Merge remote-tracking branch 'remotes/kraxel/tags/pull-gtk-20141121-1' into staging
[qemu.git] / target-ppc / fpu_helper.c
blob7f74466f32ecfca54d06fec7110388b2c40ed3b0
1 /*
2 * PowerPC floating point and SPE 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, see <http://www.gnu.org/licenses/>.
19 #include "cpu.h"
20 #include "exec/helper-proto.h"
22 /*****************************************************************************/
23 /* Floating point operations helpers */
24 uint64_t helper_float32_to_float64(CPUPPCState *env, uint32_t arg)
26 CPU_FloatU f;
27 CPU_DoubleU d;
29 f.l = arg;
30 d.d = float32_to_float64(f.f, &env->fp_status);
31 return d.ll;
34 uint32_t helper_float64_to_float32(CPUPPCState *env, uint64_t arg)
36 CPU_FloatU f;
37 CPU_DoubleU d;
39 d.ll = arg;
40 f.f = float64_to_float32(d.d, &env->fp_status);
41 return f.l;
44 static inline int isden(float64 d)
46 CPU_DoubleU u;
48 u.d = d;
50 return ((u.ll >> 52) & 0x7FF) == 0;
53 static inline int ppc_float32_get_unbiased_exp(float32 f)
55 return ((f >> 23) & 0xFF) - 127;
58 static inline int ppc_float64_get_unbiased_exp(float64 f)
60 return ((f >> 52) & 0x7FF) - 1023;
63 uint32_t helper_compute_fprf(CPUPPCState *env, uint64_t arg, uint32_t set_fprf)
65 CPU_DoubleU farg;
66 int isneg;
67 int ret;
69 farg.ll = arg;
70 isneg = float64_is_neg(farg.d);
71 if (unlikely(float64_is_any_nan(farg.d))) {
72 if (float64_is_signaling_nan(farg.d)) {
73 /* Signaling NaN: flags are undefined */
74 ret = 0x00;
75 } else {
76 /* Quiet NaN */
77 ret = 0x11;
79 } else if (unlikely(float64_is_infinity(farg.d))) {
80 /* +/- infinity */
81 if (isneg) {
82 ret = 0x09;
83 } else {
84 ret = 0x05;
86 } else {
87 if (float64_is_zero(farg.d)) {
88 /* +/- zero */
89 if (isneg) {
90 ret = 0x12;
91 } else {
92 ret = 0x02;
94 } else {
95 if (isden(farg.d)) {
96 /* Denormalized numbers */
97 ret = 0x10;
98 } else {
99 /* Normalized numbers */
100 ret = 0x00;
102 if (isneg) {
103 ret |= 0x08;
104 } else {
105 ret |= 0x04;
109 if (set_fprf) {
110 /* We update FPSCR_FPRF */
111 env->fpscr &= ~(0x1F << FPSCR_FPRF);
112 env->fpscr |= ret << FPSCR_FPRF;
114 /* We just need fpcc to update Rc1 */
115 return ret & 0xF;
118 /* Floating-point invalid operations exception */
119 static inline uint64_t fload_invalid_op_excp(CPUPPCState *env, int op,
120 int set_fpcc)
122 CPUState *cs = CPU(ppc_env_get_cpu(env));
123 uint64_t ret = 0;
124 int ve;
126 ve = fpscr_ve;
127 switch (op) {
128 case POWERPC_EXCP_FP_VXSNAN:
129 env->fpscr |= 1 << FPSCR_VXSNAN;
130 break;
131 case POWERPC_EXCP_FP_VXSOFT:
132 env->fpscr |= 1 << FPSCR_VXSOFT;
133 break;
134 case POWERPC_EXCP_FP_VXISI:
135 /* Magnitude subtraction of infinities */
136 env->fpscr |= 1 << FPSCR_VXISI;
137 goto update_arith;
138 case POWERPC_EXCP_FP_VXIDI:
139 /* Division of infinity by infinity */
140 env->fpscr |= 1 << FPSCR_VXIDI;
141 goto update_arith;
142 case POWERPC_EXCP_FP_VXZDZ:
143 /* Division of zero by zero */
144 env->fpscr |= 1 << FPSCR_VXZDZ;
145 goto update_arith;
146 case POWERPC_EXCP_FP_VXIMZ:
147 /* Multiplication of zero by infinity */
148 env->fpscr |= 1 << FPSCR_VXIMZ;
149 goto update_arith;
150 case POWERPC_EXCP_FP_VXVC:
151 /* Ordered comparison of NaN */
152 env->fpscr |= 1 << FPSCR_VXVC;
153 if (set_fpcc) {
154 env->fpscr &= ~(0xF << FPSCR_FPCC);
155 env->fpscr |= 0x11 << FPSCR_FPCC;
157 /* We must update the target FPR before raising the exception */
158 if (ve != 0) {
159 cs->exception_index = POWERPC_EXCP_PROGRAM;
160 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_VXVC;
161 /* Update the floating-point enabled exception summary */
162 env->fpscr |= 1 << FPSCR_FEX;
163 /* Exception is differed */
164 ve = 0;
166 break;
167 case POWERPC_EXCP_FP_VXSQRT:
168 /* Square root of a negative number */
169 env->fpscr |= 1 << FPSCR_VXSQRT;
170 update_arith:
171 env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI));
172 if (ve == 0) {
173 /* Set the result to quiet NaN */
174 ret = 0x7FF8000000000000ULL;
175 if (set_fpcc) {
176 env->fpscr &= ~(0xF << FPSCR_FPCC);
177 env->fpscr |= 0x11 << FPSCR_FPCC;
180 break;
181 case POWERPC_EXCP_FP_VXCVI:
182 /* Invalid conversion */
183 env->fpscr |= 1 << FPSCR_VXCVI;
184 env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI));
185 if (ve == 0) {
186 /* Set the result to quiet NaN */
187 ret = 0x7FF8000000000000ULL;
188 if (set_fpcc) {
189 env->fpscr &= ~(0xF << FPSCR_FPCC);
190 env->fpscr |= 0x11 << FPSCR_FPCC;
193 break;
195 /* Update the floating-point invalid operation summary */
196 env->fpscr |= 1 << FPSCR_VX;
197 /* Update the floating-point exception summary */
198 env->fpscr |= 1 << FPSCR_FX;
199 if (ve != 0) {
200 /* Update the floating-point enabled exception summary */
201 env->fpscr |= 1 << FPSCR_FEX;
202 if (msr_fe0 != 0 || msr_fe1 != 0) {
203 helper_raise_exception_err(env, POWERPC_EXCP_PROGRAM,
204 POWERPC_EXCP_FP | op);
207 return ret;
210 static inline void float_zero_divide_excp(CPUPPCState *env)
212 env->fpscr |= 1 << FPSCR_ZX;
213 env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI));
214 /* Update the floating-point exception summary */
215 env->fpscr |= 1 << FPSCR_FX;
216 if (fpscr_ze != 0) {
217 /* Update the floating-point enabled exception summary */
218 env->fpscr |= 1 << FPSCR_FEX;
219 if (msr_fe0 != 0 || msr_fe1 != 0) {
220 helper_raise_exception_err(env, POWERPC_EXCP_PROGRAM,
221 POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX);
226 static inline void float_overflow_excp(CPUPPCState *env)
228 CPUState *cs = CPU(ppc_env_get_cpu(env));
230 env->fpscr |= 1 << FPSCR_OX;
231 /* Update the floating-point exception summary */
232 env->fpscr |= 1 << FPSCR_FX;
233 if (fpscr_oe != 0) {
234 /* XXX: should adjust the result */
235 /* Update the floating-point enabled exception summary */
236 env->fpscr |= 1 << FPSCR_FEX;
237 /* We must update the target FPR before raising the exception */
238 cs->exception_index = POWERPC_EXCP_PROGRAM;
239 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX;
240 } else {
241 env->fpscr |= 1 << FPSCR_XX;
242 env->fpscr |= 1 << FPSCR_FI;
246 static inline void float_underflow_excp(CPUPPCState *env)
248 CPUState *cs = CPU(ppc_env_get_cpu(env));
250 env->fpscr |= 1 << FPSCR_UX;
251 /* Update the floating-point exception summary */
252 env->fpscr |= 1 << FPSCR_FX;
253 if (fpscr_ue != 0) {
254 /* XXX: should adjust the result */
255 /* Update the floating-point enabled exception summary */
256 env->fpscr |= 1 << FPSCR_FEX;
257 /* We must update the target FPR before raising the exception */
258 cs->exception_index = POWERPC_EXCP_PROGRAM;
259 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX;
263 static inline void float_inexact_excp(CPUPPCState *env)
265 CPUState *cs = CPU(ppc_env_get_cpu(env));
267 env->fpscr |= 1 << FPSCR_XX;
268 /* Update the floating-point exception summary */
269 env->fpscr |= 1 << FPSCR_FX;
270 if (fpscr_xe != 0) {
271 /* Update the floating-point enabled exception summary */
272 env->fpscr |= 1 << FPSCR_FEX;
273 /* We must update the target FPR before raising the exception */
274 cs->exception_index = POWERPC_EXCP_PROGRAM;
275 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX;
279 static inline void fpscr_set_rounding_mode(CPUPPCState *env)
281 int rnd_type;
283 /* Set rounding mode */
284 switch (fpscr_rn) {
285 case 0:
286 /* Best approximation (round to nearest) */
287 rnd_type = float_round_nearest_even;
288 break;
289 case 1:
290 /* Smaller magnitude (round toward zero) */
291 rnd_type = float_round_to_zero;
292 break;
293 case 2:
294 /* Round toward +infinite */
295 rnd_type = float_round_up;
296 break;
297 default:
298 case 3:
299 /* Round toward -infinite */
300 rnd_type = float_round_down;
301 break;
303 set_float_rounding_mode(rnd_type, &env->fp_status);
306 void helper_fpscr_clrbit(CPUPPCState *env, uint32_t bit)
308 int prev;
310 prev = (env->fpscr >> bit) & 1;
311 env->fpscr &= ~(1 << bit);
312 if (prev == 1) {
313 switch (bit) {
314 case FPSCR_RN1:
315 case FPSCR_RN:
316 fpscr_set_rounding_mode(env);
317 break;
318 default:
319 break;
324 void helper_fpscr_setbit(CPUPPCState *env, uint32_t bit)
326 CPUState *cs = CPU(ppc_env_get_cpu(env));
327 int prev;
329 prev = (env->fpscr >> bit) & 1;
330 env->fpscr |= 1 << bit;
331 if (prev == 0) {
332 switch (bit) {
333 case FPSCR_VX:
334 env->fpscr |= 1 << FPSCR_FX;
335 if (fpscr_ve) {
336 goto raise_ve;
338 break;
339 case FPSCR_OX:
340 env->fpscr |= 1 << FPSCR_FX;
341 if (fpscr_oe) {
342 goto raise_oe;
344 break;
345 case FPSCR_UX:
346 env->fpscr |= 1 << FPSCR_FX;
347 if (fpscr_ue) {
348 goto raise_ue;
350 break;
351 case FPSCR_ZX:
352 env->fpscr |= 1 << FPSCR_FX;
353 if (fpscr_ze) {
354 goto raise_ze;
356 break;
357 case FPSCR_XX:
358 env->fpscr |= 1 << FPSCR_FX;
359 if (fpscr_xe) {
360 goto raise_xe;
362 break;
363 case FPSCR_VXSNAN:
364 case FPSCR_VXISI:
365 case FPSCR_VXIDI:
366 case FPSCR_VXZDZ:
367 case FPSCR_VXIMZ:
368 case FPSCR_VXVC:
369 case FPSCR_VXSOFT:
370 case FPSCR_VXSQRT:
371 case FPSCR_VXCVI:
372 env->fpscr |= 1 << FPSCR_VX;
373 env->fpscr |= 1 << FPSCR_FX;
374 if (fpscr_ve != 0) {
375 goto raise_ve;
377 break;
378 case FPSCR_VE:
379 if (fpscr_vx != 0) {
380 raise_ve:
381 env->error_code = POWERPC_EXCP_FP;
382 if (fpscr_vxsnan) {
383 env->error_code |= POWERPC_EXCP_FP_VXSNAN;
385 if (fpscr_vxisi) {
386 env->error_code |= POWERPC_EXCP_FP_VXISI;
388 if (fpscr_vxidi) {
389 env->error_code |= POWERPC_EXCP_FP_VXIDI;
391 if (fpscr_vxzdz) {
392 env->error_code |= POWERPC_EXCP_FP_VXZDZ;
394 if (fpscr_vximz) {
395 env->error_code |= POWERPC_EXCP_FP_VXIMZ;
397 if (fpscr_vxvc) {
398 env->error_code |= POWERPC_EXCP_FP_VXVC;
400 if (fpscr_vxsoft) {
401 env->error_code |= POWERPC_EXCP_FP_VXSOFT;
403 if (fpscr_vxsqrt) {
404 env->error_code |= POWERPC_EXCP_FP_VXSQRT;
406 if (fpscr_vxcvi) {
407 env->error_code |= POWERPC_EXCP_FP_VXCVI;
409 goto raise_excp;
411 break;
412 case FPSCR_OE:
413 if (fpscr_ox != 0) {
414 raise_oe:
415 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX;
416 goto raise_excp;
418 break;
419 case FPSCR_UE:
420 if (fpscr_ux != 0) {
421 raise_ue:
422 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX;
423 goto raise_excp;
425 break;
426 case FPSCR_ZE:
427 if (fpscr_zx != 0) {
428 raise_ze:
429 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX;
430 goto raise_excp;
432 break;
433 case FPSCR_XE:
434 if (fpscr_xx != 0) {
435 raise_xe:
436 env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX;
437 goto raise_excp;
439 break;
440 case FPSCR_RN1:
441 case FPSCR_RN:
442 fpscr_set_rounding_mode(env);
443 break;
444 default:
445 break;
446 raise_excp:
447 /* Update the floating-point enabled exception summary */
448 env->fpscr |= 1 << FPSCR_FEX;
449 /* We have to update Rc1 before raising the exception */
450 cs->exception_index = POWERPC_EXCP_PROGRAM;
451 break;
456 void helper_store_fpscr(CPUPPCState *env, uint64_t arg, uint32_t mask)
458 CPUState *cs = CPU(ppc_env_get_cpu(env));
459 target_ulong prev, new;
460 int i;
462 prev = env->fpscr;
463 new = (target_ulong)arg;
464 new &= ~0x60000000LL;
465 new |= prev & 0x60000000LL;
466 for (i = 0; i < sizeof(target_ulong) * 2; i++) {
467 if (mask & (1 << i)) {
468 env->fpscr &= ~(0xFLL << (4 * i));
469 env->fpscr |= new & (0xFLL << (4 * i));
472 /* Update VX and FEX */
473 if (fpscr_ix != 0) {
474 env->fpscr |= 1 << FPSCR_VX;
475 } else {
476 env->fpscr &= ~(1 << FPSCR_VX);
478 if ((fpscr_ex & fpscr_eex) != 0) {
479 env->fpscr |= 1 << FPSCR_FEX;
480 cs->exception_index = POWERPC_EXCP_PROGRAM;
481 /* XXX: we should compute it properly */
482 env->error_code = POWERPC_EXCP_FP;
483 } else {
484 env->fpscr &= ~(1 << FPSCR_FEX);
486 fpscr_set_rounding_mode(env);
489 void store_fpscr(CPUPPCState *env, uint64_t arg, uint32_t mask)
491 helper_store_fpscr(env, arg, mask);
494 void helper_float_check_status(CPUPPCState *env)
496 CPUState *cs = CPU(ppc_env_get_cpu(env));
497 int status = get_float_exception_flags(&env->fp_status);
499 if (status & float_flag_divbyzero) {
500 float_zero_divide_excp(env);
501 } else if (status & float_flag_overflow) {
502 float_overflow_excp(env);
503 } else if (status & float_flag_underflow) {
504 float_underflow_excp(env);
505 } else if (status & float_flag_inexact) {
506 float_inexact_excp(env);
509 if (cs->exception_index == POWERPC_EXCP_PROGRAM &&
510 (env->error_code & POWERPC_EXCP_FP)) {
511 /* Differred floating-point exception after target FPR update */
512 if (msr_fe0 != 0 || msr_fe1 != 0) {
513 helper_raise_exception_err(env, cs->exception_index,
514 env->error_code);
519 void helper_reset_fpstatus(CPUPPCState *env)
521 set_float_exception_flags(0, &env->fp_status);
524 /* fadd - fadd. */
525 uint64_t helper_fadd(CPUPPCState *env, uint64_t arg1, uint64_t arg2)
527 CPU_DoubleU farg1, farg2;
529 farg1.ll = arg1;
530 farg2.ll = arg2;
532 if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
533 float64_is_neg(farg1.d) != float64_is_neg(farg2.d))) {
534 /* Magnitude subtraction of infinities */
535 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1);
536 } else {
537 if (unlikely(float64_is_signaling_nan(farg1.d) ||
538 float64_is_signaling_nan(farg2.d))) {
539 /* sNaN addition */
540 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
542 farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status);
545 return farg1.ll;
548 /* fsub - fsub. */
549 uint64_t helper_fsub(CPUPPCState *env, uint64_t arg1, uint64_t arg2)
551 CPU_DoubleU farg1, farg2;
553 farg1.ll = arg1;
554 farg2.ll = arg2;
556 if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) &&
557 float64_is_neg(farg1.d) == float64_is_neg(farg2.d))) {
558 /* Magnitude subtraction of infinities */
559 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1);
560 } else {
561 if (unlikely(float64_is_signaling_nan(farg1.d) ||
562 float64_is_signaling_nan(farg2.d))) {
563 /* sNaN subtraction */
564 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
566 farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status);
569 return farg1.ll;
572 /* fmul - fmul. */
573 uint64_t helper_fmul(CPUPPCState *env, uint64_t arg1, uint64_t arg2)
575 CPU_DoubleU farg1, farg2;
577 farg1.ll = arg1;
578 farg2.ll = arg2;
580 if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
581 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
582 /* Multiplication of zero by infinity */
583 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1);
584 } else {
585 if (unlikely(float64_is_signaling_nan(farg1.d) ||
586 float64_is_signaling_nan(farg2.d))) {
587 /* sNaN multiplication */
588 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
590 farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status);
593 return farg1.ll;
596 /* fdiv - fdiv. */
597 uint64_t helper_fdiv(CPUPPCState *env, uint64_t arg1, uint64_t arg2)
599 CPU_DoubleU farg1, farg2;
601 farg1.ll = arg1;
602 farg2.ll = arg2;
604 if (unlikely(float64_is_infinity(farg1.d) &&
605 float64_is_infinity(farg2.d))) {
606 /* Division of infinity by infinity */
607 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIDI, 1);
608 } else if (unlikely(float64_is_zero(farg1.d) && float64_is_zero(farg2.d))) {
609 /* Division of zero by zero */
610 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXZDZ, 1);
611 } else {
612 if (unlikely(float64_is_signaling_nan(farg1.d) ||
613 float64_is_signaling_nan(farg2.d))) {
614 /* sNaN division */
615 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
617 farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status);
620 return farg1.ll;
624 #define FPU_FCTI(op, cvt, nanval) \
625 uint64_t helper_##op(CPUPPCState *env, uint64_t arg) \
627 CPU_DoubleU farg; \
629 farg.ll = arg; \
630 farg.ll = float64_to_##cvt(farg.d, &env->fp_status); \
632 if (unlikely(env->fp_status.float_exception_flags)) { \
633 if (float64_is_any_nan(arg)) { \
634 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXCVI, 1); \
635 if (float64_is_signaling_nan(arg)) { \
636 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); \
638 farg.ll = nanval; \
639 } else if (env->fp_status.float_exception_flags & \
640 float_flag_invalid) { \
641 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXCVI, 1); \
643 helper_float_check_status(env); \
645 return farg.ll; \
648 FPU_FCTI(fctiw, int32, 0x80000000U)
649 FPU_FCTI(fctiwz, int32_round_to_zero, 0x80000000U)
650 FPU_FCTI(fctiwu, uint32, 0x00000000U)
651 FPU_FCTI(fctiwuz, uint32_round_to_zero, 0x00000000U)
652 FPU_FCTI(fctid, int64, 0x8000000000000000ULL)
653 FPU_FCTI(fctidz, int64_round_to_zero, 0x8000000000000000ULL)
654 FPU_FCTI(fctidu, uint64, 0x0000000000000000ULL)
655 FPU_FCTI(fctiduz, uint64_round_to_zero, 0x0000000000000000ULL)
657 #define FPU_FCFI(op, cvtr, is_single) \
658 uint64_t helper_##op(CPUPPCState *env, uint64_t arg) \
660 CPU_DoubleU farg; \
662 if (is_single) { \
663 float32 tmp = cvtr(arg, &env->fp_status); \
664 farg.d = float32_to_float64(tmp, &env->fp_status); \
665 } else { \
666 farg.d = cvtr(arg, &env->fp_status); \
668 helper_float_check_status(env); \
669 return farg.ll; \
672 FPU_FCFI(fcfid, int64_to_float64, 0)
673 FPU_FCFI(fcfids, int64_to_float32, 1)
674 FPU_FCFI(fcfidu, uint64_to_float64, 0)
675 FPU_FCFI(fcfidus, uint64_to_float32, 1)
677 static inline uint64_t do_fri(CPUPPCState *env, uint64_t arg,
678 int rounding_mode)
680 CPU_DoubleU farg;
682 farg.ll = arg;
684 if (unlikely(float64_is_signaling_nan(farg.d))) {
685 /* sNaN round */
686 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
687 farg.ll = arg | 0x0008000000000000ULL;
688 } else {
689 int inexact = get_float_exception_flags(&env->fp_status) &
690 float_flag_inexact;
691 set_float_rounding_mode(rounding_mode, &env->fp_status);
692 farg.ll = float64_round_to_int(farg.d, &env->fp_status);
693 /* Restore rounding mode from FPSCR */
694 fpscr_set_rounding_mode(env);
696 /* fri* does not set FPSCR[XX] */
697 if (!inexact) {
698 env->fp_status.float_exception_flags &= ~float_flag_inexact;
701 helper_float_check_status(env);
702 return farg.ll;
705 uint64_t helper_frin(CPUPPCState *env, uint64_t arg)
707 return do_fri(env, arg, float_round_ties_away);
710 uint64_t helper_friz(CPUPPCState *env, uint64_t arg)
712 return do_fri(env, arg, float_round_to_zero);
715 uint64_t helper_frip(CPUPPCState *env, uint64_t arg)
717 return do_fri(env, arg, float_round_up);
720 uint64_t helper_frim(CPUPPCState *env, uint64_t arg)
722 return do_fri(env, arg, float_round_down);
725 /* fmadd - fmadd. */
726 uint64_t helper_fmadd(CPUPPCState *env, uint64_t arg1, uint64_t arg2,
727 uint64_t arg3)
729 CPU_DoubleU farg1, farg2, farg3;
731 farg1.ll = arg1;
732 farg2.ll = arg2;
733 farg3.ll = arg3;
735 if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
736 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
737 /* Multiplication of zero by infinity */
738 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1);
739 } else {
740 if (unlikely(float64_is_signaling_nan(farg1.d) ||
741 float64_is_signaling_nan(farg2.d) ||
742 float64_is_signaling_nan(farg3.d))) {
743 /* sNaN operation */
744 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
746 /* This is the way the PowerPC specification defines it */
747 float128 ft0_128, ft1_128;
749 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
750 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
751 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
752 if (unlikely(float128_is_infinity(ft0_128) &&
753 float64_is_infinity(farg3.d) &&
754 float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) {
755 /* Magnitude subtraction of infinities */
756 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1);
757 } else {
758 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
759 ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status);
760 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
764 return farg1.ll;
767 /* fmsub - fmsub. */
768 uint64_t helper_fmsub(CPUPPCState *env, uint64_t arg1, uint64_t arg2,
769 uint64_t arg3)
771 CPU_DoubleU farg1, farg2, farg3;
773 farg1.ll = arg1;
774 farg2.ll = arg2;
775 farg3.ll = arg3;
777 if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
778 (float64_is_zero(farg1.d) &&
779 float64_is_infinity(farg2.d)))) {
780 /* Multiplication of zero by infinity */
781 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1);
782 } else {
783 if (unlikely(float64_is_signaling_nan(farg1.d) ||
784 float64_is_signaling_nan(farg2.d) ||
785 float64_is_signaling_nan(farg3.d))) {
786 /* sNaN operation */
787 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
789 /* This is the way the PowerPC specification defines it */
790 float128 ft0_128, ft1_128;
792 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
793 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
794 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
795 if (unlikely(float128_is_infinity(ft0_128) &&
796 float64_is_infinity(farg3.d) &&
797 float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) {
798 /* Magnitude subtraction of infinities */
799 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1);
800 } else {
801 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
802 ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status);
803 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
806 return farg1.ll;
809 /* fnmadd - fnmadd. */
810 uint64_t helper_fnmadd(CPUPPCState *env, uint64_t arg1, uint64_t arg2,
811 uint64_t arg3)
813 CPU_DoubleU farg1, farg2, farg3;
815 farg1.ll = arg1;
816 farg2.ll = arg2;
817 farg3.ll = arg3;
819 if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
820 (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
821 /* Multiplication of zero by infinity */
822 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1);
823 } else {
824 if (unlikely(float64_is_signaling_nan(farg1.d) ||
825 float64_is_signaling_nan(farg2.d) ||
826 float64_is_signaling_nan(farg3.d))) {
827 /* sNaN operation */
828 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
830 /* This is the way the PowerPC specification defines it */
831 float128 ft0_128, ft1_128;
833 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
834 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
835 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
836 if (unlikely(float128_is_infinity(ft0_128) &&
837 float64_is_infinity(farg3.d) &&
838 float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) {
839 /* Magnitude subtraction of infinities */
840 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1);
841 } else {
842 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
843 ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status);
844 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
846 if (likely(!float64_is_any_nan(farg1.d))) {
847 farg1.d = float64_chs(farg1.d);
850 return farg1.ll;
853 /* fnmsub - fnmsub. */
854 uint64_t helper_fnmsub(CPUPPCState *env, uint64_t arg1, uint64_t arg2,
855 uint64_t arg3)
857 CPU_DoubleU farg1, farg2, farg3;
859 farg1.ll = arg1;
860 farg2.ll = arg2;
861 farg3.ll = arg3;
863 if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) ||
864 (float64_is_zero(farg1.d) &&
865 float64_is_infinity(farg2.d)))) {
866 /* Multiplication of zero by infinity */
867 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1);
868 } else {
869 if (unlikely(float64_is_signaling_nan(farg1.d) ||
870 float64_is_signaling_nan(farg2.d) ||
871 float64_is_signaling_nan(farg3.d))) {
872 /* sNaN operation */
873 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
875 /* This is the way the PowerPC specification defines it */
876 float128 ft0_128, ft1_128;
878 ft0_128 = float64_to_float128(farg1.d, &env->fp_status);
879 ft1_128 = float64_to_float128(farg2.d, &env->fp_status);
880 ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status);
881 if (unlikely(float128_is_infinity(ft0_128) &&
882 float64_is_infinity(farg3.d) &&
883 float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) {
884 /* Magnitude subtraction of infinities */
885 farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1);
886 } else {
887 ft1_128 = float64_to_float128(farg3.d, &env->fp_status);
888 ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status);
889 farg1.d = float128_to_float64(ft0_128, &env->fp_status);
891 if (likely(!float64_is_any_nan(farg1.d))) {
892 farg1.d = float64_chs(farg1.d);
895 return farg1.ll;
898 /* frsp - frsp. */
899 uint64_t helper_frsp(CPUPPCState *env, uint64_t arg)
901 CPU_DoubleU farg;
902 float32 f32;
904 farg.ll = arg;
906 if (unlikely(float64_is_signaling_nan(farg.d))) {
907 /* sNaN square root */
908 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
910 f32 = float64_to_float32(farg.d, &env->fp_status);
911 farg.d = float32_to_float64(f32, &env->fp_status);
913 return farg.ll;
916 /* fsqrt - fsqrt. */
917 uint64_t helper_fsqrt(CPUPPCState *env, uint64_t arg)
919 CPU_DoubleU farg;
921 farg.ll = arg;
923 if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
924 /* Square root of a negative nonzero number */
925 farg.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, 1);
926 } else {
927 if (unlikely(float64_is_signaling_nan(farg.d))) {
928 /* sNaN square root */
929 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
931 farg.d = float64_sqrt(farg.d, &env->fp_status);
933 return farg.ll;
936 /* fre - fre. */
937 uint64_t helper_fre(CPUPPCState *env, uint64_t arg)
939 CPU_DoubleU farg;
941 farg.ll = arg;
943 if (unlikely(float64_is_signaling_nan(farg.d))) {
944 /* sNaN reciprocal */
945 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
947 farg.d = float64_div(float64_one, farg.d, &env->fp_status);
948 return farg.d;
951 /* fres - fres. */
952 uint64_t helper_fres(CPUPPCState *env, uint64_t arg)
954 CPU_DoubleU farg;
955 float32 f32;
957 farg.ll = arg;
959 if (unlikely(float64_is_signaling_nan(farg.d))) {
960 /* sNaN reciprocal */
961 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
963 farg.d = float64_div(float64_one, farg.d, &env->fp_status);
964 f32 = float64_to_float32(farg.d, &env->fp_status);
965 farg.d = float32_to_float64(f32, &env->fp_status);
967 return farg.ll;
970 /* frsqrte - frsqrte. */
971 uint64_t helper_frsqrte(CPUPPCState *env, uint64_t arg)
973 CPU_DoubleU farg;
975 farg.ll = arg;
977 if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) {
978 /* Reciprocal square root of a negative nonzero number */
979 farg.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, 1);
980 } else {
981 if (unlikely(float64_is_signaling_nan(farg.d))) {
982 /* sNaN reciprocal square root */
983 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
985 farg.d = float64_sqrt(farg.d, &env->fp_status);
986 farg.d = float64_div(float64_one, farg.d, &env->fp_status);
988 return farg.ll;
991 /* fsel - fsel. */
992 uint64_t helper_fsel(CPUPPCState *env, uint64_t arg1, uint64_t arg2,
993 uint64_t arg3)
995 CPU_DoubleU farg1;
997 farg1.ll = arg1;
999 if ((!float64_is_neg(farg1.d) || float64_is_zero(farg1.d)) &&
1000 !float64_is_any_nan(farg1.d)) {
1001 return arg2;
1002 } else {
1003 return arg3;
1007 uint32_t helper_ftdiv(uint64_t fra, uint64_t frb)
1009 int fe_flag = 0;
1010 int fg_flag = 0;
1012 if (unlikely(float64_is_infinity(fra) ||
1013 float64_is_infinity(frb) ||
1014 float64_is_zero(frb))) {
1015 fe_flag = 1;
1016 fg_flag = 1;
1017 } else {
1018 int e_a = ppc_float64_get_unbiased_exp(fra);
1019 int e_b = ppc_float64_get_unbiased_exp(frb);
1021 if (unlikely(float64_is_any_nan(fra) ||
1022 float64_is_any_nan(frb))) {
1023 fe_flag = 1;
1024 } else if ((e_b <= -1022) || (e_b >= 1021)) {
1025 fe_flag = 1;
1026 } else if (!float64_is_zero(fra) &&
1027 (((e_a - e_b) >= 1023) ||
1028 ((e_a - e_b) <= -1021) ||
1029 (e_a <= -970))) {
1030 fe_flag = 1;
1033 if (unlikely(float64_is_zero_or_denormal(frb))) {
1034 /* XB is not zero because of the above check and */
1035 /* so must be denormalized. */
1036 fg_flag = 1;
1040 return 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0);
1043 uint32_t helper_ftsqrt(uint64_t frb)
1045 int fe_flag = 0;
1046 int fg_flag = 0;
1048 if (unlikely(float64_is_infinity(frb) || float64_is_zero(frb))) {
1049 fe_flag = 1;
1050 fg_flag = 1;
1051 } else {
1052 int e_b = ppc_float64_get_unbiased_exp(frb);
1054 if (unlikely(float64_is_any_nan(frb))) {
1055 fe_flag = 1;
1056 } else if (unlikely(float64_is_zero(frb))) {
1057 fe_flag = 1;
1058 } else if (unlikely(float64_is_neg(frb))) {
1059 fe_flag = 1;
1060 } else if (!float64_is_zero(frb) && (e_b <= (-1022+52))) {
1061 fe_flag = 1;
1064 if (unlikely(float64_is_zero_or_denormal(frb))) {
1065 /* XB is not zero because of the above check and */
1066 /* therefore must be denormalized. */
1067 fg_flag = 1;
1071 return 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0);
1074 void helper_fcmpu(CPUPPCState *env, uint64_t arg1, uint64_t arg2,
1075 uint32_t crfD)
1077 CPU_DoubleU farg1, farg2;
1078 uint32_t ret = 0;
1080 farg1.ll = arg1;
1081 farg2.ll = arg2;
1083 if (unlikely(float64_is_any_nan(farg1.d) ||
1084 float64_is_any_nan(farg2.d))) {
1085 ret = 0x01UL;
1086 } else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) {
1087 ret = 0x08UL;
1088 } else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) {
1089 ret = 0x04UL;
1090 } else {
1091 ret = 0x02UL;
1094 env->fpscr &= ~(0x0F << FPSCR_FPRF);
1095 env->fpscr |= ret << FPSCR_FPRF;
1096 env->crf[crfD] = ret;
1097 if (unlikely(ret == 0x01UL
1098 && (float64_is_signaling_nan(farg1.d) ||
1099 float64_is_signaling_nan(farg2.d)))) {
1100 /* sNaN comparison */
1101 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1);
1105 void helper_fcmpo(CPUPPCState *env, uint64_t arg1, uint64_t arg2,
1106 uint32_t crfD)
1108 CPU_DoubleU farg1, farg2;
1109 uint32_t ret = 0;
1111 farg1.ll = arg1;
1112 farg2.ll = arg2;
1114 if (unlikely(float64_is_any_nan(farg1.d) ||
1115 float64_is_any_nan(farg2.d))) {
1116 ret = 0x01UL;
1117 } else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) {
1118 ret = 0x08UL;
1119 } else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) {
1120 ret = 0x04UL;
1121 } else {
1122 ret = 0x02UL;
1125 env->fpscr &= ~(0x0F << FPSCR_FPRF);
1126 env->fpscr |= ret << FPSCR_FPRF;
1127 env->crf[crfD] = ret;
1128 if (unlikely(ret == 0x01UL)) {
1129 if (float64_is_signaling_nan(farg1.d) ||
1130 float64_is_signaling_nan(farg2.d)) {
1131 /* sNaN comparison */
1132 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN |
1133 POWERPC_EXCP_FP_VXVC, 1);
1134 } else {
1135 /* qNaN comparison */
1136 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXVC, 1);
1141 /* Single-precision floating-point conversions */
1142 static inline uint32_t efscfsi(CPUPPCState *env, uint32_t val)
1144 CPU_FloatU u;
1146 u.f = int32_to_float32(val, &env->vec_status);
1148 return u.l;
1151 static inline uint32_t efscfui(CPUPPCState *env, uint32_t val)
1153 CPU_FloatU u;
1155 u.f = uint32_to_float32(val, &env->vec_status);
1157 return u.l;
1160 static inline int32_t efsctsi(CPUPPCState *env, uint32_t val)
1162 CPU_FloatU u;
1164 u.l = val;
1165 /* NaN are not treated the same way IEEE 754 does */
1166 if (unlikely(float32_is_quiet_nan(u.f))) {
1167 return 0;
1170 return float32_to_int32(u.f, &env->vec_status);
1173 static inline uint32_t efsctui(CPUPPCState *env, uint32_t val)
1175 CPU_FloatU u;
1177 u.l = val;
1178 /* NaN are not treated the same way IEEE 754 does */
1179 if (unlikely(float32_is_quiet_nan(u.f))) {
1180 return 0;
1183 return float32_to_uint32(u.f, &env->vec_status);
1186 static inline uint32_t efsctsiz(CPUPPCState *env, uint32_t val)
1188 CPU_FloatU u;
1190 u.l = val;
1191 /* NaN are not treated the same way IEEE 754 does */
1192 if (unlikely(float32_is_quiet_nan(u.f))) {
1193 return 0;
1196 return float32_to_int32_round_to_zero(u.f, &env->vec_status);
1199 static inline uint32_t efsctuiz(CPUPPCState *env, uint32_t val)
1201 CPU_FloatU u;
1203 u.l = val;
1204 /* NaN are not treated the same way IEEE 754 does */
1205 if (unlikely(float32_is_quiet_nan(u.f))) {
1206 return 0;
1209 return float32_to_uint32_round_to_zero(u.f, &env->vec_status);
1212 static inline uint32_t efscfsf(CPUPPCState *env, uint32_t val)
1214 CPU_FloatU u;
1215 float32 tmp;
1217 u.f = int32_to_float32(val, &env->vec_status);
1218 tmp = int64_to_float32(1ULL << 32, &env->vec_status);
1219 u.f = float32_div(u.f, tmp, &env->vec_status);
1221 return u.l;
1224 static inline uint32_t efscfuf(CPUPPCState *env, uint32_t val)
1226 CPU_FloatU u;
1227 float32 tmp;
1229 u.f = uint32_to_float32(val, &env->vec_status);
1230 tmp = uint64_to_float32(1ULL << 32, &env->vec_status);
1231 u.f = float32_div(u.f, tmp, &env->vec_status);
1233 return u.l;
1236 static inline uint32_t efsctsf(CPUPPCState *env, uint32_t val)
1238 CPU_FloatU u;
1239 float32 tmp;
1241 u.l = val;
1242 /* NaN are not treated the same way IEEE 754 does */
1243 if (unlikely(float32_is_quiet_nan(u.f))) {
1244 return 0;
1246 tmp = uint64_to_float32(1ULL << 32, &env->vec_status);
1247 u.f = float32_mul(u.f, tmp, &env->vec_status);
1249 return float32_to_int32(u.f, &env->vec_status);
1252 static inline uint32_t efsctuf(CPUPPCState *env, uint32_t val)
1254 CPU_FloatU u;
1255 float32 tmp;
1257 u.l = val;
1258 /* NaN are not treated the same way IEEE 754 does */
1259 if (unlikely(float32_is_quiet_nan(u.f))) {
1260 return 0;
1262 tmp = uint64_to_float32(1ULL << 32, &env->vec_status);
1263 u.f = float32_mul(u.f, tmp, &env->vec_status);
1265 return float32_to_uint32(u.f, &env->vec_status);
1268 #define HELPER_SPE_SINGLE_CONV(name) \
1269 uint32_t helper_e##name(CPUPPCState *env, uint32_t val) \
1271 return e##name(env, val); \
1273 /* efscfsi */
1274 HELPER_SPE_SINGLE_CONV(fscfsi);
1275 /* efscfui */
1276 HELPER_SPE_SINGLE_CONV(fscfui);
1277 /* efscfuf */
1278 HELPER_SPE_SINGLE_CONV(fscfuf);
1279 /* efscfsf */
1280 HELPER_SPE_SINGLE_CONV(fscfsf);
1281 /* efsctsi */
1282 HELPER_SPE_SINGLE_CONV(fsctsi);
1283 /* efsctui */
1284 HELPER_SPE_SINGLE_CONV(fsctui);
1285 /* efsctsiz */
1286 HELPER_SPE_SINGLE_CONV(fsctsiz);
1287 /* efsctuiz */
1288 HELPER_SPE_SINGLE_CONV(fsctuiz);
1289 /* efsctsf */
1290 HELPER_SPE_SINGLE_CONV(fsctsf);
1291 /* efsctuf */
1292 HELPER_SPE_SINGLE_CONV(fsctuf);
1294 #define HELPER_SPE_VECTOR_CONV(name) \
1295 uint64_t helper_ev##name(CPUPPCState *env, uint64_t val) \
1297 return ((uint64_t)e##name(env, val >> 32) << 32) | \
1298 (uint64_t)e##name(env, val); \
1300 /* evfscfsi */
1301 HELPER_SPE_VECTOR_CONV(fscfsi);
1302 /* evfscfui */
1303 HELPER_SPE_VECTOR_CONV(fscfui);
1304 /* evfscfuf */
1305 HELPER_SPE_VECTOR_CONV(fscfuf);
1306 /* evfscfsf */
1307 HELPER_SPE_VECTOR_CONV(fscfsf);
1308 /* evfsctsi */
1309 HELPER_SPE_VECTOR_CONV(fsctsi);
1310 /* evfsctui */
1311 HELPER_SPE_VECTOR_CONV(fsctui);
1312 /* evfsctsiz */
1313 HELPER_SPE_VECTOR_CONV(fsctsiz);
1314 /* evfsctuiz */
1315 HELPER_SPE_VECTOR_CONV(fsctuiz);
1316 /* evfsctsf */
1317 HELPER_SPE_VECTOR_CONV(fsctsf);
1318 /* evfsctuf */
1319 HELPER_SPE_VECTOR_CONV(fsctuf);
1321 /* Single-precision floating-point arithmetic */
1322 static inline uint32_t efsadd(CPUPPCState *env, uint32_t op1, uint32_t op2)
1324 CPU_FloatU u1, u2;
1326 u1.l = op1;
1327 u2.l = op2;
1328 u1.f = float32_add(u1.f, u2.f, &env->vec_status);
1329 return u1.l;
1332 static inline uint32_t efssub(CPUPPCState *env, uint32_t op1, uint32_t op2)
1334 CPU_FloatU u1, u2;
1336 u1.l = op1;
1337 u2.l = op2;
1338 u1.f = float32_sub(u1.f, u2.f, &env->vec_status);
1339 return u1.l;
1342 static inline uint32_t efsmul(CPUPPCState *env, uint32_t op1, uint32_t op2)
1344 CPU_FloatU u1, u2;
1346 u1.l = op1;
1347 u2.l = op2;
1348 u1.f = float32_mul(u1.f, u2.f, &env->vec_status);
1349 return u1.l;
1352 static inline uint32_t efsdiv(CPUPPCState *env, uint32_t op1, uint32_t op2)
1354 CPU_FloatU u1, u2;
1356 u1.l = op1;
1357 u2.l = op2;
1358 u1.f = float32_div(u1.f, u2.f, &env->vec_status);
1359 return u1.l;
1362 #define HELPER_SPE_SINGLE_ARITH(name) \
1363 uint32_t helper_e##name(CPUPPCState *env, uint32_t op1, uint32_t op2) \
1365 return e##name(env, op1, op2); \
1367 /* efsadd */
1368 HELPER_SPE_SINGLE_ARITH(fsadd);
1369 /* efssub */
1370 HELPER_SPE_SINGLE_ARITH(fssub);
1371 /* efsmul */
1372 HELPER_SPE_SINGLE_ARITH(fsmul);
1373 /* efsdiv */
1374 HELPER_SPE_SINGLE_ARITH(fsdiv);
1376 #define HELPER_SPE_VECTOR_ARITH(name) \
1377 uint64_t helper_ev##name(CPUPPCState *env, uint64_t op1, uint64_t op2) \
1379 return ((uint64_t)e##name(env, op1 >> 32, op2 >> 32) << 32) | \
1380 (uint64_t)e##name(env, op1, op2); \
1382 /* evfsadd */
1383 HELPER_SPE_VECTOR_ARITH(fsadd);
1384 /* evfssub */
1385 HELPER_SPE_VECTOR_ARITH(fssub);
1386 /* evfsmul */
1387 HELPER_SPE_VECTOR_ARITH(fsmul);
1388 /* evfsdiv */
1389 HELPER_SPE_VECTOR_ARITH(fsdiv);
1391 /* Single-precision floating-point comparisons */
1392 static inline uint32_t efscmplt(CPUPPCState *env, uint32_t op1, uint32_t op2)
1394 CPU_FloatU u1, u2;
1396 u1.l = op1;
1397 u2.l = op2;
1398 return float32_lt(u1.f, u2.f, &env->vec_status) ? 4 : 0;
1401 static inline uint32_t efscmpgt(CPUPPCState *env, uint32_t op1, uint32_t op2)
1403 CPU_FloatU u1, u2;
1405 u1.l = op1;
1406 u2.l = op2;
1407 return float32_le(u1.f, u2.f, &env->vec_status) ? 0 : 4;
1410 static inline uint32_t efscmpeq(CPUPPCState *env, uint32_t op1, uint32_t op2)
1412 CPU_FloatU u1, u2;
1414 u1.l = op1;
1415 u2.l = op2;
1416 return float32_eq(u1.f, u2.f, &env->vec_status) ? 4 : 0;
1419 static inline uint32_t efststlt(CPUPPCState *env, uint32_t op1, uint32_t op2)
1421 /* XXX: TODO: ignore special values (NaN, infinites, ...) */
1422 return efscmplt(env, op1, op2);
1425 static inline uint32_t efststgt(CPUPPCState *env, uint32_t op1, uint32_t op2)
1427 /* XXX: TODO: ignore special values (NaN, infinites, ...) */
1428 return efscmpgt(env, op1, op2);
1431 static inline uint32_t efststeq(CPUPPCState *env, uint32_t op1, uint32_t op2)
1433 /* XXX: TODO: ignore special values (NaN, infinites, ...) */
1434 return efscmpeq(env, op1, op2);
1437 #define HELPER_SINGLE_SPE_CMP(name) \
1438 uint32_t helper_e##name(CPUPPCState *env, uint32_t op1, uint32_t op2) \
1440 return e##name(env, op1, op2) << 2; \
1442 /* efststlt */
1443 HELPER_SINGLE_SPE_CMP(fststlt);
1444 /* efststgt */
1445 HELPER_SINGLE_SPE_CMP(fststgt);
1446 /* efststeq */
1447 HELPER_SINGLE_SPE_CMP(fststeq);
1448 /* efscmplt */
1449 HELPER_SINGLE_SPE_CMP(fscmplt);
1450 /* efscmpgt */
1451 HELPER_SINGLE_SPE_CMP(fscmpgt);
1452 /* efscmpeq */
1453 HELPER_SINGLE_SPE_CMP(fscmpeq);
1455 static inline uint32_t evcmp_merge(int t0, int t1)
1457 return (t0 << 3) | (t1 << 2) | ((t0 | t1) << 1) | (t0 & t1);
1460 #define HELPER_VECTOR_SPE_CMP(name) \
1461 uint32_t helper_ev##name(CPUPPCState *env, uint64_t op1, uint64_t op2) \
1463 return evcmp_merge(e##name(env, op1 >> 32, op2 >> 32), \
1464 e##name(env, op1, op2)); \
1466 /* evfststlt */
1467 HELPER_VECTOR_SPE_CMP(fststlt);
1468 /* evfststgt */
1469 HELPER_VECTOR_SPE_CMP(fststgt);
1470 /* evfststeq */
1471 HELPER_VECTOR_SPE_CMP(fststeq);
1472 /* evfscmplt */
1473 HELPER_VECTOR_SPE_CMP(fscmplt);
1474 /* evfscmpgt */
1475 HELPER_VECTOR_SPE_CMP(fscmpgt);
1476 /* evfscmpeq */
1477 HELPER_VECTOR_SPE_CMP(fscmpeq);
1479 /* Double-precision floating-point conversion */
1480 uint64_t helper_efdcfsi(CPUPPCState *env, uint32_t val)
1482 CPU_DoubleU u;
1484 u.d = int32_to_float64(val, &env->vec_status);
1486 return u.ll;
1489 uint64_t helper_efdcfsid(CPUPPCState *env, uint64_t val)
1491 CPU_DoubleU u;
1493 u.d = int64_to_float64(val, &env->vec_status);
1495 return u.ll;
1498 uint64_t helper_efdcfui(CPUPPCState *env, uint32_t val)
1500 CPU_DoubleU u;
1502 u.d = uint32_to_float64(val, &env->vec_status);
1504 return u.ll;
1507 uint64_t helper_efdcfuid(CPUPPCState *env, uint64_t val)
1509 CPU_DoubleU u;
1511 u.d = uint64_to_float64(val, &env->vec_status);
1513 return u.ll;
1516 uint32_t helper_efdctsi(CPUPPCState *env, uint64_t val)
1518 CPU_DoubleU u;
1520 u.ll = val;
1521 /* NaN are not treated the same way IEEE 754 does */
1522 if (unlikely(float64_is_any_nan(u.d))) {
1523 return 0;
1526 return float64_to_int32(u.d, &env->vec_status);
1529 uint32_t helper_efdctui(CPUPPCState *env, uint64_t val)
1531 CPU_DoubleU u;
1533 u.ll = val;
1534 /* NaN are not treated the same way IEEE 754 does */
1535 if (unlikely(float64_is_any_nan(u.d))) {
1536 return 0;
1539 return float64_to_uint32(u.d, &env->vec_status);
1542 uint32_t helper_efdctsiz(CPUPPCState *env, uint64_t val)
1544 CPU_DoubleU u;
1546 u.ll = val;
1547 /* NaN are not treated the same way IEEE 754 does */
1548 if (unlikely(float64_is_any_nan(u.d))) {
1549 return 0;
1552 return float64_to_int32_round_to_zero(u.d, &env->vec_status);
1555 uint64_t helper_efdctsidz(CPUPPCState *env, uint64_t val)
1557 CPU_DoubleU u;
1559 u.ll = val;
1560 /* NaN are not treated the same way IEEE 754 does */
1561 if (unlikely(float64_is_any_nan(u.d))) {
1562 return 0;
1565 return float64_to_int64_round_to_zero(u.d, &env->vec_status);
1568 uint32_t helper_efdctuiz(CPUPPCState *env, uint64_t val)
1570 CPU_DoubleU u;
1572 u.ll = val;
1573 /* NaN are not treated the same way IEEE 754 does */
1574 if (unlikely(float64_is_any_nan(u.d))) {
1575 return 0;
1578 return float64_to_uint32_round_to_zero(u.d, &env->vec_status);
1581 uint64_t helper_efdctuidz(CPUPPCState *env, uint64_t val)
1583 CPU_DoubleU u;
1585 u.ll = val;
1586 /* NaN are not treated the same way IEEE 754 does */
1587 if (unlikely(float64_is_any_nan(u.d))) {
1588 return 0;
1591 return float64_to_uint64_round_to_zero(u.d, &env->vec_status);
1594 uint64_t helper_efdcfsf(CPUPPCState *env, uint32_t val)
1596 CPU_DoubleU u;
1597 float64 tmp;
1599 u.d = int32_to_float64(val, &env->vec_status);
1600 tmp = int64_to_float64(1ULL << 32, &env->vec_status);
1601 u.d = float64_div(u.d, tmp, &env->vec_status);
1603 return u.ll;
1606 uint64_t helper_efdcfuf(CPUPPCState *env, uint32_t val)
1608 CPU_DoubleU u;
1609 float64 tmp;
1611 u.d = uint32_to_float64(val, &env->vec_status);
1612 tmp = int64_to_float64(1ULL << 32, &env->vec_status);
1613 u.d = float64_div(u.d, tmp, &env->vec_status);
1615 return u.ll;
1618 uint32_t helper_efdctsf(CPUPPCState *env, uint64_t val)
1620 CPU_DoubleU u;
1621 float64 tmp;
1623 u.ll = val;
1624 /* NaN are not treated the same way IEEE 754 does */
1625 if (unlikely(float64_is_any_nan(u.d))) {
1626 return 0;
1628 tmp = uint64_to_float64(1ULL << 32, &env->vec_status);
1629 u.d = float64_mul(u.d, tmp, &env->vec_status);
1631 return float64_to_int32(u.d, &env->vec_status);
1634 uint32_t helper_efdctuf(CPUPPCState *env, uint64_t val)
1636 CPU_DoubleU u;
1637 float64 tmp;
1639 u.ll = val;
1640 /* NaN are not treated the same way IEEE 754 does */
1641 if (unlikely(float64_is_any_nan(u.d))) {
1642 return 0;
1644 tmp = uint64_to_float64(1ULL << 32, &env->vec_status);
1645 u.d = float64_mul(u.d, tmp, &env->vec_status);
1647 return float64_to_uint32(u.d, &env->vec_status);
1650 uint32_t helper_efscfd(CPUPPCState *env, uint64_t val)
1652 CPU_DoubleU u1;
1653 CPU_FloatU u2;
1655 u1.ll = val;
1656 u2.f = float64_to_float32(u1.d, &env->vec_status);
1658 return u2.l;
1661 uint64_t helper_efdcfs(CPUPPCState *env, uint32_t val)
1663 CPU_DoubleU u2;
1664 CPU_FloatU u1;
1666 u1.l = val;
1667 u2.d = float32_to_float64(u1.f, &env->vec_status);
1669 return u2.ll;
1672 /* Double precision fixed-point arithmetic */
1673 uint64_t helper_efdadd(CPUPPCState *env, uint64_t op1, uint64_t op2)
1675 CPU_DoubleU u1, u2;
1677 u1.ll = op1;
1678 u2.ll = op2;
1679 u1.d = float64_add(u1.d, u2.d, &env->vec_status);
1680 return u1.ll;
1683 uint64_t helper_efdsub(CPUPPCState *env, uint64_t op1, uint64_t op2)
1685 CPU_DoubleU u1, u2;
1687 u1.ll = op1;
1688 u2.ll = op2;
1689 u1.d = float64_sub(u1.d, u2.d, &env->vec_status);
1690 return u1.ll;
1693 uint64_t helper_efdmul(CPUPPCState *env, uint64_t op1, uint64_t op2)
1695 CPU_DoubleU u1, u2;
1697 u1.ll = op1;
1698 u2.ll = op2;
1699 u1.d = float64_mul(u1.d, u2.d, &env->vec_status);
1700 return u1.ll;
1703 uint64_t helper_efddiv(CPUPPCState *env, uint64_t op1, uint64_t op2)
1705 CPU_DoubleU u1, u2;
1707 u1.ll = op1;
1708 u2.ll = op2;
1709 u1.d = float64_div(u1.d, u2.d, &env->vec_status);
1710 return u1.ll;
1713 /* Double precision floating point helpers */
1714 uint32_t helper_efdtstlt(CPUPPCState *env, uint64_t op1, uint64_t op2)
1716 CPU_DoubleU u1, u2;
1718 u1.ll = op1;
1719 u2.ll = op2;
1720 return float64_lt(u1.d, u2.d, &env->vec_status) ? 4 : 0;
1723 uint32_t helper_efdtstgt(CPUPPCState *env, uint64_t op1, uint64_t op2)
1725 CPU_DoubleU u1, u2;
1727 u1.ll = op1;
1728 u2.ll = op2;
1729 return float64_le(u1.d, u2.d, &env->vec_status) ? 0 : 4;
1732 uint32_t helper_efdtsteq(CPUPPCState *env, uint64_t op1, uint64_t op2)
1734 CPU_DoubleU u1, u2;
1736 u1.ll = op1;
1737 u2.ll = op2;
1738 return float64_eq_quiet(u1.d, u2.d, &env->vec_status) ? 4 : 0;
1741 uint32_t helper_efdcmplt(CPUPPCState *env, uint64_t op1, uint64_t op2)
1743 /* XXX: TODO: test special values (NaN, infinites, ...) */
1744 return helper_efdtstlt(env, op1, op2);
1747 uint32_t helper_efdcmpgt(CPUPPCState *env, uint64_t op1, uint64_t op2)
1749 /* XXX: TODO: test special values (NaN, infinites, ...) */
1750 return helper_efdtstgt(env, op1, op2);
1753 uint32_t helper_efdcmpeq(CPUPPCState *env, uint64_t op1, uint64_t op2)
1755 /* XXX: TODO: test special values (NaN, infinites, ...) */
1756 return helper_efdtsteq(env, op1, op2);
1759 #define DECODE_SPLIT(opcode, shift1, nb1, shift2, nb2) \
1760 (((((opcode) >> (shift1)) & ((1 << (nb1)) - 1)) << nb2) | \
1761 (((opcode) >> (shift2)) & ((1 << (nb2)) - 1)))
1763 #define xT(opcode) DECODE_SPLIT(opcode, 0, 1, 21, 5)
1764 #define xA(opcode) DECODE_SPLIT(opcode, 2, 1, 16, 5)
1765 #define xB(opcode) DECODE_SPLIT(opcode, 1, 1, 11, 5)
1766 #define xC(opcode) DECODE_SPLIT(opcode, 3, 1, 6, 5)
1767 #define BF(opcode) (((opcode) >> (31-8)) & 7)
1769 typedef union _ppc_vsr_t {
1770 uint64_t u64[2];
1771 uint32_t u32[4];
1772 float32 f32[4];
1773 float64 f64[2];
1774 } ppc_vsr_t;
1776 #if defined(HOST_WORDS_BIGENDIAN)
1777 #define VsrW(i) u32[i]
1778 #define VsrD(i) u64[i]
1779 #else
1780 #define VsrW(i) u32[3-(i)]
1781 #define VsrD(i) u64[1-(i)]
1782 #endif
1784 static void getVSR(int n, ppc_vsr_t *vsr, CPUPPCState *env)
1786 if (n < 32) {
1787 vsr->VsrD(0) = env->fpr[n];
1788 vsr->VsrD(1) = env->vsr[n];
1789 } else {
1790 vsr->u64[0] = env->avr[n-32].u64[0];
1791 vsr->u64[1] = env->avr[n-32].u64[1];
1795 static void putVSR(int n, ppc_vsr_t *vsr, CPUPPCState *env)
1797 if (n < 32) {
1798 env->fpr[n] = vsr->VsrD(0);
1799 env->vsr[n] = vsr->VsrD(1);
1800 } else {
1801 env->avr[n-32].u64[0] = vsr->u64[0];
1802 env->avr[n-32].u64[1] = vsr->u64[1];
1806 #define float64_to_float64(x, env) x
1809 /* VSX_ADD_SUB - VSX floating point add/subract
1810 * name - instruction mnemonic
1811 * op - operation (add or sub)
1812 * nels - number of elements (1, 2 or 4)
1813 * tp - type (float32 or float64)
1814 * fld - vsr_t field (VsrD(*) or VsrW(*))
1815 * sfprf - set FPRF
1817 #define VSX_ADD_SUB(name, op, nels, tp, fld, sfprf, r2sp) \
1818 void helper_##name(CPUPPCState *env, uint32_t opcode) \
1820 ppc_vsr_t xt, xa, xb; \
1821 int i; \
1823 getVSR(xA(opcode), &xa, env); \
1824 getVSR(xB(opcode), &xb, env); \
1825 getVSR(xT(opcode), &xt, env); \
1826 helper_reset_fpstatus(env); \
1828 for (i = 0; i < nels; i++) { \
1829 float_status tstat = env->fp_status; \
1830 set_float_exception_flags(0, &tstat); \
1831 xt.fld = tp##_##op(xa.fld, xb.fld, &tstat); \
1832 env->fp_status.float_exception_flags |= tstat.float_exception_flags; \
1834 if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \
1835 if (tp##_is_infinity(xa.fld) && tp##_is_infinity(xb.fld)) { \
1836 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, sfprf); \
1837 } else if (tp##_is_signaling_nan(xa.fld) || \
1838 tp##_is_signaling_nan(xb.fld)) { \
1839 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \
1843 if (r2sp) { \
1844 xt.fld = helper_frsp(env, xt.fld); \
1847 if (sfprf) { \
1848 helper_compute_fprf(env, xt.fld, sfprf); \
1851 putVSR(xT(opcode), &xt, env); \
1852 helper_float_check_status(env); \
1855 VSX_ADD_SUB(xsadddp, add, 1, float64, VsrD(0), 1, 0)
1856 VSX_ADD_SUB(xsaddsp, add, 1, float64, VsrD(0), 1, 1)
1857 VSX_ADD_SUB(xvadddp, add, 2, float64, VsrD(i), 0, 0)
1858 VSX_ADD_SUB(xvaddsp, add, 4, float32, VsrW(i), 0, 0)
1859 VSX_ADD_SUB(xssubdp, sub, 1, float64, VsrD(0), 1, 0)
1860 VSX_ADD_SUB(xssubsp, sub, 1, float64, VsrD(0), 1, 1)
1861 VSX_ADD_SUB(xvsubdp, sub, 2, float64, VsrD(i), 0, 0)
1862 VSX_ADD_SUB(xvsubsp, sub, 4, float32, VsrW(i), 0, 0)
1864 /* VSX_MUL - VSX floating point multiply
1865 * op - instruction mnemonic
1866 * nels - number of elements (1, 2 or 4)
1867 * tp - type (float32 or float64)
1868 * fld - vsr_t field (VsrD(*) or VsrW(*))
1869 * sfprf - set FPRF
1871 #define VSX_MUL(op, nels, tp, fld, sfprf, r2sp) \
1872 void helper_##op(CPUPPCState *env, uint32_t opcode) \
1874 ppc_vsr_t xt, xa, xb; \
1875 int i; \
1877 getVSR(xA(opcode), &xa, env); \
1878 getVSR(xB(opcode), &xb, env); \
1879 getVSR(xT(opcode), &xt, env); \
1880 helper_reset_fpstatus(env); \
1882 for (i = 0; i < nels; i++) { \
1883 float_status tstat = env->fp_status; \
1884 set_float_exception_flags(0, &tstat); \
1885 xt.fld = tp##_mul(xa.fld, xb.fld, &tstat); \
1886 env->fp_status.float_exception_flags |= tstat.float_exception_flags; \
1888 if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \
1889 if ((tp##_is_infinity(xa.fld) && tp##_is_zero(xb.fld)) || \
1890 (tp##_is_infinity(xb.fld) && tp##_is_zero(xa.fld))) { \
1891 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, sfprf); \
1892 } else if (tp##_is_signaling_nan(xa.fld) || \
1893 tp##_is_signaling_nan(xb.fld)) { \
1894 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \
1898 if (r2sp) { \
1899 xt.fld = helper_frsp(env, xt.fld); \
1902 if (sfprf) { \
1903 helper_compute_fprf(env, xt.fld, sfprf); \
1907 putVSR(xT(opcode), &xt, env); \
1908 helper_float_check_status(env); \
1911 VSX_MUL(xsmuldp, 1, float64, VsrD(0), 1, 0)
1912 VSX_MUL(xsmulsp, 1, float64, VsrD(0), 1, 1)
1913 VSX_MUL(xvmuldp, 2, float64, VsrD(i), 0, 0)
1914 VSX_MUL(xvmulsp, 4, float32, VsrW(i), 0, 0)
1916 /* VSX_DIV - VSX floating point divide
1917 * op - instruction mnemonic
1918 * nels - number of elements (1, 2 or 4)
1919 * tp - type (float32 or float64)
1920 * fld - vsr_t field (VsrD(*) or VsrW(*))
1921 * sfprf - set FPRF
1923 #define VSX_DIV(op, nels, tp, fld, sfprf, r2sp) \
1924 void helper_##op(CPUPPCState *env, uint32_t opcode) \
1926 ppc_vsr_t xt, xa, xb; \
1927 int i; \
1929 getVSR(xA(opcode), &xa, env); \
1930 getVSR(xB(opcode), &xb, env); \
1931 getVSR(xT(opcode), &xt, env); \
1932 helper_reset_fpstatus(env); \
1934 for (i = 0; i < nels; i++) { \
1935 float_status tstat = env->fp_status; \
1936 set_float_exception_flags(0, &tstat); \
1937 xt.fld = tp##_div(xa.fld, xb.fld, &tstat); \
1938 env->fp_status.float_exception_flags |= tstat.float_exception_flags; \
1940 if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \
1941 if (tp##_is_infinity(xa.fld) && tp##_is_infinity(xb.fld)) { \
1942 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIDI, sfprf); \
1943 } else if (tp##_is_zero(xa.fld) && \
1944 tp##_is_zero(xb.fld)) { \
1945 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXZDZ, sfprf); \
1946 } else if (tp##_is_signaling_nan(xa.fld) || \
1947 tp##_is_signaling_nan(xb.fld)) { \
1948 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \
1952 if (r2sp) { \
1953 xt.fld = helper_frsp(env, xt.fld); \
1956 if (sfprf) { \
1957 helper_compute_fprf(env, xt.fld, sfprf); \
1961 putVSR(xT(opcode), &xt, env); \
1962 helper_float_check_status(env); \
1965 VSX_DIV(xsdivdp, 1, float64, VsrD(0), 1, 0)
1966 VSX_DIV(xsdivsp, 1, float64, VsrD(0), 1, 1)
1967 VSX_DIV(xvdivdp, 2, float64, VsrD(i), 0, 0)
1968 VSX_DIV(xvdivsp, 4, float32, VsrW(i), 0, 0)
1970 /* VSX_RE - VSX floating point reciprocal estimate
1971 * op - instruction mnemonic
1972 * nels - number of elements (1, 2 or 4)
1973 * tp - type (float32 or float64)
1974 * fld - vsr_t field (VsrD(*) or VsrW(*))
1975 * sfprf - set FPRF
1977 #define VSX_RE(op, nels, tp, fld, sfprf, r2sp) \
1978 void helper_##op(CPUPPCState *env, uint32_t opcode) \
1980 ppc_vsr_t xt, xb; \
1981 int i; \
1983 getVSR(xB(opcode), &xb, env); \
1984 getVSR(xT(opcode), &xt, env); \
1985 helper_reset_fpstatus(env); \
1987 for (i = 0; i < nels; i++) { \
1988 if (unlikely(tp##_is_signaling_nan(xb.fld))) { \
1989 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \
1991 xt.fld = tp##_div(tp##_one, xb.fld, &env->fp_status); \
1993 if (r2sp) { \
1994 xt.fld = helper_frsp(env, xt.fld); \
1997 if (sfprf) { \
1998 helper_compute_fprf(env, xt.fld, sfprf); \
2002 putVSR(xT(opcode), &xt, env); \
2003 helper_float_check_status(env); \
2006 VSX_RE(xsredp, 1, float64, VsrD(0), 1, 0)
2007 VSX_RE(xsresp, 1, float64, VsrD(0), 1, 1)
2008 VSX_RE(xvredp, 2, float64, VsrD(i), 0, 0)
2009 VSX_RE(xvresp, 4, float32, VsrW(i), 0, 0)
2011 /* VSX_SQRT - VSX floating point square root
2012 * op - instruction mnemonic
2013 * nels - number of elements (1, 2 or 4)
2014 * tp - type (float32 or float64)
2015 * fld - vsr_t field (VsrD(*) or VsrW(*))
2016 * sfprf - set FPRF
2018 #define VSX_SQRT(op, nels, tp, fld, sfprf, r2sp) \
2019 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2021 ppc_vsr_t xt, xb; \
2022 int i; \
2024 getVSR(xB(opcode), &xb, env); \
2025 getVSR(xT(opcode), &xt, env); \
2026 helper_reset_fpstatus(env); \
2028 for (i = 0; i < nels; i++) { \
2029 float_status tstat = env->fp_status; \
2030 set_float_exception_flags(0, &tstat); \
2031 xt.fld = tp##_sqrt(xb.fld, &tstat); \
2032 env->fp_status.float_exception_flags |= tstat.float_exception_flags; \
2034 if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \
2035 if (tp##_is_neg(xb.fld) && !tp##_is_zero(xb.fld)) { \
2036 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, sfprf); \
2037 } else if (tp##_is_signaling_nan(xb.fld)) { \
2038 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \
2042 if (r2sp) { \
2043 xt.fld = helper_frsp(env, xt.fld); \
2046 if (sfprf) { \
2047 helper_compute_fprf(env, xt.fld, sfprf); \
2051 putVSR(xT(opcode), &xt, env); \
2052 helper_float_check_status(env); \
2055 VSX_SQRT(xssqrtdp, 1, float64, VsrD(0), 1, 0)
2056 VSX_SQRT(xssqrtsp, 1, float64, VsrD(0), 1, 1)
2057 VSX_SQRT(xvsqrtdp, 2, float64, VsrD(i), 0, 0)
2058 VSX_SQRT(xvsqrtsp, 4, float32, VsrW(i), 0, 0)
2060 /* VSX_RSQRTE - VSX floating point reciprocal square root estimate
2061 * op - instruction mnemonic
2062 * nels - number of elements (1, 2 or 4)
2063 * tp - type (float32 or float64)
2064 * fld - vsr_t field (VsrD(*) or VsrW(*))
2065 * sfprf - set FPRF
2067 #define VSX_RSQRTE(op, nels, tp, fld, sfprf, r2sp) \
2068 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2070 ppc_vsr_t xt, xb; \
2071 int i; \
2073 getVSR(xB(opcode), &xb, env); \
2074 getVSR(xT(opcode), &xt, env); \
2075 helper_reset_fpstatus(env); \
2077 for (i = 0; i < nels; i++) { \
2078 float_status tstat = env->fp_status; \
2079 set_float_exception_flags(0, &tstat); \
2080 xt.fld = tp##_sqrt(xb.fld, &tstat); \
2081 xt.fld = tp##_div(tp##_one, xt.fld, &tstat); \
2082 env->fp_status.float_exception_flags |= tstat.float_exception_flags; \
2084 if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \
2085 if (tp##_is_neg(xb.fld) && !tp##_is_zero(xb.fld)) { \
2086 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, sfprf); \
2087 } else if (tp##_is_signaling_nan(xb.fld)) { \
2088 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \
2092 if (r2sp) { \
2093 xt.fld = helper_frsp(env, xt.fld); \
2096 if (sfprf) { \
2097 helper_compute_fprf(env, xt.fld, sfprf); \
2101 putVSR(xT(opcode), &xt, env); \
2102 helper_float_check_status(env); \
2105 VSX_RSQRTE(xsrsqrtedp, 1, float64, VsrD(0), 1, 0)
2106 VSX_RSQRTE(xsrsqrtesp, 1, float64, VsrD(0), 1, 1)
2107 VSX_RSQRTE(xvrsqrtedp, 2, float64, VsrD(i), 0, 0)
2108 VSX_RSQRTE(xvrsqrtesp, 4, float32, VsrW(i), 0, 0)
2110 /* VSX_TDIV - VSX floating point test for divide
2111 * op - instruction mnemonic
2112 * nels - number of elements (1, 2 or 4)
2113 * tp - type (float32 or float64)
2114 * fld - vsr_t field (VsrD(*) or VsrW(*))
2115 * emin - minimum unbiased exponent
2116 * emax - maximum unbiased exponent
2117 * nbits - number of fraction bits
2119 #define VSX_TDIV(op, nels, tp, fld, emin, emax, nbits) \
2120 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2122 ppc_vsr_t xa, xb; \
2123 int i; \
2124 int fe_flag = 0; \
2125 int fg_flag = 0; \
2127 getVSR(xA(opcode), &xa, env); \
2128 getVSR(xB(opcode), &xb, env); \
2130 for (i = 0; i < nels; i++) { \
2131 if (unlikely(tp##_is_infinity(xa.fld) || \
2132 tp##_is_infinity(xb.fld) || \
2133 tp##_is_zero(xb.fld))) { \
2134 fe_flag = 1; \
2135 fg_flag = 1; \
2136 } else { \
2137 int e_a = ppc_##tp##_get_unbiased_exp(xa.fld); \
2138 int e_b = ppc_##tp##_get_unbiased_exp(xb.fld); \
2140 if (unlikely(tp##_is_any_nan(xa.fld) || \
2141 tp##_is_any_nan(xb.fld))) { \
2142 fe_flag = 1; \
2143 } else if ((e_b <= emin) || (e_b >= (emax-2))) { \
2144 fe_flag = 1; \
2145 } else if (!tp##_is_zero(xa.fld) && \
2146 (((e_a - e_b) >= emax) || \
2147 ((e_a - e_b) <= (emin+1)) || \
2148 (e_a <= (emin+nbits)))) { \
2149 fe_flag = 1; \
2152 if (unlikely(tp##_is_zero_or_denormal(xb.fld))) { \
2153 /* XB is not zero because of the above check and */ \
2154 /* so must be denormalized. */ \
2155 fg_flag = 1; \
2160 env->crf[BF(opcode)] = 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); \
2163 VSX_TDIV(xstdivdp, 1, float64, VsrD(0), -1022, 1023, 52)
2164 VSX_TDIV(xvtdivdp, 2, float64, VsrD(i), -1022, 1023, 52)
2165 VSX_TDIV(xvtdivsp, 4, float32, VsrW(i), -126, 127, 23)
2167 /* VSX_TSQRT - VSX floating point test for square root
2168 * op - instruction mnemonic
2169 * nels - number of elements (1, 2 or 4)
2170 * tp - type (float32 or float64)
2171 * fld - vsr_t field (VsrD(*) or VsrW(*))
2172 * emin - minimum unbiased exponent
2173 * emax - maximum unbiased exponent
2174 * nbits - number of fraction bits
2176 #define VSX_TSQRT(op, nels, tp, fld, emin, nbits) \
2177 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2179 ppc_vsr_t xa, xb; \
2180 int i; \
2181 int fe_flag = 0; \
2182 int fg_flag = 0; \
2184 getVSR(xA(opcode), &xa, env); \
2185 getVSR(xB(opcode), &xb, env); \
2187 for (i = 0; i < nels; i++) { \
2188 if (unlikely(tp##_is_infinity(xb.fld) || \
2189 tp##_is_zero(xb.fld))) { \
2190 fe_flag = 1; \
2191 fg_flag = 1; \
2192 } else { \
2193 int e_b = ppc_##tp##_get_unbiased_exp(xb.fld); \
2195 if (unlikely(tp##_is_any_nan(xb.fld))) { \
2196 fe_flag = 1; \
2197 } else if (unlikely(tp##_is_zero(xb.fld))) { \
2198 fe_flag = 1; \
2199 } else if (unlikely(tp##_is_neg(xb.fld))) { \
2200 fe_flag = 1; \
2201 } else if (!tp##_is_zero(xb.fld) && \
2202 (e_b <= (emin+nbits))) { \
2203 fe_flag = 1; \
2206 if (unlikely(tp##_is_zero_or_denormal(xb.fld))) { \
2207 /* XB is not zero because of the above check and */ \
2208 /* therefore must be denormalized. */ \
2209 fg_flag = 1; \
2214 env->crf[BF(opcode)] = 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); \
2217 VSX_TSQRT(xstsqrtdp, 1, float64, VsrD(0), -1022, 52)
2218 VSX_TSQRT(xvtsqrtdp, 2, float64, VsrD(i), -1022, 52)
2219 VSX_TSQRT(xvtsqrtsp, 4, float32, VsrW(i), -126, 23)
2221 /* VSX_MADD - VSX floating point muliply/add variations
2222 * op - instruction mnemonic
2223 * nels - number of elements (1, 2 or 4)
2224 * tp - type (float32 or float64)
2225 * fld - vsr_t field (VsrD(*) or VsrW(*))
2226 * maddflgs - flags for the float*muladd routine that control the
2227 * various forms (madd, msub, nmadd, nmsub)
2228 * afrm - A form (1=A, 0=M)
2229 * sfprf - set FPRF
2231 #define VSX_MADD(op, nels, tp, fld, maddflgs, afrm, sfprf, r2sp) \
2232 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2234 ppc_vsr_t xt_in, xa, xb, xt_out; \
2235 ppc_vsr_t *b, *c; \
2236 int i; \
2238 if (afrm) { /* AxB + T */ \
2239 b = &xb; \
2240 c = &xt_in; \
2241 } else { /* AxT + B */ \
2242 b = &xt_in; \
2243 c = &xb; \
2246 getVSR(xA(opcode), &xa, env); \
2247 getVSR(xB(opcode), &xb, env); \
2248 getVSR(xT(opcode), &xt_in, env); \
2250 xt_out = xt_in; \
2252 helper_reset_fpstatus(env); \
2254 for (i = 0; i < nels; i++) { \
2255 float_status tstat = env->fp_status; \
2256 set_float_exception_flags(0, &tstat); \
2257 if (r2sp && (tstat.float_rounding_mode == float_round_nearest_even)) {\
2258 /* Avoid double rounding errors by rounding the intermediate */ \
2259 /* result to odd. */ \
2260 set_float_rounding_mode(float_round_to_zero, &tstat); \
2261 xt_out.fld = tp##_muladd(xa.fld, b->fld, c->fld, \
2262 maddflgs, &tstat); \
2263 xt_out.fld |= (get_float_exception_flags(&tstat) & \
2264 float_flag_inexact) != 0; \
2265 } else { \
2266 xt_out.fld = tp##_muladd(xa.fld, b->fld, c->fld, \
2267 maddflgs, &tstat); \
2269 env->fp_status.float_exception_flags |= tstat.float_exception_flags; \
2271 if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \
2272 if (tp##_is_signaling_nan(xa.fld) || \
2273 tp##_is_signaling_nan(b->fld) || \
2274 tp##_is_signaling_nan(c->fld)) { \
2275 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \
2276 tstat.float_exception_flags &= ~float_flag_invalid; \
2278 if ((tp##_is_infinity(xa.fld) && tp##_is_zero(b->fld)) || \
2279 (tp##_is_zero(xa.fld) && tp##_is_infinity(b->fld))) { \
2280 xt_out.fld = float64_to_##tp(fload_invalid_op_excp(env, \
2281 POWERPC_EXCP_FP_VXIMZ, sfprf), &env->fp_status); \
2282 tstat.float_exception_flags &= ~float_flag_invalid; \
2284 if ((tstat.float_exception_flags & float_flag_invalid) && \
2285 ((tp##_is_infinity(xa.fld) || \
2286 tp##_is_infinity(b->fld)) && \
2287 tp##_is_infinity(c->fld))) { \
2288 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, sfprf); \
2292 if (r2sp) { \
2293 xt_out.fld = helper_frsp(env, xt_out.fld); \
2296 if (sfprf) { \
2297 helper_compute_fprf(env, xt_out.fld, sfprf); \
2300 putVSR(xT(opcode), &xt_out, env); \
2301 helper_float_check_status(env); \
2304 #define MADD_FLGS 0
2305 #define MSUB_FLGS float_muladd_negate_c
2306 #define NMADD_FLGS float_muladd_negate_result
2307 #define NMSUB_FLGS (float_muladd_negate_c | float_muladd_negate_result)
2309 VSX_MADD(xsmaddadp, 1, float64, VsrD(0), MADD_FLGS, 1, 1, 0)
2310 VSX_MADD(xsmaddmdp, 1, float64, VsrD(0), MADD_FLGS, 0, 1, 0)
2311 VSX_MADD(xsmsubadp, 1, float64, VsrD(0), MSUB_FLGS, 1, 1, 0)
2312 VSX_MADD(xsmsubmdp, 1, float64, VsrD(0), MSUB_FLGS, 0, 1, 0)
2313 VSX_MADD(xsnmaddadp, 1, float64, VsrD(0), NMADD_FLGS, 1, 1, 0)
2314 VSX_MADD(xsnmaddmdp, 1, float64, VsrD(0), NMADD_FLGS, 0, 1, 0)
2315 VSX_MADD(xsnmsubadp, 1, float64, VsrD(0), NMSUB_FLGS, 1, 1, 0)
2316 VSX_MADD(xsnmsubmdp, 1, float64, VsrD(0), NMSUB_FLGS, 0, 1, 0)
2318 VSX_MADD(xsmaddasp, 1, float64, VsrD(0), MADD_FLGS, 1, 1, 1)
2319 VSX_MADD(xsmaddmsp, 1, float64, VsrD(0), MADD_FLGS, 0, 1, 1)
2320 VSX_MADD(xsmsubasp, 1, float64, VsrD(0), MSUB_FLGS, 1, 1, 1)
2321 VSX_MADD(xsmsubmsp, 1, float64, VsrD(0), MSUB_FLGS, 0, 1, 1)
2322 VSX_MADD(xsnmaddasp, 1, float64, VsrD(0), NMADD_FLGS, 1, 1, 1)
2323 VSX_MADD(xsnmaddmsp, 1, float64, VsrD(0), NMADD_FLGS, 0, 1, 1)
2324 VSX_MADD(xsnmsubasp, 1, float64, VsrD(0), NMSUB_FLGS, 1, 1, 1)
2325 VSX_MADD(xsnmsubmsp, 1, float64, VsrD(0), NMSUB_FLGS, 0, 1, 1)
2327 VSX_MADD(xvmaddadp, 2, float64, VsrD(i), MADD_FLGS, 1, 0, 0)
2328 VSX_MADD(xvmaddmdp, 2, float64, VsrD(i), MADD_FLGS, 0, 0, 0)
2329 VSX_MADD(xvmsubadp, 2, float64, VsrD(i), MSUB_FLGS, 1, 0, 0)
2330 VSX_MADD(xvmsubmdp, 2, float64, VsrD(i), MSUB_FLGS, 0, 0, 0)
2331 VSX_MADD(xvnmaddadp, 2, float64, VsrD(i), NMADD_FLGS, 1, 0, 0)
2332 VSX_MADD(xvnmaddmdp, 2, float64, VsrD(i), NMADD_FLGS, 0, 0, 0)
2333 VSX_MADD(xvnmsubadp, 2, float64, VsrD(i), NMSUB_FLGS, 1, 0, 0)
2334 VSX_MADD(xvnmsubmdp, 2, float64, VsrD(i), NMSUB_FLGS, 0, 0, 0)
2336 VSX_MADD(xvmaddasp, 4, float32, VsrW(i), MADD_FLGS, 1, 0, 0)
2337 VSX_MADD(xvmaddmsp, 4, float32, VsrW(i), MADD_FLGS, 0, 0, 0)
2338 VSX_MADD(xvmsubasp, 4, float32, VsrW(i), MSUB_FLGS, 1, 0, 0)
2339 VSX_MADD(xvmsubmsp, 4, float32, VsrW(i), MSUB_FLGS, 0, 0, 0)
2340 VSX_MADD(xvnmaddasp, 4, float32, VsrW(i), NMADD_FLGS, 1, 0, 0)
2341 VSX_MADD(xvnmaddmsp, 4, float32, VsrW(i), NMADD_FLGS, 0, 0, 0)
2342 VSX_MADD(xvnmsubasp, 4, float32, VsrW(i), NMSUB_FLGS, 1, 0, 0)
2343 VSX_MADD(xvnmsubmsp, 4, float32, VsrW(i), NMSUB_FLGS, 0, 0, 0)
2345 #define VSX_SCALAR_CMP(op, ordered) \
2346 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2348 ppc_vsr_t xa, xb; \
2349 uint32_t cc = 0; \
2351 getVSR(xA(opcode), &xa, env); \
2352 getVSR(xB(opcode), &xb, env); \
2354 if (unlikely(float64_is_any_nan(xa.VsrD(0)) || \
2355 float64_is_any_nan(xb.VsrD(0)))) { \
2356 if (float64_is_signaling_nan(xa.VsrD(0)) || \
2357 float64_is_signaling_nan(xb.VsrD(0))) { \
2358 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \
2360 if (ordered) { \
2361 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXVC, 0); \
2363 cc = 1; \
2364 } else { \
2365 if (float64_lt(xa.VsrD(0), xb.VsrD(0), &env->fp_status)) { \
2366 cc = 8; \
2367 } else if (!float64_le(xa.VsrD(0), xb.VsrD(0), \
2368 &env->fp_status)) { \
2369 cc = 4; \
2370 } else { \
2371 cc = 2; \
2375 env->fpscr &= ~(0x0F << FPSCR_FPRF); \
2376 env->fpscr |= cc << FPSCR_FPRF; \
2377 env->crf[BF(opcode)] = cc; \
2379 helper_float_check_status(env); \
2382 VSX_SCALAR_CMP(xscmpodp, 1)
2383 VSX_SCALAR_CMP(xscmpudp, 0)
2385 #define float64_snan_to_qnan(x) ((x) | 0x0008000000000000ULL)
2386 #define float32_snan_to_qnan(x) ((x) | 0x00400000)
2388 /* VSX_MAX_MIN - VSX floating point maximum/minimum
2389 * name - instruction mnemonic
2390 * op - operation (max or min)
2391 * nels - number of elements (1, 2 or 4)
2392 * tp - type (float32 or float64)
2393 * fld - vsr_t field (VsrD(*) or VsrW(*))
2395 #define VSX_MAX_MIN(name, op, nels, tp, fld) \
2396 void helper_##name(CPUPPCState *env, uint32_t opcode) \
2398 ppc_vsr_t xt, xa, xb; \
2399 int i; \
2401 getVSR(xA(opcode), &xa, env); \
2402 getVSR(xB(opcode), &xb, env); \
2403 getVSR(xT(opcode), &xt, env); \
2405 for (i = 0; i < nels; i++) { \
2406 xt.fld = tp##_##op(xa.fld, xb.fld, &env->fp_status); \
2407 if (unlikely(tp##_is_signaling_nan(xa.fld) || \
2408 tp##_is_signaling_nan(xb.fld))) { \
2409 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \
2413 putVSR(xT(opcode), &xt, env); \
2414 helper_float_check_status(env); \
2417 VSX_MAX_MIN(xsmaxdp, maxnum, 1, float64, VsrD(0))
2418 VSX_MAX_MIN(xvmaxdp, maxnum, 2, float64, VsrD(i))
2419 VSX_MAX_MIN(xvmaxsp, maxnum, 4, float32, VsrW(i))
2420 VSX_MAX_MIN(xsmindp, minnum, 1, float64, VsrD(0))
2421 VSX_MAX_MIN(xvmindp, minnum, 2, float64, VsrD(i))
2422 VSX_MAX_MIN(xvminsp, minnum, 4, float32, VsrW(i))
2424 /* VSX_CMP - VSX floating point compare
2425 * op - instruction mnemonic
2426 * nels - number of elements (1, 2 or 4)
2427 * tp - type (float32 or float64)
2428 * fld - vsr_t field (VsrD(*) or VsrW(*))
2429 * cmp - comparison operation
2430 * svxvc - set VXVC bit
2432 #define VSX_CMP(op, nels, tp, fld, cmp, svxvc) \
2433 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2435 ppc_vsr_t xt, xa, xb; \
2436 int i; \
2437 int all_true = 1; \
2438 int all_false = 1; \
2440 getVSR(xA(opcode), &xa, env); \
2441 getVSR(xB(opcode), &xb, env); \
2442 getVSR(xT(opcode), &xt, env); \
2444 for (i = 0; i < nels; i++) { \
2445 if (unlikely(tp##_is_any_nan(xa.fld) || \
2446 tp##_is_any_nan(xb.fld))) { \
2447 if (tp##_is_signaling_nan(xa.fld) || \
2448 tp##_is_signaling_nan(xb.fld)) { \
2449 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \
2451 if (svxvc) { \
2452 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXVC, 0); \
2454 xt.fld = 0; \
2455 all_true = 0; \
2456 } else { \
2457 if (tp##_##cmp(xb.fld, xa.fld, &env->fp_status) == 1) { \
2458 xt.fld = -1; \
2459 all_false = 0; \
2460 } else { \
2461 xt.fld = 0; \
2462 all_true = 0; \
2467 putVSR(xT(opcode), &xt, env); \
2468 if ((opcode >> (31-21)) & 1) { \
2469 env->crf[6] = (all_true ? 0x8 : 0) | (all_false ? 0x2 : 0); \
2471 helper_float_check_status(env); \
2474 VSX_CMP(xvcmpeqdp, 2, float64, VsrD(i), eq, 0)
2475 VSX_CMP(xvcmpgedp, 2, float64, VsrD(i), le, 1)
2476 VSX_CMP(xvcmpgtdp, 2, float64, VsrD(i), lt, 1)
2477 VSX_CMP(xvcmpeqsp, 4, float32, VsrW(i), eq, 0)
2478 VSX_CMP(xvcmpgesp, 4, float32, VsrW(i), le, 1)
2479 VSX_CMP(xvcmpgtsp, 4, float32, VsrW(i), lt, 1)
2481 /* VSX_CVT_FP_TO_FP - VSX floating point/floating point conversion
2482 * op - instruction mnemonic
2483 * nels - number of elements (1, 2 or 4)
2484 * stp - source type (float32 or float64)
2485 * ttp - target type (float32 or float64)
2486 * sfld - source vsr_t field
2487 * tfld - target vsr_t field (f32 or f64)
2488 * sfprf - set FPRF
2490 #define VSX_CVT_FP_TO_FP(op, nels, stp, ttp, sfld, tfld, sfprf) \
2491 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2493 ppc_vsr_t xt, xb; \
2494 int i; \
2496 getVSR(xB(opcode), &xb, env); \
2497 getVSR(xT(opcode), &xt, env); \
2499 for (i = 0; i < nels; i++) { \
2500 xt.tfld = stp##_to_##ttp(xb.sfld, &env->fp_status); \
2501 if (unlikely(stp##_is_signaling_nan(xb.sfld))) { \
2502 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \
2503 xt.tfld = ttp##_snan_to_qnan(xt.tfld); \
2505 if (sfprf) { \
2506 helper_compute_fprf(env, ttp##_to_float64(xt.tfld, \
2507 &env->fp_status), sfprf); \
2511 putVSR(xT(opcode), &xt, env); \
2512 helper_float_check_status(env); \
2515 VSX_CVT_FP_TO_FP(xscvdpsp, 1, float64, float32, VsrD(0), VsrW(0), 1)
2516 VSX_CVT_FP_TO_FP(xscvspdp, 1, float32, float64, VsrW(0), VsrD(0), 1)
2517 VSX_CVT_FP_TO_FP(xvcvdpsp, 2, float64, float32, VsrD(i), VsrW(2*i), 0)
2518 VSX_CVT_FP_TO_FP(xvcvspdp, 2, float32, float64, VsrW(2*i), VsrD(i), 0)
2520 uint64_t helper_xscvdpspn(CPUPPCState *env, uint64_t xb)
2522 float_status tstat = env->fp_status;
2523 set_float_exception_flags(0, &tstat);
2525 return (uint64_t)float64_to_float32(xb, &tstat) << 32;
2528 uint64_t helper_xscvspdpn(CPUPPCState *env, uint64_t xb)
2530 float_status tstat = env->fp_status;
2531 set_float_exception_flags(0, &tstat);
2533 return float32_to_float64(xb >> 32, &tstat);
2536 /* VSX_CVT_FP_TO_INT - VSX floating point to integer conversion
2537 * op - instruction mnemonic
2538 * nels - number of elements (1, 2 or 4)
2539 * stp - source type (float32 or float64)
2540 * ttp - target type (int32, uint32, int64 or uint64)
2541 * sfld - source vsr_t field
2542 * tfld - target vsr_t field
2543 * rnan - resulting NaN
2545 #define VSX_CVT_FP_TO_INT(op, nels, stp, ttp, sfld, tfld, rnan) \
2546 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2548 ppc_vsr_t xt, xb; \
2549 int i; \
2551 getVSR(xB(opcode), &xb, env); \
2552 getVSR(xT(opcode), &xt, env); \
2554 for (i = 0; i < nels; i++) { \
2555 if (unlikely(stp##_is_any_nan(xb.sfld))) { \
2556 if (stp##_is_signaling_nan(xb.sfld)) { \
2557 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \
2559 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXCVI, 0); \
2560 xt.tfld = rnan; \
2561 } else { \
2562 xt.tfld = stp##_to_##ttp##_round_to_zero(xb.sfld, \
2563 &env->fp_status); \
2564 if (env->fp_status.float_exception_flags & float_flag_invalid) { \
2565 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXCVI, 0); \
2570 putVSR(xT(opcode), &xt, env); \
2571 helper_float_check_status(env); \
2574 VSX_CVT_FP_TO_INT(xscvdpsxds, 1, float64, int64, VsrD(0), VsrD(0), \
2575 0x8000000000000000ULL)
2576 VSX_CVT_FP_TO_INT(xscvdpsxws, 1, float64, int32, VsrD(0), VsrW(1), \
2577 0x80000000U)
2578 VSX_CVT_FP_TO_INT(xscvdpuxds, 1, float64, uint64, VsrD(0), VsrD(0), 0ULL)
2579 VSX_CVT_FP_TO_INT(xscvdpuxws, 1, float64, uint32, VsrD(0), VsrW(1), 0U)
2580 VSX_CVT_FP_TO_INT(xvcvdpsxds, 2, float64, int64, VsrD(i), VsrD(i), \
2581 0x8000000000000000ULL)
2582 VSX_CVT_FP_TO_INT(xvcvdpsxws, 2, float64, int32, VsrD(i), VsrW(2*i), \
2583 0x80000000U)
2584 VSX_CVT_FP_TO_INT(xvcvdpuxds, 2, float64, uint64, VsrD(i), VsrD(i), 0ULL)
2585 VSX_CVT_FP_TO_INT(xvcvdpuxws, 2, float64, uint32, VsrD(i), VsrW(2*i), 0U)
2586 VSX_CVT_FP_TO_INT(xvcvspsxds, 2, float32, int64, VsrW(2*i), VsrD(i), \
2587 0x8000000000000000ULL)
2588 VSX_CVT_FP_TO_INT(xvcvspsxws, 4, float32, int32, VsrW(i), VsrW(i), 0x80000000U)
2589 VSX_CVT_FP_TO_INT(xvcvspuxds, 2, float32, uint64, VsrW(2*i), VsrD(i), 0ULL)
2590 VSX_CVT_FP_TO_INT(xvcvspuxws, 4, float32, uint32, VsrW(i), VsrW(i), 0U)
2592 /* VSX_CVT_INT_TO_FP - VSX integer to floating point conversion
2593 * op - instruction mnemonic
2594 * nels - number of elements (1, 2 or 4)
2595 * stp - source type (int32, uint32, int64 or uint64)
2596 * ttp - target type (float32 or float64)
2597 * sfld - source vsr_t field
2598 * tfld - target vsr_t field
2599 * jdef - definition of the j index (i or 2*i)
2600 * sfprf - set FPRF
2602 #define VSX_CVT_INT_TO_FP(op, nels, stp, ttp, sfld, tfld, sfprf, r2sp) \
2603 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2605 ppc_vsr_t xt, xb; \
2606 int i; \
2608 getVSR(xB(opcode), &xb, env); \
2609 getVSR(xT(opcode), &xt, env); \
2611 for (i = 0; i < nels; i++) { \
2612 xt.tfld = stp##_to_##ttp(xb.sfld, &env->fp_status); \
2613 if (r2sp) { \
2614 xt.tfld = helper_frsp(env, xt.tfld); \
2616 if (sfprf) { \
2617 helper_compute_fprf(env, xt.tfld, sfprf); \
2621 putVSR(xT(opcode), &xt, env); \
2622 helper_float_check_status(env); \
2625 VSX_CVT_INT_TO_FP(xscvsxddp, 1, int64, float64, VsrD(0), VsrD(0), 1, 0)
2626 VSX_CVT_INT_TO_FP(xscvuxddp, 1, uint64, float64, VsrD(0), VsrD(0), 1, 0)
2627 VSX_CVT_INT_TO_FP(xscvsxdsp, 1, int64, float64, VsrD(0), VsrD(0), 1, 1)
2628 VSX_CVT_INT_TO_FP(xscvuxdsp, 1, uint64, float64, VsrD(0), VsrD(0), 1, 1)
2629 VSX_CVT_INT_TO_FP(xvcvsxddp, 2, int64, float64, VsrD(i), VsrD(i), 0, 0)
2630 VSX_CVT_INT_TO_FP(xvcvuxddp, 2, uint64, float64, VsrD(i), VsrD(i), 0, 0)
2631 VSX_CVT_INT_TO_FP(xvcvsxwdp, 2, int32, float64, VsrW(2*i), VsrD(i), 0, 0)
2632 VSX_CVT_INT_TO_FP(xvcvuxwdp, 2, uint64, float64, VsrW(2*i), VsrD(i), 0, 0)
2633 VSX_CVT_INT_TO_FP(xvcvsxdsp, 2, int64, float32, VsrD(i), VsrW(2*i), 0, 0)
2634 VSX_CVT_INT_TO_FP(xvcvuxdsp, 2, uint64, float32, VsrD(i), VsrW(2*i), 0, 0)
2635 VSX_CVT_INT_TO_FP(xvcvsxwsp, 4, int32, float32, VsrW(i), VsrW(i), 0, 0)
2636 VSX_CVT_INT_TO_FP(xvcvuxwsp, 4, uint32, float32, VsrW(i), VsrW(i), 0, 0)
2638 /* For "use current rounding mode", define a value that will not be one of
2639 * the existing rounding model enums.
2641 #define FLOAT_ROUND_CURRENT (float_round_nearest_even + float_round_down + \
2642 float_round_up + float_round_to_zero)
2644 /* VSX_ROUND - VSX floating point round
2645 * op - instruction mnemonic
2646 * nels - number of elements (1, 2 or 4)
2647 * tp - type (float32 or float64)
2648 * fld - vsr_t field (VsrD(*) or VsrW(*))
2649 * rmode - rounding mode
2650 * sfprf - set FPRF
2652 #define VSX_ROUND(op, nels, tp, fld, rmode, sfprf) \
2653 void helper_##op(CPUPPCState *env, uint32_t opcode) \
2655 ppc_vsr_t xt, xb; \
2656 int i; \
2657 getVSR(xB(opcode), &xb, env); \
2658 getVSR(xT(opcode), &xt, env); \
2660 if (rmode != FLOAT_ROUND_CURRENT) { \
2661 set_float_rounding_mode(rmode, &env->fp_status); \
2664 for (i = 0; i < nels; i++) { \
2665 if (unlikely(tp##_is_signaling_nan(xb.fld))) { \
2666 fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \
2667 xt.fld = tp##_snan_to_qnan(xb.fld); \
2668 } else { \
2669 xt.fld = tp##_round_to_int(xb.fld, &env->fp_status); \
2671 if (sfprf) { \
2672 helper_compute_fprf(env, xt.fld, sfprf); \
2676 /* If this is not a "use current rounding mode" instruction, \
2677 * then inhibit setting of the XX bit and restore rounding \
2678 * mode from FPSCR */ \
2679 if (rmode != FLOAT_ROUND_CURRENT) { \
2680 fpscr_set_rounding_mode(env); \
2681 env->fp_status.float_exception_flags &= ~float_flag_inexact; \
2684 putVSR(xT(opcode), &xt, env); \
2685 helper_float_check_status(env); \
2688 VSX_ROUND(xsrdpi, 1, float64, VsrD(0), float_round_nearest_even, 1)
2689 VSX_ROUND(xsrdpic, 1, float64, VsrD(0), FLOAT_ROUND_CURRENT, 1)
2690 VSX_ROUND(xsrdpim, 1, float64, VsrD(0), float_round_down, 1)
2691 VSX_ROUND(xsrdpip, 1, float64, VsrD(0), float_round_up, 1)
2692 VSX_ROUND(xsrdpiz, 1, float64, VsrD(0), float_round_to_zero, 1)
2694 VSX_ROUND(xvrdpi, 2, float64, VsrD(i), float_round_nearest_even, 0)
2695 VSX_ROUND(xvrdpic, 2, float64, VsrD(i), FLOAT_ROUND_CURRENT, 0)
2696 VSX_ROUND(xvrdpim, 2, float64, VsrD(i), float_round_down, 0)
2697 VSX_ROUND(xvrdpip, 2, float64, VsrD(i), float_round_up, 0)
2698 VSX_ROUND(xvrdpiz, 2, float64, VsrD(i), float_round_to_zero, 0)
2700 VSX_ROUND(xvrspi, 4, float32, VsrW(i), float_round_nearest_even, 0)
2701 VSX_ROUND(xvrspic, 4, float32, VsrW(i), FLOAT_ROUND_CURRENT, 0)
2702 VSX_ROUND(xvrspim, 4, float32, VsrW(i), float_round_down, 0)
2703 VSX_ROUND(xvrspip, 4, float32, VsrW(i), float_round_up, 0)
2704 VSX_ROUND(xvrspiz, 4, float32, VsrW(i), float_round_to_zero, 0)
2706 uint64_t helper_xsrsp(CPUPPCState *env, uint64_t xb)
2708 helper_reset_fpstatus(env);
2710 uint64_t xt = helper_frsp(env, xb);
2712 helper_compute_fprf(env, xt, 1);
2713 helper_float_check_status(env);
2714 return xt;