
| 76 | d->doubleword[0] = tmp; |
| 77 | } |
| 78 | } |
| 79 | |
| 80 | static void s390_vec_shr(S390Vector *d, const S390Vector *a, uint64_t count) |
| 81 | { |
| 82 | uint64_t tmp; |
| 83 | |
| 84 | g_assert(count < 128); |
| 85 | if (count == 0) { |
| 86 | d->doubleword[0] = a->doubleword[0]; |
| 87 | d->doubleword[1] = a->doubleword[1]; |
| 88 | } else if (count == 64) { |
| 89 | d->doubleword[1] = a->doubleword[0]; |
| 90 | d->doubleword[0] = 0; |
| 91 | } else if (count < 64) { |
| 92 | tmp = a->doubleword[1] >> count; |
| 93 | d->doubleword[1] = deposit64(tmp, 64 - count, count, a->doubleword[0]); |
| 94 | d->doubleword[0] = a->doubleword[0] >> count; |
| 95 | } else { |
| 96 | d->doubleword[1] = a->doubleword[0] >> (count - 64); |
| 97 | d->doubleword[0] = 0; |
| 98 | } |
| 99 | } |
| 100 | #define DEF_VAVG(BITS) \ |
| 101 | void HELPER(gvec_vavg##BITS)(void *v1, const void *v2, const void *v3, \ |
| 102 | uint32_t desc) \ |
| 103 | { \ |
| 104 | int i; \ |
| 105 | \ |
| 106 | for (i = 0; i < (128 / BITS); i++) { \ |
| 107 | const int32_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, i); \ |
| 108 | const int32_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, i); \ |
| 109 | \ |
| 110 | s390_vec_write_element##BITS(v1, i, (a + b + 1) >> 1); \ |
| 111 | } \ |
| 112 | } |
| 113 | DEF_VAVG(8) |
| 114 | DEF_VAVG(16) |
| 115 | |
| 116 | #define DEF_VAVGL(BITS) \ |
| 117 | void HELPER(gvec_vavgl##BITS)(void *v1, const void *v2, const void *v3, \ |
| 118 | uint32_t desc) \ |
| 119 | { \ |
| 120 | int i; \ |
| 121 | \ |
| 122 | for (i = 0; i < (128 / BITS); i++) { \ |
| 123 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 124 | const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ |
| 125 | \ |
| 126 | s390_vec_write_element##BITS(v1, i, (a + b + 1) >> 1); \ |
| 127 | } \ |
| 128 | } |
| 129 | DEF_VAVGL(8) |
| 130 | DEF_VAVGL(16) |
| 131 | |
| 132 | #define DEF_VCLZ(BITS) \ |
| 133 | void HELPER(gvec_vclz##BITS)(void *v1, const void *v2, uint32_t desc) \ |
| 134 | { \ |
| 135 | int i; \ |
| 136 | \ |
| 137 | for (i = 0; i < (128 / BITS); i++) { \ |
| 138 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 139 | \ |
| 140 | s390_vec_write_element##BITS(v1, i, clz32(a) - 32 + BITS); \ |
| 141 | } \ |
| 142 | } |
| 143 | DEF_VCLZ(8) |
| 144 | DEF_VCLZ(16) |
| 145 | |
| 146 | #define DEF_VCTZ(BITS) \ |
| 147 | void HELPER(gvec_vctz##BITS)(void *v1, const void *v2, uint32_t desc) \ |
| 148 | { \ |
| 149 | int i; \ |
| 150 | \ |
| 151 | for (i = 0; i < (128 / BITS); i++) { \ |
| 152 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 153 | \ |
| 154 | s390_vec_write_element##BITS(v1, i, a ? ctz32(a) : BITS); \ |
| 155 | } \ |
| 156 | } |
| 157 | DEF_VCTZ(8) |
| 158 | DEF_VCTZ(16) |
| 159 | |
| 160 | /* like binary multiplication, but XOR instead of addition */ |
| 161 | |
| 162 | /* |
| 163 | * There is no carry across the two doublewords, so their order does |
| 164 | * not matter. Nor is there partial overlap between registers. |
| 165 | */ |
| 166 | static inline uint64_t do_gfma8(uint64_t n, uint64_t m, uint64_t a) |
| 167 | { |
| 168 | return clmul_8x4_even(n, m) ^ clmul_8x4_odd(n, m) ^ a; |
| 169 | } |
| 170 | |
| 171 | void HELPER(gvec_vgfm8)(void *v1, const void *v2, const void *v3, uint32_t d) |
| 172 | { |
| 173 | uint64_t *q1 = v1; |
| 174 | const uint64_t *q2 = v2, *q3 = v3; |
| 175 | |
| 176 | q1[0] = do_gfma8(q2[0], q3[0], 0); |
| 177 | q1[1] = do_gfma8(q2[1], q3[1], 0); |
| 178 | } |
| 179 | |
| 180 | void HELPER(gvec_vgfma8)(void *v1, const void *v2, const void *v3, |
| 181 | const void *v4, uint32_t desc) |
| 182 | { |
| 183 | uint64_t *q1 = v1; |
| 184 | const uint64_t *q2 = v2, *q3 = v3, *q4 = v4; |
| 185 | |
| 186 | q1[0] = do_gfma8(q2[0], q3[0], q4[0]); |
| 187 | q1[1] = do_gfma8(q2[1], q3[1], q4[1]); |
| 188 | } |
| 189 | |
| 190 | static inline uint64_t do_gfma16(uint64_t n, uint64_t m, uint64_t a) |
| 191 | { |
| 192 | return clmul_16x2_even(n, m) ^ clmul_16x2_odd(n, m) ^ a; |
| 193 | } |
| 194 | |
| 195 | void HELPER(gvec_vgfm16)(void *v1, const void *v2, const void *v3, uint32_t d) |
| 196 | { |
| 197 | uint64_t *q1 = v1; |
| 198 | const uint64_t *q2 = v2, *q3 = v3; |
| 199 | |
| 200 | q1[0] = do_gfma16(q2[0], q3[0], 0); |
| 201 | q1[1] = do_gfma16(q2[1], q3[1], 0); |
| 202 | } |
| 203 | |
| 204 | void HELPER(gvec_vgfma16)(void *v1, const void *v2, const void *v3, |
| 205 | const void *v4, uint32_t d) |
| 206 | { |
| 207 | uint64_t *q1 = v1; |
| 208 | const uint64_t *q2 = v2, *q3 = v3, *q4 = v4; |
| 209 | |
| 210 | q1[0] = do_gfma16(q2[0], q3[0], q4[0]); |
| 211 | q1[1] = do_gfma16(q2[1], q3[1], q4[1]); |
| 212 | } |
| 213 | |
| 214 | static inline uint64_t do_gfma32(uint64_t n, uint64_t m, uint64_t a) |
| 215 | { |
| 216 | return clmul_32(n, m) ^ clmul_32(n >> 32, m >> 32) ^ a; |
| 217 | } |
| 218 | |
| 219 | void HELPER(gvec_vgfm32)(void *v1, const void *v2, const void *v3, uint32_t d) |
| 220 | { |
| 221 | uint64_t *q1 = v1; |
| 222 | const uint64_t *q2 = v2, *q3 = v3; |
| 223 | |
| 224 | q1[0] = do_gfma32(q2[0], q3[0], 0); |
| 225 | q1[1] = do_gfma32(q2[1], q3[1], 0); |
| 226 | } |
| 227 | |
| 228 | void HELPER(gvec_vgfma32)(void *v1, const void *v2, const void *v3, |
| 229 | const void *v4, uint32_t d) |
| 230 | { |
| 231 | uint64_t *q1 = v1; |
| 232 | const uint64_t *q2 = v2, *q3 = v3, *q4 = v4; |
| 233 | |
| 234 | q1[0] = do_gfma32(q2[0], q3[0], q4[0]); |
| 235 | q1[1] = do_gfma32(q2[1], q3[1], q4[1]); |
| 236 | } |
| 237 | |
| 238 | void HELPER(gvec_vgfm64)(void *v1, const void *v2, const void *v3, |
| 239 | uint32_t desc) |
| 240 | { |
| 241 | uint64_t *q1 = v1; |
| 242 | const uint64_t *q2 = v2, *q3 = v3; |
| 243 | Int128 r; |
| 244 | |
| 245 | r = int128_xor(clmul_64(q2[0], q3[0]), clmul_64(q2[1], q3[1])); |
| 246 | q1[0] = int128_gethi(r); |
| 247 | q1[1] = int128_getlo(r); |
| 248 | } |
| 249 | |
| 250 | void HELPER(gvec_vgfma64)(void *v1, const void *v2, const void *v3, |
| 251 | const void *v4, uint32_t desc) |
| 252 | { |
| 253 | uint64_t *q1 = v1; |
| 254 | const uint64_t *q2 = v2, *q3 = v3, *q4 = v4; |
| 255 | Int128 r; |
| 256 | |
| 257 | r = int128_xor(clmul_64(q2[0], q3[0]), clmul_64(q2[1], q3[1])); |
| 258 | q1[0] = q4[0] ^ int128_gethi(r); |
| 259 | q1[1] = q4[1] ^ int128_getlo(r); |
| 260 | } |
| 261 | |
| 262 | #define DEF_VMAL(BITS) \ |
| 263 | void HELPER(gvec_vmal##BITS)(void *v1, const void *v2, const void *v3, \ |
| 264 | const void *v4, uint32_t desc) \ |
| 265 | { \ |
| 266 | int i; \ |
| 267 | \ |
| 268 | for (i = 0; i < (128 / BITS); i++) { \ |
| 269 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 270 | const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ |
| 271 | const uint##BITS##_t c = s390_vec_read_element##BITS(v4, i); \ |
| 272 | \ |
| 273 | s390_vec_write_element##BITS(v1, i, a * b + c); \ |
| 274 | } \ |
| 275 | } |
| 276 | DEF_VMAL(8) |
| 277 | DEF_VMAL(16) |
| 278 | |
| 279 | #define DEF_VMAH(BITS) \ |
| 280 | void HELPER(gvec_vmah##BITS)(void *v1, const void *v2, const void *v3, \ |
| 281 | const void *v4, uint32_t desc) \ |
| 282 | { \ |
| 283 | int i; \ |
| 284 | \ |
| 285 | for (i = 0; i < (128 / BITS); i++) { \ |
| 286 | const int32_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, i); \ |
| 287 | const int32_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, i); \ |
| 288 | const int32_t c = (int##BITS##_t)s390_vec_read_element##BITS(v4, i); \ |
| 289 | \ |
| 290 | s390_vec_write_element##BITS(v1, i, (a * b + c) >> BITS); \ |
| 291 | } \ |
| 292 | } |
| 293 | DEF_VMAH(8) |
| 294 | DEF_VMAH(16) |
| 295 | |
| 296 | #define DEF_VMALH(BITS) \ |
| 297 | void HELPER(gvec_vmalh##BITS)(void *v1, const void *v2, const void *v3, \ |
| 298 | const void *v4, uint32_t desc) \ |
| 299 | { \ |
| 300 | int i; \ |
| 301 | \ |
| 302 | for (i = 0; i < (128 / BITS); i++) { \ |
| 303 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 304 | const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ |
| 305 | const uint##BITS##_t c = s390_vec_read_element##BITS(v4, i); \ |
| 306 | \ |
| 307 | s390_vec_write_element##BITS(v1, i, (a * b + c) >> BITS); \ |
| 308 | } \ |
| 309 | } |
| 310 | DEF_VMALH(8) |
| 311 | DEF_VMALH(16) |
| 312 | |
| 313 | #define DEF_VMAE(BITS, TBITS) \ |
| 314 | void HELPER(gvec_vmae##BITS)(void *v1, const void *v2, const void *v3, \ |
| 315 | const void *v4, uint32_t desc) \ |
| 316 | { \ |
| 317 | int i, j; \ |
| 318 | \ |
| 319 | for (i = 0, j = 0; i < (128 / TBITS); i++, j += 2) { \ |
| 320 | int##TBITS##_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, j); \ |
| 321 | int##TBITS##_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, j); \ |
| 322 | int##TBITS##_t c = s390_vec_read_element##TBITS(v4, i); \ |
| 323 | \ |
| 324 | s390_vec_write_element##TBITS(v1, i, a * b + c); \ |
| 325 | } \ |
| 326 | } |
| 327 | DEF_VMAE(8, 16) |
| 328 | DEF_VMAE(16, 32) |
| 329 | DEF_VMAE(32, 64) |
| 330 | |
| 331 | #define DEF_VMALE(BITS, TBITS) \ |
| 332 | void HELPER(gvec_vmale##BITS)(void *v1, const void *v2, const void *v3, \ |
| 333 | const void *v4, uint32_t desc) \ |
| 334 | { \ |
| 335 | int i, j; \ |
| 336 | \ |
| 337 | for (i = 0, j = 0; i < (128 / TBITS); i++, j += 2) { \ |
| 338 | uint##TBITS##_t a = s390_vec_read_element##BITS(v2, j); \ |
| 339 | uint##TBITS##_t b = s390_vec_read_element##BITS(v3, j); \ |
| 340 | uint##TBITS##_t c = s390_vec_read_element##TBITS(v4, i); \ |
| 341 | \ |
| 342 | s390_vec_write_element##TBITS(v1, i, a * b + c); \ |
| 343 | } \ |
| 344 | } |
| 345 | DEF_VMALE(8, 16) |
| 346 | DEF_VMALE(16, 32) |
| 347 | DEF_VMALE(32, 64) |
| 348 | |
| 349 | #define DEF_VMAO(BITS, TBITS) \ |
| 350 | void HELPER(gvec_vmao##BITS)(void *v1, const void *v2, const void *v3, \ |
| 351 | const void *v4, uint32_t desc) \ |
| 352 | { \ |
| 353 | int i, j; \ |
| 354 | \ |
| 355 | for (i = 0, j = 1; i < (128 / TBITS); i++, j += 2) { \ |
| 356 | int##TBITS##_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, j); \ |
| 357 | int##TBITS##_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, j); \ |
| 358 | int##TBITS##_t c = s390_vec_read_element##TBITS(v4, i); \ |
| 359 | \ |
| 360 | s390_vec_write_element##TBITS(v1, i, a * b + c); \ |
| 361 | } \ |
| 362 | } |
| 363 | DEF_VMAO(8, 16) |
| 364 | DEF_VMAO(16, 32) |
| 365 | DEF_VMAO(32, 64) |
| 366 | |
| 367 | #define DEF_VMALO(BITS, TBITS) \ |
| 368 | void HELPER(gvec_vmalo##BITS)(void *v1, const void *v2, const void *v3, \ |
| 369 | const void *v4, uint32_t desc) \ |
| 370 | { \ |
| 371 | int i, j; \ |
| 372 | \ |
| 373 | for (i = 0, j = 1; i < (128 / TBITS); i++, j += 2) { \ |
| 374 | uint##TBITS##_t a = s390_vec_read_element##BITS(v2, j); \ |
| 375 | uint##TBITS##_t b = s390_vec_read_element##BITS(v3, j); \ |
| 376 | uint##TBITS##_t c = s390_vec_read_element##TBITS(v4, i); \ |
| 377 | \ |
| 378 | s390_vec_write_element##TBITS(v1, i, a * b + c); \ |
| 379 | } \ |
| 380 | } |
| 381 | DEF_VMALO(8, 16) |
| 382 | DEF_VMALO(16, 32) |
| 383 | DEF_VMALO(32, 64) |
| 384 | |
| 385 | #define DEF_VMH(BITS) \ |
| 386 | void HELPER(gvec_vmh##BITS)(void *v1, const void *v2, const void *v3, \ |
| 387 | uint32_t desc) \ |
| 388 | { \ |
| 389 | int i; \ |
| 390 | \ |
| 391 | for (i = 0; i < (128 / BITS); i++) { \ |
| 392 | const int32_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, i); \ |
| 393 | const int32_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, i); \ |
| 394 | \ |
| 395 | s390_vec_write_element##BITS(v1, i, (a * b) >> BITS); \ |
| 396 | } \ |
| 397 | } |
| 398 | DEF_VMH(8) |
| 399 | DEF_VMH(16) |
| 400 | |
| 401 | #define DEF_VMLH(BITS) \ |
| 402 | void HELPER(gvec_vmlh##BITS)(void *v1, const void *v2, const void *v3, \ |
| 403 | uint32_t desc) \ |
| 404 | { \ |
| 405 | int i; \ |
| 406 | \ |
| 407 | for (i = 0; i < (128 / BITS); i++) { \ |
| 408 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 409 | const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ |
| 410 | \ |
| 411 | s390_vec_write_element##BITS(v1, i, (a * b) >> BITS); \ |
| 412 | } \ |
| 413 | } |
| 414 | DEF_VMLH(8) |
| 415 | DEF_VMLH(16) |
| 416 | |
| 417 | #define DEF_VME(BITS, TBITS) \ |
| 418 | void HELPER(gvec_vme##BITS)(void *v1, const void *v2, const void *v3, \ |
| 419 | uint32_t desc) \ |
| 420 | { \ |
| 421 | int i, j; \ |
| 422 | \ |
| 423 | for (i = 0, j = 0; i < (128 / TBITS); i++, j += 2) { \ |
| 424 | int##TBITS##_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, j); \ |
| 425 | int##TBITS##_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, j); \ |
| 426 | \ |
| 427 | s390_vec_write_element##TBITS(v1, i, a * b); \ |
| 428 | } \ |
| 429 | } |
| 430 | DEF_VME(8, 16) |
| 431 | DEF_VME(16, 32) |
| 432 | DEF_VME(32, 64) |
| 433 | |
| 434 | #define DEF_VMLE(BITS, TBITS) \ |
| 435 | void HELPER(gvec_vmle##BITS)(void *v1, const void *v2, const void *v3, \ |
| 436 | uint32_t desc) \ |
| 437 | { \ |
| 438 | int i, j; \ |
| 439 | \ |
| 440 | for (i = 0, j = 0; i < (128 / TBITS); i++, j += 2) { \ |
| 441 | const uint##TBITS##_t a = s390_vec_read_element##BITS(v2, j); \ |
| 442 | const uint##TBITS##_t b = s390_vec_read_element##BITS(v3, j); \ |
| 443 | \ |
| 444 | s390_vec_write_element##TBITS(v1, i, a * b); \ |
| 445 | } \ |
| 446 | } |
| 447 | DEF_VMLE(8, 16) |
| 448 | DEF_VMLE(16, 32) |
| 449 | DEF_VMLE(32, 64) |
| 450 | |
| 451 | #define DEF_VMO(BITS, TBITS) \ |
| 452 | void HELPER(gvec_vmo##BITS)(void *v1, const void *v2, const void *v3, \ |
| 453 | uint32_t desc) \ |
| 454 | { \ |
| 455 | int i, j; \ |
| 456 | \ |
| 457 | for (i = 0, j = 1; i < (128 / TBITS); i++, j += 2) { \ |
| 458 | int##TBITS##_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, j); \ |
| 459 | int##TBITS##_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, j); \ |
| 460 | \ |
| 461 | s390_vec_write_element##TBITS(v1, i, a * b); \ |
| 462 | } \ |
| 463 | } |
| 464 | DEF_VMO(8, 16) |
| 465 | DEF_VMO(16, 32) |
| 466 | DEF_VMO(32, 64) |
| 467 | |
| 468 | #define DEF_VMLO(BITS, TBITS) \ |
| 469 | void HELPER(gvec_vmlo##BITS)(void *v1, const void *v2, const void *v3, \ |
| 470 | uint32_t desc) \ |
| 471 | { \ |
| 472 | int i, j; \ |
| 473 | \ |
| 474 | for (i = 0, j = 1; i < (128 / TBITS); i++, j += 2) { \ |
| 475 | const uint##TBITS##_t a = s390_vec_read_element##BITS(v2, j); \ |
| 476 | const uint##TBITS##_t b = s390_vec_read_element##BITS(v3, j); \ |
| 477 | \ |
| 478 | s390_vec_write_element##TBITS(v1, i, a * b); \ |
| 479 | } \ |
| 480 | } |
| 481 | DEF_VMLO(8, 16) |
| 482 | DEF_VMLO(16, 32) |
| 483 | DEF_VMLO(32, 64) |
| 484 | |
| 485 | #define DEF_VPOPCT(BITS) \ |
| 486 | void HELPER(gvec_vpopct##BITS)(void *v1, const void *v2, uint32_t desc) \ |
| 487 | { \ |
| 488 | int i; \ |
| 489 | \ |
| 490 | for (i = 0; i < (128 / BITS); i++) { \ |
| 491 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 492 | \ |
| 493 | s390_vec_write_element##BITS(v1, i, ctpop32(a)); \ |
| 494 | } \ |
| 495 | } |
| 496 | DEF_VPOPCT(8) |
| 497 | DEF_VPOPCT(16) |
| 498 | |
| 499 | #define DEF_VERIM(BITS) \ |
| 500 | void HELPER(gvec_verim##BITS)(void *v1, const void *v2, const void *v3, \ |
| 501 | uint32_t desc) \ |
| 502 | { \ |
| 503 | const uint8_t count = simd_data(desc); \ |
| 504 | int i; \ |
| 505 | \ |
| 506 | for (i = 0; i < (128 / BITS); i++) { \ |
| 507 | const uint##BITS##_t a = s390_vec_read_element##BITS(v1, i); \ |
| 508 | const uint##BITS##_t b = s390_vec_read_element##BITS(v2, i); \ |
| 509 | const uint##BITS##_t mask = s390_vec_read_element##BITS(v3, i); \ |
| 510 | const uint##BITS##_t d = (a & ~mask) | (rol##BITS(b, count) & mask); \ |
| 511 | \ |
| 512 | s390_vec_write_element##BITS(v1, i, d); \ |
| 513 | } \ |
| 514 | } |
| 515 | DEF_VERIM(8) |
| 516 | DEF_VERIM(16) |
| 517 | |
| 518 | void HELPER(gvec_vsl)(void *v1, const void *v2, uint64_t count, |
| 519 | uint32_t desc) |
| 520 | { |
| 521 | s390_vec_shl(v1, v2, count); |
| 522 | } |
| 523 | |
| 524 | void HELPER(gvec_vsl_ve2)(void *v1, const void *v2, const void *v3, |
| 525 | uint32_t desc) |
| 526 | { |
| 527 | S390Vector tmp; |
| 528 | uint32_t sh, e0, e1 = 0; |
| 529 | int i; |
| 530 | |
| 531 | for (i = 15; i >= 0; --i, e1 = e0) { |
| 532 | e0 = s390_vec_read_element8(v2, i); |
| 533 | sh = s390_vec_read_element8(v3, i) & 7; |
| 534 | |
| 535 | s390_vec_write_element8(&tmp, i, rol32(e0 | (e1 << 24), sh)); |
| 536 | } |
| 537 | |
| 538 | *(S390Vector *)v1 = tmp; |
| 539 | } |
| 540 | |
| 541 | void HELPER(gvec_vsra)(void *v1, const void *v2, uint64_t count, |
| 542 | uint32_t desc) |
| 543 | { |
| 544 | s390_vec_sar(v1, v2, count); |
| 545 | } |
| 546 | |
| 547 | void HELPER(gvec_vsra_ve2)(void *v1, const void *v2, const void *v3, |
| 548 | uint32_t desc) |
| 549 | { |
| 550 | S390Vector tmp; |
| 551 | uint32_t sh, e0, e1 = 0; |
| 552 | int i = 0; |
| 553 | |
| 554 | /* Byte 0 is special only. */ |
| 555 | e0 = (int32_t)(int8_t)s390_vec_read_element8(v2, i); |
| 556 | sh = s390_vec_read_element8(v3, i) & 7; |
| 557 | s390_vec_write_element8(&tmp, i, e0 >> sh); |
| 558 | |
| 559 | e1 = e0; |
| 560 | for (i = 1; i < 16; ++i, e1 = e0) { |
| 561 | e0 = s390_vec_read_element8(v2, i); |
| 562 | sh = s390_vec_read_element8(v3, i) & 7; |
| 563 | s390_vec_write_element8(&tmp, i, (e0 | e1 << 8) >> sh); |
| 564 | } |
| 565 | |
| 566 | *(S390Vector *)v1 = tmp; |
| 567 | } |
| 568 | |
| 569 | void HELPER(gvec_vsrl)(void *v1, const void *v2, uint64_t count, |
| 570 | uint32_t desc) |
| 571 | { |
| 572 | s390_vec_shr(v1, v2, count); |
| 573 | } |
| 574 | |
| 575 | void HELPER(gvec_vsrl_ve2)(void *v1, const void *v2, const void *v3, |
| 576 | uint32_t desc) |
| 577 | { |
| 578 | S390Vector tmp; |
| 579 | uint32_t sh, e0, e1 = 0; |
| 580 | |
| 581 | for (int i = 0; i < 16; ++i, e1 = e0) { |
| 582 | e0 = s390_vec_read_element8(v2, i); |
| 583 | sh = s390_vec_read_element8(v3, i) & 7; |
| 584 | |
| 585 | s390_vec_write_element8(&tmp, i, (e0 | (e1 << 8)) >> sh); |
| 586 | } |
| 587 | |
| 588 | *(S390Vector *)v1 = tmp; |
| 589 | } |
| 590 | |
| 591 | #define DEF_VSCBI(BITS) \ |
| 592 | void HELPER(gvec_vscbi##BITS)(void *v1, const void *v2, const void *v3, \ |
| 593 | uint32_t desc) \ |
| 594 | { \ |
| 595 | int i; \ |
| 596 | \ |
| 597 | for (i = 0; i < (128 / BITS); i++) { \ |
| 598 | const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ |
| 599 | const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ |
| 600 | \ |
| 601 | s390_vec_write_element##BITS(v1, i, a >= b); \ |
| 602 | } \ |
| 603 | } |
| 604 | DEF_VSCBI(8) |
| 605 | DEF_VSCBI(16) |
| 606 | |
| 607 | void HELPER(gvec_vtm)(void *v1, const void *v2, CPUS390XState *env, |
| 608 | uint32_t desc) |
| 609 | { |
| 610 | S390Vector tmp; |
| 611 | |
| 612 | s390_vec_and(&tmp, v1, v2); |
| 613 | if (s390_vec_is_zero(&tmp)) { |
| 614 | /* Selected bits all zeros; or all mask bits zero */ |
| 615 | env->cc_op = 0; |
| 616 | } else if (s390_vec_equal(&tmp, v2)) { |
| 617 | /* Selected bits all ones */ |
| 618 | env->cc_op = 3; |
| 619 | } else { |
| 620 | /* Selected bits a mix of zeros and ones */ |
| 621 | env->cc_op = 1; |
| 622 | } |
| 623 | } |