1/* 2 * translate/vmx-impl.c 3 * 4 * Altivec/VMX translation 5 */ 6 7/*** Altivec vector extension ***/ 8/* Altivec registers moves */ 9 10static inline TCGv_ptr gen_avr_ptr(int reg) 11{ 12 TCGv_ptr r = tcg_temp_new_ptr(); 13 tcg_gen_addi_ptr(r, cpu_env, avr_full_offset(reg)); 14 return r; 15} 16 17#define GEN_VR_LDX(name, opc2, opc3) \ 18static void glue(gen_, name)(DisasContext *ctx) \ 19{ \ 20 TCGv EA; \ 21 TCGv_i64 avr; \ 22 if (unlikely(!ctx->altivec_enabled)) { \ 23 gen_exception(ctx, POWERPC_EXCP_VPU); \ 24 return; \ 25 } \ 26 gen_set_access_type(ctx, ACCESS_INT); \ 27 avr = tcg_temp_new_i64(); \ 28 EA = tcg_temp_new(); \ 29 gen_addr_reg_index(ctx, EA); \ 30 tcg_gen_andi_tl(EA, EA, ~0xf); \ 31 /* \ 32 * We only need to swap high and low halves. gen_qemu_ld64_i64 \ 33 * does necessary 64-bit byteswap already. \ 34 */ \ 35 if (ctx->le_mode) { \ 36 gen_qemu_ld64_i64(ctx, avr, EA); \ 37 set_avr64(rD(ctx->opcode), avr, false); \ 38 tcg_gen_addi_tl(EA, EA, 8); \ 39 gen_qemu_ld64_i64(ctx, avr, EA); \ 40 set_avr64(rD(ctx->opcode), avr, true); \ 41 } else { \ 42 gen_qemu_ld64_i64(ctx, avr, EA); \ 43 set_avr64(rD(ctx->opcode), avr, true); \ 44 tcg_gen_addi_tl(EA, EA, 8); \ 45 gen_qemu_ld64_i64(ctx, avr, EA); \ 46 set_avr64(rD(ctx->opcode), avr, false); \ 47 } \ 48 tcg_temp_free(EA); \ 49 tcg_temp_free_i64(avr); \ 50} 51 52#define GEN_VR_STX(name, opc2, opc3) \ 53static void gen_st##name(DisasContext *ctx) \ 54{ \ 55 TCGv EA; \ 56 TCGv_i64 avr; \ 57 if (unlikely(!ctx->altivec_enabled)) { \ 58 gen_exception(ctx, POWERPC_EXCP_VPU); \ 59 return; \ 60 } \ 61 gen_set_access_type(ctx, ACCESS_INT); \ 62 avr = tcg_temp_new_i64(); \ 63 EA = tcg_temp_new(); \ 64 gen_addr_reg_index(ctx, EA); \ 65 tcg_gen_andi_tl(EA, EA, ~0xf); \ 66 /* \ 67 * We only need to swap high and low halves. gen_qemu_st64_i64 \ 68 * does necessary 64-bit byteswap already. \ 69 */ \ 70 if (ctx->le_mode) { \ 71 get_avr64(avr, rD(ctx->opcode), false); \ 72 gen_qemu_st64_i64(ctx, avr, EA); \ 73 tcg_gen_addi_tl(EA, EA, 8); \ 74 get_avr64(avr, rD(ctx->opcode), true); \ 75 gen_qemu_st64_i64(ctx, avr, EA); \ 76 } else { \ 77 get_avr64(avr, rD(ctx->opcode), true); \ 78 gen_qemu_st64_i64(ctx, avr, EA); \ 79 tcg_gen_addi_tl(EA, EA, 8); \ 80 get_avr64(avr, rD(ctx->opcode), false); \ 81 gen_qemu_st64_i64(ctx, avr, EA); \ 82 } \ 83 tcg_temp_free(EA); \ 84 tcg_temp_free_i64(avr); \ 85} 86 87#define GEN_VR_LVE(name, opc2, opc3, size) \ 88static void gen_lve##name(DisasContext *ctx) \ 89 { \ 90 TCGv EA; \ 91 TCGv_ptr rs; \ 92 if (unlikely(!ctx->altivec_enabled)) { \ 93 gen_exception(ctx, POWERPC_EXCP_VPU); \ 94 return; \ 95 } \ 96 gen_set_access_type(ctx, ACCESS_INT); \ 97 EA = tcg_temp_new(); \ 98 gen_addr_reg_index(ctx, EA); \ 99 if (size > 1) { \ 100 tcg_gen_andi_tl(EA, EA, ~(size - 1)); \ 101 } \ 102 rs = gen_avr_ptr(rS(ctx->opcode)); \ 103 gen_helper_lve##name(cpu_env, rs, EA); \ 104 tcg_temp_free(EA); \ 105 tcg_temp_free_ptr(rs); \ 106 } 107 108#define GEN_VR_STVE(name, opc2, opc3, size) \ 109static void gen_stve##name(DisasContext *ctx) \ 110 { \ 111 TCGv EA; \ 112 TCGv_ptr rs; \ 113 if (unlikely(!ctx->altivec_enabled)) { \ 114 gen_exception(ctx, POWERPC_EXCP_VPU); \ 115 return; \ 116 } \ 117 gen_set_access_type(ctx, ACCESS_INT); \ 118 EA = tcg_temp_new(); \ 119 gen_addr_reg_index(ctx, EA); \ 120 if (size > 1) { \ 121 tcg_gen_andi_tl(EA, EA, ~(size - 1)); \ 122 } \ 123 rs = gen_avr_ptr(rS(ctx->opcode)); \ 124 gen_helper_stve##name(cpu_env, rs, EA); \ 125 tcg_temp_free(EA); \ 126 tcg_temp_free_ptr(rs); \ 127 } 128 129GEN_VR_LDX(lvx, 0x07, 0x03); 130/* As we don't emulate the cache, lvxl is stricly equivalent to lvx */ 131GEN_VR_LDX(lvxl, 0x07, 0x0B); 132 133GEN_VR_LVE(bx, 0x07, 0x00, 1); 134GEN_VR_LVE(hx, 0x07, 0x01, 2); 135GEN_VR_LVE(wx, 0x07, 0x02, 4); 136 137GEN_VR_STX(svx, 0x07, 0x07); 138/* As we don't emulate the cache, stvxl is stricly equivalent to stvx */ 139GEN_VR_STX(svxl, 0x07, 0x0F); 140 141GEN_VR_STVE(bx, 0x07, 0x04, 1); 142GEN_VR_STVE(hx, 0x07, 0x05, 2); 143GEN_VR_STVE(wx, 0x07, 0x06, 4); 144 145static void gen_mfvscr(DisasContext *ctx) 146{ 147 TCGv_i32 t; 148 TCGv_i64 avr; 149 if (unlikely(!ctx->altivec_enabled)) { 150 gen_exception(ctx, POWERPC_EXCP_VPU); 151 return; 152 } 153 avr = tcg_temp_new_i64(); 154 tcg_gen_movi_i64(avr, 0); 155 set_avr64(rD(ctx->opcode), avr, true); 156 t = tcg_temp_new_i32(); 157 gen_helper_mfvscr(t, cpu_env); 158 tcg_gen_extu_i32_i64(avr, t); 159 set_avr64(rD(ctx->opcode), avr, false); 160 tcg_temp_free_i32(t); 161 tcg_temp_free_i64(avr); 162} 163 164static void gen_mtvscr(DisasContext *ctx) 165{ 166 TCGv_i32 val; 167 int bofs; 168 169 if (unlikely(!ctx->altivec_enabled)) { 170 gen_exception(ctx, POWERPC_EXCP_VPU); 171 return; 172 } 173 174 val = tcg_temp_new_i32(); 175 bofs = avr_full_offset(rB(ctx->opcode)); 176#if HOST_BIG_ENDIAN 177 bofs += 3 * 4; 178#endif 179 180 tcg_gen_ld_i32(val, cpu_env, bofs); 181 gen_helper_mtvscr(cpu_env, val); 182 tcg_temp_free_i32(val); 183} 184 185#define GEN_VX_VMUL10(name, add_cin, ret_carry) \ 186static void glue(gen_, name)(DisasContext *ctx) \ 187{ \ 188 TCGv_i64 t0; \ 189 TCGv_i64 t1; \ 190 TCGv_i64 t2; \ 191 TCGv_i64 avr; \ 192 TCGv_i64 ten, z; \ 193 \ 194 if (unlikely(!ctx->altivec_enabled)) { \ 195 gen_exception(ctx, POWERPC_EXCP_VPU); \ 196 return; \ 197 } \ 198 \ 199 t0 = tcg_temp_new_i64(); \ 200 t1 = tcg_temp_new_i64(); \ 201 t2 = tcg_temp_new_i64(); \ 202 avr = tcg_temp_new_i64(); \ 203 ten = tcg_const_i64(10); \ 204 z = tcg_const_i64(0); \ 205 \ 206 if (add_cin) { \ 207 get_avr64(avr, rA(ctx->opcode), false); \ 208 tcg_gen_mulu2_i64(t0, t1, avr, ten); \ 209 get_avr64(avr, rB(ctx->opcode), false); \ 210 tcg_gen_andi_i64(t2, avr, 0xF); \ 211 tcg_gen_add2_i64(avr, t2, t0, t1, t2, z); \ 212 set_avr64(rD(ctx->opcode), avr, false); \ 213 } else { \ 214 get_avr64(avr, rA(ctx->opcode), false); \ 215 tcg_gen_mulu2_i64(avr, t2, avr, ten); \ 216 set_avr64(rD(ctx->opcode), avr, false); \ 217 } \ 218 \ 219 if (ret_carry) { \ 220 get_avr64(avr, rA(ctx->opcode), true); \ 221 tcg_gen_mulu2_i64(t0, t1, avr, ten); \ 222 tcg_gen_add2_i64(t0, avr, t0, t1, t2, z); \ 223 set_avr64(rD(ctx->opcode), avr, false); \ 224 set_avr64(rD(ctx->opcode), z, true); \ 225 } else { \ 226 get_avr64(avr, rA(ctx->opcode), true); \ 227 tcg_gen_mul_i64(t0, avr, ten); \ 228 tcg_gen_add_i64(avr, t0, t2); \ 229 set_avr64(rD(ctx->opcode), avr, true); \ 230 } \ 231 \ 232 tcg_temp_free_i64(t0); \ 233 tcg_temp_free_i64(t1); \ 234 tcg_temp_free_i64(t2); \ 235 tcg_temp_free_i64(avr); \ 236 tcg_temp_free_i64(ten); \ 237 tcg_temp_free_i64(z); \ 238} \ 239 240GEN_VX_VMUL10(vmul10uq, 0, 0); 241GEN_VX_VMUL10(vmul10euq, 1, 0); 242GEN_VX_VMUL10(vmul10cuq, 0, 1); 243GEN_VX_VMUL10(vmul10ecuq, 1, 1); 244 245#define GEN_VXFORM_V(name, vece, tcg_op, opc2, opc3) \ 246static void glue(gen_, name)(DisasContext *ctx) \ 247{ \ 248 if (unlikely(!ctx->altivec_enabled)) { \ 249 gen_exception(ctx, POWERPC_EXCP_VPU); \ 250 return; \ 251 } \ 252 \ 253 tcg_op(vece, \ 254 avr_full_offset(rD(ctx->opcode)), \ 255 avr_full_offset(rA(ctx->opcode)), \ 256 avr_full_offset(rB(ctx->opcode)), \ 257 16, 16); \ 258} 259 260/* Logical operations */ 261GEN_VXFORM_V(vand, MO_64, tcg_gen_gvec_and, 2, 16); 262GEN_VXFORM_V(vandc, MO_64, tcg_gen_gvec_andc, 2, 17); 263GEN_VXFORM_V(vor, MO_64, tcg_gen_gvec_or, 2, 18); 264GEN_VXFORM_V(vxor, MO_64, tcg_gen_gvec_xor, 2, 19); 265GEN_VXFORM_V(vnor, MO_64, tcg_gen_gvec_nor, 2, 20); 266GEN_VXFORM_V(veqv, MO_64, tcg_gen_gvec_eqv, 2, 26); 267GEN_VXFORM_V(vnand, MO_64, tcg_gen_gvec_nand, 2, 22); 268GEN_VXFORM_V(vorc, MO_64, tcg_gen_gvec_orc, 2, 21); 269 270#define GEN_VXFORM(name, opc2, opc3) \ 271static void glue(gen_, name)(DisasContext *ctx) \ 272{ \ 273 TCGv_ptr ra, rb, rd; \ 274 if (unlikely(!ctx->altivec_enabled)) { \ 275 gen_exception(ctx, POWERPC_EXCP_VPU); \ 276 return; \ 277 } \ 278 ra = gen_avr_ptr(rA(ctx->opcode)); \ 279 rb = gen_avr_ptr(rB(ctx->opcode)); \ 280 rd = gen_avr_ptr(rD(ctx->opcode)); \ 281 gen_helper_##name(rd, ra, rb); \ 282 tcg_temp_free_ptr(ra); \ 283 tcg_temp_free_ptr(rb); \ 284 tcg_temp_free_ptr(rd); \ 285} 286 287#define GEN_VXFORM_TRANS(name, opc2, opc3) \ 288static void glue(gen_, name)(DisasContext *ctx) \ 289{ \ 290 if (unlikely(!ctx->altivec_enabled)) { \ 291 gen_exception(ctx, POWERPC_EXCP_VPU); \ 292 return; \ 293 } \ 294 trans_##name(ctx); \ 295} 296 297#define GEN_VXFORM_ENV(name, opc2, opc3) \ 298static void glue(gen_, name)(DisasContext *ctx) \ 299{ \ 300 TCGv_ptr ra, rb, rd; \ 301 if (unlikely(!ctx->altivec_enabled)) { \ 302 gen_exception(ctx, POWERPC_EXCP_VPU); \ 303 return; \ 304 } \ 305 ra = gen_avr_ptr(rA(ctx->opcode)); \ 306 rb = gen_avr_ptr(rB(ctx->opcode)); \ 307 rd = gen_avr_ptr(rD(ctx->opcode)); \ 308 gen_helper_##name(cpu_env, rd, ra, rb); \ 309 tcg_temp_free_ptr(ra); \ 310 tcg_temp_free_ptr(rb); \ 311 tcg_temp_free_ptr(rd); \ 312} 313 314#define GEN_VXFORM3(name, opc2, opc3) \ 315static void glue(gen_, name)(DisasContext *ctx) \ 316{ \ 317 TCGv_ptr ra, rb, rc, rd; \ 318 if (unlikely(!ctx->altivec_enabled)) { \ 319 gen_exception(ctx, POWERPC_EXCP_VPU); \ 320 return; \ 321 } \ 322 ra = gen_avr_ptr(rA(ctx->opcode)); \ 323 rb = gen_avr_ptr(rB(ctx->opcode)); \ 324 rc = gen_avr_ptr(rC(ctx->opcode)); \ 325 rd = gen_avr_ptr(rD(ctx->opcode)); \ 326 gen_helper_##name(rd, ra, rb, rc); \ 327 tcg_temp_free_ptr(ra); \ 328 tcg_temp_free_ptr(rb); \ 329 tcg_temp_free_ptr(rc); \ 330 tcg_temp_free_ptr(rd); \ 331} 332 333/* 334 * Support for Altivec instruction pairs that use bit 31 (Rc) as 335 * an opcode bit. In general, these pairs come from different 336 * versions of the ISA, so we must also support a pair of flags for 337 * each instruction. 338 */ 339#define GEN_VXFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1) \ 340static void glue(gen_, name0##_##name1)(DisasContext *ctx) \ 341{ \ 342 if ((Rc(ctx->opcode) == 0) && \ 343 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \ 344 gen_##name0(ctx); \ 345 } else if ((Rc(ctx->opcode) == 1) && \ 346 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \ 347 gen_##name1(ctx); \ 348 } else { \ 349 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \ 350 } \ 351} 352 353/* 354 * We use this macro if one instruction is realized with direct 355 * translation, and second one with helper. 356 */ 357#define GEN_VXFORM_TRANS_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1)\ 358static void glue(gen_, name0##_##name1)(DisasContext *ctx) \ 359{ \ 360 if ((Rc(ctx->opcode) == 0) && \ 361 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \ 362 if (unlikely(!ctx->altivec_enabled)) { \ 363 gen_exception(ctx, POWERPC_EXCP_VPU); \ 364 return; \ 365 } \ 366 trans_##name0(ctx); \ 367 } else if ((Rc(ctx->opcode) == 1) && \ 368 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \ 369 gen_##name1(ctx); \ 370 } else { \ 371 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \ 372 } \ 373} 374 375/* Adds support to provide invalid mask */ 376#define GEN_VXFORM_DUAL_EXT(name0, flg0, flg2_0, inval0, \ 377 name1, flg1, flg2_1, inval1) \ 378static void glue(gen_, name0##_##name1)(DisasContext *ctx) \ 379{ \ 380 if ((Rc(ctx->opcode) == 0) && \ 381 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0)) && \ 382 !(ctx->opcode & inval0)) { \ 383 gen_##name0(ctx); \ 384 } else if ((Rc(ctx->opcode) == 1) && \ 385 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1)) && \ 386 !(ctx->opcode & inval1)) { \ 387 gen_##name1(ctx); \ 388 } else { \ 389 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \ 390 } \ 391} 392 393#define GEN_VXFORM_HETRO(name, opc2, opc3) \ 394static void glue(gen_, name)(DisasContext *ctx) \ 395{ \ 396 TCGv_ptr rb; \ 397 if (unlikely(!ctx->altivec_enabled)) { \ 398 gen_exception(ctx, POWERPC_EXCP_VPU); \ 399 return; \ 400 } \ 401 rb = gen_avr_ptr(rB(ctx->opcode)); \ 402 gen_helper_##name(cpu_gpr[rD(ctx->opcode)], cpu_gpr[rA(ctx->opcode)], rb); \ 403 tcg_temp_free_ptr(rb); \ 404} 405 406GEN_VXFORM_V(vaddubm, MO_8, tcg_gen_gvec_add, 0, 0); 407GEN_VXFORM_DUAL_EXT(vaddubm, PPC_ALTIVEC, PPC_NONE, 0, \ 408 vmul10cuq, PPC_NONE, PPC2_ISA300, 0x0000F800) 409GEN_VXFORM_V(vadduhm, MO_16, tcg_gen_gvec_add, 0, 1); 410GEN_VXFORM_DUAL(vadduhm, PPC_ALTIVEC, PPC_NONE, \ 411 vmul10ecuq, PPC_NONE, PPC2_ISA300) 412GEN_VXFORM_V(vadduwm, MO_32, tcg_gen_gvec_add, 0, 2); 413GEN_VXFORM_V(vaddudm, MO_64, tcg_gen_gvec_add, 0, 3); 414GEN_VXFORM_V(vsububm, MO_8, tcg_gen_gvec_sub, 0, 16); 415GEN_VXFORM_V(vsubuhm, MO_16, tcg_gen_gvec_sub, 0, 17); 416GEN_VXFORM_V(vsubuwm, MO_32, tcg_gen_gvec_sub, 0, 18); 417GEN_VXFORM_V(vsubudm, MO_64, tcg_gen_gvec_sub, 0, 19); 418GEN_VXFORM_V(vmaxub, MO_8, tcg_gen_gvec_umax, 1, 0); 419GEN_VXFORM_V(vmaxuh, MO_16, tcg_gen_gvec_umax, 1, 1); 420GEN_VXFORM_V(vmaxuw, MO_32, tcg_gen_gvec_umax, 1, 2); 421GEN_VXFORM_V(vmaxud, MO_64, tcg_gen_gvec_umax, 1, 3); 422GEN_VXFORM_V(vmaxsb, MO_8, tcg_gen_gvec_smax, 1, 4); 423GEN_VXFORM_V(vmaxsh, MO_16, tcg_gen_gvec_smax, 1, 5); 424GEN_VXFORM_V(vmaxsw, MO_32, tcg_gen_gvec_smax, 1, 6); 425GEN_VXFORM_V(vmaxsd, MO_64, tcg_gen_gvec_smax, 1, 7); 426GEN_VXFORM_V(vminub, MO_8, tcg_gen_gvec_umin, 1, 8); 427GEN_VXFORM_V(vminuh, MO_16, tcg_gen_gvec_umin, 1, 9); 428GEN_VXFORM_V(vminuw, MO_32, tcg_gen_gvec_umin, 1, 10); 429GEN_VXFORM_V(vminud, MO_64, tcg_gen_gvec_umin, 1, 11); 430GEN_VXFORM_V(vminsb, MO_8, tcg_gen_gvec_smin, 1, 12); 431GEN_VXFORM_V(vminsh, MO_16, tcg_gen_gvec_smin, 1, 13); 432GEN_VXFORM_V(vminsw, MO_32, tcg_gen_gvec_smin, 1, 14); 433GEN_VXFORM_V(vminsd, MO_64, tcg_gen_gvec_smin, 1, 15); 434GEN_VXFORM(vavgub, 1, 16); 435GEN_VXFORM(vabsdub, 1, 16); 436GEN_VXFORM_DUAL(vavgub, PPC_ALTIVEC, PPC_NONE, \ 437 vabsdub, PPC_NONE, PPC2_ISA300) 438GEN_VXFORM(vavguh, 1, 17); 439GEN_VXFORM(vabsduh, 1, 17); 440GEN_VXFORM_DUAL(vavguh, PPC_ALTIVEC, PPC_NONE, \ 441 vabsduh, PPC_NONE, PPC2_ISA300) 442GEN_VXFORM(vavguw, 1, 18); 443GEN_VXFORM(vabsduw, 1, 18); 444GEN_VXFORM_DUAL(vavguw, PPC_ALTIVEC, PPC_NONE, \ 445 vabsduw, PPC_NONE, PPC2_ISA300) 446GEN_VXFORM(vavgsb, 1, 20); 447GEN_VXFORM(vavgsh, 1, 21); 448GEN_VXFORM(vavgsw, 1, 22); 449GEN_VXFORM(vmrghb, 6, 0); 450GEN_VXFORM(vmrghh, 6, 1); 451GEN_VXFORM(vmrghw, 6, 2); 452GEN_VXFORM(vmrglb, 6, 4); 453GEN_VXFORM(vmrglh, 6, 5); 454GEN_VXFORM(vmrglw, 6, 6); 455 456static void trans_vmrgew(DisasContext *ctx) 457{ 458 int VT = rD(ctx->opcode); 459 int VA = rA(ctx->opcode); 460 int VB = rB(ctx->opcode); 461 TCGv_i64 tmp = tcg_temp_new_i64(); 462 TCGv_i64 avr = tcg_temp_new_i64(); 463 464 get_avr64(avr, VB, true); 465 tcg_gen_shri_i64(tmp, avr, 32); 466 get_avr64(avr, VA, true); 467 tcg_gen_deposit_i64(avr, avr, tmp, 0, 32); 468 set_avr64(VT, avr, true); 469 470 get_avr64(avr, VB, false); 471 tcg_gen_shri_i64(tmp, avr, 32); 472 get_avr64(avr, VA, false); 473 tcg_gen_deposit_i64(avr, avr, tmp, 0, 32); 474 set_avr64(VT, avr, false); 475 476 tcg_temp_free_i64(tmp); 477 tcg_temp_free_i64(avr); 478} 479 480static void trans_vmrgow(DisasContext *ctx) 481{ 482 int VT = rD(ctx->opcode); 483 int VA = rA(ctx->opcode); 484 int VB = rB(ctx->opcode); 485 TCGv_i64 t0 = tcg_temp_new_i64(); 486 TCGv_i64 t1 = tcg_temp_new_i64(); 487 TCGv_i64 avr = tcg_temp_new_i64(); 488 489 get_avr64(t0, VB, true); 490 get_avr64(t1, VA, true); 491 tcg_gen_deposit_i64(avr, t0, t1, 32, 32); 492 set_avr64(VT, avr, true); 493 494 get_avr64(t0, VB, false); 495 get_avr64(t1, VA, false); 496 tcg_gen_deposit_i64(avr, t0, t1, 32, 32); 497 set_avr64(VT, avr, false); 498 499 tcg_temp_free_i64(t0); 500 tcg_temp_free_i64(t1); 501 tcg_temp_free_i64(avr); 502} 503 504/* 505 * lvsl VRT,RA,RB - Load Vector for Shift Left 506 * 507 * Let the EA be the sum (rA|0)+(rB). Let sh=EA[28–31]. 508 * Let X be the 32-byte value 0x00 || 0x01 || 0x02 || ... || 0x1E || 0x1F. 509 * Bytes sh:sh+15 of X are placed into vD. 510 */ 511static void trans_lvsl(DisasContext *ctx) 512{ 513 int VT = rD(ctx->opcode); 514 TCGv_i64 result = tcg_temp_new_i64(); 515 TCGv_i64 sh = tcg_temp_new_i64(); 516 TCGv EA = tcg_temp_new(); 517 518 /* Get sh(from description) by anding EA with 0xf. */ 519 gen_addr_reg_index(ctx, EA); 520 tcg_gen_extu_tl_i64(sh, EA); 521 tcg_gen_andi_i64(sh, sh, 0xfULL); 522 523 /* 524 * Create bytes sh:sh+7 of X(from description) and place them in 525 * higher doubleword of vD. 526 */ 527 tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL); 528 tcg_gen_addi_i64(result, sh, 0x0001020304050607ull); 529 set_avr64(VT, result, true); 530 /* 531 * Create bytes sh+8:sh+15 of X(from description) and place them in 532 * lower doubleword of vD. 533 */ 534 tcg_gen_addi_i64(result, sh, 0x08090a0b0c0d0e0fULL); 535 set_avr64(VT, result, false); 536 537 tcg_temp_free_i64(result); 538 tcg_temp_free_i64(sh); 539 tcg_temp_free(EA); 540} 541 542/* 543 * lvsr VRT,RA,RB - Load Vector for Shift Right 544 * 545 * Let the EA be the sum (rA|0)+(rB). Let sh=EA[28–31]. 546 * Let X be the 32-byte value 0x00 || 0x01 || 0x02 || ... || 0x1E || 0x1F. 547 * Bytes (16-sh):(31-sh) of X are placed into vD. 548 */ 549static void trans_lvsr(DisasContext *ctx) 550{ 551 int VT = rD(ctx->opcode); 552 TCGv_i64 result = tcg_temp_new_i64(); 553 TCGv_i64 sh = tcg_temp_new_i64(); 554 TCGv EA = tcg_temp_new(); 555 556 557 /* Get sh(from description) by anding EA with 0xf. */ 558 gen_addr_reg_index(ctx, EA); 559 tcg_gen_extu_tl_i64(sh, EA); 560 tcg_gen_andi_i64(sh, sh, 0xfULL); 561 562 /* 563 * Create bytes (16-sh):(23-sh) of X(from description) and place them in 564 * higher doubleword of vD. 565 */ 566 tcg_gen_muli_i64(sh, sh, 0x0101010101010101ULL); 567 tcg_gen_subfi_i64(result, 0x1011121314151617ULL, sh); 568 set_avr64(VT, result, true); 569 /* 570 * Create bytes (24-sh):(32-sh) of X(from description) and place them in 571 * lower doubleword of vD. 572 */ 573 tcg_gen_subfi_i64(result, 0x18191a1b1c1d1e1fULL, sh); 574 set_avr64(VT, result, false); 575 576 tcg_temp_free_i64(result); 577 tcg_temp_free_i64(sh); 578 tcg_temp_free(EA); 579} 580 581/* 582 * vsl VRT,VRA,VRB - Vector Shift Left 583 * 584 * Shifting left 128 bit value of vA by value specified in bits 125-127 of vB. 585 * Lowest 3 bits in each byte element of register vB must be identical or 586 * result is undefined. 587 */ 588static void trans_vsl(DisasContext *ctx) 589{ 590 int VT = rD(ctx->opcode); 591 int VA = rA(ctx->opcode); 592 int VB = rB(ctx->opcode); 593 TCGv_i64 avr = tcg_temp_new_i64(); 594 TCGv_i64 sh = tcg_temp_new_i64(); 595 TCGv_i64 carry = tcg_temp_new_i64(); 596 TCGv_i64 tmp = tcg_temp_new_i64(); 597 598 /* Place bits 125-127 of vB in 'sh'. */ 599 get_avr64(avr, VB, false); 600 tcg_gen_andi_i64(sh, avr, 0x07ULL); 601 602 /* 603 * Save highest 'sh' bits of lower doubleword element of vA in variable 604 * 'carry' and perform shift on lower doubleword. 605 */ 606 get_avr64(avr, VA, false); 607 tcg_gen_subfi_i64(tmp, 32, sh); 608 tcg_gen_shri_i64(carry, avr, 32); 609 tcg_gen_shr_i64(carry, carry, tmp); 610 tcg_gen_shl_i64(avr, avr, sh); 611 set_avr64(VT, avr, false); 612 613 /* 614 * Perform shift on higher doubleword element of vA and replace lowest 615 * 'sh' bits with 'carry'. 616 */ 617 get_avr64(avr, VA, true); 618 tcg_gen_shl_i64(avr, avr, sh); 619 tcg_gen_or_i64(avr, avr, carry); 620 set_avr64(VT, avr, true); 621 622 tcg_temp_free_i64(avr); 623 tcg_temp_free_i64(sh); 624 tcg_temp_free_i64(carry); 625 tcg_temp_free_i64(tmp); 626} 627 628/* 629 * vsr VRT,VRA,VRB - Vector Shift Right 630 * 631 * Shifting right 128 bit value of vA by value specified in bits 125-127 of vB. 632 * Lowest 3 bits in each byte element of register vB must be identical or 633 * result is undefined. 634 */ 635static void trans_vsr(DisasContext *ctx) 636{ 637 int VT = rD(ctx->opcode); 638 int VA = rA(ctx->opcode); 639 int VB = rB(ctx->opcode); 640 TCGv_i64 avr = tcg_temp_new_i64(); 641 TCGv_i64 sh = tcg_temp_new_i64(); 642 TCGv_i64 carry = tcg_temp_new_i64(); 643 TCGv_i64 tmp = tcg_temp_new_i64(); 644 645 /* Place bits 125-127 of vB in 'sh'. */ 646 get_avr64(avr, VB, false); 647 tcg_gen_andi_i64(sh, avr, 0x07ULL); 648 649 /* 650 * Save lowest 'sh' bits of higher doubleword element of vA in variable 651 * 'carry' and perform shift on higher doubleword. 652 */ 653 get_avr64(avr, VA, true); 654 tcg_gen_subfi_i64(tmp, 32, sh); 655 tcg_gen_shli_i64(carry, avr, 32); 656 tcg_gen_shl_i64(carry, carry, tmp); 657 tcg_gen_shr_i64(avr, avr, sh); 658 set_avr64(VT, avr, true); 659 /* 660 * Perform shift on lower doubleword element of vA and replace highest 661 * 'sh' bits with 'carry'. 662 */ 663 get_avr64(avr, VA, false); 664 tcg_gen_shr_i64(avr, avr, sh); 665 tcg_gen_or_i64(avr, avr, carry); 666 set_avr64(VT, avr, false); 667 668 tcg_temp_free_i64(avr); 669 tcg_temp_free_i64(sh); 670 tcg_temp_free_i64(carry); 671 tcg_temp_free_i64(tmp); 672} 673 674/* 675 * vgbbd VRT,VRB - Vector Gather Bits by Bytes by Doubleword 676 * 677 * All ith bits (i in range 1 to 8) of each byte of doubleword element in source 678 * register are concatenated and placed into ith byte of appropriate doubleword 679 * element in destination register. 680 * 681 * Following solution is done for both doubleword elements of source register 682 * in parallel, in order to reduce the number of instructions needed(that's why 683 * arrays are used): 684 * First, both doubleword elements of source register vB are placed in 685 * appropriate element of array avr. Bits are gathered in 2x8 iterations(2 for 686 * loops). In first iteration bit 1 of byte 1, bit 2 of byte 2,... bit 8 of 687 * byte 8 are in their final spots so avr[i], i={0,1} can be and-ed with 688 * tcg_mask. For every following iteration, both avr[i] and tcg_mask variables 689 * have to be shifted right for 7 and 8 places, respectively, in order to get 690 * bit 1 of byte 2, bit 2 of byte 3.. bit 7 of byte 8 in their final spots so 691 * shifted avr values(saved in tmp) can be and-ed with new value of tcg_mask... 692 * After first 8 iteration(first loop), all the first bits are in their final 693 * places, all second bits but second bit from eight byte are in their places... 694 * only 1 eight bit from eight byte is in it's place). In second loop we do all 695 * operations symmetrically, in order to get other half of bits in their final 696 * spots. Results for first and second doubleword elements are saved in 697 * result[0] and result[1] respectively. In the end those results are saved in 698 * appropriate doubleword element of destination register vD. 699 */ 700static void trans_vgbbd(DisasContext *ctx) 701{ 702 int VT = rD(ctx->opcode); 703 int VB = rB(ctx->opcode); 704 TCGv_i64 tmp = tcg_temp_new_i64(); 705 uint64_t mask = 0x8040201008040201ULL; 706 int i, j; 707 708 TCGv_i64 result[2]; 709 result[0] = tcg_temp_new_i64(); 710 result[1] = tcg_temp_new_i64(); 711 TCGv_i64 avr[2]; 712 avr[0] = tcg_temp_new_i64(); 713 avr[1] = tcg_temp_new_i64(); 714 TCGv_i64 tcg_mask = tcg_temp_new_i64(); 715 716 tcg_gen_movi_i64(tcg_mask, mask); 717 for (j = 0; j < 2; j++) { 718 get_avr64(avr[j], VB, j); 719 tcg_gen_and_i64(result[j], avr[j], tcg_mask); 720 } 721 for (i = 1; i < 8; i++) { 722 tcg_gen_movi_i64(tcg_mask, mask >> (i * 8)); 723 for (j = 0; j < 2; j++) { 724 tcg_gen_shri_i64(tmp, avr[j], i * 7); 725 tcg_gen_and_i64(tmp, tmp, tcg_mask); 726 tcg_gen_or_i64(result[j], result[j], tmp); 727 } 728 } 729 for (i = 1; i < 8; i++) { 730 tcg_gen_movi_i64(tcg_mask, mask << (i * 8)); 731 for (j = 0; j < 2; j++) { 732 tcg_gen_shli_i64(tmp, avr[j], i * 7); 733 tcg_gen_and_i64(tmp, tmp, tcg_mask); 734 tcg_gen_or_i64(result[j], result[j], tmp); 735 } 736 } 737 for (j = 0; j < 2; j++) { 738 set_avr64(VT, result[j], j); 739 } 740 741 tcg_temp_free_i64(tmp); 742 tcg_temp_free_i64(tcg_mask); 743 tcg_temp_free_i64(result[0]); 744 tcg_temp_free_i64(result[1]); 745 tcg_temp_free_i64(avr[0]); 746 tcg_temp_free_i64(avr[1]); 747} 748 749/* 750 * vclzw VRT,VRB - Vector Count Leading Zeros Word 751 * 752 * Counting the number of leading zero bits of each word element in source 753 * register and placing result in appropriate word element of destination 754 * register. 755 */ 756static void trans_vclzw(DisasContext *ctx) 757{ 758 int VT = rD(ctx->opcode); 759 int VB = rB(ctx->opcode); 760 TCGv_i32 tmp = tcg_temp_new_i32(); 761 int i; 762 763 /* Perform count for every word element using tcg_gen_clzi_i32. */ 764 for (i = 0; i < 4; i++) { 765 tcg_gen_ld_i32(tmp, cpu_env, 766 offsetof(CPUPPCState, vsr[32 + VB].u64[0]) + i * 4); 767 tcg_gen_clzi_i32(tmp, tmp, 32); 768 tcg_gen_st_i32(tmp, cpu_env, 769 offsetof(CPUPPCState, vsr[32 + VT].u64[0]) + i * 4); 770 } 771 772 tcg_temp_free_i32(tmp); 773} 774 775/* 776 * vclzd VRT,VRB - Vector Count Leading Zeros Doubleword 777 * 778 * Counting the number of leading zero bits of each doubleword element in source 779 * register and placing result in appropriate doubleword element of destination 780 * register. 781 */ 782static void trans_vclzd(DisasContext *ctx) 783{ 784 int VT = rD(ctx->opcode); 785 int VB = rB(ctx->opcode); 786 TCGv_i64 avr = tcg_temp_new_i64(); 787 788 /* high doubleword */ 789 get_avr64(avr, VB, true); 790 tcg_gen_clzi_i64(avr, avr, 64); 791 set_avr64(VT, avr, true); 792 793 /* low doubleword */ 794 get_avr64(avr, VB, false); 795 tcg_gen_clzi_i64(avr, avr, 64); 796 set_avr64(VT, avr, false); 797 798 tcg_temp_free_i64(avr); 799} 800 801GEN_VXFORM_V(vmuluwm, MO_32, tcg_gen_gvec_mul, 4, 2); 802GEN_VXFORM(vsrv, 2, 28); 803GEN_VXFORM(vslv, 2, 29); 804GEN_VXFORM(vslo, 6, 16); 805GEN_VXFORM(vsro, 6, 17); 806GEN_VXFORM(vaddcuw, 0, 6); 807GEN_VXFORM(vsubcuw, 0, 22); 808 809static bool do_vector_gvec3_VX(DisasContext *ctx, arg_VX *a, int vece, 810 void (*gen_gvec)(unsigned, uint32_t, uint32_t, 811 uint32_t, uint32_t, uint32_t)) 812{ 813 REQUIRE_VECTOR(ctx); 814 815 gen_gvec(vece, avr_full_offset(a->vrt), avr_full_offset(a->vra), 816 avr_full_offset(a->vrb), 16, 16); 817 818 return true; 819} 820 821TRANS_FLAGS(ALTIVEC, VSLB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_shlv); 822TRANS_FLAGS(ALTIVEC, VSLH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_shlv); 823TRANS_FLAGS(ALTIVEC, VSLW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_shlv); 824TRANS_FLAGS2(ALTIVEC_207, VSLD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_shlv); 825 826TRANS_FLAGS(ALTIVEC, VSRB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_shrv); 827TRANS_FLAGS(ALTIVEC, VSRH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_shrv); 828TRANS_FLAGS(ALTIVEC, VSRW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_shrv); 829TRANS_FLAGS2(ALTIVEC_207, VSRD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_shrv); 830 831TRANS_FLAGS(ALTIVEC, VSRAB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_sarv); 832TRANS_FLAGS(ALTIVEC, VSRAH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_sarv); 833TRANS_FLAGS(ALTIVEC, VSRAW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_sarv); 834TRANS_FLAGS2(ALTIVEC_207, VSRAD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_sarv); 835 836TRANS_FLAGS(ALTIVEC, VRLB, do_vector_gvec3_VX, MO_8, tcg_gen_gvec_rotlv) 837TRANS_FLAGS(ALTIVEC, VRLH, do_vector_gvec3_VX, MO_16, tcg_gen_gvec_rotlv) 838TRANS_FLAGS(ALTIVEC, VRLW, do_vector_gvec3_VX, MO_32, tcg_gen_gvec_rotlv) 839TRANS_FLAGS2(ALTIVEC_207, VRLD, do_vector_gvec3_VX, MO_64, tcg_gen_gvec_rotlv) 840 841static TCGv_vec do_vrl_mask_vec(unsigned vece, TCGv_vec vrb) 842{ 843 TCGv_vec t0 = tcg_temp_new_vec_matching(vrb), 844 t1 = tcg_temp_new_vec_matching(vrb), 845 t2 = tcg_temp_new_vec_matching(vrb), 846 ones = tcg_constant_vec_matching(vrb, vece, -1); 847 848 /* Extract b and e */ 849 tcg_gen_dupi_vec(vece, t2, (8 << vece) - 1); 850 851 tcg_gen_shri_vec(vece, t0, vrb, 16); 852 tcg_gen_and_vec(vece, t0, t0, t2); 853 854 tcg_gen_shri_vec(vece, t1, vrb, 8); 855 tcg_gen_and_vec(vece, t1, t1, t2); 856 857 /* Compare b and e to negate the mask where begin > end */ 858 tcg_gen_cmp_vec(TCG_COND_GT, vece, t2, t0, t1); 859 860 /* Create the mask with (~0 >> b) ^ ((~0 >> e) >> 1) */ 861 tcg_gen_shrv_vec(vece, t0, ones, t0); 862 tcg_gen_shrv_vec(vece, t1, ones, t1); 863 tcg_gen_shri_vec(vece, t1, t1, 1); 864 tcg_gen_xor_vec(vece, t0, t0, t1); 865 866 /* negate the mask */ 867 tcg_gen_xor_vec(vece, t0, t0, t2); 868 869 tcg_temp_free_vec(t1); 870 tcg_temp_free_vec(t2); 871 872 return t0; 873} 874 875static void gen_vrlnm_vec(unsigned vece, TCGv_vec vrt, TCGv_vec vra, 876 TCGv_vec vrb) 877{ 878 TCGv_vec mask, n = tcg_temp_new_vec_matching(vrt); 879 880 /* Create the mask */ 881 mask = do_vrl_mask_vec(vece, vrb); 882 883 /* Extract n */ 884 tcg_gen_dupi_vec(vece, n, (8 << vece) - 1); 885 tcg_gen_and_vec(vece, n, vrb, n); 886 887 /* Rotate and mask */ 888 tcg_gen_rotlv_vec(vece, vrt, vra, n); 889 tcg_gen_and_vec(vece, vrt, vrt, mask); 890 891 tcg_temp_free_vec(n); 892 tcg_temp_free_vec(mask); 893} 894 895static bool do_vrlnm(DisasContext *ctx, arg_VX *a, int vece) 896{ 897 static const TCGOpcode vecop_list[] = { 898 INDEX_op_cmp_vec, INDEX_op_rotlv_vec, INDEX_op_sari_vec, 899 INDEX_op_shli_vec, INDEX_op_shri_vec, INDEX_op_shrv_vec, 0 900 }; 901 static const GVecGen3 ops[2] = { 902 { 903 .fniv = gen_vrlnm_vec, 904 .fno = gen_helper_VRLWNM, 905 .opt_opc = vecop_list, 906 .load_dest = true, 907 .vece = MO_32 908 }, 909 { 910 .fniv = gen_vrlnm_vec, 911 .fno = gen_helper_VRLDNM, 912 .opt_opc = vecop_list, 913 .load_dest = true, 914 .vece = MO_64 915 } 916 }; 917 918 REQUIRE_INSNS_FLAGS2(ctx, ISA300); 919 REQUIRE_VSX(ctx); 920 921 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra), 922 avr_full_offset(a->vrb), 16, 16, &ops[vece - 2]); 923 924 return true; 925} 926 927TRANS(VRLWNM, do_vrlnm, MO_32) 928TRANS(VRLDNM, do_vrlnm, MO_64) 929 930static void gen_vrlmi_vec(unsigned vece, TCGv_vec vrt, TCGv_vec vra, 931 TCGv_vec vrb) 932{ 933 TCGv_vec mask, n = tcg_temp_new_vec_matching(vrt), 934 tmp = tcg_temp_new_vec_matching(vrt); 935 936 /* Create the mask */ 937 mask = do_vrl_mask_vec(vece, vrb); 938 939 /* Extract n */ 940 tcg_gen_dupi_vec(vece, n, (8 << vece) - 1); 941 tcg_gen_and_vec(vece, n, vrb, n); 942 943 /* Rotate and insert */ 944 tcg_gen_rotlv_vec(vece, tmp, vra, n); 945 tcg_gen_bitsel_vec(vece, vrt, mask, tmp, vrt); 946 947 tcg_temp_free_vec(n); 948 tcg_temp_free_vec(tmp); 949 tcg_temp_free_vec(mask); 950} 951 952static bool do_vrlmi(DisasContext *ctx, arg_VX *a, int vece) 953{ 954 static const TCGOpcode vecop_list[] = { 955 INDEX_op_cmp_vec, INDEX_op_rotlv_vec, INDEX_op_sari_vec, 956 INDEX_op_shli_vec, INDEX_op_shri_vec, INDEX_op_shrv_vec, 0 957 }; 958 static const GVecGen3 ops[2] = { 959 { 960 .fniv = gen_vrlmi_vec, 961 .fno = gen_helper_VRLWMI, 962 .opt_opc = vecop_list, 963 .load_dest = true, 964 .vece = MO_32 965 }, 966 { 967 .fniv = gen_vrlnm_vec, 968 .fno = gen_helper_VRLDMI, 969 .opt_opc = vecop_list, 970 .load_dest = true, 971 .vece = MO_64 972 } 973 }; 974 975 REQUIRE_INSNS_FLAGS2(ctx, ISA300); 976 REQUIRE_VSX(ctx); 977 978 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra), 979 avr_full_offset(a->vrb), 16, 16, &ops[vece - 2]); 980 981 return true; 982} 983 984TRANS(VRLWMI, do_vrlmi, MO_32) 985TRANS(VRLDMI, do_vrlmi, MO_64) 986 987static bool do_vector_shift_quad(DisasContext *ctx, arg_VX *a, bool right, 988 bool alg) 989{ 990 TCGv_i64 hi, lo, t0, t1, n, zero = tcg_constant_i64(0); 991 992 REQUIRE_VECTOR(ctx); 993 994 n = tcg_temp_new_i64(); 995 hi = tcg_temp_new_i64(); 996 lo = tcg_temp_new_i64(); 997 t0 = tcg_temp_new_i64(); 998 t1 = tcg_const_i64(0); 999 1000 get_avr64(lo, a->vra, false); 1001 get_avr64(hi, a->vra, true); 1002 1003 get_avr64(n, a->vrb, true); 1004 1005 tcg_gen_andi_i64(t0, n, 64); 1006 if (right) { 1007 tcg_gen_movcond_i64(TCG_COND_NE, lo, t0, zero, hi, lo); 1008 if (alg) { 1009 tcg_gen_sari_i64(t1, lo, 63); 1010 } 1011 tcg_gen_movcond_i64(TCG_COND_NE, hi, t0, zero, t1, hi); 1012 } else { 1013 tcg_gen_movcond_i64(TCG_COND_NE, hi, t0, zero, lo, hi); 1014 tcg_gen_movcond_i64(TCG_COND_NE, lo, t0, zero, zero, lo); 1015 } 1016 tcg_gen_andi_i64(n, n, 0x3F); 1017 1018 if (right) { 1019 if (alg) { 1020 tcg_gen_sar_i64(t0, hi, n); 1021 } else { 1022 tcg_gen_shr_i64(t0, hi, n); 1023 } 1024 } else { 1025 tcg_gen_shl_i64(t0, lo, n); 1026 } 1027 set_avr64(a->vrt, t0, right); 1028 1029 if (right) { 1030 tcg_gen_shr_i64(lo, lo, n); 1031 } else { 1032 tcg_gen_shl_i64(hi, hi, n); 1033 } 1034 tcg_gen_xori_i64(n, n, 63); 1035 if (right) { 1036 tcg_gen_shl_i64(hi, hi, n); 1037 tcg_gen_shli_i64(hi, hi, 1); 1038 } else { 1039 tcg_gen_shr_i64(lo, lo, n); 1040 tcg_gen_shri_i64(lo, lo, 1); 1041 } 1042 tcg_gen_or_i64(hi, hi, lo); 1043 set_avr64(a->vrt, hi, !right); 1044 1045 tcg_temp_free_i64(hi); 1046 tcg_temp_free_i64(lo); 1047 tcg_temp_free_i64(t0); 1048 tcg_temp_free_i64(t1); 1049 tcg_temp_free_i64(n); 1050 1051 return true; 1052} 1053 1054TRANS_FLAGS2(ISA310, VSLQ, do_vector_shift_quad, false, false); 1055TRANS_FLAGS2(ISA310, VSRQ, do_vector_shift_quad, true, false); 1056TRANS_FLAGS2(ISA310, VSRAQ, do_vector_shift_quad, true, true); 1057 1058static void do_vrlq_mask(TCGv_i64 mh, TCGv_i64 ml, TCGv_i64 b, TCGv_i64 e) 1059{ 1060 TCGv_i64 th, tl, t0, t1, zero = tcg_constant_i64(0), 1061 ones = tcg_constant_i64(-1); 1062 1063 th = tcg_temp_new_i64(); 1064 tl = tcg_temp_new_i64(); 1065 t0 = tcg_temp_new_i64(); 1066 t1 = tcg_temp_new_i64(); 1067 1068 /* m = ~0 >> b */ 1069 tcg_gen_andi_i64(t0, b, 64); 1070 tcg_gen_movcond_i64(TCG_COND_NE, t1, t0, zero, zero, ones); 1071 tcg_gen_andi_i64(t0, b, 0x3F); 1072 tcg_gen_shr_i64(mh, t1, t0); 1073 tcg_gen_shr_i64(ml, ones, t0); 1074 tcg_gen_xori_i64(t0, t0, 63); 1075 tcg_gen_shl_i64(t1, t1, t0); 1076 tcg_gen_shli_i64(t1, t1, 1); 1077 tcg_gen_or_i64(ml, t1, ml); 1078 1079 /* t = ~0 >> e */ 1080 tcg_gen_andi_i64(t0, e, 64); 1081 tcg_gen_movcond_i64(TCG_COND_NE, t1, t0, zero, zero, ones); 1082 tcg_gen_andi_i64(t0, e, 0x3F); 1083 tcg_gen_shr_i64(th, t1, t0); 1084 tcg_gen_shr_i64(tl, ones, t0); 1085 tcg_gen_xori_i64(t0, t0, 63); 1086 tcg_gen_shl_i64(t1, t1, t0); 1087 tcg_gen_shli_i64(t1, t1, 1); 1088 tcg_gen_or_i64(tl, t1, tl); 1089 1090 /* t = t >> 1 */ 1091 tcg_gen_extract2_i64(tl, tl, th, 1); 1092 tcg_gen_shri_i64(th, th, 1); 1093 1094 /* m = m ^ t */ 1095 tcg_gen_xor_i64(mh, mh, th); 1096 tcg_gen_xor_i64(ml, ml, tl); 1097 1098 /* Negate the mask if begin > end */ 1099 tcg_gen_movcond_i64(TCG_COND_GT, t0, b, e, ones, zero); 1100 1101 tcg_gen_xor_i64(mh, mh, t0); 1102 tcg_gen_xor_i64(ml, ml, t0); 1103 1104 tcg_temp_free_i64(th); 1105 tcg_temp_free_i64(tl); 1106 tcg_temp_free_i64(t0); 1107 tcg_temp_free_i64(t1); 1108} 1109 1110static bool do_vector_rotl_quad(DisasContext *ctx, arg_VX *a, bool mask, 1111 bool insert) 1112{ 1113 TCGv_i64 ah, al, vrb, n, t0, t1, zero = tcg_constant_i64(0); 1114 1115 REQUIRE_VECTOR(ctx); 1116 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 1117 1118 ah = tcg_temp_new_i64(); 1119 al = tcg_temp_new_i64(); 1120 vrb = tcg_temp_new_i64(); 1121 n = tcg_temp_new_i64(); 1122 t0 = tcg_temp_new_i64(); 1123 t1 = tcg_temp_new_i64(); 1124 1125 get_avr64(ah, a->vra, true); 1126 get_avr64(al, a->vra, false); 1127 get_avr64(vrb, a->vrb, true); 1128 1129 tcg_gen_mov_i64(t0, ah); 1130 tcg_gen_andi_i64(t1, vrb, 64); 1131 tcg_gen_movcond_i64(TCG_COND_NE, ah, t1, zero, al, ah); 1132 tcg_gen_movcond_i64(TCG_COND_NE, al, t1, zero, t0, al); 1133 tcg_gen_andi_i64(n, vrb, 0x3F); 1134 1135 tcg_gen_shl_i64(t0, ah, n); 1136 tcg_gen_shl_i64(t1, al, n); 1137 1138 tcg_gen_xori_i64(n, n, 63); 1139 1140 tcg_gen_shr_i64(al, al, n); 1141 tcg_gen_shri_i64(al, al, 1); 1142 tcg_gen_or_i64(t0, al, t0); 1143 1144 tcg_gen_shr_i64(ah, ah, n); 1145 tcg_gen_shri_i64(ah, ah, 1); 1146 tcg_gen_or_i64(t1, ah, t1); 1147 1148 if (mask || insert) { 1149 tcg_gen_extract_i64(n, vrb, 8, 7); 1150 tcg_gen_extract_i64(vrb, vrb, 16, 7); 1151 1152 do_vrlq_mask(ah, al, vrb, n); 1153 1154 tcg_gen_and_i64(t0, t0, ah); 1155 tcg_gen_and_i64(t1, t1, al); 1156 1157 if (insert) { 1158 get_avr64(n, a->vrt, true); 1159 get_avr64(vrb, a->vrt, false); 1160 tcg_gen_andc_i64(n, n, ah); 1161 tcg_gen_andc_i64(vrb, vrb, al); 1162 tcg_gen_or_i64(t0, t0, n); 1163 tcg_gen_or_i64(t1, t1, vrb); 1164 } 1165 } 1166 1167 set_avr64(a->vrt, t0, true); 1168 set_avr64(a->vrt, t1, false); 1169 1170 tcg_temp_free_i64(ah); 1171 tcg_temp_free_i64(al); 1172 tcg_temp_free_i64(vrb); 1173 tcg_temp_free_i64(n); 1174 tcg_temp_free_i64(t0); 1175 tcg_temp_free_i64(t1); 1176 1177 return true; 1178} 1179 1180TRANS(VRLQ, do_vector_rotl_quad, false, false) 1181TRANS(VRLQNM, do_vector_rotl_quad, true, false) 1182TRANS(VRLQMI, do_vector_rotl_quad, false, true) 1183 1184#define GEN_VXFORM_SAT(NAME, VECE, NORM, SAT, OPC2, OPC3) \ 1185static void glue(glue(gen_, NAME), _vec)(unsigned vece, TCGv_vec t, \ 1186 TCGv_vec sat, TCGv_vec a, \ 1187 TCGv_vec b) \ 1188{ \ 1189 TCGv_vec x = tcg_temp_new_vec_matching(t); \ 1190 glue(glue(tcg_gen_, NORM), _vec)(VECE, x, a, b); \ 1191 glue(glue(tcg_gen_, SAT), _vec)(VECE, t, a, b); \ 1192 tcg_gen_cmp_vec(TCG_COND_NE, VECE, x, x, t); \ 1193 tcg_gen_or_vec(VECE, sat, sat, x); \ 1194 tcg_temp_free_vec(x); \ 1195} \ 1196static void glue(gen_, NAME)(DisasContext *ctx) \ 1197{ \ 1198 static const TCGOpcode vecop_list[] = { \ 1199 glue(glue(INDEX_op_, NORM), _vec), \ 1200 glue(glue(INDEX_op_, SAT), _vec), \ 1201 INDEX_op_cmp_vec, 0 \ 1202 }; \ 1203 static const GVecGen4 g = { \ 1204 .fniv = glue(glue(gen_, NAME), _vec), \ 1205 .fno = glue(gen_helper_, NAME), \ 1206 .opt_opc = vecop_list, \ 1207 .write_aofs = true, \ 1208 .vece = VECE, \ 1209 }; \ 1210 if (unlikely(!ctx->altivec_enabled)) { \ 1211 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1212 return; \ 1213 } \ 1214 tcg_gen_gvec_4(avr_full_offset(rD(ctx->opcode)), \ 1215 offsetof(CPUPPCState, vscr_sat), \ 1216 avr_full_offset(rA(ctx->opcode)), \ 1217 avr_full_offset(rB(ctx->opcode)), \ 1218 16, 16, &g); \ 1219} 1220 1221GEN_VXFORM_SAT(vaddubs, MO_8, add, usadd, 0, 8); 1222GEN_VXFORM_DUAL_EXT(vaddubs, PPC_ALTIVEC, PPC_NONE, 0, \ 1223 vmul10uq, PPC_NONE, PPC2_ISA300, 0x0000F800) 1224GEN_VXFORM_SAT(vadduhs, MO_16, add, usadd, 0, 9); 1225GEN_VXFORM_DUAL(vadduhs, PPC_ALTIVEC, PPC_NONE, \ 1226 vmul10euq, PPC_NONE, PPC2_ISA300) 1227GEN_VXFORM_SAT(vadduws, MO_32, add, usadd, 0, 10); 1228GEN_VXFORM_SAT(vaddsbs, MO_8, add, ssadd, 0, 12); 1229GEN_VXFORM_SAT(vaddshs, MO_16, add, ssadd, 0, 13); 1230GEN_VXFORM_SAT(vaddsws, MO_32, add, ssadd, 0, 14); 1231GEN_VXFORM_SAT(vsububs, MO_8, sub, ussub, 0, 24); 1232GEN_VXFORM_SAT(vsubuhs, MO_16, sub, ussub, 0, 25); 1233GEN_VXFORM_SAT(vsubuws, MO_32, sub, ussub, 0, 26); 1234GEN_VXFORM_SAT(vsubsbs, MO_8, sub, sssub, 0, 28); 1235GEN_VXFORM_SAT(vsubshs, MO_16, sub, sssub, 0, 29); 1236GEN_VXFORM_SAT(vsubsws, MO_32, sub, sssub, 0, 30); 1237GEN_VXFORM(vadduqm, 0, 4); 1238GEN_VXFORM(vaddcuq, 0, 5); 1239GEN_VXFORM3(vaddeuqm, 30, 0); 1240GEN_VXFORM3(vaddecuq, 30, 0); 1241GEN_VXFORM_DUAL(vaddeuqm, PPC_NONE, PPC2_ALTIVEC_207, \ 1242 vaddecuq, PPC_NONE, PPC2_ALTIVEC_207) 1243GEN_VXFORM(vsubuqm, 0, 20); 1244GEN_VXFORM(vsubcuq, 0, 21); 1245GEN_VXFORM3(vsubeuqm, 31, 0); 1246GEN_VXFORM3(vsubecuq, 31, 0); 1247GEN_VXFORM_DUAL(vsubeuqm, PPC_NONE, PPC2_ALTIVEC_207, \ 1248 vsubecuq, PPC_NONE, PPC2_ALTIVEC_207) 1249GEN_VXFORM_TRANS(vsl, 2, 7); 1250GEN_VXFORM_TRANS(vsr, 2, 11); 1251GEN_VXFORM_ENV(vpkuhum, 7, 0); 1252GEN_VXFORM_ENV(vpkuwum, 7, 1); 1253GEN_VXFORM_ENV(vpkudum, 7, 17); 1254GEN_VXFORM_ENV(vpkuhus, 7, 2); 1255GEN_VXFORM_ENV(vpkuwus, 7, 3); 1256GEN_VXFORM_ENV(vpkudus, 7, 19); 1257GEN_VXFORM_ENV(vpkshus, 7, 4); 1258GEN_VXFORM_ENV(vpkswus, 7, 5); 1259GEN_VXFORM_ENV(vpksdus, 7, 21); 1260GEN_VXFORM_ENV(vpkshss, 7, 6); 1261GEN_VXFORM_ENV(vpkswss, 7, 7); 1262GEN_VXFORM_ENV(vpksdss, 7, 23); 1263GEN_VXFORM(vpkpx, 7, 12); 1264GEN_VXFORM_ENV(vsum4ubs, 4, 24); 1265GEN_VXFORM_ENV(vsum4sbs, 4, 28); 1266GEN_VXFORM_ENV(vsum4shs, 4, 25); 1267GEN_VXFORM_ENV(vsum2sws, 4, 26); 1268GEN_VXFORM_ENV(vsumsws, 4, 30); 1269GEN_VXFORM_ENV(vaddfp, 5, 0); 1270GEN_VXFORM_ENV(vsubfp, 5, 1); 1271GEN_VXFORM_ENV(vmaxfp, 5, 16); 1272GEN_VXFORM_ENV(vminfp, 5, 17); 1273GEN_VXFORM_HETRO(vextublx, 6, 24) 1274GEN_VXFORM_HETRO(vextuhlx, 6, 25) 1275GEN_VXFORM_HETRO(vextuwlx, 6, 26) 1276GEN_VXFORM_TRANS_DUAL(vmrgow, PPC_NONE, PPC2_ALTIVEC_207, 1277 vextuwlx, PPC_NONE, PPC2_ISA300) 1278GEN_VXFORM_HETRO(vextubrx, 6, 28) 1279GEN_VXFORM_HETRO(vextuhrx, 6, 29) 1280GEN_VXFORM_HETRO(vextuwrx, 6, 30) 1281GEN_VXFORM_TRANS(lvsl, 6, 31) 1282GEN_VXFORM_TRANS(lvsr, 6, 32) 1283GEN_VXFORM_TRANS_DUAL(vmrgew, PPC_NONE, PPC2_ALTIVEC_207, 1284 vextuwrx, PPC_NONE, PPC2_ISA300) 1285 1286#define GEN_VXRFORM1(opname, name, str, opc2, opc3) \ 1287static void glue(gen_, name)(DisasContext *ctx) \ 1288 { \ 1289 TCGv_ptr ra, rb, rd; \ 1290 if (unlikely(!ctx->altivec_enabled)) { \ 1291 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1292 return; \ 1293 } \ 1294 ra = gen_avr_ptr(rA(ctx->opcode)); \ 1295 rb = gen_avr_ptr(rB(ctx->opcode)); \ 1296 rd = gen_avr_ptr(rD(ctx->opcode)); \ 1297 gen_helper_##opname(cpu_env, rd, ra, rb); \ 1298 tcg_temp_free_ptr(ra); \ 1299 tcg_temp_free_ptr(rb); \ 1300 tcg_temp_free_ptr(rd); \ 1301 } 1302 1303#define GEN_VXRFORM(name, opc2, opc3) \ 1304 GEN_VXRFORM1(name, name, #name, opc2, opc3) \ 1305 GEN_VXRFORM1(name##_dot, name##_, #name ".", opc2, (opc3 | (0x1 << 4))) 1306 1307/* 1308 * Support for Altivec instructions that use bit 31 (Rc) as an opcode 1309 * bit but also use bit 21 as an actual Rc bit. In general, thse pairs 1310 * come from different versions of the ISA, so we must also support a 1311 * pair of flags for each instruction. 1312 */ 1313#define GEN_VXRFORM_DUAL(name0, flg0, flg2_0, name1, flg1, flg2_1) \ 1314static void glue(gen_, name0##_##name1)(DisasContext *ctx) \ 1315{ \ 1316 if ((Rc(ctx->opcode) == 0) && \ 1317 ((ctx->insns_flags & flg0) || (ctx->insns_flags2 & flg2_0))) { \ 1318 if (Rc21(ctx->opcode) == 0) { \ 1319 gen_##name0(ctx); \ 1320 } else { \ 1321 gen_##name0##_(ctx); \ 1322 } \ 1323 } else if ((Rc(ctx->opcode) == 1) && \ 1324 ((ctx->insns_flags & flg1) || (ctx->insns_flags2 & flg2_1))) { \ 1325 if (Rc21(ctx->opcode) == 0) { \ 1326 gen_##name1(ctx); \ 1327 } else { \ 1328 gen_##name1##_(ctx); \ 1329 } \ 1330 } else { \ 1331 gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL); \ 1332 } \ 1333} 1334 1335static void do_vcmp_rc(int vrt) 1336{ 1337 TCGv_i64 tmp, set, clr; 1338 1339 tmp = tcg_temp_new_i64(); 1340 set = tcg_temp_new_i64(); 1341 clr = tcg_temp_new_i64(); 1342 1343 get_avr64(tmp, vrt, true); 1344 tcg_gen_mov_i64(set, tmp); 1345 get_avr64(tmp, vrt, false); 1346 tcg_gen_or_i64(clr, set, tmp); 1347 tcg_gen_and_i64(set, set, tmp); 1348 1349 tcg_gen_setcondi_i64(TCG_COND_EQ, clr, clr, 0); 1350 tcg_gen_shli_i64(clr, clr, 1); 1351 1352 tcg_gen_setcondi_i64(TCG_COND_EQ, set, set, -1); 1353 tcg_gen_shli_i64(set, set, 3); 1354 1355 tcg_gen_or_i64(tmp, set, clr); 1356 tcg_gen_extrl_i64_i32(cpu_crf[6], tmp); 1357 1358 tcg_temp_free_i64(tmp); 1359 tcg_temp_free_i64(set); 1360 tcg_temp_free_i64(clr); 1361} 1362 1363static bool do_vcmp(DisasContext *ctx, arg_VC *a, TCGCond cond, int vece) 1364{ 1365 REQUIRE_VECTOR(ctx); 1366 1367 tcg_gen_gvec_cmp(cond, vece, avr_full_offset(a->vrt), 1368 avr_full_offset(a->vra), avr_full_offset(a->vrb), 16, 16); 1369 1370 if (a->rc) { 1371 do_vcmp_rc(a->vrt); 1372 } 1373 1374 return true; 1375} 1376 1377TRANS_FLAGS(ALTIVEC, VCMPEQUB, do_vcmp, TCG_COND_EQ, MO_8) 1378TRANS_FLAGS(ALTIVEC, VCMPEQUH, do_vcmp, TCG_COND_EQ, MO_16) 1379TRANS_FLAGS(ALTIVEC, VCMPEQUW, do_vcmp, TCG_COND_EQ, MO_32) 1380TRANS_FLAGS2(ALTIVEC_207, VCMPEQUD, do_vcmp, TCG_COND_EQ, MO_64) 1381 1382TRANS_FLAGS(ALTIVEC, VCMPGTSB, do_vcmp, TCG_COND_GT, MO_8) 1383TRANS_FLAGS(ALTIVEC, VCMPGTSH, do_vcmp, TCG_COND_GT, MO_16) 1384TRANS_FLAGS(ALTIVEC, VCMPGTSW, do_vcmp, TCG_COND_GT, MO_32) 1385TRANS_FLAGS2(ALTIVEC_207, VCMPGTSD, do_vcmp, TCG_COND_GT, MO_64) 1386TRANS_FLAGS(ALTIVEC, VCMPGTUB, do_vcmp, TCG_COND_GTU, MO_8) 1387TRANS_FLAGS(ALTIVEC, VCMPGTUH, do_vcmp, TCG_COND_GTU, MO_16) 1388TRANS_FLAGS(ALTIVEC, VCMPGTUW, do_vcmp, TCG_COND_GTU, MO_32) 1389TRANS_FLAGS2(ALTIVEC_207, VCMPGTUD, do_vcmp, TCG_COND_GTU, MO_64) 1390 1391TRANS_FLAGS2(ISA300, VCMPNEB, do_vcmp, TCG_COND_NE, MO_8) 1392TRANS_FLAGS2(ISA300, VCMPNEH, do_vcmp, TCG_COND_NE, MO_16) 1393TRANS_FLAGS2(ISA300, VCMPNEW, do_vcmp, TCG_COND_NE, MO_32) 1394 1395static void gen_vcmpnez_vec(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b) 1396{ 1397 TCGv_vec t0, t1, zero; 1398 1399 t0 = tcg_temp_new_vec_matching(t); 1400 t1 = tcg_temp_new_vec_matching(t); 1401 zero = tcg_constant_vec_matching(t, vece, 0); 1402 1403 tcg_gen_cmp_vec(TCG_COND_EQ, vece, t0, a, zero); 1404 tcg_gen_cmp_vec(TCG_COND_EQ, vece, t1, b, zero); 1405 tcg_gen_cmp_vec(TCG_COND_NE, vece, t, a, b); 1406 1407 tcg_gen_or_vec(vece, t, t, t0); 1408 tcg_gen_or_vec(vece, t, t, t1); 1409 1410 tcg_temp_free_vec(t0); 1411 tcg_temp_free_vec(t1); 1412} 1413 1414static bool do_vcmpnez(DisasContext *ctx, arg_VC *a, int vece) 1415{ 1416 static const TCGOpcode vecop_list[] = { 1417 INDEX_op_cmp_vec, 0 1418 }; 1419 static const GVecGen3 ops[3] = { 1420 { 1421 .fniv = gen_vcmpnez_vec, 1422 .fno = gen_helper_VCMPNEZB, 1423 .opt_opc = vecop_list, 1424 .vece = MO_8 1425 }, 1426 { 1427 .fniv = gen_vcmpnez_vec, 1428 .fno = gen_helper_VCMPNEZH, 1429 .opt_opc = vecop_list, 1430 .vece = MO_16 1431 }, 1432 { 1433 .fniv = gen_vcmpnez_vec, 1434 .fno = gen_helper_VCMPNEZW, 1435 .opt_opc = vecop_list, 1436 .vece = MO_32 1437 } 1438 }; 1439 1440 REQUIRE_INSNS_FLAGS2(ctx, ISA300); 1441 REQUIRE_VECTOR(ctx); 1442 1443 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra), 1444 avr_full_offset(a->vrb), 16, 16, &ops[vece]); 1445 1446 if (a->rc) { 1447 do_vcmp_rc(a->vrt); 1448 } 1449 1450 return true; 1451} 1452 1453TRANS(VCMPNEZB, do_vcmpnez, MO_8) 1454TRANS(VCMPNEZH, do_vcmpnez, MO_16) 1455TRANS(VCMPNEZW, do_vcmpnez, MO_32) 1456 1457static bool trans_VCMPEQUQ(DisasContext *ctx, arg_VC *a) 1458{ 1459 TCGv_i64 t0, t1, t2; 1460 1461 t0 = tcg_temp_new_i64(); 1462 t1 = tcg_temp_new_i64(); 1463 t2 = tcg_temp_new_i64(); 1464 1465 get_avr64(t0, a->vra, true); 1466 get_avr64(t1, a->vrb, true); 1467 tcg_gen_xor_i64(t2, t0, t1); 1468 1469 get_avr64(t0, a->vra, false); 1470 get_avr64(t1, a->vrb, false); 1471 tcg_gen_xor_i64(t1, t0, t1); 1472 1473 tcg_gen_or_i64(t1, t1, t2); 1474 tcg_gen_setcondi_i64(TCG_COND_EQ, t1, t1, 0); 1475 tcg_gen_neg_i64(t1, t1); 1476 1477 set_avr64(a->vrt, t1, true); 1478 set_avr64(a->vrt, t1, false); 1479 1480 if (a->rc) { 1481 tcg_gen_extrl_i64_i32(cpu_crf[6], t1); 1482 tcg_gen_andi_i32(cpu_crf[6], cpu_crf[6], 0xa); 1483 tcg_gen_xori_i32(cpu_crf[6], cpu_crf[6], 0x2); 1484 } 1485 1486 tcg_temp_free_i64(t0); 1487 tcg_temp_free_i64(t1); 1488 tcg_temp_free_i64(t2); 1489 1490 return true; 1491} 1492 1493static bool do_vcmpgtq(DisasContext *ctx, arg_VC *a, bool sign) 1494{ 1495 TCGv_i64 t0, t1, t2; 1496 1497 t0 = tcg_temp_new_i64(); 1498 t1 = tcg_temp_new_i64(); 1499 t2 = tcg_temp_new_i64(); 1500 1501 get_avr64(t0, a->vra, false); 1502 get_avr64(t1, a->vrb, false); 1503 tcg_gen_setcond_i64(TCG_COND_GTU, t2, t0, t1); 1504 1505 get_avr64(t0, a->vra, true); 1506 get_avr64(t1, a->vrb, true); 1507 tcg_gen_movcond_i64(TCG_COND_EQ, t2, t0, t1, t2, tcg_constant_i64(0)); 1508 tcg_gen_setcond_i64(sign ? TCG_COND_GT : TCG_COND_GTU, t1, t0, t1); 1509 1510 tcg_gen_or_i64(t1, t1, t2); 1511 tcg_gen_neg_i64(t1, t1); 1512 1513 set_avr64(a->vrt, t1, true); 1514 set_avr64(a->vrt, t1, false); 1515 1516 if (a->rc) { 1517 tcg_gen_extrl_i64_i32(cpu_crf[6], t1); 1518 tcg_gen_andi_i32(cpu_crf[6], cpu_crf[6], 0xa); 1519 tcg_gen_xori_i32(cpu_crf[6], cpu_crf[6], 0x2); 1520 } 1521 1522 tcg_temp_free_i64(t0); 1523 tcg_temp_free_i64(t1); 1524 tcg_temp_free_i64(t2); 1525 1526 return true; 1527} 1528 1529TRANS(VCMPGTSQ, do_vcmpgtq, true) 1530TRANS(VCMPGTUQ, do_vcmpgtq, false) 1531 1532static bool do_vcmpq(DisasContext *ctx, arg_VX_bf *a, bool sign) 1533{ 1534 TCGv_i64 vra, vrb; 1535 TCGLabel *gt, *lt, *done; 1536 1537 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 1538 REQUIRE_VECTOR(ctx); 1539 1540 vra = tcg_temp_local_new_i64(); 1541 vrb = tcg_temp_local_new_i64(); 1542 gt = gen_new_label(); 1543 lt = gen_new_label(); 1544 done = gen_new_label(); 1545 1546 get_avr64(vra, a->vra, true); 1547 get_avr64(vrb, a->vrb, true); 1548 tcg_gen_brcond_i64((sign ? TCG_COND_GT : TCG_COND_GTU), vra, vrb, gt); 1549 tcg_gen_brcond_i64((sign ? TCG_COND_LT : TCG_COND_LTU), vra, vrb, lt); 1550 1551 get_avr64(vra, a->vra, false); 1552 get_avr64(vrb, a->vrb, false); 1553 tcg_gen_brcond_i64(TCG_COND_GTU, vra, vrb, gt); 1554 tcg_gen_brcond_i64(TCG_COND_LTU, vra, vrb, lt); 1555 1556 tcg_gen_movi_i32(cpu_crf[a->bf], CRF_EQ); 1557 tcg_gen_br(done); 1558 1559 gen_set_label(gt); 1560 tcg_gen_movi_i32(cpu_crf[a->bf], CRF_GT); 1561 tcg_gen_br(done); 1562 1563 gen_set_label(lt); 1564 tcg_gen_movi_i32(cpu_crf[a->bf], CRF_LT); 1565 tcg_gen_br(done); 1566 1567 gen_set_label(done); 1568 tcg_temp_free_i64(vra); 1569 tcg_temp_free_i64(vrb); 1570 1571 return true; 1572} 1573 1574TRANS(VCMPSQ, do_vcmpq, true) 1575TRANS(VCMPUQ, do_vcmpq, false) 1576 1577GEN_VXRFORM(vcmpeqfp, 3, 3) 1578GEN_VXRFORM(vcmpgefp, 3, 7) 1579GEN_VXRFORM(vcmpgtfp, 3, 11) 1580GEN_VXRFORM(vcmpbfp, 3, 15) 1581 1582static void gen_vsplti(DisasContext *ctx, int vece) 1583{ 1584 int simm; 1585 1586 if (unlikely(!ctx->altivec_enabled)) { 1587 gen_exception(ctx, POWERPC_EXCP_VPU); 1588 return; 1589 } 1590 1591 simm = SIMM5(ctx->opcode); 1592 tcg_gen_gvec_dup_imm(vece, avr_full_offset(rD(ctx->opcode)), 16, 16, simm); 1593} 1594 1595#define GEN_VXFORM_VSPLTI(name, vece, opc2, opc3) \ 1596static void glue(gen_, name)(DisasContext *ctx) { gen_vsplti(ctx, vece); } 1597 1598GEN_VXFORM_VSPLTI(vspltisb, MO_8, 6, 12); 1599GEN_VXFORM_VSPLTI(vspltish, MO_16, 6, 13); 1600GEN_VXFORM_VSPLTI(vspltisw, MO_32, 6, 14); 1601 1602#define GEN_VXFORM_NOA(name, opc2, opc3) \ 1603static void glue(gen_, name)(DisasContext *ctx) \ 1604 { \ 1605 TCGv_ptr rb, rd; \ 1606 if (unlikely(!ctx->altivec_enabled)) { \ 1607 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1608 return; \ 1609 } \ 1610 rb = gen_avr_ptr(rB(ctx->opcode)); \ 1611 rd = gen_avr_ptr(rD(ctx->opcode)); \ 1612 gen_helper_##name(rd, rb); \ 1613 tcg_temp_free_ptr(rb); \ 1614 tcg_temp_free_ptr(rd); \ 1615 } 1616 1617#define GEN_VXFORM_NOA_ENV(name, opc2, opc3) \ 1618static void glue(gen_, name)(DisasContext *ctx) \ 1619 { \ 1620 TCGv_ptr rb, rd; \ 1621 \ 1622 if (unlikely(!ctx->altivec_enabled)) { \ 1623 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1624 return; \ 1625 } \ 1626 rb = gen_avr_ptr(rB(ctx->opcode)); \ 1627 rd = gen_avr_ptr(rD(ctx->opcode)); \ 1628 gen_helper_##name(cpu_env, rd, rb); \ 1629 tcg_temp_free_ptr(rb); \ 1630 tcg_temp_free_ptr(rd); \ 1631 } 1632 1633#define GEN_VXFORM_NOA_2(name, opc2, opc3, opc4) \ 1634static void glue(gen_, name)(DisasContext *ctx) \ 1635 { \ 1636 TCGv_ptr rb, rd; \ 1637 if (unlikely(!ctx->altivec_enabled)) { \ 1638 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1639 return; \ 1640 } \ 1641 rb = gen_avr_ptr(rB(ctx->opcode)); \ 1642 rd = gen_avr_ptr(rD(ctx->opcode)); \ 1643 gen_helper_##name(rd, rb); \ 1644 tcg_temp_free_ptr(rb); \ 1645 tcg_temp_free_ptr(rd); \ 1646 } 1647 1648#define GEN_VXFORM_NOA_3(name, opc2, opc3, opc4) \ 1649static void glue(gen_, name)(DisasContext *ctx) \ 1650 { \ 1651 TCGv_ptr rb; \ 1652 if (unlikely(!ctx->altivec_enabled)) { \ 1653 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1654 return; \ 1655 } \ 1656 rb = gen_avr_ptr(rB(ctx->opcode)); \ 1657 gen_helper_##name(cpu_gpr[rD(ctx->opcode)], rb); \ 1658 tcg_temp_free_ptr(rb); \ 1659 } 1660GEN_VXFORM_NOA(vupkhsb, 7, 8); 1661GEN_VXFORM_NOA(vupkhsh, 7, 9); 1662GEN_VXFORM_NOA(vupkhsw, 7, 25); 1663GEN_VXFORM_NOA(vupklsb, 7, 10); 1664GEN_VXFORM_NOA(vupklsh, 7, 11); 1665GEN_VXFORM_NOA(vupklsw, 7, 27); 1666GEN_VXFORM_NOA(vupkhpx, 7, 13); 1667GEN_VXFORM_NOA(vupklpx, 7, 15); 1668GEN_VXFORM_NOA_ENV(vrefp, 5, 4); 1669GEN_VXFORM_NOA_ENV(vrsqrtefp, 5, 5); 1670GEN_VXFORM_NOA_ENV(vexptefp, 5, 6); 1671GEN_VXFORM_NOA_ENV(vlogefp, 5, 7); 1672GEN_VXFORM_NOA_ENV(vrfim, 5, 11); 1673GEN_VXFORM_NOA_ENV(vrfin, 5, 8); 1674GEN_VXFORM_NOA_ENV(vrfip, 5, 10); 1675GEN_VXFORM_NOA_ENV(vrfiz, 5, 9); 1676GEN_VXFORM_NOA(vprtybw, 1, 24); 1677GEN_VXFORM_NOA(vprtybd, 1, 24); 1678GEN_VXFORM_NOA(vprtybq, 1, 24); 1679 1680static void gen_vsplt(DisasContext *ctx, int vece) 1681{ 1682 int uimm, dofs, bofs; 1683 1684 if (unlikely(!ctx->altivec_enabled)) { 1685 gen_exception(ctx, POWERPC_EXCP_VPU); 1686 return; 1687 } 1688 1689 uimm = UIMM5(ctx->opcode); 1690 bofs = avr_full_offset(rB(ctx->opcode)); 1691 dofs = avr_full_offset(rD(ctx->opcode)); 1692 1693 /* Experimental testing shows that hardware masks the immediate. */ 1694 bofs += (uimm << vece) & 15; 1695#if !HOST_BIG_ENDIAN 1696 bofs ^= 15; 1697 bofs &= ~((1 << vece) - 1); 1698#endif 1699 1700 tcg_gen_gvec_dup_mem(vece, dofs, bofs, 16, 16); 1701} 1702 1703#define GEN_VXFORM_VSPLT(name, vece, opc2, opc3) \ 1704static void glue(gen_, name)(DisasContext *ctx) { gen_vsplt(ctx, vece); } 1705 1706#define GEN_VXFORM_UIMM_ENV(name, opc2, opc3) \ 1707static void glue(gen_, name)(DisasContext *ctx) \ 1708 { \ 1709 TCGv_ptr rb, rd; \ 1710 TCGv_i32 uimm; \ 1711 \ 1712 if (unlikely(!ctx->altivec_enabled)) { \ 1713 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1714 return; \ 1715 } \ 1716 uimm = tcg_const_i32(UIMM5(ctx->opcode)); \ 1717 rb = gen_avr_ptr(rB(ctx->opcode)); \ 1718 rd = gen_avr_ptr(rD(ctx->opcode)); \ 1719 gen_helper_##name(cpu_env, rd, rb, uimm); \ 1720 tcg_temp_free_i32(uimm); \ 1721 tcg_temp_free_ptr(rb); \ 1722 tcg_temp_free_ptr(rd); \ 1723 } 1724 1725#define GEN_VXFORM_UIMM_SPLAT(name, opc2, opc3, splat_max) \ 1726static void glue(gen_, name)(DisasContext *ctx) \ 1727 { \ 1728 TCGv_ptr rb, rd; \ 1729 uint8_t uimm = UIMM4(ctx->opcode); \ 1730 TCGv_i32 t0; \ 1731 if (unlikely(!ctx->altivec_enabled)) { \ 1732 gen_exception(ctx, POWERPC_EXCP_VPU); \ 1733 return; \ 1734 } \ 1735 if (uimm > splat_max) { \ 1736 uimm = 0; \ 1737 } \ 1738 t0 = tcg_temp_new_i32(); \ 1739 tcg_gen_movi_i32(t0, uimm); \ 1740 rb = gen_avr_ptr(rB(ctx->opcode)); \ 1741 rd = gen_avr_ptr(rD(ctx->opcode)); \ 1742 gen_helper_##name(rd, rb, t0); \ 1743 tcg_temp_free_i32(t0); \ 1744 tcg_temp_free_ptr(rb); \ 1745 tcg_temp_free_ptr(rd); \ 1746 } 1747 1748GEN_VXFORM_VSPLT(vspltb, MO_8, 6, 8); 1749GEN_VXFORM_VSPLT(vsplth, MO_16, 6, 9); 1750GEN_VXFORM_VSPLT(vspltw, MO_32, 6, 10); 1751GEN_VXFORM_UIMM_SPLAT(vextractub, 6, 8, 15); 1752GEN_VXFORM_UIMM_SPLAT(vextractuh, 6, 9, 14); 1753GEN_VXFORM_UIMM_SPLAT(vextractuw, 6, 10, 12); 1754GEN_VXFORM_UIMM_SPLAT(vextractd, 6, 11, 8); 1755GEN_VXFORM_UIMM_ENV(vcfux, 5, 12); 1756GEN_VXFORM_UIMM_ENV(vcfsx, 5, 13); 1757GEN_VXFORM_UIMM_ENV(vctuxs, 5, 14); 1758GEN_VXFORM_UIMM_ENV(vctsxs, 5, 15); 1759GEN_VXFORM_DUAL(vspltb, PPC_ALTIVEC, PPC_NONE, 1760 vextractub, PPC_NONE, PPC2_ISA300); 1761GEN_VXFORM_DUAL(vsplth, PPC_ALTIVEC, PPC_NONE, 1762 vextractuh, PPC_NONE, PPC2_ISA300); 1763GEN_VXFORM_DUAL(vspltw, PPC_ALTIVEC, PPC_NONE, 1764 vextractuw, PPC_NONE, PPC2_ISA300); 1765 1766static bool trans_VGNB(DisasContext *ctx, arg_VX_n *a) 1767{ 1768 /* 1769 * Similar to do_vextractm, we'll use a sequence of mask-shift-or operations 1770 * to gather the bits. The masks can be created with 1771 * 1772 * uint64_t mask(uint64_t n, uint64_t step) 1773 * { 1774 * uint64_t p = ((1UL << (1UL << step)) - 1UL) << ((n - 1UL) << step), 1775 * plen = n << step, m = 0; 1776 * for(int i = 0; i < 64/plen; i++) { 1777 * m |= p; 1778 * m = ror64(m, plen); 1779 * } 1780 * p >>= plen * DIV_ROUND_UP(64, plen) - 64; 1781 * return m | p; 1782 * } 1783 * 1784 * But since there are few values of N, we'll use a lookup table to avoid 1785 * these calculations at runtime. 1786 */ 1787 static const uint64_t mask[6][5] = { 1788 { 1789 0xAAAAAAAAAAAAAAAAULL, 0xccccccccccccccccULL, 0xf0f0f0f0f0f0f0f0ULL, 1790 0xff00ff00ff00ff00ULL, 0xffff0000ffff0000ULL 1791 }, 1792 { 1793 0x9249249249249249ULL, 0xC30C30C30C30C30CULL, 0xF00F00F00F00F00FULL, 1794 0xFF0000FF0000FF00ULL, 0xFFFF00000000FFFFULL 1795 }, 1796 { 1797 /* For N >= 4, some mask operations can be elided */ 1798 0x8888888888888888ULL, 0, 0xf000f000f000f000ULL, 0, 1799 0xFFFF000000000000ULL 1800 }, 1801 { 1802 0x8421084210842108ULL, 0, 0xF0000F0000F0000FULL, 0, 0 1803 }, 1804 { 1805 0x8208208208208208ULL, 0, 0xF00000F00000F000ULL, 0, 0 1806 }, 1807 { 1808 0x8102040810204081ULL, 0, 0xF000000F000000F0ULL, 0, 0 1809 } 1810 }; 1811 uint64_t m; 1812 int i, sh, nbits = DIV_ROUND_UP(64, a->n); 1813 TCGv_i64 hi, lo, t0, t1; 1814 1815 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 1816 REQUIRE_VECTOR(ctx); 1817 1818 if (a->n < 2) { 1819 /* 1820 * "N can be any value between 2 and 7, inclusive." Otherwise, the 1821 * result is undefined, so we don't need to change RT. Also, N > 7 is 1822 * impossible since the immediate field is 3 bits only. 1823 */ 1824 return true; 1825 } 1826 1827 hi = tcg_temp_new_i64(); 1828 lo = tcg_temp_new_i64(); 1829 t0 = tcg_temp_new_i64(); 1830 t1 = tcg_temp_new_i64(); 1831 1832 get_avr64(hi, a->vrb, true); 1833 get_avr64(lo, a->vrb, false); 1834 1835 /* Align the lower doubleword so we can use the same mask */ 1836 tcg_gen_shli_i64(lo, lo, a->n * nbits - 64); 1837 1838 /* 1839 * Starting from the most significant bit, gather every Nth bit with a 1840 * sequence of mask-shift-or operation. E.g.: for N=3 1841 * AxxBxxCxxDxxExxFxxGxxHxxIxxJxxKxxLxxMxxNxxOxxPxxQxxRxxSxxTxxUxxV 1842 * & rep(0b100) 1843 * A..B..C..D..E..F..G..H..I..J..K..L..M..N..O..P..Q..R..S..T..U..V 1844 * << 2 1845 * .B..C..D..E..F..G..H..I..J..K..L..M..N..O..P..Q..R..S..T..U..V.. 1846 * | 1847 * AB.BC.CD.DE.EF.FG.GH.HI.IJ.JK.KL.LM.MN.NO.OP.PQ.QR.RS.ST.TU.UV.V 1848 * & rep(0b110000) 1849 * AB....CD....EF....GH....IJ....KL....MN....OP....QR....ST....UV.. 1850 * << 4 1851 * ..CD....EF....GH....IJ....KL....MN....OP....QR....ST....UV...... 1852 * | 1853 * ABCD..CDEF..EFGH..GHIJ..IJKL..KLMN..MNOP..OPQR..QRST..STUV..UV.. 1854 * & rep(0b111100000000) 1855 * ABCD........EFGH........IJKL........MNOP........QRST........UV.. 1856 * << 8 1857 * ....EFGH........IJKL........MNOP........QRST........UV.......... 1858 * | 1859 * ABCDEFGH....EFGHIJKL....IJKLMNOP....MNOPQRST....QRSTUV......UV.. 1860 * & rep(0b111111110000000000000000) 1861 * ABCDEFGH................IJKLMNOP................QRSTUV.......... 1862 * << 16 1863 * ........IJKLMNOP................QRSTUV.......................... 1864 * | 1865 * ABCDEFGHIJKLMNOP........IJKLMNOPQRSTUV..........QRSTUV.......... 1866 * & rep(0b111111111111111100000000000000000000000000000000) 1867 * ABCDEFGHIJKLMNOP................................QRSTUV.......... 1868 * << 32 1869 * ................QRSTUV.......................................... 1870 * | 1871 * ABCDEFGHIJKLMNOPQRSTUV..........................QRSTUV.......... 1872 */ 1873 for (i = 0, sh = a->n - 1; i < 5; i++, sh <<= 1) { 1874 m = mask[a->n - 2][i]; 1875 if (m) { 1876 tcg_gen_andi_i64(hi, hi, m); 1877 tcg_gen_andi_i64(lo, lo, m); 1878 } 1879 if (sh < 64) { 1880 tcg_gen_shli_i64(t0, hi, sh); 1881 tcg_gen_shli_i64(t1, lo, sh); 1882 tcg_gen_or_i64(hi, t0, hi); 1883 tcg_gen_or_i64(lo, t1, lo); 1884 } 1885 } 1886 1887 tcg_gen_andi_i64(hi, hi, ~(~0ULL >> nbits)); 1888 tcg_gen_andi_i64(lo, lo, ~(~0ULL >> nbits)); 1889 tcg_gen_shri_i64(lo, lo, nbits); 1890 tcg_gen_or_i64(hi, hi, lo); 1891 tcg_gen_trunc_i64_tl(cpu_gpr[a->rt], hi); 1892 1893 tcg_temp_free_i64(hi); 1894 tcg_temp_free_i64(lo); 1895 tcg_temp_free_i64(t0); 1896 tcg_temp_free_i64(t1); 1897 1898 return true; 1899} 1900 1901static bool do_vextdx(DisasContext *ctx, arg_VA *a, int size, bool right, 1902 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv)) 1903{ 1904 TCGv_ptr vrt, vra, vrb; 1905 TCGv rc; 1906 1907 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 1908 REQUIRE_VECTOR(ctx); 1909 1910 vrt = gen_avr_ptr(a->vrt); 1911 vra = gen_avr_ptr(a->vra); 1912 vrb = gen_avr_ptr(a->vrb); 1913 rc = tcg_temp_new(); 1914 1915 tcg_gen_andi_tl(rc, cpu_gpr[a->rc], 0x1F); 1916 if (right) { 1917 tcg_gen_subfi_tl(rc, 32 - size, rc); 1918 } 1919 gen_helper(cpu_env, vrt, vra, vrb, rc); 1920 1921 tcg_temp_free_ptr(vrt); 1922 tcg_temp_free_ptr(vra); 1923 tcg_temp_free_ptr(vrb); 1924 tcg_temp_free(rc); 1925 return true; 1926} 1927 1928TRANS(VEXTDUBVLX, do_vextdx, 1, false, gen_helper_VEXTDUBVLX) 1929TRANS(VEXTDUHVLX, do_vextdx, 2, false, gen_helper_VEXTDUHVLX) 1930TRANS(VEXTDUWVLX, do_vextdx, 4, false, gen_helper_VEXTDUWVLX) 1931TRANS(VEXTDDVLX, do_vextdx, 8, false, gen_helper_VEXTDDVLX) 1932 1933TRANS(VEXTDUBVRX, do_vextdx, 1, true, gen_helper_VEXTDUBVLX) 1934TRANS(VEXTDUHVRX, do_vextdx, 2, true, gen_helper_VEXTDUHVLX) 1935TRANS(VEXTDUWVRX, do_vextdx, 4, true, gen_helper_VEXTDUWVLX) 1936TRANS(VEXTDDVRX, do_vextdx, 8, true, gen_helper_VEXTDDVLX) 1937 1938static bool do_vinsx(DisasContext *ctx, int vrt, int size, bool right, TCGv ra, 1939 TCGv_i64 rb, void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv)) 1940{ 1941 TCGv_ptr t; 1942 TCGv idx; 1943 1944 t = gen_avr_ptr(vrt); 1945 idx = tcg_temp_new(); 1946 1947 tcg_gen_andi_tl(idx, ra, 0xF); 1948 if (right) { 1949 tcg_gen_subfi_tl(idx, 16 - size, idx); 1950 } 1951 1952 gen_helper(cpu_env, t, rb, idx); 1953 1954 tcg_temp_free_ptr(t); 1955 tcg_temp_free(idx); 1956 1957 return true; 1958} 1959 1960static bool do_vinsvx(DisasContext *ctx, int vrt, int size, bool right, TCGv ra, 1961 int vrb, void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv)) 1962{ 1963 bool ok; 1964 TCGv_i64 val; 1965 1966 val = tcg_temp_new_i64(); 1967 get_avr64(val, vrb, true); 1968 ok = do_vinsx(ctx, vrt, size, right, ra, val, gen_helper); 1969 1970 tcg_temp_free_i64(val); 1971 return ok; 1972} 1973 1974static bool do_vinsx_VX(DisasContext *ctx, arg_VX *a, int size, bool right, 1975 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv)) 1976{ 1977 bool ok; 1978 TCGv_i64 val; 1979 1980 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 1981 REQUIRE_VECTOR(ctx); 1982 1983 val = tcg_temp_new_i64(); 1984 tcg_gen_extu_tl_i64(val, cpu_gpr[a->vrb]); 1985 1986 ok = do_vinsx(ctx, a->vrt, size, right, cpu_gpr[a->vra], val, gen_helper); 1987 1988 tcg_temp_free_i64(val); 1989 return ok; 1990} 1991 1992static bool do_vinsvx_VX(DisasContext *ctx, arg_VX *a, int size, bool right, 1993 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv)) 1994{ 1995 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 1996 REQUIRE_VECTOR(ctx); 1997 1998 return do_vinsvx(ctx, a->vrt, size, right, cpu_gpr[a->vra], a->vrb, 1999 gen_helper); 2000} 2001 2002static bool do_vins_VX_uim4(DisasContext *ctx, arg_VX_uim4 *a, int size, 2003 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv)) 2004{ 2005 bool ok; 2006 TCGv_i64 val; 2007 2008 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2009 REQUIRE_VECTOR(ctx); 2010 2011 if (a->uim > (16 - size)) { 2012 /* 2013 * PowerISA v3.1 says that the resulting value is undefined in this 2014 * case, so just log a guest error and leave VRT unchanged. The 2015 * real hardware would do a partial insert, e.g. if VRT is zeroed and 2016 * RB is 0x12345678, executing "vinsw VRT,RB,14" results in 2017 * VRT = 0x0000...00001234, but we don't bother to reproduce this 2018 * behavior as software shouldn't rely on it. 2019 */ 2020 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for VINS* at" 2021 " 0x" TARGET_FMT_lx ", UIM = %d > %d\n", ctx->cia, a->uim, 2022 16 - size); 2023 return true; 2024 } 2025 2026 val = tcg_temp_new_i64(); 2027 tcg_gen_extu_tl_i64(val, cpu_gpr[a->vrb]); 2028 2029 ok = do_vinsx(ctx, a->vrt, size, false, tcg_constant_tl(a->uim), val, 2030 gen_helper); 2031 2032 tcg_temp_free_i64(val); 2033 return ok; 2034} 2035 2036static bool do_vinsert_VX_uim4(DisasContext *ctx, arg_VX_uim4 *a, int size, 2037 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_i64, TCGv)) 2038{ 2039 REQUIRE_INSNS_FLAGS2(ctx, ISA300); 2040 REQUIRE_VECTOR(ctx); 2041 2042 if (a->uim > (16 - size)) { 2043 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for VINSERT* at" 2044 " 0x" TARGET_FMT_lx ", UIM = %d > %d\n", ctx->cia, a->uim, 2045 16 - size); 2046 return true; 2047 } 2048 2049 return do_vinsvx(ctx, a->vrt, size, false, tcg_constant_tl(a->uim), a->vrb, 2050 gen_helper); 2051} 2052 2053TRANS(VINSBLX, do_vinsx_VX, 1, false, gen_helper_VINSBLX) 2054TRANS(VINSHLX, do_vinsx_VX, 2, false, gen_helper_VINSHLX) 2055TRANS(VINSWLX, do_vinsx_VX, 4, false, gen_helper_VINSWLX) 2056TRANS(VINSDLX, do_vinsx_VX, 8, false, gen_helper_VINSDLX) 2057 2058TRANS(VINSBRX, do_vinsx_VX, 1, true, gen_helper_VINSBLX) 2059TRANS(VINSHRX, do_vinsx_VX, 2, true, gen_helper_VINSHLX) 2060TRANS(VINSWRX, do_vinsx_VX, 4, true, gen_helper_VINSWLX) 2061TRANS(VINSDRX, do_vinsx_VX, 8, true, gen_helper_VINSDLX) 2062 2063TRANS(VINSW, do_vins_VX_uim4, 4, gen_helper_VINSWLX) 2064TRANS(VINSD, do_vins_VX_uim4, 8, gen_helper_VINSDLX) 2065 2066TRANS(VINSBVLX, do_vinsvx_VX, 1, false, gen_helper_VINSBLX) 2067TRANS(VINSHVLX, do_vinsvx_VX, 2, false, gen_helper_VINSHLX) 2068TRANS(VINSWVLX, do_vinsvx_VX, 4, false, gen_helper_VINSWLX) 2069 2070TRANS(VINSBVRX, do_vinsvx_VX, 1, true, gen_helper_VINSBLX) 2071TRANS(VINSHVRX, do_vinsvx_VX, 2, true, gen_helper_VINSHLX) 2072TRANS(VINSWVRX, do_vinsvx_VX, 4, true, gen_helper_VINSWLX) 2073 2074TRANS(VINSERTB, do_vinsert_VX_uim4, 1, gen_helper_VINSBLX) 2075TRANS(VINSERTH, do_vinsert_VX_uim4, 2, gen_helper_VINSHLX) 2076TRANS(VINSERTW, do_vinsert_VX_uim4, 4, gen_helper_VINSWLX) 2077TRANS(VINSERTD, do_vinsert_VX_uim4, 8, gen_helper_VINSDLX) 2078 2079static void gen_vsldoi(DisasContext *ctx) 2080{ 2081 TCGv_ptr ra, rb, rd; 2082 TCGv_i32 sh; 2083 if (unlikely(!ctx->altivec_enabled)) { 2084 gen_exception(ctx, POWERPC_EXCP_VPU); 2085 return; 2086 } 2087 ra = gen_avr_ptr(rA(ctx->opcode)); 2088 rb = gen_avr_ptr(rB(ctx->opcode)); 2089 rd = gen_avr_ptr(rD(ctx->opcode)); 2090 sh = tcg_const_i32(VSH(ctx->opcode)); 2091 gen_helper_vsldoi(rd, ra, rb, sh); 2092 tcg_temp_free_ptr(ra); 2093 tcg_temp_free_ptr(rb); 2094 tcg_temp_free_ptr(rd); 2095 tcg_temp_free_i32(sh); 2096} 2097 2098static bool trans_VSLDBI(DisasContext *ctx, arg_VN *a) 2099{ 2100 TCGv_i64 t0, t1, t2; 2101 2102 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2103 REQUIRE_VECTOR(ctx); 2104 2105 t0 = tcg_temp_new_i64(); 2106 t1 = tcg_temp_new_i64(); 2107 2108 get_avr64(t0, a->vra, true); 2109 get_avr64(t1, a->vra, false); 2110 2111 if (a->sh != 0) { 2112 t2 = tcg_temp_new_i64(); 2113 2114 get_avr64(t2, a->vrb, true); 2115 2116 tcg_gen_extract2_i64(t0, t1, t0, 64 - a->sh); 2117 tcg_gen_extract2_i64(t1, t2, t1, 64 - a->sh); 2118 2119 tcg_temp_free_i64(t2); 2120 } 2121 2122 set_avr64(a->vrt, t0, true); 2123 set_avr64(a->vrt, t1, false); 2124 2125 tcg_temp_free_i64(t0); 2126 tcg_temp_free_i64(t1); 2127 2128 return true; 2129} 2130 2131static bool trans_VSRDBI(DisasContext *ctx, arg_VN *a) 2132{ 2133 TCGv_i64 t2, t1, t0; 2134 2135 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2136 REQUIRE_VECTOR(ctx); 2137 2138 t0 = tcg_temp_new_i64(); 2139 t1 = tcg_temp_new_i64(); 2140 2141 get_avr64(t0, a->vrb, false); 2142 get_avr64(t1, a->vrb, true); 2143 2144 if (a->sh != 0) { 2145 t2 = tcg_temp_new_i64(); 2146 2147 get_avr64(t2, a->vra, false); 2148 2149 tcg_gen_extract2_i64(t0, t0, t1, a->sh); 2150 tcg_gen_extract2_i64(t1, t1, t2, a->sh); 2151 2152 tcg_temp_free_i64(t2); 2153 } 2154 2155 set_avr64(a->vrt, t0, false); 2156 set_avr64(a->vrt, t1, true); 2157 2158 tcg_temp_free_i64(t0); 2159 tcg_temp_free_i64(t1); 2160 2161 return true; 2162} 2163 2164static bool do_vexpand(DisasContext *ctx, arg_VX_tb *a, unsigned vece) 2165{ 2166 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2167 REQUIRE_VECTOR(ctx); 2168 2169 tcg_gen_gvec_sari(vece, avr_full_offset(a->vrt), avr_full_offset(a->vrb), 2170 (8 << vece) - 1, 16, 16); 2171 2172 return true; 2173} 2174 2175TRANS(VEXPANDBM, do_vexpand, MO_8) 2176TRANS(VEXPANDHM, do_vexpand, MO_16) 2177TRANS(VEXPANDWM, do_vexpand, MO_32) 2178TRANS(VEXPANDDM, do_vexpand, MO_64) 2179 2180static bool trans_VEXPANDQM(DisasContext *ctx, arg_VX_tb *a) 2181{ 2182 TCGv_i64 tmp; 2183 2184 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2185 REQUIRE_VECTOR(ctx); 2186 2187 tmp = tcg_temp_new_i64(); 2188 2189 get_avr64(tmp, a->vrb, true); 2190 tcg_gen_sari_i64(tmp, tmp, 63); 2191 set_avr64(a->vrt, tmp, false); 2192 set_avr64(a->vrt, tmp, true); 2193 2194 tcg_temp_free_i64(tmp); 2195 return true; 2196} 2197 2198static bool do_vextractm(DisasContext *ctx, arg_VX_tb *a, unsigned vece) 2199{ 2200 const uint64_t elem_width = 8 << vece, elem_count_half = 8 >> vece, 2201 mask = dup_const(vece, 1 << (elem_width - 1)); 2202 uint64_t i, j; 2203 TCGv_i64 lo, hi, t0, t1; 2204 2205 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2206 REQUIRE_VECTOR(ctx); 2207 2208 hi = tcg_temp_new_i64(); 2209 lo = tcg_temp_new_i64(); 2210 t0 = tcg_temp_new_i64(); 2211 t1 = tcg_temp_new_i64(); 2212 2213 get_avr64(lo, a->vrb, false); 2214 get_avr64(hi, a->vrb, true); 2215 2216 tcg_gen_andi_i64(lo, lo, mask); 2217 tcg_gen_andi_i64(hi, hi, mask); 2218 2219 /* 2220 * Gather the most significant bit of each element in the highest element 2221 * element. E.g. for bytes: 2222 * aXXXXXXXbXXXXXXXcXXXXXXXdXXXXXXXeXXXXXXXfXXXXXXXgXXXXXXXhXXXXXXX 2223 * & dup(1 << (elem_width - 1)) 2224 * a0000000b0000000c0000000d0000000e0000000f0000000g0000000h0000000 2225 * << 32 - 4 2226 * 0000e0000000f0000000g0000000h00000000000000000000000000000000000 2227 * | 2228 * a000e000b000f000c000g000d000h000e0000000f0000000g0000000h0000000 2229 * << 16 - 2 2230 * 00c000g000d000h000e0000000f0000000g0000000h000000000000000000000 2231 * | 2232 * a0c0e0g0b0d0f0h0c0e0g000d0f0h000e0g00000f0h00000g0000000h0000000 2233 * << 8 - 1 2234 * 0b0d0f0h0c0e0g000d0f0h000e0g00000f0h00000g0000000h00000000000000 2235 * | 2236 * abcdefghbcdefgh0cdefgh00defgh000efgh0000fgh00000gh000000h0000000 2237 */ 2238 for (i = elem_count_half / 2, j = 32; i > 0; i >>= 1, j >>= 1) { 2239 tcg_gen_shli_i64(t0, hi, j - i); 2240 tcg_gen_shli_i64(t1, lo, j - i); 2241 tcg_gen_or_i64(hi, hi, t0); 2242 tcg_gen_or_i64(lo, lo, t1); 2243 } 2244 2245 tcg_gen_shri_i64(hi, hi, 64 - elem_count_half); 2246 tcg_gen_extract2_i64(lo, lo, hi, 64 - elem_count_half); 2247 tcg_gen_trunc_i64_tl(cpu_gpr[a->vrt], lo); 2248 2249 tcg_temp_free_i64(hi); 2250 tcg_temp_free_i64(lo); 2251 tcg_temp_free_i64(t0); 2252 tcg_temp_free_i64(t1); 2253 2254 return true; 2255} 2256 2257TRANS(VEXTRACTBM, do_vextractm, MO_8) 2258TRANS(VEXTRACTHM, do_vextractm, MO_16) 2259TRANS(VEXTRACTWM, do_vextractm, MO_32) 2260TRANS(VEXTRACTDM, do_vextractm, MO_64) 2261 2262static bool trans_VEXTRACTQM(DisasContext *ctx, arg_VX_tb *a) 2263{ 2264 TCGv_i64 tmp; 2265 2266 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2267 REQUIRE_VECTOR(ctx); 2268 2269 tmp = tcg_temp_new_i64(); 2270 2271 get_avr64(tmp, a->vrb, true); 2272 tcg_gen_shri_i64(tmp, tmp, 63); 2273 tcg_gen_trunc_i64_tl(cpu_gpr[a->vrt], tmp); 2274 2275 tcg_temp_free_i64(tmp); 2276 2277 return true; 2278} 2279 2280static bool do_mtvsrm(DisasContext *ctx, arg_VX_tb *a, unsigned vece) 2281{ 2282 const uint64_t elem_width = 8 << vece, elem_count_half = 8 >> vece; 2283 uint64_t c; 2284 int i, j; 2285 TCGv_i64 hi, lo, t0, t1; 2286 2287 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2288 REQUIRE_VECTOR(ctx); 2289 2290 hi = tcg_temp_new_i64(); 2291 lo = tcg_temp_new_i64(); 2292 t0 = tcg_temp_new_i64(); 2293 t1 = tcg_temp_new_i64(); 2294 2295 tcg_gen_extu_tl_i64(t0, cpu_gpr[a->vrb]); 2296 tcg_gen_extract_i64(hi, t0, elem_count_half, elem_count_half); 2297 tcg_gen_extract_i64(lo, t0, 0, elem_count_half); 2298 2299 /* 2300 * Spread the bits into their respective elements. 2301 * E.g. for bytes: 2302 * 00000000000000000000000000000000000000000000000000000000abcdefgh 2303 * << 32 - 4 2304 * 0000000000000000000000000000abcdefgh0000000000000000000000000000 2305 * | 2306 * 0000000000000000000000000000abcdefgh00000000000000000000abcdefgh 2307 * << 16 - 2 2308 * 00000000000000abcdefgh00000000000000000000abcdefgh00000000000000 2309 * | 2310 * 00000000000000abcdefgh000000abcdefgh000000abcdefgh000000abcdefgh 2311 * << 8 - 1 2312 * 0000000abcdefgh000000abcdefgh000000abcdefgh000000abcdefgh0000000 2313 * | 2314 * 0000000abcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgXbcdefgh 2315 * & dup(1) 2316 * 0000000a0000000b0000000c0000000d0000000e0000000f0000000g0000000h 2317 * * 0xff 2318 * aaaaaaaabbbbbbbbccccccccddddddddeeeeeeeeffffffffgggggggghhhhhhhh 2319 */ 2320 for (i = elem_count_half / 2, j = 32; i > 0; i >>= 1, j >>= 1) { 2321 tcg_gen_shli_i64(t0, hi, j - i); 2322 tcg_gen_shli_i64(t1, lo, j - i); 2323 tcg_gen_or_i64(hi, hi, t0); 2324 tcg_gen_or_i64(lo, lo, t1); 2325 } 2326 2327 c = dup_const(vece, 1); 2328 tcg_gen_andi_i64(hi, hi, c); 2329 tcg_gen_andi_i64(lo, lo, c); 2330 2331 c = MAKE_64BIT_MASK(0, elem_width); 2332 tcg_gen_muli_i64(hi, hi, c); 2333 tcg_gen_muli_i64(lo, lo, c); 2334 2335 set_avr64(a->vrt, lo, false); 2336 set_avr64(a->vrt, hi, true); 2337 2338 tcg_temp_free_i64(hi); 2339 tcg_temp_free_i64(lo); 2340 tcg_temp_free_i64(t0); 2341 tcg_temp_free_i64(t1); 2342 2343 return true; 2344} 2345 2346TRANS(MTVSRBM, do_mtvsrm, MO_8) 2347TRANS(MTVSRHM, do_mtvsrm, MO_16) 2348TRANS(MTVSRWM, do_mtvsrm, MO_32) 2349TRANS(MTVSRDM, do_mtvsrm, MO_64) 2350 2351static bool trans_MTVSRQM(DisasContext *ctx, arg_VX_tb *a) 2352{ 2353 TCGv_i64 tmp; 2354 2355 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2356 REQUIRE_VECTOR(ctx); 2357 2358 tmp = tcg_temp_new_i64(); 2359 2360 tcg_gen_ext_tl_i64(tmp, cpu_gpr[a->vrb]); 2361 tcg_gen_sextract_i64(tmp, tmp, 0, 1); 2362 set_avr64(a->vrt, tmp, false); 2363 set_avr64(a->vrt, tmp, true); 2364 2365 tcg_temp_free_i64(tmp); 2366 2367 return true; 2368} 2369 2370static bool trans_MTVSRBMI(DisasContext *ctx, arg_DX_b *a) 2371{ 2372 const uint64_t mask = dup_const(MO_8, 1); 2373 uint64_t hi, lo; 2374 2375 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2376 REQUIRE_VECTOR(ctx); 2377 2378 hi = extract16(a->b, 8, 8); 2379 lo = extract16(a->b, 0, 8); 2380 2381 for (int i = 4, j = 32; i > 0; i >>= 1, j >>= 1) { 2382 hi |= hi << (j - i); 2383 lo |= lo << (j - i); 2384 } 2385 2386 hi = (hi & mask) * 0xFF; 2387 lo = (lo & mask) * 0xFF; 2388 2389 set_avr64(a->vrt, tcg_constant_i64(hi), true); 2390 set_avr64(a->vrt, tcg_constant_i64(lo), false); 2391 2392 return true; 2393} 2394 2395static bool do_vcntmb(DisasContext *ctx, arg_VX_mp *a, int vece) 2396{ 2397 TCGv_i64 rt, vrb, mask; 2398 rt = tcg_const_i64(0); 2399 vrb = tcg_temp_new_i64(); 2400 mask = tcg_constant_i64(dup_const(vece, 1ULL << ((8 << vece) - 1))); 2401 2402 for (int i = 0; i < 2; i++) { 2403 get_avr64(vrb, a->vrb, i); 2404 if (a->mp) { 2405 tcg_gen_and_i64(vrb, mask, vrb); 2406 } else { 2407 tcg_gen_andc_i64(vrb, mask, vrb); 2408 } 2409 tcg_gen_ctpop_i64(vrb, vrb); 2410 tcg_gen_add_i64(rt, rt, vrb); 2411 } 2412 2413 tcg_gen_shli_i64(rt, rt, TARGET_LONG_BITS - 8 + vece); 2414 tcg_gen_trunc_i64_tl(cpu_gpr[a->rt], rt); 2415 2416 tcg_temp_free_i64(vrb); 2417 tcg_temp_free_i64(rt); 2418 2419 return true; 2420} 2421 2422TRANS(VCNTMBB, do_vcntmb, MO_8) 2423TRANS(VCNTMBH, do_vcntmb, MO_16) 2424TRANS(VCNTMBW, do_vcntmb, MO_32) 2425TRANS(VCNTMBD, do_vcntmb, MO_64) 2426 2427static bool do_vstri(DisasContext *ctx, arg_VX_tb_rc *a, 2428 void (*gen_helper)(TCGv_i32, TCGv_ptr, TCGv_ptr)) 2429{ 2430 TCGv_ptr vrt, vrb; 2431 2432 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2433 REQUIRE_VECTOR(ctx); 2434 2435 vrt = gen_avr_ptr(a->vrt); 2436 vrb = gen_avr_ptr(a->vrb); 2437 2438 if (a->rc) { 2439 gen_helper(cpu_crf[6], vrt, vrb); 2440 } else { 2441 TCGv_i32 discard = tcg_temp_new_i32(); 2442 gen_helper(discard, vrt, vrb); 2443 tcg_temp_free_i32(discard); 2444 } 2445 2446 tcg_temp_free_ptr(vrt); 2447 tcg_temp_free_ptr(vrb); 2448 2449 return true; 2450} 2451 2452TRANS(VSTRIBL, do_vstri, gen_helper_VSTRIBL) 2453TRANS(VSTRIBR, do_vstri, gen_helper_VSTRIBR) 2454TRANS(VSTRIHL, do_vstri, gen_helper_VSTRIHL) 2455TRANS(VSTRIHR, do_vstri, gen_helper_VSTRIHR) 2456 2457static bool do_vclrb(DisasContext *ctx, arg_VX *a, bool right) 2458{ 2459 TCGv_i64 rb, mh, ml, tmp, 2460 ones = tcg_constant_i64(-1), 2461 zero = tcg_constant_i64(0); 2462 2463 rb = tcg_temp_new_i64(); 2464 mh = tcg_temp_new_i64(); 2465 ml = tcg_temp_new_i64(); 2466 tmp = tcg_temp_new_i64(); 2467 2468 tcg_gen_extu_tl_i64(rb, cpu_gpr[a->vrb]); 2469 tcg_gen_andi_i64(tmp, rb, 7); 2470 tcg_gen_shli_i64(tmp, tmp, 3); 2471 if (right) { 2472 tcg_gen_shr_i64(tmp, ones, tmp); 2473 } else { 2474 tcg_gen_shl_i64(tmp, ones, tmp); 2475 } 2476 tcg_gen_not_i64(tmp, tmp); 2477 2478 if (right) { 2479 tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(8), 2480 tmp, ones); 2481 tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(8), 2482 zero, tmp); 2483 tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(16), 2484 ml, ones); 2485 } else { 2486 tcg_gen_movcond_i64(TCG_COND_LTU, ml, rb, tcg_constant_i64(8), 2487 tmp, ones); 2488 tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(8), 2489 zero, tmp); 2490 tcg_gen_movcond_i64(TCG_COND_LTU, mh, rb, tcg_constant_i64(16), 2491 mh, ones); 2492 } 2493 2494 get_avr64(tmp, a->vra, true); 2495 tcg_gen_and_i64(tmp, tmp, mh); 2496 set_avr64(a->vrt, tmp, true); 2497 2498 get_avr64(tmp, a->vra, false); 2499 tcg_gen_and_i64(tmp, tmp, ml); 2500 set_avr64(a->vrt, tmp, false); 2501 2502 tcg_temp_free_i64(rb); 2503 tcg_temp_free_i64(mh); 2504 tcg_temp_free_i64(ml); 2505 tcg_temp_free_i64(tmp); 2506 2507 return true; 2508} 2509 2510TRANS(VCLRLB, do_vclrb, false) 2511TRANS(VCLRRB, do_vclrb, true) 2512 2513#define GEN_VAFORM_PAIRED(name0, name1, opc2) \ 2514static void glue(gen_, name0##_##name1)(DisasContext *ctx) \ 2515 { \ 2516 TCGv_ptr ra, rb, rc, rd; \ 2517 if (unlikely(!ctx->altivec_enabled)) { \ 2518 gen_exception(ctx, POWERPC_EXCP_VPU); \ 2519 return; \ 2520 } \ 2521 ra = gen_avr_ptr(rA(ctx->opcode)); \ 2522 rb = gen_avr_ptr(rB(ctx->opcode)); \ 2523 rc = gen_avr_ptr(rC(ctx->opcode)); \ 2524 rd = gen_avr_ptr(rD(ctx->opcode)); \ 2525 if (Rc(ctx->opcode)) { \ 2526 gen_helper_##name1(cpu_env, rd, ra, rb, rc); \ 2527 } else { \ 2528 gen_helper_##name0(cpu_env, rd, ra, rb, rc); \ 2529 } \ 2530 tcg_temp_free_ptr(ra); \ 2531 tcg_temp_free_ptr(rb); \ 2532 tcg_temp_free_ptr(rc); \ 2533 tcg_temp_free_ptr(rd); \ 2534 } 2535 2536GEN_VAFORM_PAIRED(vmhaddshs, vmhraddshs, 16) 2537 2538static void gen_vmladduhm(DisasContext *ctx) 2539{ 2540 TCGv_ptr ra, rb, rc, rd; 2541 if (unlikely(!ctx->altivec_enabled)) { 2542 gen_exception(ctx, POWERPC_EXCP_VPU); 2543 return; 2544 } 2545 ra = gen_avr_ptr(rA(ctx->opcode)); 2546 rb = gen_avr_ptr(rB(ctx->opcode)); 2547 rc = gen_avr_ptr(rC(ctx->opcode)); 2548 rd = gen_avr_ptr(rD(ctx->opcode)); 2549 gen_helper_vmladduhm(rd, ra, rb, rc); 2550 tcg_temp_free_ptr(ra); 2551 tcg_temp_free_ptr(rb); 2552 tcg_temp_free_ptr(rc); 2553 tcg_temp_free_ptr(rd); 2554} 2555 2556static bool trans_VPERM(DisasContext *ctx, arg_VA *a) 2557{ 2558 TCGv_ptr vrt, vra, vrb, vrc; 2559 2560 REQUIRE_INSNS_FLAGS(ctx, ALTIVEC); 2561 REQUIRE_VECTOR(ctx); 2562 2563 vrt = gen_avr_ptr(a->vrt); 2564 vra = gen_avr_ptr(a->vra); 2565 vrb = gen_avr_ptr(a->vrb); 2566 vrc = gen_avr_ptr(a->rc); 2567 2568 gen_helper_VPERM(vrt, vra, vrb, vrc); 2569 2570 tcg_temp_free_ptr(vrt); 2571 tcg_temp_free_ptr(vra); 2572 tcg_temp_free_ptr(vrb); 2573 tcg_temp_free_ptr(vrc); 2574 2575 return true; 2576} 2577 2578static bool trans_VPERMR(DisasContext *ctx, arg_VA *a) 2579{ 2580 TCGv_ptr vrt, vra, vrb, vrc; 2581 2582 REQUIRE_INSNS_FLAGS2(ctx, ISA300); 2583 REQUIRE_VECTOR(ctx); 2584 2585 vrt = gen_avr_ptr(a->vrt); 2586 vra = gen_avr_ptr(a->vra); 2587 vrb = gen_avr_ptr(a->vrb); 2588 vrc = gen_avr_ptr(a->rc); 2589 2590 gen_helper_VPERMR(vrt, vra, vrb, vrc); 2591 2592 tcg_temp_free_ptr(vrt); 2593 tcg_temp_free_ptr(vra); 2594 tcg_temp_free_ptr(vrb); 2595 tcg_temp_free_ptr(vrc); 2596 2597 return true; 2598} 2599 2600static bool trans_VSEL(DisasContext *ctx, arg_VA *a) 2601{ 2602 REQUIRE_INSNS_FLAGS(ctx, ALTIVEC); 2603 REQUIRE_VECTOR(ctx); 2604 2605 tcg_gen_gvec_bitsel(MO_64, avr_full_offset(a->vrt), avr_full_offset(a->rc), 2606 avr_full_offset(a->vrb), avr_full_offset(a->vra), 2607 16, 16); 2608 2609 return true; 2610} 2611 2612GEN_VAFORM_PAIRED(vmsumubm, vmsummbm, 18) 2613GEN_VAFORM_PAIRED(vmsumuhm, vmsumuhs, 19) 2614GEN_VAFORM_PAIRED(vmsumshm, vmsumshs, 20) 2615GEN_VAFORM_PAIRED(vmaddfp, vnmsubfp, 23) 2616 2617GEN_VXFORM_NOA(vclzb, 1, 28) 2618GEN_VXFORM_NOA(vclzh, 1, 29) 2619GEN_VXFORM_TRANS(vclzw, 1, 30) 2620GEN_VXFORM_TRANS(vclzd, 1, 31) 2621GEN_VXFORM_NOA_2(vnegw, 1, 24, 6) 2622GEN_VXFORM_NOA_2(vnegd, 1, 24, 7) 2623 2624static void gen_vexts_i64(TCGv_i64 t, TCGv_i64 b, int64_t s) 2625{ 2626 tcg_gen_sextract_i64(t, b, 0, 64 - s); 2627} 2628 2629static void gen_vexts_i32(TCGv_i32 t, TCGv_i32 b, int32_t s) 2630{ 2631 tcg_gen_sextract_i32(t, b, 0, 32 - s); 2632} 2633 2634static void gen_vexts_vec(unsigned vece, TCGv_vec t, TCGv_vec b, int64_t s) 2635{ 2636 tcg_gen_shli_vec(vece, t, b, s); 2637 tcg_gen_sari_vec(vece, t, t, s); 2638} 2639 2640static bool do_vexts(DisasContext *ctx, arg_VX_tb *a, unsigned vece, int64_t s) 2641{ 2642 static const TCGOpcode vecop_list[] = { 2643 INDEX_op_shli_vec, INDEX_op_sari_vec, 0 2644 }; 2645 2646 static const GVecGen2i op[2] = { 2647 { 2648 .fni4 = gen_vexts_i32, 2649 .fniv = gen_vexts_vec, 2650 .opt_opc = vecop_list, 2651 .vece = MO_32 2652 }, 2653 { 2654 .fni8 = gen_vexts_i64, 2655 .fniv = gen_vexts_vec, 2656 .opt_opc = vecop_list, 2657 .vece = MO_64 2658 }, 2659 }; 2660 2661 REQUIRE_INSNS_FLAGS2(ctx, ISA300); 2662 REQUIRE_VECTOR(ctx); 2663 2664 tcg_gen_gvec_2i(avr_full_offset(a->vrt), avr_full_offset(a->vrb), 2665 16, 16, s, &op[vece - MO_32]); 2666 2667 return true; 2668} 2669 2670TRANS(VEXTSB2W, do_vexts, MO_32, 24); 2671TRANS(VEXTSH2W, do_vexts, MO_32, 16); 2672TRANS(VEXTSB2D, do_vexts, MO_64, 56); 2673TRANS(VEXTSH2D, do_vexts, MO_64, 48); 2674TRANS(VEXTSW2D, do_vexts, MO_64, 32); 2675 2676static bool trans_VEXTSD2Q(DisasContext *ctx, arg_VX_tb *a) 2677{ 2678 TCGv_i64 tmp; 2679 2680 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2681 REQUIRE_VECTOR(ctx); 2682 2683 tmp = tcg_temp_new_i64(); 2684 2685 get_avr64(tmp, a->vrb, false); 2686 set_avr64(a->vrt, tmp, false); 2687 tcg_gen_sari_i64(tmp, tmp, 63); 2688 set_avr64(a->vrt, tmp, true); 2689 2690 tcg_temp_free_i64(tmp); 2691 return true; 2692} 2693 2694GEN_VXFORM_NOA_2(vctzb, 1, 24, 28) 2695GEN_VXFORM_NOA_2(vctzh, 1, 24, 29) 2696GEN_VXFORM_NOA_2(vctzw, 1, 24, 30) 2697GEN_VXFORM_NOA_2(vctzd, 1, 24, 31) 2698GEN_VXFORM_NOA_3(vclzlsbb, 1, 24, 0) 2699GEN_VXFORM_NOA_3(vctzlsbb, 1, 24, 1) 2700GEN_VXFORM_NOA(vpopcntb, 1, 28) 2701GEN_VXFORM_NOA(vpopcnth, 1, 29) 2702GEN_VXFORM_NOA(vpopcntw, 1, 30) 2703GEN_VXFORM_NOA(vpopcntd, 1, 31) 2704GEN_VXFORM_DUAL(vclzb, PPC_NONE, PPC2_ALTIVEC_207, \ 2705 vpopcntb, PPC_NONE, PPC2_ALTIVEC_207) 2706GEN_VXFORM_DUAL(vclzh, PPC_NONE, PPC2_ALTIVEC_207, \ 2707 vpopcnth, PPC_NONE, PPC2_ALTIVEC_207) 2708GEN_VXFORM_DUAL(vclzw, PPC_NONE, PPC2_ALTIVEC_207, \ 2709 vpopcntw, PPC_NONE, PPC2_ALTIVEC_207) 2710GEN_VXFORM_DUAL(vclzd, PPC_NONE, PPC2_ALTIVEC_207, \ 2711 vpopcntd, PPC_NONE, PPC2_ALTIVEC_207) 2712GEN_VXFORM(vbpermd, 6, 23); 2713GEN_VXFORM(vbpermq, 6, 21); 2714GEN_VXFORM_TRANS(vgbbd, 6, 20); 2715GEN_VXFORM(vpmsumb, 4, 16) 2716GEN_VXFORM(vpmsumh, 4, 17) 2717GEN_VXFORM(vpmsumw, 4, 18) 2718GEN_VXFORM(vpmsumd, 4, 19) 2719 2720#define GEN_BCD(op) \ 2721static void gen_##op(DisasContext *ctx) \ 2722{ \ 2723 TCGv_ptr ra, rb, rd; \ 2724 TCGv_i32 ps; \ 2725 \ 2726 if (unlikely(!ctx->altivec_enabled)) { \ 2727 gen_exception(ctx, POWERPC_EXCP_VPU); \ 2728 return; \ 2729 } \ 2730 \ 2731 ra = gen_avr_ptr(rA(ctx->opcode)); \ 2732 rb = gen_avr_ptr(rB(ctx->opcode)); \ 2733 rd = gen_avr_ptr(rD(ctx->opcode)); \ 2734 \ 2735 ps = tcg_const_i32((ctx->opcode & 0x200) != 0); \ 2736 \ 2737 gen_helper_##op(cpu_crf[6], rd, ra, rb, ps); \ 2738 \ 2739 tcg_temp_free_ptr(ra); \ 2740 tcg_temp_free_ptr(rb); \ 2741 tcg_temp_free_ptr(rd); \ 2742 tcg_temp_free_i32(ps); \ 2743} 2744 2745#define GEN_BCD2(op) \ 2746static void gen_##op(DisasContext *ctx) \ 2747{ \ 2748 TCGv_ptr rd, rb; \ 2749 TCGv_i32 ps; \ 2750 \ 2751 if (unlikely(!ctx->altivec_enabled)) { \ 2752 gen_exception(ctx, POWERPC_EXCP_VPU); \ 2753 return; \ 2754 } \ 2755 \ 2756 rb = gen_avr_ptr(rB(ctx->opcode)); \ 2757 rd = gen_avr_ptr(rD(ctx->opcode)); \ 2758 \ 2759 ps = tcg_const_i32((ctx->opcode & 0x200) != 0); \ 2760 \ 2761 gen_helper_##op(cpu_crf[6], rd, rb, ps); \ 2762 \ 2763 tcg_temp_free_ptr(rb); \ 2764 tcg_temp_free_ptr(rd); \ 2765 tcg_temp_free_i32(ps); \ 2766} 2767 2768GEN_BCD(bcdadd) 2769GEN_BCD(bcdsub) 2770GEN_BCD2(bcdcfn) 2771GEN_BCD2(bcdctn) 2772GEN_BCD2(bcdcfz) 2773GEN_BCD2(bcdctz) 2774GEN_BCD2(bcdcfsq) 2775GEN_BCD2(bcdctsq) 2776GEN_BCD2(bcdsetsgn) 2777GEN_BCD(bcdcpsgn); 2778GEN_BCD(bcds); 2779GEN_BCD(bcdus); 2780GEN_BCD(bcdsr); 2781GEN_BCD(bcdtrunc); 2782GEN_BCD(bcdutrunc); 2783 2784static void gen_xpnd04_1(DisasContext *ctx) 2785{ 2786 switch (opc4(ctx->opcode)) { 2787 case 0: 2788 gen_bcdctsq(ctx); 2789 break; 2790 case 2: 2791 gen_bcdcfsq(ctx); 2792 break; 2793 case 4: 2794 gen_bcdctz(ctx); 2795 break; 2796 case 5: 2797 gen_bcdctn(ctx); 2798 break; 2799 case 6: 2800 gen_bcdcfz(ctx); 2801 break; 2802 case 7: 2803 gen_bcdcfn(ctx); 2804 break; 2805 case 31: 2806 gen_bcdsetsgn(ctx); 2807 break; 2808 default: 2809 gen_invalid(ctx); 2810 break; 2811 } 2812} 2813 2814static void gen_xpnd04_2(DisasContext *ctx) 2815{ 2816 switch (opc4(ctx->opcode)) { 2817 case 0: 2818 gen_bcdctsq(ctx); 2819 break; 2820 case 2: 2821 gen_bcdcfsq(ctx); 2822 break; 2823 case 4: 2824 gen_bcdctz(ctx); 2825 break; 2826 case 6: 2827 gen_bcdcfz(ctx); 2828 break; 2829 case 7: 2830 gen_bcdcfn(ctx); 2831 break; 2832 case 31: 2833 gen_bcdsetsgn(ctx); 2834 break; 2835 default: 2836 gen_invalid(ctx); 2837 break; 2838 } 2839} 2840 2841 2842GEN_VXFORM_DUAL(vsubcuw, PPC_ALTIVEC, PPC_NONE, \ 2843 xpnd04_1, PPC_NONE, PPC2_ISA300) 2844GEN_VXFORM_DUAL(vsubsws, PPC_ALTIVEC, PPC_NONE, \ 2845 xpnd04_2, PPC_NONE, PPC2_ISA300) 2846 2847GEN_VXFORM_DUAL(vsububm, PPC_ALTIVEC, PPC_NONE, \ 2848 bcdadd, PPC_NONE, PPC2_ALTIVEC_207) 2849GEN_VXFORM_DUAL(vsububs, PPC_ALTIVEC, PPC_NONE, \ 2850 bcdadd, PPC_NONE, PPC2_ALTIVEC_207) 2851GEN_VXFORM_DUAL(vsubuhm, PPC_ALTIVEC, PPC_NONE, \ 2852 bcdsub, PPC_NONE, PPC2_ALTIVEC_207) 2853GEN_VXFORM_DUAL(vsubuhs, PPC_ALTIVEC, PPC_NONE, \ 2854 bcdsub, PPC_NONE, PPC2_ALTIVEC_207) 2855GEN_VXFORM_DUAL(vaddshs, PPC_ALTIVEC, PPC_NONE, \ 2856 bcdcpsgn, PPC_NONE, PPC2_ISA300) 2857GEN_VXFORM_DUAL(vsubudm, PPC2_ALTIVEC_207, PPC_NONE, \ 2858 bcds, PPC_NONE, PPC2_ISA300) 2859GEN_VXFORM_DUAL(vsubuwm, PPC_ALTIVEC, PPC_NONE, \ 2860 bcdus, PPC_NONE, PPC2_ISA300) 2861GEN_VXFORM_DUAL(vsubsbs, PPC_ALTIVEC, PPC_NONE, \ 2862 bcdtrunc, PPC_NONE, PPC2_ISA300) 2863GEN_VXFORM_DUAL(vsubuqm, PPC2_ALTIVEC_207, PPC_NONE, \ 2864 bcdtrunc, PPC_NONE, PPC2_ISA300) 2865GEN_VXFORM_DUAL(vsubcuq, PPC2_ALTIVEC_207, PPC_NONE, \ 2866 bcdutrunc, PPC_NONE, PPC2_ISA300) 2867 2868 2869static void gen_vsbox(DisasContext *ctx) 2870{ 2871 TCGv_ptr ra, rd; 2872 if (unlikely(!ctx->altivec_enabled)) { 2873 gen_exception(ctx, POWERPC_EXCP_VPU); 2874 return; 2875 } 2876 ra = gen_avr_ptr(rA(ctx->opcode)); 2877 rd = gen_avr_ptr(rD(ctx->opcode)); 2878 gen_helper_vsbox(rd, ra); 2879 tcg_temp_free_ptr(ra); 2880 tcg_temp_free_ptr(rd); 2881} 2882 2883GEN_VXFORM(vcipher, 4, 20) 2884GEN_VXFORM(vcipherlast, 4, 20) 2885GEN_VXFORM(vncipher, 4, 21) 2886GEN_VXFORM(vncipherlast, 4, 21) 2887 2888GEN_VXFORM_DUAL(vcipher, PPC_NONE, PPC2_ALTIVEC_207, 2889 vcipherlast, PPC_NONE, PPC2_ALTIVEC_207) 2890GEN_VXFORM_DUAL(vncipher, PPC_NONE, PPC2_ALTIVEC_207, 2891 vncipherlast, PPC_NONE, PPC2_ALTIVEC_207) 2892 2893#define VSHASIGMA(op) \ 2894static void gen_##op(DisasContext *ctx) \ 2895{ \ 2896 TCGv_ptr ra, rd; \ 2897 TCGv_i32 st_six; \ 2898 if (unlikely(!ctx->altivec_enabled)) { \ 2899 gen_exception(ctx, POWERPC_EXCP_VPU); \ 2900 return; \ 2901 } \ 2902 ra = gen_avr_ptr(rA(ctx->opcode)); \ 2903 rd = gen_avr_ptr(rD(ctx->opcode)); \ 2904 st_six = tcg_const_i32(rB(ctx->opcode)); \ 2905 gen_helper_##op(rd, ra, st_six); \ 2906 tcg_temp_free_ptr(ra); \ 2907 tcg_temp_free_ptr(rd); \ 2908 tcg_temp_free_i32(st_six); \ 2909} 2910 2911VSHASIGMA(vshasigmaw) 2912VSHASIGMA(vshasigmad) 2913 2914GEN_VXFORM3(vpermxor, 22, 0xFF) 2915GEN_VXFORM_DUAL(vsldoi, PPC_ALTIVEC, PPC_NONE, 2916 vpermxor, PPC_NONE, PPC2_ALTIVEC_207) 2917 2918static bool trans_VCFUGED(DisasContext *ctx, arg_VX *a) 2919{ 2920 static const GVecGen3 g = { 2921 .fni8 = gen_helper_CFUGED, 2922 .vece = MO_64, 2923 }; 2924 2925 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2926 REQUIRE_VECTOR(ctx); 2927 2928 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra), 2929 avr_full_offset(a->vrb), 16, 16, &g); 2930 2931 return true; 2932} 2933 2934static bool trans_VCLZDM(DisasContext *ctx, arg_VX *a) 2935{ 2936 static const GVecGen3i g = { 2937 .fni8 = do_cntzdm, 2938 .vece = MO_64, 2939 }; 2940 2941 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2942 REQUIRE_VECTOR(ctx); 2943 2944 tcg_gen_gvec_3i(avr_full_offset(a->vrt), avr_full_offset(a->vra), 2945 avr_full_offset(a->vrb), 16, 16, false, &g); 2946 2947 return true; 2948} 2949 2950static bool trans_VCTZDM(DisasContext *ctx, arg_VX *a) 2951{ 2952 static const GVecGen3i g = { 2953 .fni8 = do_cntzdm, 2954 .vece = MO_64, 2955 }; 2956 2957 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2958 REQUIRE_VECTOR(ctx); 2959 2960 tcg_gen_gvec_3i(avr_full_offset(a->vrt), avr_full_offset(a->vra), 2961 avr_full_offset(a->vrb), 16, 16, true, &g); 2962 2963 return true; 2964} 2965 2966static bool trans_VPDEPD(DisasContext *ctx, arg_VX *a) 2967{ 2968 static const GVecGen3 g = { 2969 .fni8 = gen_helper_PDEPD, 2970 .vece = MO_64, 2971 }; 2972 2973 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2974 REQUIRE_VECTOR(ctx); 2975 2976 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra), 2977 avr_full_offset(a->vrb), 16, 16, &g); 2978 2979 return true; 2980} 2981 2982static bool trans_VPEXTD(DisasContext *ctx, arg_VX *a) 2983{ 2984 static const GVecGen3 g = { 2985 .fni8 = gen_helper_PEXTD, 2986 .vece = MO_64, 2987 }; 2988 2989 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 2990 REQUIRE_VECTOR(ctx); 2991 2992 tcg_gen_gvec_3(avr_full_offset(a->vrt), avr_full_offset(a->vra), 2993 avr_full_offset(a->vrb), 16, 16, &g); 2994 2995 return true; 2996} 2997 2998static bool trans_VMSUMUDM(DisasContext *ctx, arg_VA *a) 2999{ 3000 TCGv_i64 rl, rh, src1, src2; 3001 int dw; 3002 3003 REQUIRE_INSNS_FLAGS2(ctx, ISA300); 3004 REQUIRE_VECTOR(ctx); 3005 3006 rh = tcg_temp_new_i64(); 3007 rl = tcg_temp_new_i64(); 3008 src1 = tcg_temp_new_i64(); 3009 src2 = tcg_temp_new_i64(); 3010 3011 get_avr64(rl, a->rc, false); 3012 get_avr64(rh, a->rc, true); 3013 3014 for (dw = 0; dw < 2; dw++) { 3015 get_avr64(src1, a->vra, dw); 3016 get_avr64(src2, a->vrb, dw); 3017 tcg_gen_mulu2_i64(src1, src2, src1, src2); 3018 tcg_gen_add2_i64(rl, rh, rl, rh, src1, src2); 3019 } 3020 3021 set_avr64(a->vrt, rl, false); 3022 set_avr64(a->vrt, rh, true); 3023 3024 tcg_temp_free_i64(rl); 3025 tcg_temp_free_i64(rh); 3026 tcg_temp_free_i64(src1); 3027 tcg_temp_free_i64(src2); 3028 3029 return true; 3030} 3031 3032static bool trans_VMSUMCUD(DisasContext *ctx, arg_VA *a) 3033{ 3034 TCGv_i64 tmp0, tmp1, prod1h, prod1l, prod0h, prod0l, zero; 3035 3036 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 3037 REQUIRE_VECTOR(ctx); 3038 3039 tmp0 = tcg_temp_new_i64(); 3040 tmp1 = tcg_temp_new_i64(); 3041 prod1h = tcg_temp_new_i64(); 3042 prod1l = tcg_temp_new_i64(); 3043 prod0h = tcg_temp_new_i64(); 3044 prod0l = tcg_temp_new_i64(); 3045 zero = tcg_constant_i64(0); 3046 3047 /* prod1 = vsr[vra+32].dw[1] * vsr[vrb+32].dw[1] */ 3048 get_avr64(tmp0, a->vra, false); 3049 get_avr64(tmp1, a->vrb, false); 3050 tcg_gen_mulu2_i64(prod1l, prod1h, tmp0, tmp1); 3051 3052 /* prod0 = vsr[vra+32].dw[0] * vsr[vrb+32].dw[0] */ 3053 get_avr64(tmp0, a->vra, true); 3054 get_avr64(tmp1, a->vrb, true); 3055 tcg_gen_mulu2_i64(prod0l, prod0h, tmp0, tmp1); 3056 3057 /* Sum lower 64-bits elements */ 3058 get_avr64(tmp1, a->rc, false); 3059 tcg_gen_add2_i64(tmp1, tmp0, tmp1, zero, prod1l, zero); 3060 tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod0l, zero); 3061 3062 /* 3063 * Discard lower 64-bits, leaving the carry into bit 64. 3064 * Then sum the higher 64-bit elements. 3065 */ 3066 get_avr64(tmp1, a->rc, true); 3067 tcg_gen_add2_i64(tmp1, tmp0, tmp0, zero, tmp1, zero); 3068 tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod1h, zero); 3069 tcg_gen_add2_i64(tmp1, tmp0, tmp1, tmp0, prod0h, zero); 3070 3071 /* Discard 64 more bits to complete the CHOP128(temp >> 128) */ 3072 set_avr64(a->vrt, tmp0, false); 3073 set_avr64(a->vrt, zero, true); 3074 3075 tcg_temp_free_i64(tmp0); 3076 tcg_temp_free_i64(tmp1); 3077 tcg_temp_free_i64(prod1h); 3078 tcg_temp_free_i64(prod1l); 3079 tcg_temp_free_i64(prod0h); 3080 tcg_temp_free_i64(prod0l); 3081 3082 return true; 3083} 3084 3085static bool do_vx_helper(DisasContext *ctx, arg_VX *a, 3086 void (*gen_helper)(TCGv_ptr, TCGv_ptr, TCGv_ptr)) 3087{ 3088 TCGv_ptr ra, rb, rd; 3089 REQUIRE_VECTOR(ctx); 3090 3091 ra = gen_avr_ptr(a->vra); 3092 rb = gen_avr_ptr(a->vrb); 3093 rd = gen_avr_ptr(a->vrt); 3094 gen_helper(rd, ra, rb); 3095 tcg_temp_free_ptr(ra); 3096 tcg_temp_free_ptr(rb); 3097 tcg_temp_free_ptr(rd); 3098 3099 return true; 3100} 3101 3102static bool do_vx_vmuleo(DisasContext *ctx, arg_VX *a, bool even, 3103 void (*gen_mul)(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_i64)) 3104{ 3105 TCGv_i64 vra, vrb, vrt0, vrt1; 3106 REQUIRE_VECTOR(ctx); 3107 3108 vra = tcg_temp_new_i64(); 3109 vrb = tcg_temp_new_i64(); 3110 vrt0 = tcg_temp_new_i64(); 3111 vrt1 = tcg_temp_new_i64(); 3112 3113 get_avr64(vra, a->vra, even); 3114 get_avr64(vrb, a->vrb, even); 3115 gen_mul(vrt0, vrt1, vra, vrb); 3116 set_avr64(a->vrt, vrt0, false); 3117 set_avr64(a->vrt, vrt1, true); 3118 3119 tcg_temp_free_i64(vra); 3120 tcg_temp_free_i64(vrb); 3121 tcg_temp_free_i64(vrt0); 3122 tcg_temp_free_i64(vrt1); 3123 3124 return true; 3125} 3126 3127static bool trans_VMULLD(DisasContext *ctx, arg_VX *a) 3128{ 3129 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 3130 REQUIRE_VECTOR(ctx); 3131 3132 tcg_gen_gvec_mul(MO_64, avr_full_offset(a->vrt), avr_full_offset(a->vra), 3133 avr_full_offset(a->vrb), 16, 16); 3134 3135 return true; 3136} 3137 3138TRANS_FLAGS(ALTIVEC, VMULESB, do_vx_helper, gen_helper_VMULESB) 3139TRANS_FLAGS(ALTIVEC, VMULOSB, do_vx_helper, gen_helper_VMULOSB) 3140TRANS_FLAGS(ALTIVEC, VMULEUB, do_vx_helper, gen_helper_VMULEUB) 3141TRANS_FLAGS(ALTIVEC, VMULOUB, do_vx_helper, gen_helper_VMULOUB) 3142TRANS_FLAGS(ALTIVEC, VMULESH, do_vx_helper, gen_helper_VMULESH) 3143TRANS_FLAGS(ALTIVEC, VMULOSH, do_vx_helper, gen_helper_VMULOSH) 3144TRANS_FLAGS(ALTIVEC, VMULEUH, do_vx_helper, gen_helper_VMULEUH) 3145TRANS_FLAGS(ALTIVEC, VMULOUH, do_vx_helper, gen_helper_VMULOUH) 3146TRANS_FLAGS2(ALTIVEC_207, VMULESW, do_vx_helper, gen_helper_VMULESW) 3147TRANS_FLAGS2(ALTIVEC_207, VMULOSW, do_vx_helper, gen_helper_VMULOSW) 3148TRANS_FLAGS2(ALTIVEC_207, VMULEUW, do_vx_helper, gen_helper_VMULEUW) 3149TRANS_FLAGS2(ALTIVEC_207, VMULOUW, do_vx_helper, gen_helper_VMULOUW) 3150TRANS_FLAGS2(ISA310, VMULESD, do_vx_vmuleo, true , tcg_gen_muls2_i64) 3151TRANS_FLAGS2(ISA310, VMULOSD, do_vx_vmuleo, false, tcg_gen_muls2_i64) 3152TRANS_FLAGS2(ISA310, VMULEUD, do_vx_vmuleo, true , tcg_gen_mulu2_i64) 3153TRANS_FLAGS2(ISA310, VMULOUD, do_vx_vmuleo, false, tcg_gen_mulu2_i64) 3154 3155static void do_vx_vmulhw_i64(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b, bool sign) 3156{ 3157 TCGv_i64 hh, lh, temp; 3158 3159 hh = tcg_temp_new_i64(); 3160 lh = tcg_temp_new_i64(); 3161 temp = tcg_temp_new_i64(); 3162 3163 if (sign) { 3164 tcg_gen_ext32s_i64(lh, a); 3165 tcg_gen_ext32s_i64(temp, b); 3166 } else { 3167 tcg_gen_ext32u_i64(lh, a); 3168 tcg_gen_ext32u_i64(temp, b); 3169 } 3170 tcg_gen_mul_i64(lh, lh, temp); 3171 3172 if (sign) { 3173 tcg_gen_sari_i64(hh, a, 32); 3174 tcg_gen_sari_i64(temp, b, 32); 3175 } else { 3176 tcg_gen_shri_i64(hh, a, 32); 3177 tcg_gen_shri_i64(temp, b, 32); 3178 } 3179 tcg_gen_mul_i64(hh, hh, temp); 3180 3181 tcg_gen_shri_i64(lh, lh, 32); 3182 tcg_gen_deposit_i64(t, hh, lh, 0, 32); 3183 3184 tcg_temp_free_i64(hh); 3185 tcg_temp_free_i64(lh); 3186 tcg_temp_free_i64(temp); 3187} 3188 3189static void do_vx_vmulhd_i64(TCGv_i64 t, TCGv_i64 a, TCGv_i64 b, bool sign) 3190{ 3191 TCGv_i64 tlow; 3192 3193 tlow = tcg_temp_new_i64(); 3194 if (sign) { 3195 tcg_gen_muls2_i64(tlow, t, a, b); 3196 } else { 3197 tcg_gen_mulu2_i64(tlow, t, a, b); 3198 } 3199 3200 tcg_temp_free_i64(tlow); 3201} 3202 3203static bool do_vx_mulh(DisasContext *ctx, arg_VX *a, bool sign, 3204 void (*func)(TCGv_i64, TCGv_i64, TCGv_i64, bool)) 3205{ 3206 REQUIRE_INSNS_FLAGS2(ctx, ISA310); 3207 REQUIRE_VECTOR(ctx); 3208 3209 TCGv_i64 vra, vrb, vrt; 3210 int i; 3211 3212 vra = tcg_temp_new_i64(); 3213 vrb = tcg_temp_new_i64(); 3214 vrt = tcg_temp_new_i64(); 3215 3216 for (i = 0; i < 2; i++) { 3217 get_avr64(vra, a->vra, i); 3218 get_avr64(vrb, a->vrb, i); 3219 get_avr64(vrt, a->vrt, i); 3220 3221 func(vrt, vra, vrb, sign); 3222 3223 set_avr64(a->vrt, vrt, i); 3224 } 3225 3226 tcg_temp_free_i64(vra); 3227 tcg_temp_free_i64(vrb); 3228 tcg_temp_free_i64(vrt); 3229 3230 return true; 3231 3232} 3233 3234TRANS(VMULHSW, do_vx_mulh, true , do_vx_vmulhw_i64) 3235TRANS(VMULHSD, do_vx_mulh, true , do_vx_vmulhd_i64) 3236TRANS(VMULHUW, do_vx_mulh, false, do_vx_vmulhw_i64) 3237TRANS(VMULHUD, do_vx_mulh, false, do_vx_vmulhd_i64) 3238 3239#undef GEN_VR_LDX 3240#undef GEN_VR_STX 3241#undef GEN_VR_LVE 3242#undef GEN_VR_STVE 3243 3244#undef GEN_VX_LOGICAL 3245#undef GEN_VX_LOGICAL_207 3246#undef GEN_VXFORM 3247#undef GEN_VXFORM_207 3248#undef GEN_VXFORM_DUAL 3249#undef GEN_VXRFORM_DUAL 3250#undef GEN_VXRFORM1 3251#undef GEN_VXRFORM 3252#undef GEN_VXFORM_VSPLTI 3253#undef GEN_VXFORM_NOA 3254#undef GEN_VXFORM_UIMM 3255#undef GEN_VAFORM_PAIRED 3256 3257#undef GEN_BCD2 3258