Switch Mac99 to OpenBIOS
[qemu/mini2440/sniper_sniper_test.git] / target-ppc / op_helper.c
blob715f86bb697864d6ffc3dc667f8b489389602a09
1 /*
2 * PowerPC emulation helpers for qemu.
4 * Copyright (c) 2003-2007 Jocelyn Mayer
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301 USA
20 #include <string.h>
21 #include "exec.h"
22 #include "host-utils.h"
23 #include "helper.h"
25 #include "helper_regs.h"
27 //#define DEBUG_OP
28 //#define DEBUG_EXCEPTIONS
29 //#define DEBUG_SOFTWARE_TLB
31 #ifdef DEBUG_SOFTWARE_TLB
32 # define LOG_SWTLB(...) qemu_log(__VA_ARGS__)
33 #else
34 # define LOG_SWTLB(...) do { } while (0)
35 #endif
38 /*****************************************************************************/
39 /* Exceptions processing helpers */
41 void helper_raise_exception_err (uint32_t exception, uint32_t error_code)
43 #if 0
44 printf("Raise exception %3x code : %d\n", exception, error_code);
45 #endif
46 env->exception_index = exception;
47 env->error_code = error_code;
48 cpu_loop_exit();
51 void helper_raise_exception (uint32_t exception)
53 helper_raise_exception_err(exception, 0);
56 /*****************************************************************************/
57 /* Registers load and stores */
58 target_ulong helper_load_cr (void)
60 return (env->crf[0] << 28) |
61 (env->crf[1] << 24) |
62 (env->crf[2] << 20) |
63 (env->crf[3] << 16) |
64 (env->crf[4] << 12) |
65 (env->crf[5] << 8) |
66 (env->crf[6] << 4) |
67 (env->crf[7] << 0);
70 void helper_store_cr (target_ulong val, uint32_t mask)
72 int i, sh;
74 for (i = 0, sh = 7; i < 8; i++, sh--) {
75 if (mask & (1 << sh))
76 env->crf[i] = (val >> (sh * 4)) & 0xFUL;
80 /*****************************************************************************/
81 /* SPR accesses */
82 void helper_load_dump_spr (uint32_t sprn)
84 qemu_log("Read SPR %d %03x => " ADDRX "\n",
85 sprn, sprn, env->spr[sprn]);
88 void helper_store_dump_spr (uint32_t sprn)
90 qemu_log("Write SPR %d %03x <= " ADDRX "\n",
91 sprn, sprn, env->spr[sprn]);
94 target_ulong helper_load_tbl (void)
96 return cpu_ppc_load_tbl(env);
99 target_ulong helper_load_tbu (void)
101 return cpu_ppc_load_tbu(env);
104 target_ulong helper_load_atbl (void)
106 return cpu_ppc_load_atbl(env);
109 target_ulong helper_load_atbu (void)
111 return cpu_ppc_load_atbu(env);
114 target_ulong helper_load_601_rtcl (void)
116 return cpu_ppc601_load_rtcl(env);
119 target_ulong helper_load_601_rtcu (void)
121 return cpu_ppc601_load_rtcu(env);
124 #if !defined(CONFIG_USER_ONLY)
125 #if defined (TARGET_PPC64)
126 void helper_store_asr (target_ulong val)
128 ppc_store_asr(env, val);
130 #endif
132 void helper_store_sdr1 (target_ulong val)
134 ppc_store_sdr1(env, val);
137 void helper_store_tbl (target_ulong val)
139 cpu_ppc_store_tbl(env, val);
142 void helper_store_tbu (target_ulong val)
144 cpu_ppc_store_tbu(env, val);
147 void helper_store_atbl (target_ulong val)
149 cpu_ppc_store_atbl(env, val);
152 void helper_store_atbu (target_ulong val)
154 cpu_ppc_store_atbu(env, val);
157 void helper_store_601_rtcl (target_ulong val)
159 cpu_ppc601_store_rtcl(env, val);
162 void helper_store_601_rtcu (target_ulong val)
164 cpu_ppc601_store_rtcu(env, val);
167 target_ulong helper_load_decr (void)
169 return cpu_ppc_load_decr(env);
172 void helper_store_decr (target_ulong val)
174 cpu_ppc_store_decr(env, val);
177 void helper_store_hid0_601 (target_ulong val)
179 target_ulong hid0;
181 hid0 = env->spr[SPR_HID0];
182 if ((val ^ hid0) & 0x00000008) {
183 /* Change current endianness */
184 env->hflags &= ~(1 << MSR_LE);
185 env->hflags_nmsr &= ~(1 << MSR_LE);
186 env->hflags_nmsr |= (1 << MSR_LE) & (((val >> 3) & 1) << MSR_LE);
187 env->hflags |= env->hflags_nmsr;
188 qemu_log("%s: set endianness to %c => " ADDRX "\n",
189 __func__, val & 0x8 ? 'l' : 'b', env->hflags);
191 env->spr[SPR_HID0] = (uint32_t)val;
194 void helper_store_403_pbr (uint32_t num, target_ulong value)
196 if (likely(env->pb[num] != value)) {
197 env->pb[num] = value;
198 /* Should be optimized */
199 tlb_flush(env, 1);
203 target_ulong helper_load_40x_pit (void)
205 return load_40x_pit(env);
208 void helper_store_40x_pit (target_ulong val)
210 store_40x_pit(env, val);
213 void helper_store_40x_dbcr0 (target_ulong val)
215 store_40x_dbcr0(env, val);
218 void helper_store_40x_sler (target_ulong val)
220 store_40x_sler(env, val);
223 void helper_store_booke_tcr (target_ulong val)
225 store_booke_tcr(env, val);
228 void helper_store_booke_tsr (target_ulong val)
230 store_booke_tsr(env, val);
233 void helper_store_ibatu (uint32_t nr, target_ulong val)
235 ppc_store_ibatu(env, nr, val);
238 void helper_store_ibatl (uint32_t nr, target_ulong val)
240 ppc_store_ibatl(env, nr, val);
243 void helper_store_dbatu (uint32_t nr, target_ulong val)
245 ppc_store_dbatu(env, nr, val);
248 void helper_store_dbatl (uint32_t nr, target_ulong val)
250 ppc_store_dbatl(env, nr, val);
253 void helper_store_601_batl (uint32_t nr, target_ulong val)
255 ppc_store_ibatl_601(env, nr, val);
258 void helper_store_601_batu (uint32_t nr, target_ulong val)
260 ppc_store_ibatu_601(env, nr, val);
262 #endif
264 /*****************************************************************************/
265 /* Memory load and stores */
267 static always_inline target_ulong addr_add(target_ulong addr, target_long arg)
269 #if defined(TARGET_PPC64)
270 if (!msr_sf)
271 return (uint32_t)(addr + arg);
272 else
273 #endif
274 return addr + arg;
277 void helper_lmw (target_ulong addr, uint32_t reg)
279 for (; reg < 32; reg++) {
280 if (msr_le)
281 env->gpr[reg] = bswap32(ldl(addr));
282 else
283 env->gpr[reg] = ldl(addr);
284 addr = addr_add(addr, 4);
288 void helper_stmw (target_ulong addr, uint32_t reg)
290 for (; reg < 32; reg++) {
291 if (msr_le)
292 stl(addr, bswap32((uint32_t)env->gpr[reg]));
293 else
294 stl(addr, (uint32_t)env->gpr[reg]);
295 addr = addr_add(addr, 4);
299 void helper_lsw(target_ulong addr, uint32_t nb, uint32_t reg)
301 int sh;
302 for (; nb > 3; nb -= 4) {
303 env->gpr[reg] = ldl(addr);
304 reg = (reg + 1) % 32;
305 addr = addr_add(addr, 4);
307 if (unlikely(nb > 0)) {
308 env->gpr[reg] = 0;
309 for (sh = 24; nb > 0; nb--, sh -= 8) {
310 env->gpr[reg] |= ldub(addr) << sh;
311 addr = addr_add(addr, 1);
315 /* PPC32 specification says we must generate an exception if
316 * rA is in the range of registers to be loaded.
317 * In an other hand, IBM says this is valid, but rA won't be loaded.
318 * For now, I'll follow the spec...
320 void helper_lswx(target_ulong addr, uint32_t reg, uint32_t ra, uint32_t rb)
322 if (likely(xer_bc != 0)) {
323 if (unlikely((ra != 0 && reg < ra && (reg + xer_bc) > ra) ||
324 (reg < rb && (reg + xer_bc) > rb))) {
325 helper_raise_exception_err(POWERPC_EXCP_PROGRAM,
326 POWERPC_EXCP_INVAL |
327 POWERPC_EXCP_INVAL_LSWX);
328 } else {
329 helper_lsw(addr, xer_bc, reg);
334 void helper_stsw(target_ulong addr, uint32_t nb, uint32_t reg)
336 int sh;
337 for (; nb > 3; nb -= 4) {
338 stl(addr, env->gpr[reg]);
339 reg = (reg + 1) % 32;
340 addr = addr_add(addr, 4);
342 if (unlikely(nb > 0)) {
343 for (sh = 24; nb > 0; nb--, sh -= 8) {
344 stb(addr, (env->gpr[reg] >> sh) & 0xFF);
345 addr = addr_add(addr, 1);
350 static void do_dcbz(target_ulong addr, int dcache_line_size)
352 addr &= ~(dcache_line_size - 1);
353 int i;
354 for (i = 0 ; i < dcache_line_size ; i += 4) {
355 stl(addr + i , 0);
357 if (env->reserve == addr)
358 env->reserve = (target_ulong)-1ULL;
361 void helper_dcbz(target_ulong addr)
363 do_dcbz(addr, env->dcache_line_size);
366 void helper_dcbz_970(target_ulong addr)
368 if (((env->spr[SPR_970_HID5] >> 7) & 0x3) == 1)
369 do_dcbz(addr, 32);
370 else
371 do_dcbz(addr, env->dcache_line_size);
374 void helper_icbi(target_ulong addr)
376 uint32_t tmp;
378 addr &= ~(env->dcache_line_size - 1);
379 /* Invalidate one cache line :
380 * PowerPC specification says this is to be treated like a load
381 * (not a fetch) by the MMU. To be sure it will be so,
382 * do the load "by hand".
384 tmp = ldl(addr);
385 tb_invalidate_page_range(addr, addr + env->icache_line_size);
388 // XXX: to be tested
389 target_ulong helper_lscbx (target_ulong addr, uint32_t reg, uint32_t ra, uint32_t rb)
391 int i, c, d;
392 d = 24;
393 for (i = 0; i < xer_bc; i++) {
394 c = ldub(addr);
395 addr = addr_add(addr, 1);
396 /* ra (if not 0) and rb are never modified */
397 if (likely(reg != rb && (ra == 0 || reg != ra))) {
398 env->gpr[reg] = (env->gpr[reg] & ~(0xFF << d)) | (c << d);
400 if (unlikely(c == xer_cmp))
401 break;
402 if (likely(d != 0)) {
403 d -= 8;
404 } else {
405 d = 24;
406 reg++;
407 reg = reg & 0x1F;
410 return i;
413 /*****************************************************************************/
414 /* Fixed point operations helpers */
415 #if defined(TARGET_PPC64)
417 /* multiply high word */
418 uint64_t helper_mulhd (uint64_t arg1, uint64_t arg2)
420 uint64_t tl, th;
422 muls64(&tl, &th, arg1, arg2);
423 return th;
426 /* multiply high word unsigned */
427 uint64_t helper_mulhdu (uint64_t arg1, uint64_t arg2)
429 uint64_t tl, th;
431 mulu64(&tl, &th, arg1, arg2);
432 return th;
435 uint64_t helper_mulldo (uint64_t arg1, uint64_t arg2)
437 int64_t th;
438 uint64_t tl;
440 muls64(&tl, (uint64_t *)&th, arg1, arg2);
441 /* If th != 0 && th != -1, then we had an overflow */
442 if (likely((uint64_t)(th + 1) <= 1)) {
443 env->xer &= ~(1 << XER_OV);
444 } else {
445 env->xer |= (1 << XER_OV) | (1 << XER_SO);
447 return (int64_t)tl;
449 #endif
451 target_ulong helper_cntlzw (target_ulong t)
453 return clz32(t);
456 #if defined(TARGET_PPC64)
457 target_ulong helper_cntlzd (target_ulong t)
459 return clz64(t);
461 #endif
463 /* shift right arithmetic helper */
464 target_ulong helper_sraw (target_ulong value, target_ulong shift)
466 int32_t ret;
468 if (likely(!(shift & 0x20))) {
469 if (likely((uint32_t)shift != 0)) {
470 shift &= 0x1f;
471 ret = (int32_t)value >> shift;
472 if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) {
473 env->xer &= ~(1 << XER_CA);
474 } else {
475 env->xer |= (1 << XER_CA);
477 } else {
478 ret = (int32_t)value;
479 env->xer &= ~(1 << XER_CA);
481 } else {
482 ret = (int32_t)value >> 31;
483 if (ret) {
484 env->xer |= (1 << XER_CA);
485 } else {
486 env->xer &= ~(1 << XER_CA);
489 return (target_long)ret;
492 #if defined(TARGET_PPC64)
493 target_ulong helper_srad (target_ulong value, target_ulong shift)
495 int64_t ret;
497 if (likely(!(shift & 0x40))) {
498 if (likely((uint64_t)shift != 0)) {
499 shift &= 0x3f;
500 ret = (int64_t)value >> shift;
501 if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) {
502 env->xer &= ~(1 << XER_CA);
503 } else {
504 env->xer |= (1 << XER_CA);
506 } else {
507 ret = (int64_t)value;
508 env->xer &= ~(1 << XER_CA);
510 } else {
511 ret = (int64_t)value >> 63;
512 if (ret) {
513 env->xer |= (1 << XER_CA);
514 } else {
515 env->xer &= ~(1 << XER_CA);
518 return ret;
520 #endif
522 target_ulong helper_popcntb (target_ulong val)
524 val = (val & 0x55555555) + ((val >> 1) & 0x55555555);
525 val = (val & 0x33333333) + ((val >> 2) & 0x33333333);
526 val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f);
527 return val;
530 #if defined(TARGET_PPC64)
531 target_ulong helper_popcntb_64 (target_ulong val)
533 val = (val & 0x5555555555555555ULL) + ((val >> 1) & 0x5555555555555555ULL);
534 val = (val & 0x3333333333333333ULL) + ((val >> 2) & 0x3333333333333333ULL);
535 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) & 0x0f0f0f0f0f0f0f0fULL);
536 return val;
538 #endif
540 /*****************************************************************************/
541 /* Floating point operations helpers */
542 uint64_t helper_float32_to_float64(uint32_t arg)
544 CPU_FloatU f;
545 CPU_DoubleU d;
546 f.l = arg;
547 d.d = float32_to_float64(f.f, &env->fp_status);
548 return d.ll;
551 uint32_t helper_float64_to_float32(uint64_t arg)
553 CPU_FloatU f;
554 CPU_DoubleU d;
555 d.ll = arg;
556 f.f = float64_to_float32(d.d, &env->fp_status);
557 return f.l;
560 static always_inline int isden (float64 d)
562 CPU_DoubleU u;
564 u.d = d;
566 return ((u.ll >> 52) & 0x7FF) == 0;
569 uint32_t helper_compute_fprf (uint64_t arg, uint32_t set_fprf)
571 CPU_DoubleU farg;
572 int isneg;
573 int ret;
574 farg.ll = arg;
575 isneg = float64_is_neg(farg.d);
576 if (unlikely(float64_is_nan(farg.d))) {
577 if (float64_is_signaling_nan(farg.d)) {
578 /* Signaling NaN: flags are undefined */
579 ret = 0x00;
580 } else {
581 /* Quiet NaN */
582 ret = 0x11;
584 } else if (unlikely(float64_is_infinity(farg.d))) {
585 /* +/- infinity */
586 if (isneg)
587 ret = 0x09;
588 else
589 ret = 0x05;
590 } else {
591 if (float64_is_zero(farg.d)) {
592 /* +/- zero */
593 if (isneg)
594 ret = 0x12;
595 else
596 ret = 0x02;
597 } else {
598 if (isden(farg.d)) {
599 /* Denormalized numbers */
600 ret = 0x10;
601 } else {
602 /* Normalized numbers */
603 ret = 0x00;
605 if (isneg) {
606 ret |= 0x08;
607 } else {
608 ret |= 0x04;
612 if (set_fprf) {
613 /* We update FPSCR_FPRF */
614 env->fpscr &= ~(0x1F << FPSCR_FPRF);
615 env->fpscr |= ret << FPSCR_FPRF;
617 /* We just need fpcc to update Rc1 */
618 return ret & 0xF;
621 /* Floating-point invalid operations exception */
622 static always_inline uint64_t fload_invalid_op_excp (int op)
624 uint64_t ret = 0;
625 int ve;
627 ve = fpscr_ve;
628 switch (op) {
629 case POWERPC_EXCP_FP_VXSNAN:
630 env->fpscr |= 1 << FPSCR_VXSNAN;
631 break;
632 case POWERPC_EXCP_FP_VXSOFT:
633 env->fpscr |= 1 << FPSCR_VXSOFT;
634 break;
635 case POWERPC_EXCP_FP_VXISI:
636 /* Magnitude subtraction of infinities */
637 env->fpscr |= 1 << FPSCR_VXISI;
638 goto update_arith;
639 case POWERPC_EXCP_FP_VXIDI:
640 /* Division of infinity by infinity */
641 env->fpscr |= 1 << FPSCR_VXIDI;
642 goto update_arith;
643 case POWERPC_EXCP_FP_VXZDZ:
644 /* Division of zero by zero */
645 env->fpscr |= 1 << FPSCR_VXZDZ;
646 goto update_arith;
647 case POWERPC_EXCP_FP_VXIMZ:
648 /* Multiplication of zero by infinity */
649 env->fpscr |= 1 << FPSCR_VXIMZ;
650 goto update_arith;
651 case POWERPC_EXCP_FP_VXVC:
652 /* Ordered comparison of NaN */
653 env->fpscr |= 1 << FPSCR_VXVC;
654 env->fpscr &= ~(0xF << FPSCR_FPCC);
655 env->fpscr |= 0x11 << FPSCR_FPCC;
656 /* We must update the target FPR before raising the exception */
657 if (ve != 0) {
658 env->exception_index = POWERPC_EXCP_PROGRAM;
659 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_VXVC;
660 /* Update the floating-point enabled exception summary */
661 env->fpscr |= 1 << FPSCR_FEX;
662 /* Exception is differed */
663 ve = 0;
665 break;
666 case POWERPC_EXCP_FP_VXSQRT:
667 /* Square root of a negative number */
668 env->fpscr |= 1 << FPSCR_VXSQRT;
669 update_arith:
670 env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI));
671 if (ve == 0) {
672 /* Set the result to quiet NaN */
673 ret = 0xFFF8000000000000ULL;
674 env->fpscr &= ~(0xF << FPSCR_FPCC);
675 env->fpscr |= 0x11 << FPSCR_FPCC;
677 break;
678 case POWERPC_EXCP_FP_VXCVI:
679 /* Invalid conversion */
680 env->fpscr |= 1 << FPSCR_VXCVI;
681 env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI));
682 if (ve == 0) {
683 /* Set the result to quiet NaN */
684 ret = 0xFFF8000000000000ULL;
685 env->fpscr &= ~(0xF << FPSCR_FPCC);
686 env->fpscr |= 0x11 << FPSCR_FPCC;
688 break;
690 /* Update the floating-point invalid operation summary */
691 env->fpscr |= 1 << FPSCR_VX;
692 /* Update the floating-point exception summary */
693 env->fpscr |= 1 << FPSCR_FX;
694 if (ve != 0) {
695 /* Update the floating-point enabled exception summary */
696 env->fpscr |= 1 << FPSCR_FEX;
697 if (msr_fe0 != 0 || msr_fe1 != 0)
698 helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_FP | op);
700 return ret;
703 static always_inline void float_zero_divide_excp (void)
705 env->fpscr |= 1 << FPSCR_ZX;
706 env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI));
707 /* Update the floating-point exception summary */
708 env->fpscr |= 1 << FPSCR_FX;
709 if (fpscr_ze != 0) {
710 /* Update the floating-point enabled exception summary */
711 env->fpscr |= 1 << FPSCR_FEX;
712 if (msr_fe0 != 0 || msr_fe1 != 0) {
713 helper_raise_exception_err(POWERPC_EXCP_PROGRAM,
714 POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX);
719 static always_inline void float_overflow_excp (void)
721 env->fpscr |= 1 << FPSCR_OX;
722 /* Update the floating-point exception summary */
723 env->fpscr |= 1 << FPSCR_FX;
724 if (fpscr_oe != 0) {
725 /* XXX: should adjust the result */
726 /* Update the floating-point enabled exception summary */
727 env->fpscr |= 1 << FPSCR_FEX;
728 /* We must update the target FPR before raising the exception */
729 env->exception_index = POWERPC_EXCP_PROGRAM;
730 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX;
731 } else {
732 env->fpscr |= 1 << FPSCR_XX;
733 env->fpscr |= 1 << FPSCR_FI;
737 static always_inline void float_underflow_excp (void)
739 env->fpscr |= 1 << FPSCR_UX;
740 /* Update the floating-point exception summary */
741 env->fpscr |= 1 << FPSCR_FX;
742 if (fpscr_ue != 0) {
743 /* XXX: should adjust the result */
744 /* Update the floating-point enabled exception summary */
745 env->fpscr |= 1 << FPSCR_FEX;
746 /* We must update the target FPR before raising the exception */
747 env->exception_index = POWERPC_EXCP_PROGRAM;
748 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX;
752 static always_inline void float_inexact_excp (void)
754 env->fpscr |= 1 << FPSCR_XX;
755 /* Update the floating-point exception summary */
756 env->fpscr |= 1 << FPSCR_FX;
757 if (fpscr_xe != 0) {
758 /* Update the floating-point enabled exception summary */
759 env->fpscr |= 1 << FPSCR_FEX;
760 /* We must update the target FPR before raising the exception */
761 env->exception_index = POWERPC_EXCP_PROGRAM;
762 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX;
766 static always_inline void fpscr_set_rounding_mode (void)
768 int rnd_type;
770 /* Set rounding mode */
771 switch (fpscr_rn) {
772 case 0:
773 /* Best approximation (round to nearest) */
774 rnd_type = float_round_nearest_even;
775 break;
776 case 1:
777 /* Smaller magnitude (round toward zero) */
778 rnd_type = float_round_to_zero;
779 break;
780 case 2:
781 /* Round toward +infinite */
782 rnd_type = float_round_up;
783 break;
784 default:
785 case 3:
786 /* Round toward -infinite */
787 rnd_type = float_round_down;
788 break;
790 set_float_rounding_mode(rnd_type, &env->fp_status);
793 void helper_fpscr_clrbit (uint32_t bit)
795 int prev;
797 prev = (env->fpscr >> bit) & 1;
798 env->fpscr &= ~(1 << bit);
799 if (prev == 1) {
800 switch (bit) {
801 case FPSCR_RN1:
802 case FPSCR_RN:
803 fpscr_set_rounding_mode();
804 break;
805 default:
806 break;
811 void helper_fpscr_setbit (uint32_t bit)
813 int prev;
815 prev = (env->fpscr >> bit) & 1;
816 env->fpscr |= 1 << bit;
817 if (prev == 0) {
818 switch (bit) {
819 case FPSCR_VX:
820 env->fpscr |= 1 << FPSCR_FX;
821 if (fpscr_ve)
822 goto raise_ve;
823 case FPSCR_OX:
824 env->fpscr |= 1 << FPSCR_FX;
825 if (fpscr_oe)
826 goto raise_oe;
827 break;
828 case FPSCR_UX:
829 env->fpscr |= 1 << FPSCR_FX;
830 if (fpscr_ue)
831 goto raise_ue;
832 break;
833 case FPSCR_ZX:
834 env->fpscr |= 1 << FPSCR_FX;
835 if (fpscr_ze)
836 goto raise_ze;
837 break;
838 case FPSCR_XX:
839 env->fpscr |= 1 << FPSCR_FX;
840 if (fpscr_xe)
841 goto raise_xe;
842 break;
843 case FPSCR_VXSNAN:
844 case FPSCR_VXISI:
845 case FPSCR_VXIDI:
846 case FPSCR_VXZDZ:
847 case FPSCR_VXIMZ:
848 case FPSCR_VXVC:
849 case FPSCR_VXSOFT:
850 case FPSCR_VXSQRT:
851 case FPSCR_VXCVI:
852 env->fpscr |= 1 << FPSCR_VX;
853 env->fpscr |= 1 << FPSCR_FX;
854 if (fpscr_ve != 0)
855 goto raise_ve;
856 break;
857 case FPSCR_VE:
858 if (fpscr_vx != 0) {
859 raise_ve:
860 env->error_code = POWERPC_EXCP_FP;
861 if (fpscr_vxsnan)
862 env->error_code |= POWERPC_EXCP_FP_VXSNAN;
863 if (fpscr_vxisi)
864 env->error_code |= POWERPC_EXCP_FP_VXISI;
865 if (fpscr_vxidi)
866 env->error_code |= POWERPC_EXCP_FP_VXIDI;
867 if (fpscr_vxzdz)
868 env->error_code |= POWERPC_EXCP_FP_VXZDZ;
869 if (fpscr_vximz)
870 env->error_code |= POWERPC_EXCP_FP_VXIMZ;
871 if (fpscr_vxvc)
872 env->error_code |= POWERPC_EXCP_FP_VXVC;
873 if (fpscr_vxsoft)
874 env->error_code |= POWERPC_EXCP_FP_VXSOFT;
875 if (fpscr_vxsqrt)
876 env->error_code |= POWERPC_EXCP_FP_VXSQRT;
877 if (fpscr_vxcvi)
878 env->error_code |= POWERPC_EXCP_FP_VXCVI;
879 goto raise_excp;
881 break;
882 case FPSCR_OE:
883 if (fpscr_ox != 0) {
884 raise_oe:
885 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX;
886 goto raise_excp;
888 break;
889 case FPSCR_UE:
890 if (fpscr_ux != 0) {
891 raise_ue:
892 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX;
893 goto raise_excp;
895 break;
896 case FPSCR_ZE:
897 if (fpscr_zx != 0) {
898 raise_ze:
899 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX;
900 goto raise_excp;
902 break;
903 case FPSCR_XE:
904 if (fpscr_xx != 0) {
905 raise_xe:
906 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX;
907 goto raise_excp;
909 break;
910 case FPSCR_RN1:
911 case FPSCR_RN:
912 fpscr_set_rounding_mode();
913 break;
914 default:
915 break;
916 raise_excp:
917 /* Update the floating-point enabled exception summary */
918 env->fpscr |= 1 << FPSCR_FEX;
919 /* We have to update Rc1 before raising the exception */
920 env->exception_index = POWERPC_EXCP_PROGRAM;
921 break;
926 void helper_store_fpscr (uint64_t arg, uint32_t mask)
929 * We use only the 32 LSB of the incoming fpr
931 uint32_t prev, new;
932 int i;
934 prev = env->fpscr;
935 new = (uint32_t)arg;
936 new &= ~0x60000000;
937 new |= prev & 0x60000000;
938 for (i = 0; i < 8; i++) {
939 if (mask & (1 << i)) {
940 env->fpscr &= ~(0xF << (4 * i));
941 env->fpscr |= new & (0xF << (4 * i));
944 /* Update VX and FEX */
945 if (fpscr_ix != 0)
946 env->fpscr |= 1 << FPSCR_VX;
947 else
948 env->fpscr &= ~(1 << FPSCR_VX);
949 if ((fpscr_ex & fpscr_eex) != 0) {
950 env->fpscr |= 1 << FPSCR_FEX;
951 env->exception_index = POWERPC_EXCP_PROGRAM;
952 /* XXX: we should compute it properly */
953 env->error_code = POWERPC_EXCP_FP;
955 else
956 env->fpscr &= ~(1 << FPSCR_FEX);
957 fpscr_set_rounding_mode();
960 void helper_float_check_status (void)
962 #ifdef CONFIG_SOFTFLOAT
963 if (env->exception_index == POWERPC_EXCP_PROGRAM &&
964 (env->error_code & POWERPC_EXCP_FP)) {
965 /* Differred floating-point exception after target FPR update */
966 if (msr_fe0 != 0 || msr_fe1 != 0)
967 helper_raise_exception_err(env->exception_index, env->error_code);
968 } else {
969 int status = get_float_exception_flags(&env->fp_status);
970 if (status & float_flag_divbyzero) {
971 float_zero_divide_excp();
972 } else if (status & float_flag_overflow) {
973 float_overflow_excp();
974 } else if (status & float_flag_underflow) {
975 float_underflow_excp();
976 } else if (status & float_flag_inexact) {
977 float_inexact_excp();
980 #else
981 if (env->exception_index == POWERPC_EXCP_PROGRAM &&
982 (env->error_code & POWERPC_EXCP_FP)) {
983 /* Differred floating-point exception after target FPR update */
984 if (msr_fe0 != 0 || msr_fe1 != 0)
985 helper_raise_exception_err(env->exception_index, env->error_code);
987 #endif
990 #ifdef CONFIG_SOFTFLOAT
991 void helper_reset_fpstatus (void)
993 set_float_exception_flags(0, &env->fp_status);
995 #endif
997 /* fadd - fadd. */
998 uint64_t helper_fadd (uint64_t arg1, uint64_t arg2)
1000 CPU_DoubleU farg1, farg2;
1002 farg1.ll = arg1;
1003 farg2.ll = arg2;
1004 #if USE_PRECISE_EMULATION
1005 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1006 float64_is_signaling_nan(farg2.d))) {
1007 /* sNaN addition */
1008 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1009 } else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
1010 float64_is_neg(farg1.d) != float64_is_neg(farg2.d))) {
1011 /* Magnitude subtraction of infinities */
1012 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
1013 } else {
1014 farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status);
1016 #else
1017 farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status);
1018 #endif
1019 return farg1.ll;
1022 /* fsub - fsub. */
1023 uint64_t helper_fsub (uint64_t arg1, uint64_t arg2)
1025 CPU_DoubleU farg1, farg2;
1027 farg1.ll = arg1;
1028 farg2.ll = arg2;
1029 #if USE_PRECISE_EMULATION
1031 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1032 float64_is_signaling_nan(farg2.d))) {
1033 /* sNaN subtraction */
1034 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1035 } else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
1036 float64_is_neg(farg1.d) == float64_is_neg(farg2.d))) {
1037 /* Magnitude subtraction of infinities */
1038 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
1039 } else {
1040 farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status);
1043 #else
1044 farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status);
1045 #endif
1046 return farg1.ll;
1049 /* fmul - fmul. */
1050 uint64_t helper_fmul (uint64_t arg1, uint64_t arg2)
1052 CPU_DoubleU farg1, farg2;
1054 farg1.ll = arg1;
1055 farg2.ll = arg2;
1056 #if USE_PRECISE_EMULATION
1057 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1058 float64_is_signaling_nan(farg2.d))) {
1059 /* sNaN multiplication */
1060 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1061 } else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
1062 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
1063 /* Multiplication of zero by infinity */
1064 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
1065 } else {
1066 farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
1068 #else
1069 farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
1070 #endif
1071 return farg1.ll;
1074 /* fdiv - fdiv. */
1075 uint64_t helper_fdiv (uint64_t arg1, uint64_t arg2)
1077 CPU_DoubleU farg1, farg2;
1079 farg1.ll = arg1;
1080 farg2.ll = arg2;
1081 #if USE_PRECISE_EMULATION
1082 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1083 float64_is_signaling_nan(farg2.d))) {
1084 /* sNaN division */
1085 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1086 } else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d))) {
1087 /* Division of infinity by infinity */
1088 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIDI);
1089 } else if (unlikely(float64_is_zero(farg1.d) && float64_is_zero(farg2.d))) {
1090 /* Division of zero by zero */
1091 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXZDZ);
1092 } else {
1093 farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status);
1095 #else
1096 farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status);
1097 #endif
1098 return farg1.ll;
1101 /* fabs */
1102 uint64_t helper_fabs (uint64_t arg)
1104 CPU_DoubleU farg;
1106 farg.ll = arg;
1107 farg.d = float64_abs(farg.d);
1108 return farg.ll;
1111 /* fnabs */
1112 uint64_t helper_fnabs (uint64_t arg)
1114 CPU_DoubleU farg;
1116 farg.ll = arg;
1117 farg.d = float64_abs(farg.d);
1118 farg.d = float64_chs(farg.d);
1119 return farg.ll;
1122 /* fneg */
1123 uint64_t helper_fneg (uint64_t arg)
1125 CPU_DoubleU farg;
1127 farg.ll = arg;
1128 farg.d = float64_chs(farg.d);
1129 return farg.ll;
1132 /* fctiw - fctiw. */
1133 uint64_t helper_fctiw (uint64_t arg)
1135 CPU_DoubleU farg;
1136 farg.ll = arg;
1138 if (unlikely(float64_is_signaling_nan(farg.d))) {
1139 /* sNaN conversion */
1140 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI);
1141 } else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) {
1142 /* qNan / infinity conversion */
1143 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI);
1144 } else {
1145 farg.ll = float64_to_int32(farg.d, &env->fp_status);
1146 #if USE_PRECISE_EMULATION
1147 /* XXX: higher bits are not supposed to be significant.
1148 * to make tests easier, return the same as a real PowerPC 750
1150 farg.ll |= 0xFFF80000ULL << 32;
1151 #endif
1153 return farg.ll;
1156 /* fctiwz - fctiwz. */
1157 uint64_t helper_fctiwz (uint64_t arg)
1159 CPU_DoubleU farg;
1160 farg.ll = arg;
1162 if (unlikely(float64_is_signaling_nan(farg.d))) {
1163 /* sNaN conversion */
1164 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI);
1165 } else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) {
1166 /* qNan / infinity conversion */
1167 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI);
1168 } else {
1169 farg.ll = float64_to_int32_round_to_zero(farg.d, &env->fp_status);
1170 #if USE_PRECISE_EMULATION
1171 /* XXX: higher bits are not supposed to be significant.
1172 * to make tests easier, return the same as a real PowerPC 750
1174 farg.ll |= 0xFFF80000ULL << 32;
1175 #endif
1177 return farg.ll;
1180 #if defined(TARGET_PPC64)
1181 /* fcfid - fcfid. */
1182 uint64_t helper_fcfid (uint64_t arg)
1184 CPU_DoubleU farg;
1185 farg.d = int64_to_float64(arg, &env->fp_status);
1186 return farg.ll;
1189 /* fctid - fctid. */
1190 uint64_t helper_fctid (uint64_t arg)
1192 CPU_DoubleU farg;
1193 farg.ll = arg;
1195 if (unlikely(float64_is_signaling_nan(farg.d))) {
1196 /* sNaN conversion */
1197 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI);
1198 } else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) {
1199 /* qNan / infinity conversion */
1200 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI);
1201 } else {
1202 farg.ll = float64_to_int64(farg.d, &env->fp_status);
1204 return farg.ll;
1207 /* fctidz - fctidz. */
1208 uint64_t helper_fctidz (uint64_t arg)
1210 CPU_DoubleU farg;
1211 farg.ll = arg;
1213 if (unlikely(float64_is_signaling_nan(farg.d))) {
1214 /* sNaN conversion */
1215 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI);
1216 } else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) {
1217 /* qNan / infinity conversion */
1218 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI);
1219 } else {
1220 farg.ll = float64_to_int64_round_to_zero(farg.d, &env->fp_status);
1222 return farg.ll;
1225 #endif
1227 static always_inline uint64_t do_fri (uint64_t arg, int rounding_mode)
1229 CPU_DoubleU farg;
1230 farg.ll = arg;
1232 if (unlikely(float64_is_signaling_nan(farg.d))) {
1233 /* sNaN round */
1234 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI);
1235 } else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) {
1236 /* qNan / infinity round */
1237 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI);
1238 } else {
1239 set_float_rounding_mode(rounding_mode, &env->fp_status);
1240 farg.ll = float64_round_to_int(farg.d, &env->fp_status);
1241 /* Restore rounding mode from FPSCR */
1242 fpscr_set_rounding_mode();
1244 return farg.ll;
1247 uint64_t helper_frin (uint64_t arg)
1249 return do_fri(arg, float_round_nearest_even);
1252 uint64_t helper_friz (uint64_t arg)
1254 return do_fri(arg, float_round_to_zero);
1257 uint64_t helper_frip (uint64_t arg)
1259 return do_fri(arg, float_round_up);
1262 uint64_t helper_frim (uint64_t arg)
1264 return do_fri(arg, float_round_down);
1267 /* fmadd - fmadd. */
1268 uint64_t helper_fmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3)
1270 CPU_DoubleU farg1, farg2, farg3;
1272 farg1.ll = arg1;
1273 farg2.ll = arg2;
1274 farg3.ll = arg3;
1275 #if USE_PRECISE_EMULATION
1276 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1277 float64_is_signaling_nan(farg2.d) ||
1278 float64_is_signaling_nan(farg3.d))) {
1279 /* sNaN operation */
1280 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1281 } else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
1282 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
1283 /* Multiplication of zero by infinity */
1284 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
1285 } else {
1286 #ifdef FLOAT128
1287 /* This is the way the PowerPC specification defines it */
1288 float128 ft0_128, ft1_128;
1290 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
1291 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
1292 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
1293 if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
1294 float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) {
1295 /* Magnitude subtraction of infinities */
1296 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
1297 } else {
1298 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
1299 ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status);
1300 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
1302 #else
1303 /* This is OK on x86 hosts */
1304 farg1.d = (farg1.d * farg2.d) + farg3.d;
1305 #endif
1307 #else
1308 farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
1309 farg1.d = float64_add(farg1.d, farg3.d, &env->fp_status);
1310 #endif
1311 return farg1.ll;
1314 /* fmsub - fmsub. */
1315 uint64_t helper_fmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3)
1317 CPU_DoubleU farg1, farg2, farg3;
1319 farg1.ll = arg1;
1320 farg2.ll = arg2;
1321 farg3.ll = arg3;
1322 #if USE_PRECISE_EMULATION
1323 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1324 float64_is_signaling_nan(farg2.d) ||
1325 float64_is_signaling_nan(farg3.d))) {
1326 /* sNaN operation */
1327 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1328 } else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
1329 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
1330 /* Multiplication of zero by infinity */
1331 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
1332 } else {
1333 #ifdef FLOAT128
1334 /* This is the way the PowerPC specification defines it */
1335 float128 ft0_128, ft1_128;
1337 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
1338 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
1339 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
1340 if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
1341 float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) {
1342 /* Magnitude subtraction of infinities */
1343 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
1344 } else {
1345 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
1346 ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status);
1347 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
1349 #else
1350 /* This is OK on x86 hosts */
1351 farg1.d = (farg1.d * farg2.d) - farg3.d;
1352 #endif
1354 #else
1355 farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
1356 farg1.d = float64_sub(farg1.d, farg3.d, &env->fp_status);
1357 #endif
1358 return farg1.ll;
1361 /* fnmadd - fnmadd. */
1362 uint64_t helper_fnmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3)
1364 CPU_DoubleU farg1, farg2, farg3;
1366 farg1.ll = arg1;
1367 farg2.ll = arg2;
1368 farg3.ll = arg3;
1370 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1371 float64_is_signaling_nan(farg2.d) ||
1372 float64_is_signaling_nan(farg3.d))) {
1373 /* sNaN operation */
1374 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1375 } else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
1376 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
1377 /* Multiplication of zero by infinity */
1378 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
1379 } else {
1380 #if USE_PRECISE_EMULATION
1381 #ifdef FLOAT128
1382 /* This is the way the PowerPC specification defines it */
1383 float128 ft0_128, ft1_128;
1385 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
1386 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
1387 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
1388 if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
1389 float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) {
1390 /* Magnitude subtraction of infinities */
1391 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
1392 } else {
1393 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
1394 ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status);
1395 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
1397 #else
1398 /* This is OK on x86 hosts */
1399 farg1.d = (farg1.d * farg2.d) + farg3.d;
1400 #endif
1401 #else
1402 farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
1403 farg1.d = float64_add(farg1.d, farg3.d, &env->fp_status);
1404 #endif
1405 if (likely(!float64_is_nan(farg1.d)))
1406 farg1.d = float64_chs(farg1.d);
1408 return farg1.ll;
1411 /* fnmsub - fnmsub. */
1412 uint64_t helper_fnmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3)
1414 CPU_DoubleU farg1, farg2, farg3;
1416 farg1.ll = arg1;
1417 farg2.ll = arg2;
1418 farg3.ll = arg3;
1420 if (unlikely(float64_is_signaling_nan(farg1.d) ||
1421 float64_is_signaling_nan(farg2.d) ||
1422 float64_is_signaling_nan(farg3.d))) {
1423 /* sNaN operation */
1424 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1425 } else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
1426 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
1427 /* Multiplication of zero by infinity */
1428 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ);
1429 } else {
1430 #if USE_PRECISE_EMULATION
1431 #ifdef FLOAT128
1432 /* This is the way the PowerPC specification defines it */
1433 float128 ft0_128, ft1_128;
1435 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
1436 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
1437 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
1438 if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
1439 float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) {
1440 /* Magnitude subtraction of infinities */
1441 farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI);
1442 } else {
1443 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
1444 ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status);
1445 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
1447 #else
1448 /* This is OK on x86 hosts */
1449 farg1.d = (farg1.d * farg2.d) - farg3.d;
1450 #endif
1451 #else
1452 farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
1453 farg1.d = float64_sub(farg1.d, farg3.d, &env->fp_status);
1454 #endif
1455 if (likely(!float64_is_nan(farg1.d)))
1456 farg1.d = float64_chs(farg1.d);
1458 return farg1.ll;
1461 /* frsp - frsp. */
1462 uint64_t helper_frsp (uint64_t arg)
1464 CPU_DoubleU farg;
1465 float32 f32;
1466 farg.ll = arg;
1468 #if USE_PRECISE_EMULATION
1469 if (unlikely(float64_is_signaling_nan(farg.d))) {
1470 /* sNaN square root */
1471 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1472 } else {
1473 f32 = float64_to_float32(farg.d, &env->fp_status);
1474 farg.d = float32_to_float64(f32, &env->fp_status);
1476 #else
1477 f32 = float64_to_float32(farg.d, &env->fp_status);
1478 farg.d = float32_to_float64(f32, &env->fp_status);
1479 #endif
1480 return farg.ll;
1483 /* fsqrt - fsqrt. */
1484 uint64_t helper_fsqrt (uint64_t arg)
1486 CPU_DoubleU farg;
1487 farg.ll = arg;
1489 if (unlikely(float64_is_signaling_nan(farg.d))) {
1490 /* sNaN square root */
1491 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1492 } else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
1493 /* Square root of a negative nonzero number */
1494 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT);
1495 } else {
1496 farg.d = float64_sqrt(farg.d, &env->fp_status);
1498 return farg.ll;
1501 /* fre - fre. */
1502 uint64_t helper_fre (uint64_t arg)
1504 CPU_DoubleU farg;
1505 farg.ll = arg;
1507 if (unlikely(float64_is_signaling_nan(farg.d))) {
1508 /* sNaN reciprocal */
1509 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1510 } else {
1511 farg.d = float64_div(float64_one, farg.d, &env->fp_status);
1513 return farg.d;
1516 /* fres - fres. */
1517 uint64_t helper_fres (uint64_t arg)
1519 CPU_DoubleU farg;
1520 float32 f32;
1521 farg.ll = arg;
1523 if (unlikely(float64_is_signaling_nan(farg.d))) {
1524 /* sNaN reciprocal */
1525 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1526 } else {
1527 farg.d = float64_div(float64_one, farg.d, &env->fp_status);
1528 f32 = float64_to_float32(farg.d, &env->fp_status);
1529 farg.d = float32_to_float64(f32, &env->fp_status);
1531 return farg.ll;
1534 /* frsqrte - frsqrte. */
1535 uint64_t helper_frsqrte (uint64_t arg)
1537 CPU_DoubleU farg;
1538 float32 f32;
1539 farg.ll = arg;
1541 if (unlikely(float64_is_signaling_nan(farg.d))) {
1542 /* sNaN reciprocal square root */
1543 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1544 } else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
1545 /* Reciprocal square root of a negative nonzero number */
1546 farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT);
1547 } else {
1548 farg.d = float64_sqrt(farg.d, &env->fp_status);
1549 farg.d = float64_div(float64_one, farg.d, &env->fp_status);
1550 f32 = float64_to_float32(farg.d, &env->fp_status);
1551 farg.d = float32_to_float64(f32, &env->fp_status);
1553 return farg.ll;
1556 /* fsel - fsel. */
1557 uint64_t helper_fsel (uint64_t arg1, uint64_t arg2, uint64_t arg3)
1559 CPU_DoubleU farg1;
1561 farg1.ll = arg1;
1563 if ((!float64_is_neg(farg1.d) || float64_is_zero(farg1.d)) && !float64_is_nan(farg1.d))
1564 return arg2;
1565 else
1566 return arg3;
1569 void helper_fcmpu (uint64_t arg1, uint64_t arg2, uint32_t crfD)
1571 CPU_DoubleU farg1, farg2;
1572 uint32_t ret = 0;
1573 farg1.ll = arg1;
1574 farg2.ll = arg2;
1576 if (unlikely(float64_is_nan(farg1.d) ||
1577 float64_is_nan(farg2.d))) {
1578 ret = 0x01UL;
1579 } else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) {
1580 ret = 0x08UL;
1581 } else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) {
1582 ret = 0x04UL;
1583 } else {
1584 ret = 0x02UL;
1587 env->fpscr &= ~(0x0F << FPSCR_FPRF);
1588 env->fpscr |= ret << FPSCR_FPRF;
1589 env->crf[crfD] = ret;
1590 if (unlikely(ret == 0x01UL
1591 && (float64_is_signaling_nan(farg1.d) ||
1592 float64_is_signaling_nan(farg2.d)))) {
1593 /* sNaN comparison */
1594 fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN);
1598 void helper_fcmpo (uint64_t arg1, uint64_t arg2, uint32_t crfD)
1600 CPU_DoubleU farg1, farg2;
1601 uint32_t ret = 0;
1602 farg1.ll = arg1;
1603 farg2.ll = arg2;
1605 if (unlikely(float64_is_nan(farg1.d) ||
1606 float64_is_nan(farg2.d))) {
1607 ret = 0x01UL;
1608 } else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) {
1609 ret = 0x08UL;
1610 } else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) {
1611 ret = 0x04UL;
1612 } else {
1613 ret = 0x02UL;
1616 env->fpscr &= ~(0x0F << FPSCR_FPRF);
1617 env->fpscr |= ret << FPSCR_FPRF;
1618 env->crf[crfD] = ret;
1619 if (unlikely (ret == 0x01UL)) {
1620 if (float64_is_signaling_nan(farg1.d) ||
1621 float64_is_signaling_nan(farg2.d)) {
1622 /* sNaN comparison */
1623 fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN |
1624 POWERPC_EXCP_FP_VXVC);
1625 } else {
1626 /* qNaN comparison */
1627 fload_invalid_op_excp(POWERPC_EXCP_FP_VXVC);
1632 #if !defined (CONFIG_USER_ONLY)
1633 void helper_store_msr (target_ulong val)
1635 val = hreg_store_msr(env, val, 0);
1636 if (val != 0) {
1637 env->interrupt_request |= CPU_INTERRUPT_EXITTB;
1638 helper_raise_exception(val);
1642 static always_inline void do_rfi (target_ulong nip, target_ulong msr,
1643 target_ulong msrm, int keep_msrh)
1645 #if defined(TARGET_PPC64)
1646 if (msr & (1ULL << MSR_SF)) {
1647 nip = (uint64_t)nip;
1648 msr &= (uint64_t)msrm;
1649 } else {
1650 nip = (uint32_t)nip;
1651 msr = (uint32_t)(msr & msrm);
1652 if (keep_msrh)
1653 msr |= env->msr & ~((uint64_t)0xFFFFFFFF);
1655 #else
1656 nip = (uint32_t)nip;
1657 msr &= (uint32_t)msrm;
1658 #endif
1659 /* XXX: beware: this is false if VLE is supported */
1660 env->nip = nip & ~((target_ulong)0x00000003);
1661 hreg_store_msr(env, msr, 1);
1662 #if defined (DEBUG_OP)
1663 cpu_dump_rfi(env->nip, env->msr);
1664 #endif
1665 /* No need to raise an exception here,
1666 * as rfi is always the last insn of a TB
1668 env->interrupt_request |= CPU_INTERRUPT_EXITTB;
1671 void helper_rfi (void)
1673 do_rfi(env->spr[SPR_SRR0], env->spr[SPR_SRR1],
1674 ~((target_ulong)0xFFFF0000), 1);
1677 #if defined(TARGET_PPC64)
1678 void helper_rfid (void)
1680 do_rfi(env->spr[SPR_SRR0], env->spr[SPR_SRR1],
1681 ~((target_ulong)0xFFFF0000), 0);
1684 void helper_hrfid (void)
1686 do_rfi(env->spr[SPR_HSRR0], env->spr[SPR_HSRR1],
1687 ~((target_ulong)0xFFFF0000), 0);
1689 #endif
1690 #endif
1692 void helper_tw (target_ulong arg1, target_ulong arg2, uint32_t flags)
1694 if (!likely(!(((int32_t)arg1 < (int32_t)arg2 && (flags & 0x10)) ||
1695 ((int32_t)arg1 > (int32_t)arg2 && (flags & 0x08)) ||
1696 ((int32_t)arg1 == (int32_t)arg2 && (flags & 0x04)) ||
1697 ((uint32_t)arg1 < (uint32_t)arg2 && (flags & 0x02)) ||
1698 ((uint32_t)arg1 > (uint32_t)arg2 && (flags & 0x01))))) {
1699 helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_TRAP);
1703 #if defined(TARGET_PPC64)
1704 void helper_td (target_ulong arg1, target_ulong arg2, uint32_t flags)
1706 if (!likely(!(((int64_t)arg1 < (int64_t)arg2 && (flags & 0x10)) ||
1707 ((int64_t)arg1 > (int64_t)arg2 && (flags & 0x08)) ||
1708 ((int64_t)arg1 == (int64_t)arg2 && (flags & 0x04)) ||
1709 ((uint64_t)arg1 < (uint64_t)arg2 && (flags & 0x02)) ||
1710 ((uint64_t)arg1 > (uint64_t)arg2 && (flags & 0x01)))))
1711 helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_TRAP);
1713 #endif
1715 /*****************************************************************************/
1716 /* PowerPC 601 specific instructions (POWER bridge) */
1718 target_ulong helper_clcs (uint32_t arg)
1720 switch (arg) {
1721 case 0x0CUL:
1722 /* Instruction cache line size */
1723 return env->icache_line_size;
1724 break;
1725 case 0x0DUL:
1726 /* Data cache line size */
1727 return env->dcache_line_size;
1728 break;
1729 case 0x0EUL:
1730 /* Minimum cache line size */
1731 return (env->icache_line_size < env->dcache_line_size) ?
1732 env->icache_line_size : env->dcache_line_size;
1733 break;
1734 case 0x0FUL:
1735 /* Maximum cache line size */
1736 return (env->icache_line_size > env->dcache_line_size) ?
1737 env->icache_line_size : env->dcache_line_size;
1738 break;
1739 default:
1740 /* Undefined */
1741 return 0;
1742 break;
1746 target_ulong helper_div (target_ulong arg1, target_ulong arg2)
1748 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
1750 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
1751 (int32_t)arg2 == 0) {
1752 env->spr[SPR_MQ] = 0;
1753 return INT32_MIN;
1754 } else {
1755 env->spr[SPR_MQ] = tmp % arg2;
1756 return tmp / (int32_t)arg2;
1760 target_ulong helper_divo (target_ulong arg1, target_ulong arg2)
1762 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
1764 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
1765 (int32_t)arg2 == 0) {
1766 env->xer |= (1 << XER_OV) | (1 << XER_SO);
1767 env->spr[SPR_MQ] = 0;
1768 return INT32_MIN;
1769 } else {
1770 env->spr[SPR_MQ] = tmp % arg2;
1771 tmp /= (int32_t)arg2;
1772 if ((int32_t)tmp != tmp) {
1773 env->xer |= (1 << XER_OV) | (1 << XER_SO);
1774 } else {
1775 env->xer &= ~(1 << XER_OV);
1777 return tmp;
1781 target_ulong helper_divs (target_ulong arg1, target_ulong arg2)
1783 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
1784 (int32_t)arg2 == 0) {
1785 env->spr[SPR_MQ] = 0;
1786 return INT32_MIN;
1787 } else {
1788 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
1789 return (int32_t)arg1 / (int32_t)arg2;
1793 target_ulong helper_divso (target_ulong arg1, target_ulong arg2)
1795 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
1796 (int32_t)arg2 == 0) {
1797 env->xer |= (1 << XER_OV) | (1 << XER_SO);
1798 env->spr[SPR_MQ] = 0;
1799 return INT32_MIN;
1800 } else {
1801 env->xer &= ~(1 << XER_OV);
1802 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
1803 return (int32_t)arg1 / (int32_t)arg2;
1807 #if !defined (CONFIG_USER_ONLY)
1808 target_ulong helper_rac (target_ulong addr)
1810 mmu_ctx_t ctx;
1811 int nb_BATs;
1812 target_ulong ret = 0;
1814 /* We don't have to generate many instances of this instruction,
1815 * as rac is supervisor only.
1817 /* XXX: FIX THIS: Pretend we have no BAT */
1818 nb_BATs = env->nb_BATs;
1819 env->nb_BATs = 0;
1820 if (get_physical_address(env, &ctx, addr, 0, ACCESS_INT) == 0)
1821 ret = ctx.raddr;
1822 env->nb_BATs = nb_BATs;
1823 return ret;
1826 void helper_rfsvc (void)
1828 do_rfi(env->lr, env->ctr, 0x0000FFFF, 0);
1830 #endif
1832 /*****************************************************************************/
1833 /* 602 specific instructions */
1834 /* mfrom is the most crazy instruction ever seen, imho ! */
1835 /* Real implementation uses a ROM table. Do the same */
1836 /* Extremly decomposed:
1837 * -arg / 256
1838 * return 256 * log10(10 + 1.0) + 0.5
1840 #if !defined (CONFIG_USER_ONLY)
1841 target_ulong helper_602_mfrom (target_ulong arg)
1843 if (likely(arg < 602)) {
1844 #include "mfrom_table.c"
1845 return mfrom_ROM_table[arg];
1846 } else {
1847 return 0;
1850 #endif
1852 /*****************************************************************************/
1853 /* Embedded PowerPC specific helpers */
1855 /* XXX: to be improved to check access rights when in user-mode */
1856 target_ulong helper_load_dcr (target_ulong dcrn)
1858 target_ulong val = 0;
1860 if (unlikely(env->dcr_env == NULL)) {
1861 qemu_log("No DCR environment\n");
1862 helper_raise_exception_err(POWERPC_EXCP_PROGRAM,
1863 POWERPC_EXCP_INVAL | POWERPC_EXCP_INVAL_INVAL);
1864 } else if (unlikely(ppc_dcr_read(env->dcr_env, dcrn, &val) != 0)) {
1865 qemu_log("DCR read error %d %03x\n", (int)dcrn, (int)dcrn);
1866 helper_raise_exception_err(POWERPC_EXCP_PROGRAM,
1867 POWERPC_EXCP_INVAL | POWERPC_EXCP_PRIV_REG);
1869 return val;
1872 void helper_store_dcr (target_ulong dcrn, target_ulong val)
1874 if (unlikely(env->dcr_env == NULL)) {
1875 qemu_log("No DCR environment\n");
1876 helper_raise_exception_err(POWERPC_EXCP_PROGRAM,
1877 POWERPC_EXCP_INVAL | POWERPC_EXCP_INVAL_INVAL);
1878 } else if (unlikely(ppc_dcr_write(env->dcr_env, dcrn, val) != 0)) {
1879 qemu_log("DCR write error %d %03x\n", (int)dcrn, (int)dcrn);
1880 helper_raise_exception_err(POWERPC_EXCP_PROGRAM,
1881 POWERPC_EXCP_INVAL | POWERPC_EXCP_PRIV_REG);
1885 #if !defined(CONFIG_USER_ONLY)
1886 void helper_40x_rfci (void)
1888 do_rfi(env->spr[SPR_40x_SRR2], env->spr[SPR_40x_SRR3],
1889 ~((target_ulong)0xFFFF0000), 0);
1892 void helper_rfci (void)
1894 do_rfi(env->spr[SPR_BOOKE_CSRR0], SPR_BOOKE_CSRR1,
1895 ~((target_ulong)0x3FFF0000), 0);
1898 void helper_rfdi (void)
1900 do_rfi(env->spr[SPR_BOOKE_DSRR0], SPR_BOOKE_DSRR1,
1901 ~((target_ulong)0x3FFF0000), 0);
1904 void helper_rfmci (void)
1906 do_rfi(env->spr[SPR_BOOKE_MCSRR0], SPR_BOOKE_MCSRR1,
1907 ~((target_ulong)0x3FFF0000), 0);
1909 #endif
1911 /* 440 specific */
1912 target_ulong helper_dlmzb (target_ulong high, target_ulong low, uint32_t update_Rc)
1914 target_ulong mask;
1915 int i;
1917 i = 1;
1918 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
1919 if ((high & mask) == 0) {
1920 if (update_Rc) {
1921 env->crf[0] = 0x4;
1923 goto done;
1925 i++;
1927 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
1928 if ((low & mask) == 0) {
1929 if (update_Rc) {
1930 env->crf[0] = 0x8;
1932 goto done;
1934 i++;
1936 if (update_Rc) {
1937 env->crf[0] = 0x2;
1939 done:
1940 env->xer = (env->xer & ~0x7F) | i;
1941 if (update_Rc) {
1942 env->crf[0] |= xer_so;
1944 return i;
1947 /*****************************************************************************/
1948 /* Altivec extension helpers */
1949 #if defined(WORDS_BIGENDIAN)
1950 #define HI_IDX 0
1951 #define LO_IDX 1
1952 #else
1953 #define HI_IDX 1
1954 #define LO_IDX 0
1955 #endif
1957 #if defined(WORDS_BIGENDIAN)
1958 #define VECTOR_FOR_INORDER_I(index, element) \
1959 for (index = 0; index < ARRAY_SIZE(r->element); index++)
1960 #else
1961 #define VECTOR_FOR_INORDER_I(index, element) \
1962 for (index = ARRAY_SIZE(r->element)-1; index >= 0; index--)
1963 #endif
1965 /* If X is a NaN, store the corresponding QNaN into RESULT. Otherwise,
1966 * execute the following block. */
1967 #define DO_HANDLE_NAN(result, x) \
1968 if (float32_is_nan(x) || float32_is_signaling_nan(x)) { \
1969 CPU_FloatU __f; \
1970 __f.f = x; \
1971 __f.l = __f.l | (1 << 22); /* Set QNaN bit. */ \
1972 result = __f.f; \
1973 } else
1975 #define HANDLE_NAN1(result, x) \
1976 DO_HANDLE_NAN(result, x)
1977 #define HANDLE_NAN2(result, x, y) \
1978 DO_HANDLE_NAN(result, x) DO_HANDLE_NAN(result, y)
1979 #define HANDLE_NAN3(result, x, y, z) \
1980 DO_HANDLE_NAN(result, x) DO_HANDLE_NAN(result, y) DO_HANDLE_NAN(result, z)
1982 /* Saturating arithmetic helpers. */
1983 #define SATCVT(from, to, from_type, to_type, min, max, use_min, use_max) \
1984 static always_inline to_type cvt##from##to (from_type x, int *sat) \
1986 to_type r; \
1987 if (use_min && x < min) { \
1988 r = min; \
1989 *sat = 1; \
1990 } else if (use_max && x > max) { \
1991 r = max; \
1992 *sat = 1; \
1993 } else { \
1994 r = x; \
1996 return r; \
1998 SATCVT(sh, sb, int16_t, int8_t, INT8_MIN, INT8_MAX, 1, 1)
1999 SATCVT(sw, sh, int32_t, int16_t, INT16_MIN, INT16_MAX, 1, 1)
2000 SATCVT(sd, sw, int64_t, int32_t, INT32_MIN, INT32_MAX, 1, 1)
2001 SATCVT(uh, ub, uint16_t, uint8_t, 0, UINT8_MAX, 0, 1)
2002 SATCVT(uw, uh, uint32_t, uint16_t, 0, UINT16_MAX, 0, 1)
2003 SATCVT(ud, uw, uint64_t, uint32_t, 0, UINT32_MAX, 0, 1)
2004 SATCVT(sh, ub, int16_t, uint8_t, 0, UINT8_MAX, 1, 1)
2005 SATCVT(sw, uh, int32_t, uint16_t, 0, UINT16_MAX, 1, 1)
2006 SATCVT(sd, uw, int64_t, uint32_t, 0, UINT32_MAX, 1, 1)
2007 #undef SATCVT
2009 #define LVE(name, access, swap, element) \
2010 void helper_##name (ppc_avr_t *r, target_ulong addr) \
2012 size_t n_elems = ARRAY_SIZE(r->element); \
2013 int adjust = HI_IDX*(n_elems-1); \
2014 int sh = sizeof(r->element[0]) >> 1; \
2015 int index = (addr & 0xf) >> sh; \
2016 if(msr_le) { \
2017 r->element[LO_IDX ? index : (adjust - index)] = swap(access(addr)); \
2018 } else { \
2019 r->element[LO_IDX ? index : (adjust - index)] = access(addr); \
2022 #define I(x) (x)
2023 LVE(lvebx, ldub, I, u8)
2024 LVE(lvehx, lduw, bswap16, u16)
2025 LVE(lvewx, ldl, bswap32, u32)
2026 #undef I
2027 #undef LVE
2029 void helper_lvsl (ppc_avr_t *r, target_ulong sh)
2031 int i, j = (sh & 0xf);
2033 VECTOR_FOR_INORDER_I (i, u8) {
2034 r->u8[i] = j++;
2038 void helper_lvsr (ppc_avr_t *r, target_ulong sh)
2040 int i, j = 0x10 - (sh & 0xf);
2042 VECTOR_FOR_INORDER_I (i, u8) {
2043 r->u8[i] = j++;
2047 #define STVE(name, access, swap, element) \
2048 void helper_##name (ppc_avr_t *r, target_ulong addr) \
2050 size_t n_elems = ARRAY_SIZE(r->element); \
2051 int adjust = HI_IDX*(n_elems-1); \
2052 int sh = sizeof(r->element[0]) >> 1; \
2053 int index = (addr & 0xf) >> sh; \
2054 if(msr_le) { \
2055 access(addr, swap(r->element[LO_IDX ? index : (adjust - index)])); \
2056 } else { \
2057 access(addr, r->element[LO_IDX ? index : (adjust - index)]); \
2060 #define I(x) (x)
2061 STVE(stvebx, stb, I, u8)
2062 STVE(stvehx, stw, bswap16, u16)
2063 STVE(stvewx, stl, bswap32, u32)
2064 #undef I
2065 #undef LVE
2067 void helper_mtvscr (ppc_avr_t *r)
2069 #if defined(WORDS_BIGENDIAN)
2070 env->vscr = r->u32[3];
2071 #else
2072 env->vscr = r->u32[0];
2073 #endif
2074 set_flush_to_zero(vscr_nj, &env->vec_status);
2077 void helper_vaddcuw (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2079 int i;
2080 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
2081 r->u32[i] = ~a->u32[i] < b->u32[i];
2085 #define VARITH_DO(name, op, element) \
2086 void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2088 int i; \
2089 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2090 r->element[i] = a->element[i] op b->element[i]; \
2093 #define VARITH(suffix, element) \
2094 VARITH_DO(add##suffix, +, element) \
2095 VARITH_DO(sub##suffix, -, element)
2096 VARITH(ubm, u8)
2097 VARITH(uhm, u16)
2098 VARITH(uwm, u32)
2099 #undef VARITH_DO
2100 #undef VARITH
2102 #define VARITHSAT_CASE(type, op, cvt, element) \
2104 type result = (type)a->element[i] op (type)b->element[i]; \
2105 r->element[i] = cvt(result, &sat); \
2108 #define VARITHSAT_DO(name, op, optype, cvt, element) \
2109 void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2111 int sat = 0; \
2112 int i; \
2113 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2114 switch (sizeof(r->element[0])) { \
2115 case 1: VARITHSAT_CASE(optype, op, cvt, element); break; \
2116 case 2: VARITHSAT_CASE(optype, op, cvt, element); break; \
2117 case 4: VARITHSAT_CASE(optype, op, cvt, element); break; \
2120 if (sat) { \
2121 env->vscr |= (1 << VSCR_SAT); \
2124 #define VARITHSAT_SIGNED(suffix, element, optype, cvt) \
2125 VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element) \
2126 VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element)
2127 #define VARITHSAT_UNSIGNED(suffix, element, optype, cvt) \
2128 VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element) \
2129 VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element)
2130 VARITHSAT_SIGNED(b, s8, int16_t, cvtshsb)
2131 VARITHSAT_SIGNED(h, s16, int32_t, cvtswsh)
2132 VARITHSAT_SIGNED(w, s32, int64_t, cvtsdsw)
2133 VARITHSAT_UNSIGNED(b, u8, uint16_t, cvtshub)
2134 VARITHSAT_UNSIGNED(h, u16, uint32_t, cvtswuh)
2135 VARITHSAT_UNSIGNED(w, u32, uint64_t, cvtsduw)
2136 #undef VARITHSAT_CASE
2137 #undef VARITHSAT_DO
2138 #undef VARITHSAT_SIGNED
2139 #undef VARITHSAT_UNSIGNED
2141 #define VAVG_DO(name, element, etype) \
2142 void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2144 int i; \
2145 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2146 etype x = (etype)a->element[i] + (etype)b->element[i] + 1; \
2147 r->element[i] = x >> 1; \
2151 #define VAVG(type, signed_element, signed_type, unsigned_element, unsigned_type) \
2152 VAVG_DO(avgs##type, signed_element, signed_type) \
2153 VAVG_DO(avgu##type, unsigned_element, unsigned_type)
2154 VAVG(b, s8, int16_t, u8, uint16_t)
2155 VAVG(h, s16, int32_t, u16, uint32_t)
2156 VAVG(w, s32, int64_t, u32, uint64_t)
2157 #undef VAVG_DO
2158 #undef VAVG
2160 #define VCF(suffix, cvt, element) \
2161 void helper_vcf##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t uim) \
2163 int i; \
2164 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
2165 float32 t = cvt(b->element[i], &env->vec_status); \
2166 r->f[i] = float32_scalbn (t, -uim, &env->vec_status); \
2169 VCF(ux, uint32_to_float32, u32)
2170 VCF(sx, int32_to_float32, s32)
2171 #undef VCF
2173 #define VCMP_DO(suffix, compare, element, record) \
2174 void helper_vcmp##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2176 uint32_t ones = (uint32_t)-1; \
2177 uint32_t all = ones; \
2178 uint32_t none = 0; \
2179 int i; \
2180 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2181 uint32_t result = (a->element[i] compare b->element[i] ? ones : 0x0); \
2182 switch (sizeof (a->element[0])) { \
2183 case 4: r->u32[i] = result; break; \
2184 case 2: r->u16[i] = result; break; \
2185 case 1: r->u8[i] = result; break; \
2187 all &= result; \
2188 none |= result; \
2190 if (record) { \
2191 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \
2194 #define VCMP(suffix, compare, element) \
2195 VCMP_DO(suffix, compare, element, 0) \
2196 VCMP_DO(suffix##_dot, compare, element, 1)
2197 VCMP(equb, ==, u8)
2198 VCMP(equh, ==, u16)
2199 VCMP(equw, ==, u32)
2200 VCMP(gtub, >, u8)
2201 VCMP(gtuh, >, u16)
2202 VCMP(gtuw, >, u32)
2203 VCMP(gtsb, >, s8)
2204 VCMP(gtsh, >, s16)
2205 VCMP(gtsw, >, s32)
2206 #undef VCMP_DO
2207 #undef VCMP
2209 void helper_vmhaddshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2211 int sat = 0;
2212 int i;
2214 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
2215 int32_t prod = a->s16[i] * b->s16[i];
2216 int32_t t = (int32_t)c->s16[i] + (prod >> 15);
2217 r->s16[i] = cvtswsh (t, &sat);
2220 if (sat) {
2221 env->vscr |= (1 << VSCR_SAT);
2225 void helper_vmhraddshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2227 int sat = 0;
2228 int i;
2230 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
2231 int32_t prod = a->s16[i] * b->s16[i] + 0x00004000;
2232 int32_t t = (int32_t)c->s16[i] + (prod >> 15);
2233 r->s16[i] = cvtswsh (t, &sat);
2236 if (sat) {
2237 env->vscr |= (1 << VSCR_SAT);
2241 #define VMINMAX_DO(name, compare, element) \
2242 void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2244 int i; \
2245 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2246 if (a->element[i] compare b->element[i]) { \
2247 r->element[i] = b->element[i]; \
2248 } else { \
2249 r->element[i] = a->element[i]; \
2253 #define VMINMAX(suffix, element) \
2254 VMINMAX_DO(min##suffix, >, element) \
2255 VMINMAX_DO(max##suffix, <, element)
2256 VMINMAX(sb, s8)
2257 VMINMAX(sh, s16)
2258 VMINMAX(sw, s32)
2259 VMINMAX(ub, u8)
2260 VMINMAX(uh, u16)
2261 VMINMAX(uw, u32)
2262 #undef VMINMAX_DO
2263 #undef VMINMAX
2265 void helper_vmladduhm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2267 int i;
2268 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
2269 int32_t prod = a->s16[i] * b->s16[i];
2270 r->s16[i] = (int16_t) (prod + c->s16[i]);
2274 #define VMRG_DO(name, element, highp) \
2275 void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2277 ppc_avr_t result; \
2278 int i; \
2279 size_t n_elems = ARRAY_SIZE(r->element); \
2280 for (i = 0; i < n_elems/2; i++) { \
2281 if (highp) { \
2282 result.element[i*2+HI_IDX] = a->element[i]; \
2283 result.element[i*2+LO_IDX] = b->element[i]; \
2284 } else { \
2285 result.element[n_elems - i*2 - (1+HI_IDX)] = b->element[n_elems - i - 1]; \
2286 result.element[n_elems - i*2 - (1+LO_IDX)] = a->element[n_elems - i - 1]; \
2289 *r = result; \
2291 #if defined(WORDS_BIGENDIAN)
2292 #define MRGHI 0
2293 #define MRGLO 1
2294 #else
2295 #define MRGHI 1
2296 #define MRGLO 0
2297 #endif
2298 #define VMRG(suffix, element) \
2299 VMRG_DO(mrgl##suffix, element, MRGHI) \
2300 VMRG_DO(mrgh##suffix, element, MRGLO)
2301 VMRG(b, u8)
2302 VMRG(h, u16)
2303 VMRG(w, u32)
2304 #undef VMRG_DO
2305 #undef VMRG
2306 #undef MRGHI
2307 #undef MRGLO
2309 void helper_vmsummbm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2311 int32_t prod[16];
2312 int i;
2314 for (i = 0; i < ARRAY_SIZE(r->s8); i++) {
2315 prod[i] = (int32_t)a->s8[i] * b->u8[i];
2318 VECTOR_FOR_INORDER_I(i, s32) {
2319 r->s32[i] = c->s32[i] + prod[4*i] + prod[4*i+1] + prod[4*i+2] + prod[4*i+3];
2323 void helper_vmsumshm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2325 int32_t prod[8];
2326 int i;
2328 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
2329 prod[i] = a->s16[i] * b->s16[i];
2332 VECTOR_FOR_INORDER_I(i, s32) {
2333 r->s32[i] = c->s32[i] + prod[2*i] + prod[2*i+1];
2337 void helper_vmsumshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2339 int32_t prod[8];
2340 int i;
2341 int sat = 0;
2343 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
2344 prod[i] = (int32_t)a->s16[i] * b->s16[i];
2347 VECTOR_FOR_INORDER_I (i, s32) {
2348 int64_t t = (int64_t)c->s32[i] + prod[2*i] + prod[2*i+1];
2349 r->u32[i] = cvtsdsw(t, &sat);
2352 if (sat) {
2353 env->vscr |= (1 << VSCR_SAT);
2357 void helper_vmsumubm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2359 uint16_t prod[16];
2360 int i;
2362 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
2363 prod[i] = a->u8[i] * b->u8[i];
2366 VECTOR_FOR_INORDER_I(i, u32) {
2367 r->u32[i] = c->u32[i] + prod[4*i] + prod[4*i+1] + prod[4*i+2] + prod[4*i+3];
2371 void helper_vmsumuhm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2373 uint32_t prod[8];
2374 int i;
2376 for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
2377 prod[i] = a->u16[i] * b->u16[i];
2380 VECTOR_FOR_INORDER_I(i, u32) {
2381 r->u32[i] = c->u32[i] + prod[2*i] + prod[2*i+1];
2385 void helper_vmsumuhs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2387 uint32_t prod[8];
2388 int i;
2389 int sat = 0;
2391 for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
2392 prod[i] = a->u16[i] * b->u16[i];
2395 VECTOR_FOR_INORDER_I (i, s32) {
2396 uint64_t t = (uint64_t)c->u32[i] + prod[2*i] + prod[2*i+1];
2397 r->u32[i] = cvtuduw(t, &sat);
2400 if (sat) {
2401 env->vscr |= (1 << VSCR_SAT);
2405 #define VMUL_DO(name, mul_element, prod_element, evenp) \
2406 void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2408 int i; \
2409 VECTOR_FOR_INORDER_I(i, prod_element) { \
2410 if (evenp) { \
2411 r->prod_element[i] = a->mul_element[i*2+HI_IDX] * b->mul_element[i*2+HI_IDX]; \
2412 } else { \
2413 r->prod_element[i] = a->mul_element[i*2+LO_IDX] * b->mul_element[i*2+LO_IDX]; \
2417 #define VMUL(suffix, mul_element, prod_element) \
2418 VMUL_DO(mule##suffix, mul_element, prod_element, 1) \
2419 VMUL_DO(mulo##suffix, mul_element, prod_element, 0)
2420 VMUL(sb, s8, s16)
2421 VMUL(sh, s16, s32)
2422 VMUL(ub, u8, u16)
2423 VMUL(uh, u16, u32)
2424 #undef VMUL_DO
2425 #undef VMUL
2427 void helper_vperm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2429 ppc_avr_t result;
2430 int i;
2431 VECTOR_FOR_INORDER_I (i, u8) {
2432 int s = c->u8[i] & 0x1f;
2433 #if defined(WORDS_BIGENDIAN)
2434 int index = s & 0xf;
2435 #else
2436 int index = 15 - (s & 0xf);
2437 #endif
2438 if (s & 0x10) {
2439 result.u8[i] = b->u8[index];
2440 } else {
2441 result.u8[i] = a->u8[index];
2444 *r = result;
2447 #if defined(WORDS_BIGENDIAN)
2448 #define PKBIG 1
2449 #else
2450 #define PKBIG 0
2451 #endif
2452 void helper_vpkpx (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2454 int i, j;
2455 ppc_avr_t result;
2456 #if defined(WORDS_BIGENDIAN)
2457 const ppc_avr_t *x[2] = { a, b };
2458 #else
2459 const ppc_avr_t *x[2] = { b, a };
2460 #endif
2462 VECTOR_FOR_INORDER_I (i, u64) {
2463 VECTOR_FOR_INORDER_I (j, u32){
2464 uint32_t e = x[i]->u32[j];
2465 result.u16[4*i+j] = (((e >> 9) & 0xfc00) |
2466 ((e >> 6) & 0x3e0) |
2467 ((e >> 3) & 0x1f));
2470 *r = result;
2473 #define VPK(suffix, from, to, cvt, dosat) \
2474 void helper_vpk##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2476 int i; \
2477 int sat = 0; \
2478 ppc_avr_t result; \
2479 ppc_avr_t *a0 = PKBIG ? a : b; \
2480 ppc_avr_t *a1 = PKBIG ? b : a; \
2481 VECTOR_FOR_INORDER_I (i, from) { \
2482 result.to[i] = cvt(a0->from[i], &sat); \
2483 result.to[i+ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat); \
2485 *r = result; \
2486 if (dosat && sat) { \
2487 env->vscr |= (1 << VSCR_SAT); \
2490 #define I(x, y) (x)
2491 VPK(shss, s16, s8, cvtshsb, 1)
2492 VPK(shus, s16, u8, cvtshub, 1)
2493 VPK(swss, s32, s16, cvtswsh, 1)
2494 VPK(swus, s32, u16, cvtswuh, 1)
2495 VPK(uhus, u16, u8, cvtuhub, 1)
2496 VPK(uwus, u32, u16, cvtuwuh, 1)
2497 VPK(uhum, u16, u8, I, 0)
2498 VPK(uwum, u32, u16, I, 0)
2499 #undef I
2500 #undef VPK
2501 #undef PKBIG
2503 #define VRFI(suffix, rounding) \
2504 void helper_vrfi##suffix (ppc_avr_t *r, ppc_avr_t *b) \
2506 int i; \
2507 float_status s = env->vec_status; \
2508 set_float_rounding_mode(rounding, &s); \
2509 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
2510 HANDLE_NAN1(r->f[i], b->f[i]) { \
2511 r->f[i] = float32_round_to_int (b->f[i], &s); \
2515 VRFI(n, float_round_nearest_even)
2516 VRFI(m, float_round_down)
2517 VRFI(p, float_round_up)
2518 VRFI(z, float_round_to_zero)
2519 #undef VRFI
2521 #define VROTATE(suffix, element) \
2522 void helper_vrl##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2524 int i; \
2525 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2526 unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \
2527 unsigned int shift = b->element[i] & mask; \
2528 r->element[i] = (a->element[i] << shift) | (a->element[i] >> (sizeof(a->element[0]) * 8 - shift)); \
2531 VROTATE(b, u8)
2532 VROTATE(h, u16)
2533 VROTATE(w, u32)
2534 #undef VROTATE
2536 void helper_vsel (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2538 r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]);
2539 r->u64[1] = (a->u64[1] & ~c->u64[1]) | (b->u64[1] & c->u64[1]);
2542 void helper_vlogefp (ppc_avr_t *r, ppc_avr_t *b)
2544 int i;
2545 for (i = 0; i < ARRAY_SIZE(r->f); i++) {
2546 HANDLE_NAN1(r->f[i], b->f[i]) {
2547 r->f[i] = float32_log2(b->f[i], &env->vec_status);
2552 #if defined(WORDS_BIGENDIAN)
2553 #define LEFT 0
2554 #define RIGHT 1
2555 #else
2556 #define LEFT 1
2557 #define RIGHT 0
2558 #endif
2559 /* The specification says that the results are undefined if all of the
2560 * shift counts are not identical. We check to make sure that they are
2561 * to conform to what real hardware appears to do. */
2562 #define VSHIFT(suffix, leftp) \
2563 void helper_vs##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2565 int shift = b->u8[LO_IDX*0x15] & 0x7; \
2566 int doit = 1; \
2567 int i; \
2568 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { \
2569 doit = doit && ((b->u8[i] & 0x7) == shift); \
2571 if (doit) { \
2572 if (shift == 0) { \
2573 *r = *a; \
2574 } else if (leftp) { \
2575 uint64_t carry = a->u64[LO_IDX] >> (64 - shift); \
2576 r->u64[HI_IDX] = (a->u64[HI_IDX] << shift) | carry; \
2577 r->u64[LO_IDX] = a->u64[LO_IDX] << shift; \
2578 } else { \
2579 uint64_t carry = a->u64[HI_IDX] << (64 - shift); \
2580 r->u64[LO_IDX] = (a->u64[LO_IDX] >> shift) | carry; \
2581 r->u64[HI_IDX] = a->u64[HI_IDX] >> shift; \
2585 VSHIFT(l, LEFT)
2586 VSHIFT(r, RIGHT)
2587 #undef VSHIFT
2588 #undef LEFT
2589 #undef RIGHT
2591 #define VSL(suffix, element) \
2592 void helper_vsl##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2594 int i; \
2595 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2596 unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \
2597 unsigned int shift = b->element[i] & mask; \
2598 r->element[i] = a->element[i] << shift; \
2601 VSL(b, u8)
2602 VSL(h, u16)
2603 VSL(w, u32)
2604 #undef VSL
2606 void helper_vsldoi (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t shift)
2608 int sh = shift & 0xf;
2609 int i;
2610 ppc_avr_t result;
2612 #if defined(WORDS_BIGENDIAN)
2613 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
2614 int index = sh + i;
2615 if (index > 0xf) {
2616 result.u8[i] = b->u8[index-0x10];
2617 } else {
2618 result.u8[i] = a->u8[index];
2621 #else
2622 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
2623 int index = (16 - sh) + i;
2624 if (index > 0xf) {
2625 result.u8[i] = a->u8[index-0x10];
2626 } else {
2627 result.u8[i] = b->u8[index];
2630 #endif
2631 *r = result;
2634 void helper_vslo (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2636 int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf;
2638 #if defined (WORDS_BIGENDIAN)
2639 memmove (&r->u8[0], &a->u8[sh], 16-sh);
2640 memset (&r->u8[16-sh], 0, sh);
2641 #else
2642 memmove (&r->u8[sh], &a->u8[0], 16-sh);
2643 memset (&r->u8[0], 0, sh);
2644 #endif
2647 /* Experimental testing shows that hardware masks the immediate. */
2648 #define _SPLAT_MASKED(element) (splat & (ARRAY_SIZE(r->element) - 1))
2649 #if defined(WORDS_BIGENDIAN)
2650 #define SPLAT_ELEMENT(element) _SPLAT_MASKED(element)
2651 #else
2652 #define SPLAT_ELEMENT(element) (ARRAY_SIZE(r->element)-1 - _SPLAT_MASKED(element))
2653 #endif
2654 #define VSPLT(suffix, element) \
2655 void helper_vsplt##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t splat) \
2657 uint32_t s = b->element[SPLAT_ELEMENT(element)]; \
2658 int i; \
2659 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2660 r->element[i] = s; \
2663 VSPLT(b, u8)
2664 VSPLT(h, u16)
2665 VSPLT(w, u32)
2666 #undef VSPLT
2667 #undef SPLAT_ELEMENT
2668 #undef _SPLAT_MASKED
2670 #define VSPLTI(suffix, element, splat_type) \
2671 void helper_vspltis##suffix (ppc_avr_t *r, uint32_t splat) \
2673 splat_type x = (int8_t)(splat << 3) >> 3; \
2674 int i; \
2675 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2676 r->element[i] = x; \
2679 VSPLTI(b, s8, int8_t)
2680 VSPLTI(h, s16, int16_t)
2681 VSPLTI(w, s32, int32_t)
2682 #undef VSPLTI
2684 #define VSR(suffix, element) \
2685 void helper_vsr##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
2687 int i; \
2688 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2689 unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \
2690 unsigned int shift = b->element[i] & mask; \
2691 r->element[i] = a->element[i] >> shift; \
2694 VSR(ab, s8)
2695 VSR(ah, s16)
2696 VSR(aw, s32)
2697 VSR(b, u8)
2698 VSR(h, u16)
2699 VSR(w, u32)
2700 #undef VSR
2702 void helper_vsro (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2704 int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf;
2706 #if defined (WORDS_BIGENDIAN)
2707 memmove (&r->u8[sh], &a->u8[0], 16-sh);
2708 memset (&r->u8[0], 0, sh);
2709 #else
2710 memmove (&r->u8[0], &a->u8[sh], 16-sh);
2711 memset (&r->u8[16-sh], 0, sh);
2712 #endif
2715 void helper_vsubcuw (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2717 int i;
2718 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
2719 r->u32[i] = a->u32[i] >= b->u32[i];
2723 void helper_vsumsws (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2725 int64_t t;
2726 int i, upper;
2727 ppc_avr_t result;
2728 int sat = 0;
2730 #if defined(WORDS_BIGENDIAN)
2731 upper = ARRAY_SIZE(r->s32)-1;
2732 #else
2733 upper = 0;
2734 #endif
2735 t = (int64_t)b->s32[upper];
2736 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
2737 t += a->s32[i];
2738 result.s32[i] = 0;
2740 result.s32[upper] = cvtsdsw(t, &sat);
2741 *r = result;
2743 if (sat) {
2744 env->vscr |= (1 << VSCR_SAT);
2748 void helper_vsum2sws (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2750 int i, j, upper;
2751 ppc_avr_t result;
2752 int sat = 0;
2754 #if defined(WORDS_BIGENDIAN)
2755 upper = 1;
2756 #else
2757 upper = 0;
2758 #endif
2759 for (i = 0; i < ARRAY_SIZE(r->u64); i++) {
2760 int64_t t = (int64_t)b->s32[upper+i*2];
2761 result.u64[i] = 0;
2762 for (j = 0; j < ARRAY_SIZE(r->u64); j++) {
2763 t += a->s32[2*i+j];
2765 result.s32[upper+i*2] = cvtsdsw(t, &sat);
2768 *r = result;
2769 if (sat) {
2770 env->vscr |= (1 << VSCR_SAT);
2774 void helper_vsum4sbs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2776 int i, j;
2777 int sat = 0;
2779 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
2780 int64_t t = (int64_t)b->s32[i];
2781 for (j = 0; j < ARRAY_SIZE(r->s32); j++) {
2782 t += a->s8[4*i+j];
2784 r->s32[i] = cvtsdsw(t, &sat);
2787 if (sat) {
2788 env->vscr |= (1 << VSCR_SAT);
2792 void helper_vsum4shs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2794 int sat = 0;
2795 int i;
2797 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
2798 int64_t t = (int64_t)b->s32[i];
2799 t += a->s16[2*i] + a->s16[2*i+1];
2800 r->s32[i] = cvtsdsw(t, &sat);
2803 if (sat) {
2804 env->vscr |= (1 << VSCR_SAT);
2808 void helper_vsum4ubs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2810 int i, j;
2811 int sat = 0;
2813 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
2814 uint64_t t = (uint64_t)b->u32[i];
2815 for (j = 0; j < ARRAY_SIZE(r->u32); j++) {
2816 t += a->u8[4*i+j];
2818 r->u32[i] = cvtuduw(t, &sat);
2821 if (sat) {
2822 env->vscr |= (1 << VSCR_SAT);
2826 #if defined(WORDS_BIGENDIAN)
2827 #define UPKHI 1
2828 #define UPKLO 0
2829 #else
2830 #define UPKHI 0
2831 #define UPKLO 1
2832 #endif
2833 #define VUPKPX(suffix, hi) \
2834 void helper_vupk##suffix (ppc_avr_t *r, ppc_avr_t *b) \
2836 int i; \
2837 ppc_avr_t result; \
2838 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \
2839 uint16_t e = b->u16[hi ? i : i+4]; \
2840 uint8_t a = (e >> 15) ? 0xff : 0; \
2841 uint8_t r = (e >> 10) & 0x1f; \
2842 uint8_t g = (e >> 5) & 0x1f; \
2843 uint8_t b = e & 0x1f; \
2844 result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \
2846 *r = result; \
2848 VUPKPX(lpx, UPKLO)
2849 VUPKPX(hpx, UPKHI)
2850 #undef VUPKPX
2852 #define VUPK(suffix, unpacked, packee, hi) \
2853 void helper_vupk##suffix (ppc_avr_t *r, ppc_avr_t *b) \
2855 int i; \
2856 ppc_avr_t result; \
2857 if (hi) { \
2858 for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \
2859 result.unpacked[i] = b->packee[i]; \
2861 } else { \
2862 for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); i++) { \
2863 result.unpacked[i-ARRAY_SIZE(r->unpacked)] = b->packee[i]; \
2866 *r = result; \
2868 VUPK(hsb, s16, s8, UPKHI)
2869 VUPK(hsh, s32, s16, UPKHI)
2870 VUPK(lsb, s16, s8, UPKLO)
2871 VUPK(lsh, s32, s16, UPKLO)
2872 #undef VUPK
2873 #undef UPKHI
2874 #undef UPKLO
2876 #undef DO_HANDLE_NAN
2877 #undef HANDLE_NAN1
2878 #undef HANDLE_NAN2
2879 #undef HANDLE_NAN3
2880 #undef VECTOR_FOR_INORDER_I
2881 #undef HI_IDX
2882 #undef LO_IDX
2884 /*****************************************************************************/
2885 /* SPE extension helpers */
2886 /* Use a table to make this quicker */
2887 static uint8_t hbrev[16] = {
2888 0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE,
2889 0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF,
2892 static always_inline uint8_t byte_reverse (uint8_t val)
2894 return hbrev[val >> 4] | (hbrev[val & 0xF] << 4);
2897 static always_inline uint32_t word_reverse (uint32_t val)
2899 return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) |
2900 (byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24);
2903 #define MASKBITS 16 // Random value - to be fixed (implementation dependant)
2904 target_ulong helper_brinc (target_ulong arg1, target_ulong arg2)
2906 uint32_t a, b, d, mask;
2908 mask = UINT32_MAX >> (32 - MASKBITS);
2909 a = arg1 & mask;
2910 b = arg2 & mask;
2911 d = word_reverse(1 + word_reverse(a | ~b));
2912 return (arg1 & ~mask) | (d & b);
2915 uint32_t helper_cntlsw32 (uint32_t val)
2917 if (val & 0x80000000)
2918 return clz32(~val);
2919 else
2920 return clz32(val);
2923 uint32_t helper_cntlzw32 (uint32_t val)
2925 return clz32(val);
2928 /* Single-precision floating-point conversions */
2929 static always_inline uint32_t efscfsi (uint32_t val)
2931 CPU_FloatU u;
2933 u.f = int32_to_float32(val, &env->vec_status);
2935 return u.l;
2938 static always_inline uint32_t efscfui (uint32_t val)
2940 CPU_FloatU u;
2942 u.f = uint32_to_float32(val, &env->vec_status);
2944 return u.l;
2947 static always_inline int32_t efsctsi (uint32_t val)
2949 CPU_FloatU u;
2951 u.l = val;
2952 /* NaN are not treated the same way IEEE 754 does */
2953 if (unlikely(float32_is_nan(u.f)))
2954 return 0;
2956 return float32_to_int32(u.f, &env->vec_status);
2959 static always_inline uint32_t efsctui (uint32_t val)
2961 CPU_FloatU u;
2963 u.l = val;
2964 /* NaN are not treated the same way IEEE 754 does */
2965 if (unlikely(float32_is_nan(u.f)))
2966 return 0;
2968 return float32_to_uint32(u.f, &env->vec_status);
2971 static always_inline uint32_t efsctsiz (uint32_t val)
2973 CPU_FloatU u;
2975 u.l = val;
2976 /* NaN are not treated the same way IEEE 754 does */
2977 if (unlikely(float32_is_nan(u.f)))
2978 return 0;
2980 return float32_to_int32_round_to_zero(u.f, &env->vec_status);
2983 static always_inline uint32_t efsctuiz (uint32_t val)
2985 CPU_FloatU u;
2987 u.l = val;
2988 /* NaN are not treated the same way IEEE 754 does */
2989 if (unlikely(float32_is_nan(u.f)))
2990 return 0;
2992 return float32_to_uint32_round_to_zero(u.f, &env->vec_status);
2995 static always_inline uint32_t efscfsf (uint32_t val)
2997 CPU_FloatU u;
2998 float32 tmp;
3000 u.f = int32_to_float32(val, &env->vec_status);
3001 tmp = int64_to_float32(1ULL << 32, &env->vec_status);
3002 u.f = float32_div(u.f, tmp, &env->vec_status);
3004 return u.l;
3007 static always_inline uint32_t efscfuf (uint32_t val)
3009 CPU_FloatU u;
3010 float32 tmp;
3012 u.f = uint32_to_float32(val, &env->vec_status);
3013 tmp = uint64_to_float32(1ULL << 32, &env->vec_status);
3014 u.f = float32_div(u.f, tmp, &env->vec_status);
3016 return u.l;
3019 static always_inline uint32_t efsctsf (uint32_t val)
3021 CPU_FloatU u;
3022 float32 tmp;
3024 u.l = val;
3025 /* NaN are not treated the same way IEEE 754 does */
3026 if (unlikely(float32_is_nan(u.f)))
3027 return 0;
3028 tmp = uint64_to_float32(1ULL << 32, &env->vec_status);
3029 u.f = float32_mul(u.f, tmp, &env->vec_status);
3031 return float32_to_int32(u.f, &env->vec_status);
3034 static always_inline uint32_t efsctuf (uint32_t val)
3036 CPU_FloatU u;
3037 float32 tmp;
3039 u.l = val;
3040 /* NaN are not treated the same way IEEE 754 does */
3041 if (unlikely(float32_is_nan(u.f)))
3042 return 0;
3043 tmp = uint64_to_float32(1ULL << 32, &env->vec_status);
3044 u.f = float32_mul(u.f, tmp, &env->vec_status);
3046 return float32_to_uint32(u.f, &env->vec_status);
3049 #define HELPER_SPE_SINGLE_CONV(name) \
3050 uint32_t helper_e##name (uint32_t val) \
3052 return e##name(val); \
3054 /* efscfsi */
3055 HELPER_SPE_SINGLE_CONV(fscfsi);
3056 /* efscfui */
3057 HELPER_SPE_SINGLE_CONV(fscfui);
3058 /* efscfuf */
3059 HELPER_SPE_SINGLE_CONV(fscfuf);
3060 /* efscfsf */
3061 HELPER_SPE_SINGLE_CONV(fscfsf);
3062 /* efsctsi */
3063 HELPER_SPE_SINGLE_CONV(fsctsi);
3064 /* efsctui */
3065 HELPER_SPE_SINGLE_CONV(fsctui);
3066 /* efsctsiz */
3067 HELPER_SPE_SINGLE_CONV(fsctsiz);
3068 /* efsctuiz */
3069 HELPER_SPE_SINGLE_CONV(fsctuiz);
3070 /* efsctsf */
3071 HELPER_SPE_SINGLE_CONV(fsctsf);
3072 /* efsctuf */
3073 HELPER_SPE_SINGLE_CONV(fsctuf);
3075 #define HELPER_SPE_VECTOR_CONV(name) \
3076 uint64_t helper_ev##name (uint64_t val) \
3078 return ((uint64_t)e##name(val >> 32) << 32) | \
3079 (uint64_t)e##name(val); \
3081 /* evfscfsi */
3082 HELPER_SPE_VECTOR_CONV(fscfsi);
3083 /* evfscfui */
3084 HELPER_SPE_VECTOR_CONV(fscfui);
3085 /* evfscfuf */
3086 HELPER_SPE_VECTOR_CONV(fscfuf);
3087 /* evfscfsf */
3088 HELPER_SPE_VECTOR_CONV(fscfsf);
3089 /* evfsctsi */
3090 HELPER_SPE_VECTOR_CONV(fsctsi);
3091 /* evfsctui */
3092 HELPER_SPE_VECTOR_CONV(fsctui);
3093 /* evfsctsiz */
3094 HELPER_SPE_VECTOR_CONV(fsctsiz);
3095 /* evfsctuiz */
3096 HELPER_SPE_VECTOR_CONV(fsctuiz);
3097 /* evfsctsf */
3098 HELPER_SPE_VECTOR_CONV(fsctsf);
3099 /* evfsctuf */
3100 HELPER_SPE_VECTOR_CONV(fsctuf);
3102 /* Single-precision floating-point arithmetic */
3103 static always_inline uint32_t efsadd (uint32_t op1, uint32_t op2)
3105 CPU_FloatU u1, u2;
3106 u1.l = op1;
3107 u2.l = op2;
3108 u1.f = float32_add(u1.f, u2.f, &env->vec_status);
3109 return u1.l;
3112 static always_inline uint32_t efssub (uint32_t op1, uint32_t op2)
3114 CPU_FloatU u1, u2;
3115 u1.l = op1;
3116 u2.l = op2;
3117 u1.f = float32_sub(u1.f, u2.f, &env->vec_status);
3118 return u1.l;
3121 static always_inline uint32_t efsmul (uint32_t op1, uint32_t op2)
3123 CPU_FloatU u1, u2;
3124 u1.l = op1;
3125 u2.l = op2;
3126 u1.f = float32_mul(u1.f, u2.f, &env->vec_status);
3127 return u1.l;
3130 static always_inline uint32_t efsdiv (uint32_t op1, uint32_t op2)
3132 CPU_FloatU u1, u2;
3133 u1.l = op1;
3134 u2.l = op2;
3135 u1.f = float32_div(u1.f, u2.f, &env->vec_status);
3136 return u1.l;
3139 #define HELPER_SPE_SINGLE_ARITH(name) \
3140 uint32_t helper_e##name (uint32_t op1, uint32_t op2) \
3142 return e##name(op1, op2); \
3144 /* efsadd */
3145 HELPER_SPE_SINGLE_ARITH(fsadd);
3146 /* efssub */
3147 HELPER_SPE_SINGLE_ARITH(fssub);
3148 /* efsmul */
3149 HELPER_SPE_SINGLE_ARITH(fsmul);
3150 /* efsdiv */
3151 HELPER_SPE_SINGLE_ARITH(fsdiv);
3153 #define HELPER_SPE_VECTOR_ARITH(name) \
3154 uint64_t helper_ev##name (uint64_t op1, uint64_t op2) \
3156 return ((uint64_t)e##name(op1 >> 32, op2 >> 32) << 32) | \
3157 (uint64_t)e##name(op1, op2); \
3159 /* evfsadd */
3160 HELPER_SPE_VECTOR_ARITH(fsadd);
3161 /* evfssub */
3162 HELPER_SPE_VECTOR_ARITH(fssub);
3163 /* evfsmul */
3164 HELPER_SPE_VECTOR_ARITH(fsmul);
3165 /* evfsdiv */
3166 HELPER_SPE_VECTOR_ARITH(fsdiv);
3168 /* Single-precision floating-point comparisons */
3169 static always_inline uint32_t efststlt (uint32_t op1, uint32_t op2)
3171 CPU_FloatU u1, u2;
3172 u1.l = op1;
3173 u2.l = op2;
3174 return float32_lt(u1.f, u2.f, &env->vec_status) ? 4 : 0;
3177 static always_inline uint32_t efststgt (uint32_t op1, uint32_t op2)
3179 CPU_FloatU u1, u2;
3180 u1.l = op1;
3181 u2.l = op2;
3182 return float32_le(u1.f, u2.f, &env->vec_status) ? 0 : 4;
3185 static always_inline uint32_t efststeq (uint32_t op1, uint32_t op2)
3187 CPU_FloatU u1, u2;
3188 u1.l = op1;
3189 u2.l = op2;
3190 return float32_eq(u1.f, u2.f, &env->vec_status) ? 4 : 0;
3193 static always_inline uint32_t efscmplt (uint32_t op1, uint32_t op2)
3195 /* XXX: TODO: test special values (NaN, infinites, ...) */
3196 return efststlt(op1, op2);
3199 static always_inline uint32_t efscmpgt (uint32_t op1, uint32_t op2)
3201 /* XXX: TODO: test special values (NaN, infinites, ...) */
3202 return efststgt(op1, op2);
3205 static always_inline uint32_t efscmpeq (uint32_t op1, uint32_t op2)
3207 /* XXX: TODO: test special values (NaN, infinites, ...) */
3208 return efststeq(op1, op2);
3211 #define HELPER_SINGLE_SPE_CMP(name) \
3212 uint32_t helper_e##name (uint32_t op1, uint32_t op2) \
3214 return e##name(op1, op2) << 2; \
3216 /* efststlt */
3217 HELPER_SINGLE_SPE_CMP(fststlt);
3218 /* efststgt */
3219 HELPER_SINGLE_SPE_CMP(fststgt);
3220 /* efststeq */
3221 HELPER_SINGLE_SPE_CMP(fststeq);
3222 /* efscmplt */
3223 HELPER_SINGLE_SPE_CMP(fscmplt);
3224 /* efscmpgt */
3225 HELPER_SINGLE_SPE_CMP(fscmpgt);
3226 /* efscmpeq */
3227 HELPER_SINGLE_SPE_CMP(fscmpeq);
3229 static always_inline uint32_t evcmp_merge (int t0, int t1)
3231 return (t0 << 3) | (t1 << 2) | ((t0 | t1) << 1) | (t0 & t1);
3234 #define HELPER_VECTOR_SPE_CMP(name) \
3235 uint32_t helper_ev##name (uint64_t op1, uint64_t op2) \
3237 return evcmp_merge(e##name(op1 >> 32, op2 >> 32), e##name(op1, op2)); \
3239 /* evfststlt */
3240 HELPER_VECTOR_SPE_CMP(fststlt);
3241 /* evfststgt */
3242 HELPER_VECTOR_SPE_CMP(fststgt);
3243 /* evfststeq */
3244 HELPER_VECTOR_SPE_CMP(fststeq);
3245 /* evfscmplt */
3246 HELPER_VECTOR_SPE_CMP(fscmplt);
3247 /* evfscmpgt */
3248 HELPER_VECTOR_SPE_CMP(fscmpgt);
3249 /* evfscmpeq */
3250 HELPER_VECTOR_SPE_CMP(fscmpeq);
3252 /* Double-precision floating-point conversion */
3253 uint64_t helper_efdcfsi (uint32_t val)
3255 CPU_DoubleU u;
3257 u.d = int32_to_float64(val, &env->vec_status);
3259 return u.ll;
3262 uint64_t helper_efdcfsid (uint64_t val)
3264 CPU_DoubleU u;
3266 u.d = int64_to_float64(val, &env->vec_status);
3268 return u.ll;
3271 uint64_t helper_efdcfui (uint32_t val)
3273 CPU_DoubleU u;
3275 u.d = uint32_to_float64(val, &env->vec_status);
3277 return u.ll;
3280 uint64_t helper_efdcfuid (uint64_t val)
3282 CPU_DoubleU u;
3284 u.d = uint64_to_float64(val, &env->vec_status);
3286 return u.ll;
3289 uint32_t helper_efdctsi (uint64_t val)
3291 CPU_DoubleU u;
3293 u.ll = val;
3294 /* NaN are not treated the same way IEEE 754 does */
3295 if (unlikely(float64_is_nan(u.d)))
3296 return 0;
3298 return float64_to_int32(u.d, &env->vec_status);
3301 uint32_t helper_efdctui (uint64_t val)
3303 CPU_DoubleU u;
3305 u.ll = val;
3306 /* NaN are not treated the same way IEEE 754 does */
3307 if (unlikely(float64_is_nan(u.d)))
3308 return 0;
3310 return float64_to_uint32(u.d, &env->vec_status);
3313 uint32_t helper_efdctsiz (uint64_t val)
3315 CPU_DoubleU u;
3317 u.ll = val;
3318 /* NaN are not treated the same way IEEE 754 does */
3319 if (unlikely(float64_is_nan(u.d)))
3320 return 0;
3322 return float64_to_int32_round_to_zero(u.d, &env->vec_status);
3325 uint64_t helper_efdctsidz (uint64_t val)
3327 CPU_DoubleU u;
3329 u.ll = val;
3330 /* NaN are not treated the same way IEEE 754 does */
3331 if (unlikely(float64_is_nan(u.d)))
3332 return 0;
3334 return float64_to_int64_round_to_zero(u.d, &env->vec_status);
3337 uint32_t helper_efdctuiz (uint64_t val)
3339 CPU_DoubleU u;
3341 u.ll = val;
3342 /* NaN are not treated the same way IEEE 754 does */
3343 if (unlikely(float64_is_nan(u.d)))
3344 return 0;
3346 return float64_to_uint32_round_to_zero(u.d, &env->vec_status);
3349 uint64_t helper_efdctuidz (uint64_t val)
3351 CPU_DoubleU u;
3353 u.ll = val;
3354 /* NaN are not treated the same way IEEE 754 does */
3355 if (unlikely(float64_is_nan(u.d)))
3356 return 0;
3358 return float64_to_uint64_round_to_zero(u.d, &env->vec_status);
3361 uint64_t helper_efdcfsf (uint32_t val)
3363 CPU_DoubleU u;
3364 float64 tmp;
3366 u.d = int32_to_float64(val, &env->vec_status);
3367 tmp = int64_to_float64(1ULL << 32, &env->vec_status);
3368 u.d = float64_div(u.d, tmp, &env->vec_status);
3370 return u.ll;
3373 uint64_t helper_efdcfuf (uint32_t val)
3375 CPU_DoubleU u;
3376 float64 tmp;
3378 u.d = uint32_to_float64(val, &env->vec_status);
3379 tmp = int64_to_float64(1ULL << 32, &env->vec_status);
3380 u.d = float64_div(u.d, tmp, &env->vec_status);
3382 return u.ll;
3385 uint32_t helper_efdctsf (uint64_t val)
3387 CPU_DoubleU u;
3388 float64 tmp;
3390 u.ll = val;
3391 /* NaN are not treated the same way IEEE 754 does */
3392 if (unlikely(float64_is_nan(u.d)))
3393 return 0;
3394 tmp = uint64_to_float64(1ULL << 32, &env->vec_status);
3395 u.d = float64_mul(u.d, tmp, &env->vec_status);
3397 return float64_to_int32(u.d, &env->vec_status);
3400 uint32_t helper_efdctuf (uint64_t val)
3402 CPU_DoubleU u;
3403 float64 tmp;
3405 u.ll = val;
3406 /* NaN are not treated the same way IEEE 754 does */
3407 if (unlikely(float64_is_nan(u.d)))
3408 return 0;
3409 tmp = uint64_to_float64(1ULL << 32, &env->vec_status);
3410 u.d = float64_mul(u.d, tmp, &env->vec_status);
3412 return float64_to_uint32(u.d, &env->vec_status);
3415 uint32_t helper_efscfd (uint64_t val)
3417 CPU_DoubleU u1;
3418 CPU_FloatU u2;
3420 u1.ll = val;
3421 u2.f = float64_to_float32(u1.d, &env->vec_status);
3423 return u2.l;
3426 uint64_t helper_efdcfs (uint32_t val)
3428 CPU_DoubleU u2;
3429 CPU_FloatU u1;
3431 u1.l = val;
3432 u2.d = float32_to_float64(u1.f, &env->vec_status);
3434 return u2.ll;
3437 /* Double precision fixed-point arithmetic */
3438 uint64_t helper_efdadd (uint64_t op1, uint64_t op2)
3440 CPU_DoubleU u1, u2;
3441 u1.ll = op1;
3442 u2.ll = op2;
3443 u1.d = float64_add(u1.d, u2.d, &env->vec_status);
3444 return u1.ll;
3447 uint64_t helper_efdsub (uint64_t op1, uint64_t op2)
3449 CPU_DoubleU u1, u2;
3450 u1.ll = op1;
3451 u2.ll = op2;
3452 u1.d = float64_sub(u1.d, u2.d, &env->vec_status);
3453 return u1.ll;
3456 uint64_t helper_efdmul (uint64_t op1, uint64_t op2)
3458 CPU_DoubleU u1, u2;
3459 u1.ll = op1;
3460 u2.ll = op2;
3461 u1.d = float64_mul(u1.d, u2.d, &env->vec_status);
3462 return u1.ll;
3465 uint64_t helper_efddiv (uint64_t op1, uint64_t op2)
3467 CPU_DoubleU u1, u2;
3468 u1.ll = op1;
3469 u2.ll = op2;
3470 u1.d = float64_div(u1.d, u2.d, &env->vec_status);
3471 return u1.ll;
3474 /* Double precision floating point helpers */
3475 uint32_t helper_efdtstlt (uint64_t op1, uint64_t op2)
3477 CPU_DoubleU u1, u2;
3478 u1.ll = op1;
3479 u2.ll = op2;
3480 return float64_lt(u1.d, u2.d, &env->vec_status) ? 4 : 0;
3483 uint32_t helper_efdtstgt (uint64_t op1, uint64_t op2)
3485 CPU_DoubleU u1, u2;
3486 u1.ll = op1;
3487 u2.ll = op2;
3488 return float64_le(u1.d, u2.d, &env->vec_status) ? 0 : 4;
3491 uint32_t helper_efdtsteq (uint64_t op1, uint64_t op2)
3493 CPU_DoubleU u1, u2;
3494 u1.ll = op1;
3495 u2.ll = op2;
3496 return float64_eq(u1.d, u2.d, &env->vec_status) ? 4 : 0;
3499 uint32_t helper_efdcmplt (uint64_t op1, uint64_t op2)
3501 /* XXX: TODO: test special values (NaN, infinites, ...) */
3502 return helper_efdtstlt(op1, op2);
3505 uint32_t helper_efdcmpgt (uint64_t op1, uint64_t op2)
3507 /* XXX: TODO: test special values (NaN, infinites, ...) */
3508 return helper_efdtstgt(op1, op2);
3511 uint32_t helper_efdcmpeq (uint64_t op1, uint64_t op2)
3513 /* XXX: TODO: test special values (NaN, infinites, ...) */
3514 return helper_efdtsteq(op1, op2);
3517 /*****************************************************************************/
3518 /* Softmmu support */
3519 #if !defined (CONFIG_USER_ONLY)
3521 #define MMUSUFFIX _mmu
3523 #define SHIFT 0
3524 #include "softmmu_template.h"
3526 #define SHIFT 1
3527 #include "softmmu_template.h"
3529 #define SHIFT 2
3530 #include "softmmu_template.h"
3532 #define SHIFT 3
3533 #include "softmmu_template.h"
3535 /* try to fill the TLB and return an exception if error. If retaddr is
3536 NULL, it means that the function was called in C code (i.e. not
3537 from generated code or from helper.c) */
3538 /* XXX: fix it to restore all registers */
3539 void tlb_fill (target_ulong addr, int is_write, int mmu_idx, void *retaddr)
3541 TranslationBlock *tb;
3542 CPUState *saved_env;
3543 unsigned long pc;
3544 int ret;
3546 /* XXX: hack to restore env in all cases, even if not called from
3547 generated code */
3548 saved_env = env;
3549 env = cpu_single_env;
3550 ret = cpu_ppc_handle_mmu_fault(env, addr, is_write, mmu_idx, 1);
3551 if (unlikely(ret != 0)) {
3552 if (likely(retaddr)) {
3553 /* now we have a real cpu fault */
3554 pc = (unsigned long)retaddr;
3555 tb = tb_find_pc(pc);
3556 if (likely(tb)) {
3557 /* the PC is inside the translated code. It means that we have
3558 a virtual CPU fault */
3559 cpu_restore_state(tb, env, pc, NULL);
3562 helper_raise_exception_err(env->exception_index, env->error_code);
3564 env = saved_env;
3567 /* Segment registers load and store */
3568 target_ulong helper_load_sr (target_ulong sr_num)
3570 return env->sr[sr_num];
3573 void helper_store_sr (target_ulong sr_num, target_ulong val)
3575 ppc_store_sr(env, sr_num, val);
3578 /* SLB management */
3579 #if defined(TARGET_PPC64)
3580 target_ulong helper_load_slb (target_ulong slb_nr)
3582 return ppc_load_slb(env, slb_nr);
3585 void helper_store_slb (target_ulong slb_nr, target_ulong rs)
3587 ppc_store_slb(env, slb_nr, rs);
3590 void helper_slbia (void)
3592 ppc_slb_invalidate_all(env);
3595 void helper_slbie (target_ulong addr)
3597 ppc_slb_invalidate_one(env, addr);
3600 #endif /* defined(TARGET_PPC64) */
3602 /* TLB management */
3603 void helper_tlbia (void)
3605 ppc_tlb_invalidate_all(env);
3608 void helper_tlbie (target_ulong addr)
3610 ppc_tlb_invalidate_one(env, addr);
3613 /* Software driven TLBs management */
3614 /* PowerPC 602/603 software TLB load instructions helpers */
3615 static void do_6xx_tlb (target_ulong new_EPN, int is_code)
3617 target_ulong RPN, CMP, EPN;
3618 int way;
3620 RPN = env->spr[SPR_RPA];
3621 if (is_code) {
3622 CMP = env->spr[SPR_ICMP];
3623 EPN = env->spr[SPR_IMISS];
3624 } else {
3625 CMP = env->spr[SPR_DCMP];
3626 EPN = env->spr[SPR_DMISS];
3628 way = (env->spr[SPR_SRR1] >> 17) & 1;
3629 LOG_SWTLB("%s: EPN " ADDRX " " ADDRX " PTE0 " ADDRX
3630 " PTE1 " ADDRX " way %d\n",
3631 __func__, new_EPN, EPN, CMP, RPN, way);
3632 /* Store this TLB */
3633 ppc6xx_tlb_store(env, (uint32_t)(new_EPN & TARGET_PAGE_MASK),
3634 way, is_code, CMP, RPN);
3637 void helper_6xx_tlbd (target_ulong EPN)
3639 do_6xx_tlb(EPN, 0);
3642 void helper_6xx_tlbi (target_ulong EPN)
3644 do_6xx_tlb(EPN, 1);
3647 /* PowerPC 74xx software TLB load instructions helpers */
3648 static void do_74xx_tlb (target_ulong new_EPN, int is_code)
3650 target_ulong RPN, CMP, EPN;
3651 int way;
3653 RPN = env->spr[SPR_PTELO];
3654 CMP = env->spr[SPR_PTEHI];
3655 EPN = env->spr[SPR_TLBMISS] & ~0x3;
3656 way = env->spr[SPR_TLBMISS] & 0x3;
3657 LOG_SWTLB("%s: EPN " ADDRX " " ADDRX " PTE0 " ADDRX
3658 " PTE1 " ADDRX " way %d\n",
3659 __func__, new_EPN, EPN, CMP, RPN, way);
3660 /* Store this TLB */
3661 ppc6xx_tlb_store(env, (uint32_t)(new_EPN & TARGET_PAGE_MASK),
3662 way, is_code, CMP, RPN);
3665 void helper_74xx_tlbd (target_ulong EPN)
3667 do_74xx_tlb(EPN, 0);
3670 void helper_74xx_tlbi (target_ulong EPN)
3672 do_74xx_tlb(EPN, 1);
3675 static always_inline target_ulong booke_tlb_to_page_size (int size)
3677 return 1024 << (2 * size);
3680 static always_inline int booke_page_size_to_tlb (target_ulong page_size)
3682 int size;
3684 switch (page_size) {
3685 case 0x00000400UL:
3686 size = 0x0;
3687 break;
3688 case 0x00001000UL:
3689 size = 0x1;
3690 break;
3691 case 0x00004000UL:
3692 size = 0x2;
3693 break;
3694 case 0x00010000UL:
3695 size = 0x3;
3696 break;
3697 case 0x00040000UL:
3698 size = 0x4;
3699 break;
3700 case 0x00100000UL:
3701 size = 0x5;
3702 break;
3703 case 0x00400000UL:
3704 size = 0x6;
3705 break;
3706 case 0x01000000UL:
3707 size = 0x7;
3708 break;
3709 case 0x04000000UL:
3710 size = 0x8;
3711 break;
3712 case 0x10000000UL:
3713 size = 0x9;
3714 break;
3715 case 0x40000000UL:
3716 size = 0xA;
3717 break;
3718 #if defined (TARGET_PPC64)
3719 case 0x000100000000ULL:
3720 size = 0xB;
3721 break;
3722 case 0x000400000000ULL:
3723 size = 0xC;
3724 break;
3725 case 0x001000000000ULL:
3726 size = 0xD;
3727 break;
3728 case 0x004000000000ULL:
3729 size = 0xE;
3730 break;
3731 case 0x010000000000ULL:
3732 size = 0xF;
3733 break;
3734 #endif
3735 default:
3736 size = -1;
3737 break;
3740 return size;
3743 /* Helpers for 4xx TLB management */
3744 target_ulong helper_4xx_tlbre_lo (target_ulong entry)
3746 ppcemb_tlb_t *tlb;
3747 target_ulong ret;
3748 int size;
3750 entry &= 0x3F;
3751 tlb = &env->tlb[entry].tlbe;
3752 ret = tlb->EPN;
3753 if (tlb->prot & PAGE_VALID)
3754 ret |= 0x400;
3755 size = booke_page_size_to_tlb(tlb->size);
3756 if (size < 0 || size > 0x7)
3757 size = 1;
3758 ret |= size << 7;
3759 env->spr[SPR_40x_PID] = tlb->PID;
3760 return ret;
3763 target_ulong helper_4xx_tlbre_hi (target_ulong entry)
3765 ppcemb_tlb_t *tlb;
3766 target_ulong ret;
3768 entry &= 0x3F;
3769 tlb = &env->tlb[entry].tlbe;
3770 ret = tlb->RPN;
3771 if (tlb->prot & PAGE_EXEC)
3772 ret |= 0x200;
3773 if (tlb->prot & PAGE_WRITE)
3774 ret |= 0x100;
3775 return ret;
3778 void helper_4xx_tlbwe_hi (target_ulong entry, target_ulong val)
3780 ppcemb_tlb_t *tlb;
3781 target_ulong page, end;
3783 LOG_SWTLB("%s entry %d val " ADDRX "\n", __func__, (int)entry, val);
3784 entry &= 0x3F;
3785 tlb = &env->tlb[entry].tlbe;
3786 /* Invalidate previous TLB (if it's valid) */
3787 if (tlb->prot & PAGE_VALID) {
3788 end = tlb->EPN + tlb->size;
3789 LOG_SWTLB("%s: invalidate old TLB %d start " ADDRX
3790 " end " ADDRX "\n", __func__, (int)entry, tlb->EPN, end);
3791 for (page = tlb->EPN; page < end; page += TARGET_PAGE_SIZE)
3792 tlb_flush_page(env, page);
3794 tlb->size = booke_tlb_to_page_size((val >> 7) & 0x7);
3795 /* We cannot handle TLB size < TARGET_PAGE_SIZE.
3796 * If this ever occurs, one should use the ppcemb target instead
3797 * of the ppc or ppc64 one
3799 if ((val & 0x40) && tlb->size < TARGET_PAGE_SIZE) {
3800 cpu_abort(env, "TLB size " TARGET_FMT_lu " < %u "
3801 "are not supported (%d)\n",
3802 tlb->size, TARGET_PAGE_SIZE, (int)((val >> 7) & 0x7));
3804 tlb->EPN = val & ~(tlb->size - 1);
3805 if (val & 0x40)
3806 tlb->prot |= PAGE_VALID;
3807 else
3808 tlb->prot &= ~PAGE_VALID;
3809 if (val & 0x20) {
3810 /* XXX: TO BE FIXED */
3811 cpu_abort(env, "Little-endian TLB entries are not supported by now\n");
3813 tlb->PID = env->spr[SPR_40x_PID]; /* PID */
3814 tlb->attr = val & 0xFF;
3815 LOG_SWTLB("%s: set up TLB %d RPN " PADDRX " EPN " ADDRX
3816 " size " ADDRX " prot %c%c%c%c PID %d\n", __func__,
3817 (int)entry, tlb->RPN, tlb->EPN, tlb->size,
3818 tlb->prot & PAGE_READ ? 'r' : '-',
3819 tlb->prot & PAGE_WRITE ? 'w' : '-',
3820 tlb->prot & PAGE_EXEC ? 'x' : '-',
3821 tlb->prot & PAGE_VALID ? 'v' : '-', (int)tlb->PID);
3822 /* Invalidate new TLB (if valid) */
3823 if (tlb->prot & PAGE_VALID) {
3824 end = tlb->EPN + tlb->size;
3825 LOG_SWTLB("%s: invalidate TLB %d start " ADDRX
3826 " end " ADDRX "\n", __func__, (int)entry, tlb->EPN, end);
3827 for (page = tlb->EPN; page < end; page += TARGET_PAGE_SIZE)
3828 tlb_flush_page(env, page);
3832 void helper_4xx_tlbwe_lo (target_ulong entry, target_ulong val)
3834 ppcemb_tlb_t *tlb;
3836 LOG_SWTLB("%s entry %i val " ADDRX "\n", __func__, (int)entry, val);
3837 entry &= 0x3F;
3838 tlb = &env->tlb[entry].tlbe;
3839 tlb->RPN = val & 0xFFFFFC00;
3840 tlb->prot = PAGE_READ;
3841 if (val & 0x200)
3842 tlb->prot |= PAGE_EXEC;
3843 if (val & 0x100)
3844 tlb->prot |= PAGE_WRITE;
3845 LOG_SWTLB("%s: set up TLB %d RPN " PADDRX " EPN " ADDRX
3846 " size " ADDRX " prot %c%c%c%c PID %d\n", __func__,
3847 (int)entry, tlb->RPN, tlb->EPN, tlb->size,
3848 tlb->prot & PAGE_READ ? 'r' : '-',
3849 tlb->prot & PAGE_WRITE ? 'w' : '-',
3850 tlb->prot & PAGE_EXEC ? 'x' : '-',
3851 tlb->prot & PAGE_VALID ? 'v' : '-', (int)tlb->PID);
3854 target_ulong helper_4xx_tlbsx (target_ulong address)
3856 return ppcemb_tlb_search(env, address, env->spr[SPR_40x_PID]);
3859 /* PowerPC 440 TLB management */
3860 void helper_440_tlbwe (uint32_t word, target_ulong entry, target_ulong value)
3862 ppcemb_tlb_t *tlb;
3863 target_ulong EPN, RPN, size;
3864 int do_flush_tlbs;
3866 LOG_SWTLB("%s word %d entry %d value " ADDRX "\n",
3867 __func__, word, (int)entry, value);
3868 do_flush_tlbs = 0;
3869 entry &= 0x3F;
3870 tlb = &env->tlb[entry].tlbe;
3871 switch (word) {
3872 default:
3873 /* Just here to please gcc */
3874 case 0:
3875 EPN = value & 0xFFFFFC00;
3876 if ((tlb->prot & PAGE_VALID) && EPN != tlb->EPN)
3877 do_flush_tlbs = 1;
3878 tlb->EPN = EPN;
3879 size = booke_tlb_to_page_size((value >> 4) & 0xF);
3880 if ((tlb->prot & PAGE_VALID) && tlb->size < size)
3881 do_flush_tlbs = 1;
3882 tlb->size = size;
3883 tlb->attr &= ~0x1;
3884 tlb->attr |= (value >> 8) & 1;
3885 if (value & 0x200) {
3886 tlb->prot |= PAGE_VALID;
3887 } else {
3888 if (tlb->prot & PAGE_VALID) {
3889 tlb->prot &= ~PAGE_VALID;
3890 do_flush_tlbs = 1;
3893 tlb->PID = env->spr[SPR_440_MMUCR] & 0x000000FF;
3894 if (do_flush_tlbs)
3895 tlb_flush(env, 1);
3896 break;
3897 case 1:
3898 RPN = value & 0xFFFFFC0F;
3899 if ((tlb->prot & PAGE_VALID) && tlb->RPN != RPN)
3900 tlb_flush(env, 1);
3901 tlb->RPN = RPN;
3902 break;
3903 case 2:
3904 tlb->attr = (tlb->attr & 0x1) | (value & 0x0000FF00);
3905 tlb->prot = tlb->prot & PAGE_VALID;
3906 if (value & 0x1)
3907 tlb->prot |= PAGE_READ << 4;
3908 if (value & 0x2)
3909 tlb->prot |= PAGE_WRITE << 4;
3910 if (value & 0x4)
3911 tlb->prot |= PAGE_EXEC << 4;
3912 if (value & 0x8)
3913 tlb->prot |= PAGE_READ;
3914 if (value & 0x10)
3915 tlb->prot |= PAGE_WRITE;
3916 if (value & 0x20)
3917 tlb->prot |= PAGE_EXEC;
3918 break;
3922 target_ulong helper_440_tlbre (uint32_t word, target_ulong entry)
3924 ppcemb_tlb_t *tlb;
3925 target_ulong ret;
3926 int size;
3928 entry &= 0x3F;
3929 tlb = &env->tlb[entry].tlbe;
3930 switch (word) {
3931 default:
3932 /* Just here to please gcc */
3933 case 0:
3934 ret = tlb->EPN;
3935 size = booke_page_size_to_tlb(tlb->size);
3936 if (size < 0 || size > 0xF)
3937 size = 1;
3938 ret |= size << 4;
3939 if (tlb->attr & 0x1)
3940 ret |= 0x100;
3941 if (tlb->prot & PAGE_VALID)
3942 ret |= 0x200;
3943 env->spr[SPR_440_MMUCR] &= ~0x000000FF;
3944 env->spr[SPR_440_MMUCR] |= tlb->PID;
3945 break;
3946 case 1:
3947 ret = tlb->RPN;
3948 break;
3949 case 2:
3950 ret = tlb->attr & ~0x1;
3951 if (tlb->prot & (PAGE_READ << 4))
3952 ret |= 0x1;
3953 if (tlb->prot & (PAGE_WRITE << 4))
3954 ret |= 0x2;
3955 if (tlb->prot & (PAGE_EXEC << 4))
3956 ret |= 0x4;
3957 if (tlb->prot & PAGE_READ)
3958 ret |= 0x8;
3959 if (tlb->prot & PAGE_WRITE)
3960 ret |= 0x10;
3961 if (tlb->prot & PAGE_EXEC)
3962 ret |= 0x20;
3963 break;
3965 return ret;
3968 target_ulong helper_440_tlbsx (target_ulong address)
3970 return ppcemb_tlb_search(env, address, env->spr[SPR_440_MMUCR] & 0xFF);
3973 #endif /* !CONFIG_USER_ONLY */