]>
Commit | Line | Data |
---|---|---|
bd23cd45 BS |
1 | /* |
2 | * PowerPC floating point and SPE emulation helpers for QEMU. | |
3 | * | |
4 | * Copyright (c) 2003-2007 Jocelyn Mayer | |
5 | * | |
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. | |
10 | * | |
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. | |
15 | * | |
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/>. | |
18 | */ | |
19 | #include "cpu.h" | |
2ef6175a | 20 | #include "exec/helper-proto.h" |
bd23cd45 BS |
21 | |
22 | /*****************************************************************************/ | |
23 | /* Floating point operations helpers */ | |
8e703949 | 24 | uint64_t helper_float32_to_float64(CPUPPCState *env, uint32_t arg) |
bd23cd45 BS |
25 | { |
26 | CPU_FloatU f; | |
27 | CPU_DoubleU d; | |
28 | ||
29 | f.l = arg; | |
30 | d.d = float32_to_float64(f.f, &env->fp_status); | |
31 | return d.ll; | |
32 | } | |
33 | ||
8e703949 | 34 | uint32_t helper_float64_to_float32(CPUPPCState *env, uint64_t arg) |
bd23cd45 BS |
35 | { |
36 | CPU_FloatU f; | |
37 | CPU_DoubleU d; | |
38 | ||
39 | d.ll = arg; | |
40 | f.f = float64_to_float32(d.d, &env->fp_status); | |
41 | return f.l; | |
42 | } | |
43 | ||
44 | static inline int isden(float64 d) | |
45 | { | |
46 | CPU_DoubleU u; | |
47 | ||
48 | u.d = d; | |
49 | ||
50 | return ((u.ll >> 52) & 0x7FF) == 0; | |
51 | } | |
52 | ||
da29cb7b TM |
53 | static inline int ppc_float32_get_unbiased_exp(float32 f) |
54 | { | |
55 | return ((f >> 23) & 0xFF) - 127; | |
56 | } | |
57 | ||
58 | static inline int ppc_float64_get_unbiased_exp(float64 f) | |
59 | { | |
60 | return ((f >> 52) & 0x7FF) - 1023; | |
61 | } | |
62 | ||
8e703949 | 63 | uint32_t helper_compute_fprf(CPUPPCState *env, uint64_t arg, uint32_t set_fprf) |
bd23cd45 BS |
64 | { |
65 | CPU_DoubleU farg; | |
66 | int isneg; | |
67 | int ret; | |
68 | ||
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; | |
78 | } | |
79 | } else if (unlikely(float64_is_infinity(farg.d))) { | |
80 | /* +/- infinity */ | |
81 | if (isneg) { | |
82 | ret = 0x09; | |
83 | } else { | |
84 | ret = 0x05; | |
85 | } | |
86 | } else { | |
87 | if (float64_is_zero(farg.d)) { | |
88 | /* +/- zero */ | |
89 | if (isneg) { | |
90 | ret = 0x12; | |
91 | } else { | |
92 | ret = 0x02; | |
93 | } | |
94 | } else { | |
95 | if (isden(farg.d)) { | |
96 | /* Denormalized numbers */ | |
97 | ret = 0x10; | |
98 | } else { | |
99 | /* Normalized numbers */ | |
100 | ret = 0x00; | |
101 | } | |
102 | if (isneg) { | |
103 | ret |= 0x08; | |
104 | } else { | |
105 | ret |= 0x04; | |
106 | } | |
107 | } | |
108 | } | |
109 | if (set_fprf) { | |
110 | /* We update FPSCR_FPRF */ | |
111 | env->fpscr &= ~(0x1F << FPSCR_FPRF); | |
112 | env->fpscr |= ret << FPSCR_FPRF; | |
113 | } | |
114 | /* We just need fpcc to update Rc1 */ | |
115 | return ret & 0xF; | |
116 | } | |
117 | ||
118 | /* Floating-point invalid operations exception */ | |
59800ec8 TM |
119 | static inline uint64_t fload_invalid_op_excp(CPUPPCState *env, int op, |
120 | int set_fpcc) | |
bd23cd45 | 121 | { |
27103424 | 122 | CPUState *cs = CPU(ppc_env_get_cpu(env)); |
bd23cd45 BS |
123 | uint64_t ret = 0; |
124 | int ve; | |
125 | ||
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; | |
59800ec8 TM |
153 | if (set_fpcc) { |
154 | env->fpscr &= ~(0xF << FPSCR_FPCC); | |
155 | env->fpscr |= 0x11 << FPSCR_FPCC; | |
156 | } | |
bd23cd45 BS |
157 | /* We must update the target FPR before raising the exception */ |
158 | if (ve != 0) { | |
27103424 | 159 | cs->exception_index = POWERPC_EXCP_PROGRAM; |
bd23cd45 BS |
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; | |
165 | } | |
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; | |
59800ec8 TM |
175 | if (set_fpcc) { |
176 | env->fpscr &= ~(0xF << FPSCR_FPCC); | |
177 | env->fpscr |= 0x11 << FPSCR_FPCC; | |
178 | } | |
bd23cd45 BS |
179 | } |
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; | |
59800ec8 TM |
188 | if (set_fpcc) { |
189 | env->fpscr &= ~(0xF << FPSCR_FPCC); | |
190 | env->fpscr |= 0x11 << FPSCR_FPCC; | |
191 | } | |
bd23cd45 BS |
192 | } |
193 | break; | |
194 | } | |
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); | |
205 | } | |
206 | } | |
207 | return ret; | |
208 | } | |
209 | ||
8e703949 | 210 | static inline void float_zero_divide_excp(CPUPPCState *env) |
bd23cd45 BS |
211 | { |
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); | |
222 | } | |
223 | } | |
224 | } | |
225 | ||
8e703949 | 226 | static inline void float_overflow_excp(CPUPPCState *env) |
bd23cd45 | 227 | { |
27103424 AF |
228 | CPUState *cs = CPU(ppc_env_get_cpu(env)); |
229 | ||
bd23cd45 BS |
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 */ | |
27103424 | 238 | cs->exception_index = POWERPC_EXCP_PROGRAM; |
bd23cd45 BS |
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; | |
243 | } | |
244 | } | |
245 | ||
8e703949 | 246 | static inline void float_underflow_excp(CPUPPCState *env) |
bd23cd45 | 247 | { |
27103424 AF |
248 | CPUState *cs = CPU(ppc_env_get_cpu(env)); |
249 | ||
bd23cd45 BS |
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 */ | |
27103424 | 258 | cs->exception_index = POWERPC_EXCP_PROGRAM; |
bd23cd45 BS |
259 | env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX; |
260 | } | |
261 | } | |
262 | ||
8e703949 | 263 | static inline void float_inexact_excp(CPUPPCState *env) |
bd23cd45 | 264 | { |
27103424 AF |
265 | CPUState *cs = CPU(ppc_env_get_cpu(env)); |
266 | ||
bd23cd45 BS |
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 */ | |
27103424 | 274 | cs->exception_index = POWERPC_EXCP_PROGRAM; |
bd23cd45 BS |
275 | env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX; |
276 | } | |
277 | } | |
278 | ||
8e703949 | 279 | static inline void fpscr_set_rounding_mode(CPUPPCState *env) |
bd23cd45 BS |
280 | { |
281 | int rnd_type; | |
282 | ||
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; | |
302 | } | |
303 | set_float_rounding_mode(rnd_type, &env->fp_status); | |
304 | } | |
305 | ||
8e703949 | 306 | void helper_fpscr_clrbit(CPUPPCState *env, uint32_t bit) |
bd23cd45 BS |
307 | { |
308 | int prev; | |
309 | ||
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: | |
8e703949 | 316 | fpscr_set_rounding_mode(env); |
bd23cd45 BS |
317 | break; |
318 | default: | |
319 | break; | |
320 | } | |
321 | } | |
322 | } | |
323 | ||
8e703949 | 324 | void helper_fpscr_setbit(CPUPPCState *env, uint32_t bit) |
bd23cd45 | 325 | { |
27103424 | 326 | CPUState *cs = CPU(ppc_env_get_cpu(env)); |
bd23cd45 BS |
327 | int prev; |
328 | ||
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; | |
337 | } | |
90638255 | 338 | break; |
bd23cd45 BS |
339 | case FPSCR_OX: |
340 | env->fpscr |= 1 << FPSCR_FX; | |
341 | if (fpscr_oe) { | |
342 | goto raise_oe; | |
343 | } | |
344 | break; | |
345 | case FPSCR_UX: | |
346 | env->fpscr |= 1 << FPSCR_FX; | |
347 | if (fpscr_ue) { | |
348 | goto raise_ue; | |
349 | } | |
350 | break; | |
351 | case FPSCR_ZX: | |
352 | env->fpscr |= 1 << FPSCR_FX; | |
353 | if (fpscr_ze) { | |
354 | goto raise_ze; | |
355 | } | |
356 | break; | |
357 | case FPSCR_XX: | |
358 | env->fpscr |= 1 << FPSCR_FX; | |
359 | if (fpscr_xe) { | |
360 | goto raise_xe; | |
361 | } | |
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; | |
376 | } | |
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; | |
384 | } | |
385 | if (fpscr_vxisi) { | |
386 | env->error_code |= POWERPC_EXCP_FP_VXISI; | |
387 | } | |
388 | if (fpscr_vxidi) { | |
389 | env->error_code |= POWERPC_EXCP_FP_VXIDI; | |
390 | } | |
391 | if (fpscr_vxzdz) { | |
392 | env->error_code |= POWERPC_EXCP_FP_VXZDZ; | |
393 | } | |
394 | if (fpscr_vximz) { | |
395 | env->error_code |= POWERPC_EXCP_FP_VXIMZ; | |
396 | } | |
397 | if (fpscr_vxvc) { | |
398 | env->error_code |= POWERPC_EXCP_FP_VXVC; | |
399 | } | |
400 | if (fpscr_vxsoft) { | |
401 | env->error_code |= POWERPC_EXCP_FP_VXSOFT; | |
402 | } | |
403 | if (fpscr_vxsqrt) { | |
404 | env->error_code |= POWERPC_EXCP_FP_VXSQRT; | |
405 | } | |
406 | if (fpscr_vxcvi) { | |
407 | env->error_code |= POWERPC_EXCP_FP_VXCVI; | |
408 | } | |
409 | goto raise_excp; | |
410 | } | |
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; | |
417 | } | |
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; | |
424 | } | |
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; | |
431 | } | |
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; | |
438 | } | |
439 | break; | |
440 | case FPSCR_RN1: | |
441 | case FPSCR_RN: | |
8e703949 | 442 | fpscr_set_rounding_mode(env); |
bd23cd45 BS |
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 */ | |
27103424 | 450 | cs->exception_index = POWERPC_EXCP_PROGRAM; |
bd23cd45 BS |
451 | break; |
452 | } | |
453 | } | |
454 | } | |
455 | ||
8e703949 | 456 | void helper_store_fpscr(CPUPPCState *env, uint64_t arg, uint32_t mask) |
bd23cd45 | 457 | { |
27103424 | 458 | CPUState *cs = CPU(ppc_env_get_cpu(env)); |
7d08d856 | 459 | target_ulong prev, new; |
bd23cd45 BS |
460 | int i; |
461 | ||
462 | prev = env->fpscr; | |
7d08d856 AJ |
463 | new = (target_ulong)arg; |
464 | new &= ~0x60000000LL; | |
465 | new |= prev & 0x60000000LL; | |
466 | for (i = 0; i < sizeof(target_ulong) * 2; i++) { | |
bd23cd45 | 467 | if (mask & (1 << i)) { |
7d08d856 AJ |
468 | env->fpscr &= ~(0xFLL << (4 * i)); |
469 | env->fpscr |= new & (0xFLL << (4 * i)); | |
bd23cd45 BS |
470 | } |
471 | } | |
472 | /* Update VX and FEX */ | |
473 | if (fpscr_ix != 0) { | |
474 | env->fpscr |= 1 << FPSCR_VX; | |
475 | } else { | |
476 | env->fpscr &= ~(1 << FPSCR_VX); | |
477 | } | |
478 | if ((fpscr_ex & fpscr_eex) != 0) { | |
479 | env->fpscr |= 1 << FPSCR_FEX; | |
27103424 | 480 | cs->exception_index = POWERPC_EXCP_PROGRAM; |
bd23cd45 BS |
481 | /* XXX: we should compute it properly */ |
482 | env->error_code = POWERPC_EXCP_FP; | |
483 | } else { | |
484 | env->fpscr &= ~(1 << FPSCR_FEX); | |
485 | } | |
8e703949 | 486 | fpscr_set_rounding_mode(env); |
bd23cd45 BS |
487 | } |
488 | ||
d6478bc7 FC |
489 | void store_fpscr(CPUPPCState *env, uint64_t arg, uint32_t mask) |
490 | { | |
491 | helper_store_fpscr(env, arg, mask); | |
492 | } | |
493 | ||
8e703949 | 494 | void helper_float_check_status(CPUPPCState *env) |
bd23cd45 | 495 | { |
27103424 | 496 | CPUState *cs = CPU(ppc_env_get_cpu(env)); |
db72c9f2 TG |
497 | int status = get_float_exception_flags(&env->fp_status); |
498 | ||
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); | |
507 | } | |
508 | ||
27103424 | 509 | if (cs->exception_index == POWERPC_EXCP_PROGRAM && |
bd23cd45 BS |
510 | (env->error_code & POWERPC_EXCP_FP)) { |
511 | /* Differred floating-point exception after target FPR update */ | |
512 | if (msr_fe0 != 0 || msr_fe1 != 0) { | |
27103424 | 513 | helper_raise_exception_err(env, cs->exception_index, |
bd23cd45 BS |
514 | env->error_code); |
515 | } | |
bd23cd45 BS |
516 | } |
517 | } | |
518 | ||
8e703949 | 519 | void helper_reset_fpstatus(CPUPPCState *env) |
bd23cd45 BS |
520 | { |
521 | set_float_exception_flags(0, &env->fp_status); | |
522 | } | |
523 | ||
524 | /* fadd - fadd. */ | |
8e703949 | 525 | uint64_t helper_fadd(CPUPPCState *env, uint64_t arg1, uint64_t arg2) |
bd23cd45 BS |
526 | { |
527 | CPU_DoubleU farg1, farg2; | |
528 | ||
529 | farg1.ll = arg1; | |
530 | farg2.ll = arg2; | |
531 | ||
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 */ | |
59800ec8 | 535 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1); |
bd23cd45 BS |
536 | } else { |
537 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
538 | float64_is_signaling_nan(farg2.d))) { | |
539 | /* sNaN addition */ | |
59800ec8 | 540 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
541 | } |
542 | farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status); | |
543 | } | |
544 | ||
545 | return farg1.ll; | |
546 | } | |
547 | ||
548 | /* fsub - fsub. */ | |
8e703949 | 549 | uint64_t helper_fsub(CPUPPCState *env, uint64_t arg1, uint64_t arg2) |
bd23cd45 BS |
550 | { |
551 | CPU_DoubleU farg1, farg2; | |
552 | ||
553 | farg1.ll = arg1; | |
554 | farg2.ll = arg2; | |
555 | ||
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 */ | |
59800ec8 | 559 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1); |
bd23cd45 BS |
560 | } else { |
561 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
562 | float64_is_signaling_nan(farg2.d))) { | |
563 | /* sNaN subtraction */ | |
59800ec8 | 564 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
565 | } |
566 | farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status); | |
567 | } | |
568 | ||
569 | return farg1.ll; | |
570 | } | |
571 | ||
572 | /* fmul - fmul. */ | |
8e703949 | 573 | uint64_t helper_fmul(CPUPPCState *env, uint64_t arg1, uint64_t arg2) |
bd23cd45 BS |
574 | { |
575 | CPU_DoubleU farg1, farg2; | |
576 | ||
577 | farg1.ll = arg1; | |
578 | farg2.ll = arg2; | |
579 | ||
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 */ | |
59800ec8 | 583 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1); |
bd23cd45 BS |
584 | } else { |
585 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
586 | float64_is_signaling_nan(farg2.d))) { | |
587 | /* sNaN multiplication */ | |
59800ec8 | 588 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
589 | } |
590 | farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); | |
591 | } | |
592 | ||
593 | return farg1.ll; | |
594 | } | |
595 | ||
596 | /* fdiv - fdiv. */ | |
8e703949 | 597 | uint64_t helper_fdiv(CPUPPCState *env, uint64_t arg1, uint64_t arg2) |
bd23cd45 BS |
598 | { |
599 | CPU_DoubleU farg1, farg2; | |
600 | ||
601 | farg1.ll = arg1; | |
602 | farg2.ll = arg2; | |
603 | ||
604 | if (unlikely(float64_is_infinity(farg1.d) && | |
605 | float64_is_infinity(farg2.d))) { | |
606 | /* Division of infinity by infinity */ | |
59800ec8 | 607 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIDI, 1); |
bd23cd45 BS |
608 | } else if (unlikely(float64_is_zero(farg1.d) && float64_is_zero(farg2.d))) { |
609 | /* Division of zero by zero */ | |
59800ec8 | 610 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXZDZ, 1); |
bd23cd45 BS |
611 | } else { |
612 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
613 | float64_is_signaling_nan(farg2.d))) { | |
614 | /* sNaN division */ | |
59800ec8 | 615 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
616 | } |
617 | farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status); | |
618 | } | |
619 | ||
620 | return farg1.ll; | |
621 | } | |
622 | ||
bd23cd45 | 623 | |
fab7fe42 TM |
624 | #define FPU_FCTI(op, cvt, nanval) \ |
625 | uint64_t helper_##op(CPUPPCState *env, uint64_t arg) \ | |
626 | { \ | |
627 | CPU_DoubleU farg; \ | |
628 | \ | |
629 | farg.ll = arg; \ | |
630 | farg.ll = float64_to_##cvt(farg.d, &env->fp_status); \ | |
631 | \ | |
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); \ | |
637 | } \ | |
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); \ | |
642 | } \ | |
643 | helper_float_check_status(env); \ | |
644 | } \ | |
645 | return farg.ll; \ | |
646 | } | |
647 | ||
7dff9abe TM |
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) | |
fab7fe42 | 652 | #if defined(TARGET_PPC64) |
7dff9abe TM |
653 | FPU_FCTI(fctid, int64, 0x8000000000000000ULL) |
654 | FPU_FCTI(fctidz, int64_round_to_zero, 0x8000000000000000ULL) | |
655 | FPU_FCTI(fctidu, uint64, 0x0000000000000000ULL) | |
656 | FPU_FCTI(fctiduz, uint64_round_to_zero, 0x0000000000000000ULL) | |
fab7fe42 | 657 | #endif |
bd23cd45 BS |
658 | |
659 | #if defined(TARGET_PPC64) | |
bd23cd45 | 660 | |
28288b48 TM |
661 | #define FPU_FCFI(op, cvtr, is_single) \ |
662 | uint64_t helper_##op(CPUPPCState *env, uint64_t arg) \ | |
663 | { \ | |
664 | CPU_DoubleU farg; \ | |
665 | \ | |
666 | if (is_single) { \ | |
667 | float32 tmp = cvtr(arg, &env->fp_status); \ | |
668 | farg.d = float32_to_float64(tmp, &env->fp_status); \ | |
669 | } else { \ | |
670 | farg.d = cvtr(arg, &env->fp_status); \ | |
671 | } \ | |
672 | helper_float_check_status(env); \ | |
673 | return farg.ll; \ | |
bd23cd45 BS |
674 | } |
675 | ||
28288b48 TM |
676 | FPU_FCFI(fcfid, int64_to_float64, 0) |
677 | FPU_FCFI(fcfids, int64_to_float32, 1) | |
678 | FPU_FCFI(fcfidu, uint64_to_float64, 0) | |
679 | FPU_FCFI(fcfidus, uint64_to_float32, 1) | |
bd23cd45 BS |
680 | |
681 | #endif | |
682 | ||
8e703949 BS |
683 | static inline uint64_t do_fri(CPUPPCState *env, uint64_t arg, |
684 | int rounding_mode) | |
bd23cd45 BS |
685 | { |
686 | CPU_DoubleU farg; | |
687 | ||
688 | farg.ll = arg; | |
689 | ||
690 | if (unlikely(float64_is_signaling_nan(farg.d))) { | |
691 | /* sNaN round */ | |
c7386080 | 692 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
7dff9abe | 693 | farg.ll = arg | 0x0008000000000000ULL; |
bd23cd45 | 694 | } else { |
c7386080 TM |
695 | int inexact = get_float_exception_flags(&env->fp_status) & |
696 | float_flag_inexact; | |
bd23cd45 BS |
697 | set_float_rounding_mode(rounding_mode, &env->fp_status); |
698 | farg.ll = float64_round_to_int(farg.d, &env->fp_status); | |
699 | /* Restore rounding mode from FPSCR */ | |
8e703949 | 700 | fpscr_set_rounding_mode(env); |
c7386080 TM |
701 | |
702 | /* fri* does not set FPSCR[XX] */ | |
703 | if (!inexact) { | |
704 | env->fp_status.float_exception_flags &= ~float_flag_inexact; | |
705 | } | |
bd23cd45 | 706 | } |
c7386080 | 707 | helper_float_check_status(env); |
bd23cd45 BS |
708 | return farg.ll; |
709 | } | |
710 | ||
8e703949 | 711 | uint64_t helper_frin(CPUPPCState *env, uint64_t arg) |
bd23cd45 | 712 | { |
c7386080 | 713 | return do_fri(env, arg, float_round_ties_away); |
bd23cd45 BS |
714 | } |
715 | ||
8e703949 | 716 | uint64_t helper_friz(CPUPPCState *env, uint64_t arg) |
bd23cd45 | 717 | { |
8e703949 | 718 | return do_fri(env, arg, float_round_to_zero); |
bd23cd45 BS |
719 | } |
720 | ||
8e703949 | 721 | uint64_t helper_frip(CPUPPCState *env, uint64_t arg) |
bd23cd45 | 722 | { |
8e703949 | 723 | return do_fri(env, arg, float_round_up); |
bd23cd45 BS |
724 | } |
725 | ||
8e703949 | 726 | uint64_t helper_frim(CPUPPCState *env, uint64_t arg) |
bd23cd45 | 727 | { |
8e703949 | 728 | return do_fri(env, arg, float_round_down); |
bd23cd45 BS |
729 | } |
730 | ||
731 | /* fmadd - fmadd. */ | |
8e703949 BS |
732 | uint64_t helper_fmadd(CPUPPCState *env, uint64_t arg1, uint64_t arg2, |
733 | uint64_t arg3) | |
bd23cd45 BS |
734 | { |
735 | CPU_DoubleU farg1, farg2, farg3; | |
736 | ||
737 | farg1.ll = arg1; | |
738 | farg2.ll = arg2; | |
739 | farg3.ll = arg3; | |
740 | ||
741 | if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || | |
742 | (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) { | |
743 | /* Multiplication of zero by infinity */ | |
59800ec8 | 744 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1); |
bd23cd45 BS |
745 | } else { |
746 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
747 | float64_is_signaling_nan(farg2.d) || | |
748 | float64_is_signaling_nan(farg3.d))) { | |
749 | /* sNaN operation */ | |
59800ec8 | 750 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
751 | } |
752 | /* This is the way the PowerPC specification defines it */ | |
753 | float128 ft0_128, ft1_128; | |
754 | ||
755 | ft0_128 = float64_to_float128(farg1.d, &env->fp_status); | |
756 | ft1_128 = float64_to_float128(farg2.d, &env->fp_status); | |
757 | ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); | |
758 | if (unlikely(float128_is_infinity(ft0_128) && | |
759 | float64_is_infinity(farg3.d) && | |
760 | float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) { | |
761 | /* Magnitude subtraction of infinities */ | |
59800ec8 | 762 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1); |
bd23cd45 BS |
763 | } else { |
764 | ft1_128 = float64_to_float128(farg3.d, &env->fp_status); | |
765 | ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status); | |
766 | farg1.d = float128_to_float64(ft0_128, &env->fp_status); | |
767 | } | |
768 | } | |
769 | ||
770 | return farg1.ll; | |
771 | } | |
772 | ||
773 | /* fmsub - fmsub. */ | |
8e703949 BS |
774 | uint64_t helper_fmsub(CPUPPCState *env, uint64_t arg1, uint64_t arg2, |
775 | uint64_t arg3) | |
bd23cd45 BS |
776 | { |
777 | CPU_DoubleU farg1, farg2, farg3; | |
778 | ||
779 | farg1.ll = arg1; | |
780 | farg2.ll = arg2; | |
781 | farg3.ll = arg3; | |
782 | ||
783 | if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || | |
784 | (float64_is_zero(farg1.d) && | |
785 | float64_is_infinity(farg2.d)))) { | |
786 | /* Multiplication of zero by infinity */ | |
59800ec8 | 787 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1); |
bd23cd45 BS |
788 | } else { |
789 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
790 | float64_is_signaling_nan(farg2.d) || | |
791 | float64_is_signaling_nan(farg3.d))) { | |
792 | /* sNaN operation */ | |
59800ec8 | 793 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
794 | } |
795 | /* This is the way the PowerPC specification defines it */ | |
796 | float128 ft0_128, ft1_128; | |
797 | ||
798 | ft0_128 = float64_to_float128(farg1.d, &env->fp_status); | |
799 | ft1_128 = float64_to_float128(farg2.d, &env->fp_status); | |
800 | ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); | |
801 | if (unlikely(float128_is_infinity(ft0_128) && | |
802 | float64_is_infinity(farg3.d) && | |
803 | float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) { | |
804 | /* Magnitude subtraction of infinities */ | |
59800ec8 | 805 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1); |
bd23cd45 BS |
806 | } else { |
807 | ft1_128 = float64_to_float128(farg3.d, &env->fp_status); | |
808 | ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status); | |
809 | farg1.d = float128_to_float64(ft0_128, &env->fp_status); | |
810 | } | |
811 | } | |
812 | return farg1.ll; | |
813 | } | |
814 | ||
815 | /* fnmadd - fnmadd. */ | |
8e703949 BS |
816 | uint64_t helper_fnmadd(CPUPPCState *env, uint64_t arg1, uint64_t arg2, |
817 | uint64_t arg3) | |
bd23cd45 BS |
818 | { |
819 | CPU_DoubleU farg1, farg2, farg3; | |
820 | ||
821 | farg1.ll = arg1; | |
822 | farg2.ll = arg2; | |
823 | farg3.ll = arg3; | |
824 | ||
825 | if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || | |
826 | (float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) { | |
827 | /* Multiplication of zero by infinity */ | |
59800ec8 | 828 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1); |
bd23cd45 BS |
829 | } else { |
830 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
831 | float64_is_signaling_nan(farg2.d) || | |
832 | float64_is_signaling_nan(farg3.d))) { | |
833 | /* sNaN operation */ | |
59800ec8 | 834 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
835 | } |
836 | /* This is the way the PowerPC specification defines it */ | |
837 | float128 ft0_128, ft1_128; | |
838 | ||
839 | ft0_128 = float64_to_float128(farg1.d, &env->fp_status); | |
840 | ft1_128 = float64_to_float128(farg2.d, &env->fp_status); | |
841 | ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); | |
842 | if (unlikely(float128_is_infinity(ft0_128) && | |
843 | float64_is_infinity(farg3.d) && | |
844 | float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) { | |
845 | /* Magnitude subtraction of infinities */ | |
59800ec8 | 846 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1); |
bd23cd45 BS |
847 | } else { |
848 | ft1_128 = float64_to_float128(farg3.d, &env->fp_status); | |
849 | ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status); | |
850 | farg1.d = float128_to_float64(ft0_128, &env->fp_status); | |
851 | } | |
852 | if (likely(!float64_is_any_nan(farg1.d))) { | |
853 | farg1.d = float64_chs(farg1.d); | |
854 | } | |
855 | } | |
856 | return farg1.ll; | |
857 | } | |
858 | ||
859 | /* fnmsub - fnmsub. */ | |
8e703949 BS |
860 | uint64_t helper_fnmsub(CPUPPCState *env, uint64_t arg1, uint64_t arg2, |
861 | uint64_t arg3) | |
bd23cd45 BS |
862 | { |
863 | CPU_DoubleU farg1, farg2, farg3; | |
864 | ||
865 | farg1.ll = arg1; | |
866 | farg2.ll = arg2; | |
867 | farg3.ll = arg3; | |
868 | ||
869 | if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || | |
870 | (float64_is_zero(farg1.d) && | |
871 | float64_is_infinity(farg2.d)))) { | |
872 | /* Multiplication of zero by infinity */ | |
59800ec8 | 873 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, 1); |
bd23cd45 BS |
874 | } else { |
875 | if (unlikely(float64_is_signaling_nan(farg1.d) || | |
876 | float64_is_signaling_nan(farg2.d) || | |
877 | float64_is_signaling_nan(farg3.d))) { | |
878 | /* sNaN operation */ | |
59800ec8 | 879 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
880 | } |
881 | /* This is the way the PowerPC specification defines it */ | |
882 | float128 ft0_128, ft1_128; | |
883 | ||
884 | ft0_128 = float64_to_float128(farg1.d, &env->fp_status); | |
885 | ft1_128 = float64_to_float128(farg2.d, &env->fp_status); | |
886 | ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); | |
887 | if (unlikely(float128_is_infinity(ft0_128) && | |
888 | float64_is_infinity(farg3.d) && | |
889 | float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) { | |
890 | /* Magnitude subtraction of infinities */ | |
59800ec8 | 891 | farg1.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, 1); |
bd23cd45 BS |
892 | } else { |
893 | ft1_128 = float64_to_float128(farg3.d, &env->fp_status); | |
894 | ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status); | |
895 | farg1.d = float128_to_float64(ft0_128, &env->fp_status); | |
896 | } | |
897 | if (likely(!float64_is_any_nan(farg1.d))) { | |
898 | farg1.d = float64_chs(farg1.d); | |
899 | } | |
900 | } | |
901 | return farg1.ll; | |
902 | } | |
903 | ||
904 | /* frsp - frsp. */ | |
8e703949 | 905 | uint64_t helper_frsp(CPUPPCState *env, uint64_t arg) |
bd23cd45 BS |
906 | { |
907 | CPU_DoubleU farg; | |
908 | float32 f32; | |
909 | ||
910 | farg.ll = arg; | |
911 | ||
912 | if (unlikely(float64_is_signaling_nan(farg.d))) { | |
913 | /* sNaN square root */ | |
59800ec8 | 914 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
915 | } |
916 | f32 = float64_to_float32(farg.d, &env->fp_status); | |
917 | farg.d = float32_to_float64(f32, &env->fp_status); | |
918 | ||
919 | return farg.ll; | |
920 | } | |
921 | ||
922 | /* fsqrt - fsqrt. */ | |
8e703949 | 923 | uint64_t helper_fsqrt(CPUPPCState *env, uint64_t arg) |
bd23cd45 BS |
924 | { |
925 | CPU_DoubleU farg; | |
926 | ||
927 | farg.ll = arg; | |
928 | ||
929 | if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) { | |
930 | /* Square root of a negative nonzero number */ | |
59800ec8 | 931 | farg.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, 1); |
bd23cd45 BS |
932 | } else { |
933 | if (unlikely(float64_is_signaling_nan(farg.d))) { | |
934 | /* sNaN square root */ | |
59800ec8 | 935 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
936 | } |
937 | farg.d = float64_sqrt(farg.d, &env->fp_status); | |
938 | } | |
939 | return farg.ll; | |
940 | } | |
941 | ||
942 | /* fre - fre. */ | |
8e703949 | 943 | uint64_t helper_fre(CPUPPCState *env, uint64_t arg) |
bd23cd45 BS |
944 | { |
945 | CPU_DoubleU farg; | |
946 | ||
947 | farg.ll = arg; | |
948 | ||
949 | if (unlikely(float64_is_signaling_nan(farg.d))) { | |
950 | /* sNaN reciprocal */ | |
59800ec8 | 951 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
952 | } |
953 | farg.d = float64_div(float64_one, farg.d, &env->fp_status); | |
954 | return farg.d; | |
955 | } | |
956 | ||
957 | /* fres - fres. */ | |
8e703949 | 958 | uint64_t helper_fres(CPUPPCState *env, uint64_t arg) |
bd23cd45 BS |
959 | { |
960 | CPU_DoubleU farg; | |
961 | float32 f32; | |
962 | ||
963 | farg.ll = arg; | |
964 | ||
965 | if (unlikely(float64_is_signaling_nan(farg.d))) { | |
966 | /* sNaN reciprocal */ | |
59800ec8 | 967 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
968 | } |
969 | farg.d = float64_div(float64_one, farg.d, &env->fp_status); | |
970 | f32 = float64_to_float32(farg.d, &env->fp_status); | |
971 | farg.d = float32_to_float64(f32, &env->fp_status); | |
972 | ||
973 | return farg.ll; | |
974 | } | |
975 | ||
976 | /* frsqrte - frsqrte. */ | |
8e703949 | 977 | uint64_t helper_frsqrte(CPUPPCState *env, uint64_t arg) |
bd23cd45 BS |
978 | { |
979 | CPU_DoubleU farg; | |
980 | float32 f32; | |
981 | ||
982 | farg.ll = arg; | |
983 | ||
984 | if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) { | |
985 | /* Reciprocal square root of a negative nonzero number */ | |
59800ec8 | 986 | farg.ll = fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, 1); |
bd23cd45 BS |
987 | } else { |
988 | if (unlikely(float64_is_signaling_nan(farg.d))) { | |
989 | /* sNaN reciprocal square root */ | |
59800ec8 | 990 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
991 | } |
992 | farg.d = float64_sqrt(farg.d, &env->fp_status); | |
993 | farg.d = float64_div(float64_one, farg.d, &env->fp_status); | |
994 | f32 = float64_to_float32(farg.d, &env->fp_status); | |
995 | farg.d = float32_to_float64(f32, &env->fp_status); | |
996 | } | |
997 | return farg.ll; | |
998 | } | |
999 | ||
1000 | /* fsel - fsel. */ | |
8e703949 BS |
1001 | uint64_t helper_fsel(CPUPPCState *env, uint64_t arg1, uint64_t arg2, |
1002 | uint64_t arg3) | |
bd23cd45 BS |
1003 | { |
1004 | CPU_DoubleU farg1; | |
1005 | ||
1006 | farg1.ll = arg1; | |
1007 | ||
1008 | if ((!float64_is_neg(farg1.d) || float64_is_zero(farg1.d)) && | |
1009 | !float64_is_any_nan(farg1.d)) { | |
1010 | return arg2; | |
1011 | } else { | |
1012 | return arg3; | |
1013 | } | |
1014 | } | |
1015 | ||
da29cb7b TM |
1016 | uint32_t helper_ftdiv(uint64_t fra, uint64_t frb) |
1017 | { | |
1018 | int fe_flag = 0; | |
1019 | int fg_flag = 0; | |
1020 | ||
1021 | if (unlikely(float64_is_infinity(fra) || | |
1022 | float64_is_infinity(frb) || | |
1023 | float64_is_zero(frb))) { | |
1024 | fe_flag = 1; | |
1025 | fg_flag = 1; | |
1026 | } else { | |
1027 | int e_a = ppc_float64_get_unbiased_exp(fra); | |
1028 | int e_b = ppc_float64_get_unbiased_exp(frb); | |
1029 | ||
1030 | if (unlikely(float64_is_any_nan(fra) || | |
1031 | float64_is_any_nan(frb))) { | |
1032 | fe_flag = 1; | |
1033 | } else if ((e_b <= -1022) || (e_b >= 1021)) { | |
1034 | fe_flag = 1; | |
1035 | } else if (!float64_is_zero(fra) && | |
1036 | (((e_a - e_b) >= 1023) || | |
1037 | ((e_a - e_b) <= -1021) || | |
1038 | (e_a <= -970))) { | |
1039 | fe_flag = 1; | |
1040 | } | |
1041 | ||
1042 | if (unlikely(float64_is_zero_or_denormal(frb))) { | |
1043 | /* XB is not zero because of the above check and */ | |
1044 | /* so must be denormalized. */ | |
1045 | fg_flag = 1; | |
1046 | } | |
1047 | } | |
1048 | ||
1049 | return 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); | |
1050 | } | |
6d41d146 TM |
1051 | |
1052 | uint32_t helper_ftsqrt(uint64_t frb) | |
1053 | { | |
1054 | int fe_flag = 0; | |
1055 | int fg_flag = 0; | |
1056 | ||
1057 | if (unlikely(float64_is_infinity(frb) || float64_is_zero(frb))) { | |
1058 | fe_flag = 1; | |
1059 | fg_flag = 1; | |
1060 | } else { | |
1061 | int e_b = ppc_float64_get_unbiased_exp(frb); | |
1062 | ||
1063 | if (unlikely(float64_is_any_nan(frb))) { | |
1064 | fe_flag = 1; | |
1065 | } else if (unlikely(float64_is_zero(frb))) { | |
1066 | fe_flag = 1; | |
1067 | } else if (unlikely(float64_is_neg(frb))) { | |
1068 | fe_flag = 1; | |
1069 | } else if (!float64_is_zero(frb) && (e_b <= (-1022+52))) { | |
1070 | fe_flag = 1; | |
1071 | } | |
1072 | ||
1073 | if (unlikely(float64_is_zero_or_denormal(frb))) { | |
1074 | /* XB is not zero because of the above check and */ | |
1075 | /* therefore must be denormalized. */ | |
1076 | fg_flag = 1; | |
1077 | } | |
1078 | } | |
1079 | ||
1080 | return 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); | |
1081 | } | |
da29cb7b | 1082 | |
8e703949 BS |
1083 | void helper_fcmpu(CPUPPCState *env, uint64_t arg1, uint64_t arg2, |
1084 | uint32_t crfD) | |
bd23cd45 BS |
1085 | { |
1086 | CPU_DoubleU farg1, farg2; | |
1087 | uint32_t ret = 0; | |
1088 | ||
1089 | farg1.ll = arg1; | |
1090 | farg2.ll = arg2; | |
1091 | ||
1092 | if (unlikely(float64_is_any_nan(farg1.d) || | |
1093 | float64_is_any_nan(farg2.d))) { | |
1094 | ret = 0x01UL; | |
1095 | } else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) { | |
1096 | ret = 0x08UL; | |
1097 | } else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) { | |
1098 | ret = 0x04UL; | |
1099 | } else { | |
1100 | ret = 0x02UL; | |
1101 | } | |
1102 | ||
1103 | env->fpscr &= ~(0x0F << FPSCR_FPRF); | |
1104 | env->fpscr |= ret << FPSCR_FPRF; | |
1105 | env->crf[crfD] = ret; | |
1106 | if (unlikely(ret == 0x01UL | |
1107 | && (float64_is_signaling_nan(farg1.d) || | |
1108 | float64_is_signaling_nan(farg2.d)))) { | |
1109 | /* sNaN comparison */ | |
59800ec8 | 1110 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 1); |
bd23cd45 BS |
1111 | } |
1112 | } | |
1113 | ||
8e703949 BS |
1114 | void helper_fcmpo(CPUPPCState *env, uint64_t arg1, uint64_t arg2, |
1115 | uint32_t crfD) | |
bd23cd45 BS |
1116 | { |
1117 | CPU_DoubleU farg1, farg2; | |
1118 | uint32_t ret = 0; | |
1119 | ||
1120 | farg1.ll = arg1; | |
1121 | farg2.ll = arg2; | |
1122 | ||
1123 | if (unlikely(float64_is_any_nan(farg1.d) || | |
1124 | float64_is_any_nan(farg2.d))) { | |
1125 | ret = 0x01UL; | |
1126 | } else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) { | |
1127 | ret = 0x08UL; | |
1128 | } else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) { | |
1129 | ret = 0x04UL; | |
1130 | } else { | |
1131 | ret = 0x02UL; | |
1132 | } | |
1133 | ||
1134 | env->fpscr &= ~(0x0F << FPSCR_FPRF); | |
1135 | env->fpscr |= ret << FPSCR_FPRF; | |
1136 | env->crf[crfD] = ret; | |
1137 | if (unlikely(ret == 0x01UL)) { | |
1138 | if (float64_is_signaling_nan(farg1.d) || | |
1139 | float64_is_signaling_nan(farg2.d)) { | |
1140 | /* sNaN comparison */ | |
8e703949 | 1141 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN | |
59800ec8 | 1142 | POWERPC_EXCP_FP_VXVC, 1); |
bd23cd45 BS |
1143 | } else { |
1144 | /* qNaN comparison */ | |
59800ec8 | 1145 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXVC, 1); |
bd23cd45 BS |
1146 | } |
1147 | } | |
1148 | } | |
1149 | ||
1150 | /* Single-precision floating-point conversions */ | |
8e703949 | 1151 | static inline uint32_t efscfsi(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1152 | { |
1153 | CPU_FloatU u; | |
1154 | ||
1155 | u.f = int32_to_float32(val, &env->vec_status); | |
1156 | ||
1157 | return u.l; | |
1158 | } | |
1159 | ||
8e703949 | 1160 | static inline uint32_t efscfui(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1161 | { |
1162 | CPU_FloatU u; | |
1163 | ||
1164 | u.f = uint32_to_float32(val, &env->vec_status); | |
1165 | ||
1166 | return u.l; | |
1167 | } | |
1168 | ||
8e703949 | 1169 | static inline int32_t efsctsi(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1170 | { |
1171 | CPU_FloatU u; | |
1172 | ||
1173 | u.l = val; | |
1174 | /* NaN are not treated the same way IEEE 754 does */ | |
1175 | if (unlikely(float32_is_quiet_nan(u.f))) { | |
1176 | return 0; | |
1177 | } | |
1178 | ||
1179 | return float32_to_int32(u.f, &env->vec_status); | |
1180 | } | |
1181 | ||
8e703949 | 1182 | static inline uint32_t efsctui(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1183 | { |
1184 | CPU_FloatU u; | |
1185 | ||
1186 | u.l = val; | |
1187 | /* NaN are not treated the same way IEEE 754 does */ | |
1188 | if (unlikely(float32_is_quiet_nan(u.f))) { | |
1189 | return 0; | |
1190 | } | |
1191 | ||
1192 | return float32_to_uint32(u.f, &env->vec_status); | |
1193 | } | |
1194 | ||
8e703949 | 1195 | static inline uint32_t efsctsiz(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1196 | { |
1197 | CPU_FloatU u; | |
1198 | ||
1199 | u.l = val; | |
1200 | /* NaN are not treated the same way IEEE 754 does */ | |
1201 | if (unlikely(float32_is_quiet_nan(u.f))) { | |
1202 | return 0; | |
1203 | } | |
1204 | ||
1205 | return float32_to_int32_round_to_zero(u.f, &env->vec_status); | |
1206 | } | |
1207 | ||
8e703949 | 1208 | static inline uint32_t efsctuiz(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1209 | { |
1210 | CPU_FloatU u; | |
1211 | ||
1212 | u.l = val; | |
1213 | /* NaN are not treated the same way IEEE 754 does */ | |
1214 | if (unlikely(float32_is_quiet_nan(u.f))) { | |
1215 | return 0; | |
1216 | } | |
1217 | ||
1218 | return float32_to_uint32_round_to_zero(u.f, &env->vec_status); | |
1219 | } | |
1220 | ||
8e703949 | 1221 | static inline uint32_t efscfsf(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1222 | { |
1223 | CPU_FloatU u; | |
1224 | float32 tmp; | |
1225 | ||
1226 | u.f = int32_to_float32(val, &env->vec_status); | |
1227 | tmp = int64_to_float32(1ULL << 32, &env->vec_status); | |
1228 | u.f = float32_div(u.f, tmp, &env->vec_status); | |
1229 | ||
1230 | return u.l; | |
1231 | } | |
1232 | ||
8e703949 | 1233 | static inline uint32_t efscfuf(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1234 | { |
1235 | CPU_FloatU u; | |
1236 | float32 tmp; | |
1237 | ||
1238 | u.f = uint32_to_float32(val, &env->vec_status); | |
1239 | tmp = uint64_to_float32(1ULL << 32, &env->vec_status); | |
1240 | u.f = float32_div(u.f, tmp, &env->vec_status); | |
1241 | ||
1242 | return u.l; | |
1243 | } | |
1244 | ||
8e703949 | 1245 | static inline uint32_t efsctsf(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1246 | { |
1247 | CPU_FloatU u; | |
1248 | float32 tmp; | |
1249 | ||
1250 | u.l = val; | |
1251 | /* NaN are not treated the same way IEEE 754 does */ | |
1252 | if (unlikely(float32_is_quiet_nan(u.f))) { | |
1253 | return 0; | |
1254 | } | |
1255 | tmp = uint64_to_float32(1ULL << 32, &env->vec_status); | |
1256 | u.f = float32_mul(u.f, tmp, &env->vec_status); | |
1257 | ||
1258 | return float32_to_int32(u.f, &env->vec_status); | |
1259 | } | |
1260 | ||
8e703949 | 1261 | static inline uint32_t efsctuf(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1262 | { |
1263 | CPU_FloatU u; | |
1264 | float32 tmp; | |
1265 | ||
1266 | u.l = val; | |
1267 | /* NaN are not treated the same way IEEE 754 does */ | |
1268 | if (unlikely(float32_is_quiet_nan(u.f))) { | |
1269 | return 0; | |
1270 | } | |
1271 | tmp = uint64_to_float32(1ULL << 32, &env->vec_status); | |
1272 | u.f = float32_mul(u.f, tmp, &env->vec_status); | |
1273 | ||
1274 | return float32_to_uint32(u.f, &env->vec_status); | |
1275 | } | |
1276 | ||
8e703949 BS |
1277 | #define HELPER_SPE_SINGLE_CONV(name) \ |
1278 | uint32_t helper_e##name(CPUPPCState *env, uint32_t val) \ | |
1279 | { \ | |
1280 | return e##name(env, val); \ | |
bd23cd45 BS |
1281 | } |
1282 | /* efscfsi */ | |
1283 | HELPER_SPE_SINGLE_CONV(fscfsi); | |
1284 | /* efscfui */ | |
1285 | HELPER_SPE_SINGLE_CONV(fscfui); | |
1286 | /* efscfuf */ | |
1287 | HELPER_SPE_SINGLE_CONV(fscfuf); | |
1288 | /* efscfsf */ | |
1289 | HELPER_SPE_SINGLE_CONV(fscfsf); | |
1290 | /* efsctsi */ | |
1291 | HELPER_SPE_SINGLE_CONV(fsctsi); | |
1292 | /* efsctui */ | |
1293 | HELPER_SPE_SINGLE_CONV(fsctui); | |
1294 | /* efsctsiz */ | |
1295 | HELPER_SPE_SINGLE_CONV(fsctsiz); | |
1296 | /* efsctuiz */ | |
1297 | HELPER_SPE_SINGLE_CONV(fsctuiz); | |
1298 | /* efsctsf */ | |
1299 | HELPER_SPE_SINGLE_CONV(fsctsf); | |
1300 | /* efsctuf */ | |
1301 | HELPER_SPE_SINGLE_CONV(fsctuf); | |
1302 | ||
8e703949 BS |
1303 | #define HELPER_SPE_VECTOR_CONV(name) \ |
1304 | uint64_t helper_ev##name(CPUPPCState *env, uint64_t val) \ | |
1305 | { \ | |
1306 | return ((uint64_t)e##name(env, val >> 32) << 32) | \ | |
1307 | (uint64_t)e##name(env, val); \ | |
bd23cd45 BS |
1308 | } |
1309 | /* evfscfsi */ | |
1310 | HELPER_SPE_VECTOR_CONV(fscfsi); | |
1311 | /* evfscfui */ | |
1312 | HELPER_SPE_VECTOR_CONV(fscfui); | |
1313 | /* evfscfuf */ | |
1314 | HELPER_SPE_VECTOR_CONV(fscfuf); | |
1315 | /* evfscfsf */ | |
1316 | HELPER_SPE_VECTOR_CONV(fscfsf); | |
1317 | /* evfsctsi */ | |
1318 | HELPER_SPE_VECTOR_CONV(fsctsi); | |
1319 | /* evfsctui */ | |
1320 | HELPER_SPE_VECTOR_CONV(fsctui); | |
1321 | /* evfsctsiz */ | |
1322 | HELPER_SPE_VECTOR_CONV(fsctsiz); | |
1323 | /* evfsctuiz */ | |
1324 | HELPER_SPE_VECTOR_CONV(fsctuiz); | |
1325 | /* evfsctsf */ | |
1326 | HELPER_SPE_VECTOR_CONV(fsctsf); | |
1327 | /* evfsctuf */ | |
1328 | HELPER_SPE_VECTOR_CONV(fsctuf); | |
1329 | ||
1330 | /* Single-precision floating-point arithmetic */ | |
8e703949 | 1331 | static inline uint32_t efsadd(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1332 | { |
1333 | CPU_FloatU u1, u2; | |
1334 | ||
1335 | u1.l = op1; | |
1336 | u2.l = op2; | |
1337 | u1.f = float32_add(u1.f, u2.f, &env->vec_status); | |
1338 | return u1.l; | |
1339 | } | |
1340 | ||
8e703949 | 1341 | static inline uint32_t efssub(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1342 | { |
1343 | CPU_FloatU u1, u2; | |
1344 | ||
1345 | u1.l = op1; | |
1346 | u2.l = op2; | |
1347 | u1.f = float32_sub(u1.f, u2.f, &env->vec_status); | |
1348 | return u1.l; | |
1349 | } | |
1350 | ||
8e703949 | 1351 | static inline uint32_t efsmul(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1352 | { |
1353 | CPU_FloatU u1, u2; | |
1354 | ||
1355 | u1.l = op1; | |
1356 | u2.l = op2; | |
1357 | u1.f = float32_mul(u1.f, u2.f, &env->vec_status); | |
1358 | return u1.l; | |
1359 | } | |
1360 | ||
8e703949 | 1361 | static inline uint32_t efsdiv(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1362 | { |
1363 | CPU_FloatU u1, u2; | |
1364 | ||
1365 | u1.l = op1; | |
1366 | u2.l = op2; | |
1367 | u1.f = float32_div(u1.f, u2.f, &env->vec_status); | |
1368 | return u1.l; | |
1369 | } | |
1370 | ||
8e703949 BS |
1371 | #define HELPER_SPE_SINGLE_ARITH(name) \ |
1372 | uint32_t helper_e##name(CPUPPCState *env, uint32_t op1, uint32_t op2) \ | |
1373 | { \ | |
1374 | return e##name(env, op1, op2); \ | |
bd23cd45 BS |
1375 | } |
1376 | /* efsadd */ | |
1377 | HELPER_SPE_SINGLE_ARITH(fsadd); | |
1378 | /* efssub */ | |
1379 | HELPER_SPE_SINGLE_ARITH(fssub); | |
1380 | /* efsmul */ | |
1381 | HELPER_SPE_SINGLE_ARITH(fsmul); | |
1382 | /* efsdiv */ | |
1383 | HELPER_SPE_SINGLE_ARITH(fsdiv); | |
1384 | ||
1385 | #define HELPER_SPE_VECTOR_ARITH(name) \ | |
8e703949 | 1386 | uint64_t helper_ev##name(CPUPPCState *env, uint64_t op1, uint64_t op2) \ |
bd23cd45 | 1387 | { \ |
8e703949 BS |
1388 | return ((uint64_t)e##name(env, op1 >> 32, op2 >> 32) << 32) | \ |
1389 | (uint64_t)e##name(env, op1, op2); \ | |
bd23cd45 BS |
1390 | } |
1391 | /* evfsadd */ | |
1392 | HELPER_SPE_VECTOR_ARITH(fsadd); | |
1393 | /* evfssub */ | |
1394 | HELPER_SPE_VECTOR_ARITH(fssub); | |
1395 | /* evfsmul */ | |
1396 | HELPER_SPE_VECTOR_ARITH(fsmul); | |
1397 | /* evfsdiv */ | |
1398 | HELPER_SPE_VECTOR_ARITH(fsdiv); | |
1399 | ||
1400 | /* Single-precision floating-point comparisons */ | |
8e703949 | 1401 | static inline uint32_t efscmplt(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1402 | { |
1403 | CPU_FloatU u1, u2; | |
1404 | ||
1405 | u1.l = op1; | |
1406 | u2.l = op2; | |
1407 | return float32_lt(u1.f, u2.f, &env->vec_status) ? 4 : 0; | |
1408 | } | |
1409 | ||
8e703949 | 1410 | static inline uint32_t efscmpgt(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1411 | { |
1412 | CPU_FloatU u1, u2; | |
1413 | ||
1414 | u1.l = op1; | |
1415 | u2.l = op2; | |
1416 | return float32_le(u1.f, u2.f, &env->vec_status) ? 0 : 4; | |
1417 | } | |
1418 | ||
8e703949 | 1419 | static inline uint32_t efscmpeq(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1420 | { |
1421 | CPU_FloatU u1, u2; | |
1422 | ||
1423 | u1.l = op1; | |
1424 | u2.l = op2; | |
1425 | return float32_eq(u1.f, u2.f, &env->vec_status) ? 4 : 0; | |
1426 | } | |
1427 | ||
8e703949 | 1428 | static inline uint32_t efststlt(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1429 | { |
1430 | /* XXX: TODO: ignore special values (NaN, infinites, ...) */ | |
8e703949 | 1431 | return efscmplt(env, op1, op2); |
bd23cd45 BS |
1432 | } |
1433 | ||
8e703949 | 1434 | static inline uint32_t efststgt(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1435 | { |
1436 | /* XXX: TODO: ignore special values (NaN, infinites, ...) */ | |
8e703949 | 1437 | return efscmpgt(env, op1, op2); |
bd23cd45 BS |
1438 | } |
1439 | ||
8e703949 | 1440 | static inline uint32_t efststeq(CPUPPCState *env, uint32_t op1, uint32_t op2) |
bd23cd45 BS |
1441 | { |
1442 | /* XXX: TODO: ignore special values (NaN, infinites, ...) */ | |
8e703949 | 1443 | return efscmpeq(env, op1, op2); |
bd23cd45 BS |
1444 | } |
1445 | ||
8e703949 BS |
1446 | #define HELPER_SINGLE_SPE_CMP(name) \ |
1447 | uint32_t helper_e##name(CPUPPCState *env, uint32_t op1, uint32_t op2) \ | |
1448 | { \ | |
1449 | return e##name(env, op1, op2) << 2; \ | |
bd23cd45 BS |
1450 | } |
1451 | /* efststlt */ | |
1452 | HELPER_SINGLE_SPE_CMP(fststlt); | |
1453 | /* efststgt */ | |
1454 | HELPER_SINGLE_SPE_CMP(fststgt); | |
1455 | /* efststeq */ | |
1456 | HELPER_SINGLE_SPE_CMP(fststeq); | |
1457 | /* efscmplt */ | |
1458 | HELPER_SINGLE_SPE_CMP(fscmplt); | |
1459 | /* efscmpgt */ | |
1460 | HELPER_SINGLE_SPE_CMP(fscmpgt); | |
1461 | /* efscmpeq */ | |
1462 | HELPER_SINGLE_SPE_CMP(fscmpeq); | |
1463 | ||
1464 | static inline uint32_t evcmp_merge(int t0, int t1) | |
1465 | { | |
1466 | return (t0 << 3) | (t1 << 2) | ((t0 | t1) << 1) | (t0 & t1); | |
1467 | } | |
1468 | ||
1469 | #define HELPER_VECTOR_SPE_CMP(name) \ | |
8e703949 | 1470 | uint32_t helper_ev##name(CPUPPCState *env, uint64_t op1, uint64_t op2) \ |
bd23cd45 | 1471 | { \ |
8e703949 BS |
1472 | return evcmp_merge(e##name(env, op1 >> 32, op2 >> 32), \ |
1473 | e##name(env, op1, op2)); \ | |
bd23cd45 BS |
1474 | } |
1475 | /* evfststlt */ | |
1476 | HELPER_VECTOR_SPE_CMP(fststlt); | |
1477 | /* evfststgt */ | |
1478 | HELPER_VECTOR_SPE_CMP(fststgt); | |
1479 | /* evfststeq */ | |
1480 | HELPER_VECTOR_SPE_CMP(fststeq); | |
1481 | /* evfscmplt */ | |
1482 | HELPER_VECTOR_SPE_CMP(fscmplt); | |
1483 | /* evfscmpgt */ | |
1484 | HELPER_VECTOR_SPE_CMP(fscmpgt); | |
1485 | /* evfscmpeq */ | |
1486 | HELPER_VECTOR_SPE_CMP(fscmpeq); | |
1487 | ||
1488 | /* Double-precision floating-point conversion */ | |
8e703949 | 1489 | uint64_t helper_efdcfsi(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1490 | { |
1491 | CPU_DoubleU u; | |
1492 | ||
1493 | u.d = int32_to_float64(val, &env->vec_status); | |
1494 | ||
1495 | return u.ll; | |
1496 | } | |
1497 | ||
8e703949 | 1498 | uint64_t helper_efdcfsid(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1499 | { |
1500 | CPU_DoubleU u; | |
1501 | ||
1502 | u.d = int64_to_float64(val, &env->vec_status); | |
1503 | ||
1504 | return u.ll; | |
1505 | } | |
1506 | ||
8e703949 | 1507 | uint64_t helper_efdcfui(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1508 | { |
1509 | CPU_DoubleU u; | |
1510 | ||
1511 | u.d = uint32_to_float64(val, &env->vec_status); | |
1512 | ||
1513 | return u.ll; | |
1514 | } | |
1515 | ||
8e703949 | 1516 | uint64_t helper_efdcfuid(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1517 | { |
1518 | CPU_DoubleU u; | |
1519 | ||
1520 | u.d = uint64_to_float64(val, &env->vec_status); | |
1521 | ||
1522 | return u.ll; | |
1523 | } | |
1524 | ||
8e703949 | 1525 | uint32_t helper_efdctsi(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1526 | { |
1527 | CPU_DoubleU u; | |
1528 | ||
1529 | u.ll = val; | |
1530 | /* NaN are not treated the same way IEEE 754 does */ | |
1531 | if (unlikely(float64_is_any_nan(u.d))) { | |
1532 | return 0; | |
1533 | } | |
1534 | ||
1535 | return float64_to_int32(u.d, &env->vec_status); | |
1536 | } | |
1537 | ||
8e703949 | 1538 | uint32_t helper_efdctui(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1539 | { |
1540 | CPU_DoubleU u; | |
1541 | ||
1542 | u.ll = val; | |
1543 | /* NaN are not treated the same way IEEE 754 does */ | |
1544 | if (unlikely(float64_is_any_nan(u.d))) { | |
1545 | return 0; | |
1546 | } | |
1547 | ||
1548 | return float64_to_uint32(u.d, &env->vec_status); | |
1549 | } | |
1550 | ||
8e703949 | 1551 | uint32_t helper_efdctsiz(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1552 | { |
1553 | CPU_DoubleU u; | |
1554 | ||
1555 | u.ll = val; | |
1556 | /* NaN are not treated the same way IEEE 754 does */ | |
1557 | if (unlikely(float64_is_any_nan(u.d))) { | |
1558 | return 0; | |
1559 | } | |
1560 | ||
1561 | return float64_to_int32_round_to_zero(u.d, &env->vec_status); | |
1562 | } | |
1563 | ||
8e703949 | 1564 | uint64_t helper_efdctsidz(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1565 | { |
1566 | CPU_DoubleU u; | |
1567 | ||
1568 | u.ll = val; | |
1569 | /* NaN are not treated the same way IEEE 754 does */ | |
1570 | if (unlikely(float64_is_any_nan(u.d))) { | |
1571 | return 0; | |
1572 | } | |
1573 | ||
1574 | return float64_to_int64_round_to_zero(u.d, &env->vec_status); | |
1575 | } | |
1576 | ||
8e703949 | 1577 | uint32_t helper_efdctuiz(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1578 | { |
1579 | CPU_DoubleU u; | |
1580 | ||
1581 | u.ll = val; | |
1582 | /* NaN are not treated the same way IEEE 754 does */ | |
1583 | if (unlikely(float64_is_any_nan(u.d))) { | |
1584 | return 0; | |
1585 | } | |
1586 | ||
1587 | return float64_to_uint32_round_to_zero(u.d, &env->vec_status); | |
1588 | } | |
1589 | ||
8e703949 | 1590 | uint64_t helper_efdctuidz(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1591 | { |
1592 | CPU_DoubleU u; | |
1593 | ||
1594 | u.ll = val; | |
1595 | /* NaN are not treated the same way IEEE 754 does */ | |
1596 | if (unlikely(float64_is_any_nan(u.d))) { | |
1597 | return 0; | |
1598 | } | |
1599 | ||
1600 | return float64_to_uint64_round_to_zero(u.d, &env->vec_status); | |
1601 | } | |
1602 | ||
8e703949 | 1603 | uint64_t helper_efdcfsf(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1604 | { |
1605 | CPU_DoubleU u; | |
1606 | float64 tmp; | |
1607 | ||
1608 | u.d = int32_to_float64(val, &env->vec_status); | |
1609 | tmp = int64_to_float64(1ULL << 32, &env->vec_status); | |
1610 | u.d = float64_div(u.d, tmp, &env->vec_status); | |
1611 | ||
1612 | return u.ll; | |
1613 | } | |
1614 | ||
8e703949 | 1615 | uint64_t helper_efdcfuf(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1616 | { |
1617 | CPU_DoubleU u; | |
1618 | float64 tmp; | |
1619 | ||
1620 | u.d = uint32_to_float64(val, &env->vec_status); | |
1621 | tmp = int64_to_float64(1ULL << 32, &env->vec_status); | |
1622 | u.d = float64_div(u.d, tmp, &env->vec_status); | |
1623 | ||
1624 | return u.ll; | |
1625 | } | |
1626 | ||
8e703949 | 1627 | uint32_t helper_efdctsf(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1628 | { |
1629 | CPU_DoubleU u; | |
1630 | float64 tmp; | |
1631 | ||
1632 | u.ll = val; | |
1633 | /* NaN are not treated the same way IEEE 754 does */ | |
1634 | if (unlikely(float64_is_any_nan(u.d))) { | |
1635 | return 0; | |
1636 | } | |
1637 | tmp = uint64_to_float64(1ULL << 32, &env->vec_status); | |
1638 | u.d = float64_mul(u.d, tmp, &env->vec_status); | |
1639 | ||
1640 | return float64_to_int32(u.d, &env->vec_status); | |
1641 | } | |
1642 | ||
8e703949 | 1643 | uint32_t helper_efdctuf(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1644 | { |
1645 | CPU_DoubleU u; | |
1646 | float64 tmp; | |
1647 | ||
1648 | u.ll = val; | |
1649 | /* NaN are not treated the same way IEEE 754 does */ | |
1650 | if (unlikely(float64_is_any_nan(u.d))) { | |
1651 | return 0; | |
1652 | } | |
1653 | tmp = uint64_to_float64(1ULL << 32, &env->vec_status); | |
1654 | u.d = float64_mul(u.d, tmp, &env->vec_status); | |
1655 | ||
1656 | return float64_to_uint32(u.d, &env->vec_status); | |
1657 | } | |
1658 | ||
8e703949 | 1659 | uint32_t helper_efscfd(CPUPPCState *env, uint64_t val) |
bd23cd45 BS |
1660 | { |
1661 | CPU_DoubleU u1; | |
1662 | CPU_FloatU u2; | |
1663 | ||
1664 | u1.ll = val; | |
1665 | u2.f = float64_to_float32(u1.d, &env->vec_status); | |
1666 | ||
1667 | return u2.l; | |
1668 | } | |
1669 | ||
8e703949 | 1670 | uint64_t helper_efdcfs(CPUPPCState *env, uint32_t val) |
bd23cd45 BS |
1671 | { |
1672 | CPU_DoubleU u2; | |
1673 | CPU_FloatU u1; | |
1674 | ||
1675 | u1.l = val; | |
1676 | u2.d = float32_to_float64(u1.f, &env->vec_status); | |
1677 | ||
1678 | return u2.ll; | |
1679 | } | |
1680 | ||
1681 | /* Double precision fixed-point arithmetic */ | |
8e703949 | 1682 | uint64_t helper_efdadd(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1683 | { |
1684 | CPU_DoubleU u1, u2; | |
1685 | ||
1686 | u1.ll = op1; | |
1687 | u2.ll = op2; | |
1688 | u1.d = float64_add(u1.d, u2.d, &env->vec_status); | |
1689 | return u1.ll; | |
1690 | } | |
1691 | ||
8e703949 | 1692 | uint64_t helper_efdsub(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1693 | { |
1694 | CPU_DoubleU u1, u2; | |
1695 | ||
1696 | u1.ll = op1; | |
1697 | u2.ll = op2; | |
1698 | u1.d = float64_sub(u1.d, u2.d, &env->vec_status); | |
1699 | return u1.ll; | |
1700 | } | |
1701 | ||
8e703949 | 1702 | uint64_t helper_efdmul(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1703 | { |
1704 | CPU_DoubleU u1, u2; | |
1705 | ||
1706 | u1.ll = op1; | |
1707 | u2.ll = op2; | |
1708 | u1.d = float64_mul(u1.d, u2.d, &env->vec_status); | |
1709 | return u1.ll; | |
1710 | } | |
1711 | ||
8e703949 | 1712 | uint64_t helper_efddiv(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1713 | { |
1714 | CPU_DoubleU u1, u2; | |
1715 | ||
1716 | u1.ll = op1; | |
1717 | u2.ll = op2; | |
1718 | u1.d = float64_div(u1.d, u2.d, &env->vec_status); | |
1719 | return u1.ll; | |
1720 | } | |
1721 | ||
1722 | /* Double precision floating point helpers */ | |
8e703949 | 1723 | uint32_t helper_efdtstlt(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1724 | { |
1725 | CPU_DoubleU u1, u2; | |
1726 | ||
1727 | u1.ll = op1; | |
1728 | u2.ll = op2; | |
1729 | return float64_lt(u1.d, u2.d, &env->vec_status) ? 4 : 0; | |
1730 | } | |
1731 | ||
8e703949 | 1732 | uint32_t helper_efdtstgt(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1733 | { |
1734 | CPU_DoubleU u1, u2; | |
1735 | ||
1736 | u1.ll = op1; | |
1737 | u2.ll = op2; | |
1738 | return float64_le(u1.d, u2.d, &env->vec_status) ? 0 : 4; | |
1739 | } | |
1740 | ||
8e703949 | 1741 | uint32_t helper_efdtsteq(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1742 | { |
1743 | CPU_DoubleU u1, u2; | |
1744 | ||
1745 | u1.ll = op1; | |
1746 | u2.ll = op2; | |
1747 | return float64_eq_quiet(u1.d, u2.d, &env->vec_status) ? 4 : 0; | |
1748 | } | |
1749 | ||
8e703949 | 1750 | uint32_t helper_efdcmplt(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1751 | { |
1752 | /* XXX: TODO: test special values (NaN, infinites, ...) */ | |
8e703949 | 1753 | return helper_efdtstlt(env, op1, op2); |
bd23cd45 BS |
1754 | } |
1755 | ||
8e703949 | 1756 | uint32_t helper_efdcmpgt(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1757 | { |
1758 | /* XXX: TODO: test special values (NaN, infinites, ...) */ | |
8e703949 | 1759 | return helper_efdtstgt(env, op1, op2); |
bd23cd45 BS |
1760 | } |
1761 | ||
8e703949 | 1762 | uint32_t helper_efdcmpeq(CPUPPCState *env, uint64_t op1, uint64_t op2) |
bd23cd45 BS |
1763 | { |
1764 | /* XXX: TODO: test special values (NaN, infinites, ...) */ | |
8e703949 | 1765 | return helper_efdtsteq(env, op1, op2); |
bd23cd45 | 1766 | } |
3c3cbbdc TM |
1767 | |
1768 | #define DECODE_SPLIT(opcode, shift1, nb1, shift2, nb2) \ | |
1769 | (((((opcode) >> (shift1)) & ((1 << (nb1)) - 1)) << nb2) | \ | |
1770 | (((opcode) >> (shift2)) & ((1 << (nb2)) - 1))) | |
1771 | ||
1772 | #define xT(opcode) DECODE_SPLIT(opcode, 0, 1, 21, 5) | |
1773 | #define xA(opcode) DECODE_SPLIT(opcode, 2, 1, 16, 5) | |
1774 | #define xB(opcode) DECODE_SPLIT(opcode, 1, 1, 11, 5) | |
1775 | #define xC(opcode) DECODE_SPLIT(opcode, 3, 1, 6, 5) | |
1776 | #define BF(opcode) (((opcode) >> (31-8)) & 7) | |
1777 | ||
1778 | typedef union _ppc_vsr_t { | |
1779 | uint64_t u64[2]; | |
1780 | uint32_t u32[4]; | |
1781 | float32 f32[4]; | |
1782 | float64 f64[2]; | |
1783 | } ppc_vsr_t; | |
1784 | ||
80189035 TM |
1785 | #if defined(HOST_WORDS_BIGENDIAN) |
1786 | #define VsrW(i) u32[i] | |
1787 | #define VsrD(i) u64[i] | |
1788 | #else | |
1789 | #define VsrW(i) u32[3-(i)] | |
1790 | #define VsrD(i) u64[1-(i)] | |
1791 | #endif | |
1792 | ||
3c3cbbdc TM |
1793 | static void getVSR(int n, ppc_vsr_t *vsr, CPUPPCState *env) |
1794 | { | |
1795 | if (n < 32) { | |
d359db00 TM |
1796 | vsr->VsrD(0) = env->fpr[n]; |
1797 | vsr->VsrD(1) = env->vsr[n]; | |
3c3cbbdc TM |
1798 | } else { |
1799 | vsr->u64[0] = env->avr[n-32].u64[0]; | |
1800 | vsr->u64[1] = env->avr[n-32].u64[1]; | |
1801 | } | |
1802 | } | |
1803 | ||
1804 | static void putVSR(int n, ppc_vsr_t *vsr, CPUPPCState *env) | |
1805 | { | |
1806 | if (n < 32) { | |
d359db00 TM |
1807 | env->fpr[n] = vsr->VsrD(0); |
1808 | env->vsr[n] = vsr->VsrD(1); | |
3c3cbbdc TM |
1809 | } else { |
1810 | env->avr[n-32].u64[0] = vsr->u64[0]; | |
1811 | env->avr[n-32].u64[1] = vsr->u64[1]; | |
1812 | } | |
1813 | } | |
1814 | ||
1815 | #define float64_to_float64(x, env) x | |
ee6e02c0 TM |
1816 | |
1817 | ||
1818 | /* VSX_ADD_SUB - VSX floating point add/subract | |
1819 | * name - instruction mnemonic | |
1820 | * op - operation (add or sub) | |
1821 | * nels - number of elements (1, 2 or 4) | |
1822 | * tp - type (float32 or float64) | |
bcb7652e | 1823 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
ee6e02c0 TM |
1824 | * sfprf - set FPRF |
1825 | */ | |
3fd0aadf | 1826 | #define VSX_ADD_SUB(name, op, nels, tp, fld, sfprf, r2sp) \ |
ee6e02c0 TM |
1827 | void helper_##name(CPUPPCState *env, uint32_t opcode) \ |
1828 | { \ | |
1829 | ppc_vsr_t xt, xa, xb; \ | |
1830 | int i; \ | |
1831 | \ | |
1832 | getVSR(xA(opcode), &xa, env); \ | |
1833 | getVSR(xB(opcode), &xb, env); \ | |
1834 | getVSR(xT(opcode), &xt, env); \ | |
1835 | helper_reset_fpstatus(env); \ | |
1836 | \ | |
1837 | for (i = 0; i < nels; i++) { \ | |
1838 | float_status tstat = env->fp_status; \ | |
1839 | set_float_exception_flags(0, &tstat); \ | |
bcb7652e | 1840 | xt.fld = tp##_##op(xa.fld, xb.fld, &tstat); \ |
ee6e02c0 TM |
1841 | env->fp_status.float_exception_flags |= tstat.float_exception_flags; \ |
1842 | \ | |
1843 | if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \ | |
bcb7652e | 1844 | if (tp##_is_infinity(xa.fld) && tp##_is_infinity(xb.fld)) { \ |
ee6e02c0 | 1845 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, sfprf); \ |
bcb7652e TM |
1846 | } else if (tp##_is_signaling_nan(xa.fld) || \ |
1847 | tp##_is_signaling_nan(xb.fld)) { \ | |
ee6e02c0 TM |
1848 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \ |
1849 | } \ | |
1850 | } \ | |
1851 | \ | |
3fd0aadf | 1852 | if (r2sp) { \ |
bcb7652e | 1853 | xt.fld = helper_frsp(env, xt.fld); \ |
3fd0aadf TM |
1854 | } \ |
1855 | \ | |
ee6e02c0 | 1856 | if (sfprf) { \ |
bcb7652e | 1857 | helper_compute_fprf(env, xt.fld, sfprf); \ |
ee6e02c0 TM |
1858 | } \ |
1859 | } \ | |
1860 | putVSR(xT(opcode), &xt, env); \ | |
1861 | helper_float_check_status(env); \ | |
1862 | } | |
1863 | ||
bcb7652e TM |
1864 | VSX_ADD_SUB(xsadddp, add, 1, float64, VsrD(0), 1, 0) |
1865 | VSX_ADD_SUB(xsaddsp, add, 1, float64, VsrD(0), 1, 1) | |
1866 | VSX_ADD_SUB(xvadddp, add, 2, float64, VsrD(i), 0, 0) | |
1867 | VSX_ADD_SUB(xvaddsp, add, 4, float32, VsrW(i), 0, 0) | |
1868 | VSX_ADD_SUB(xssubdp, sub, 1, float64, VsrD(0), 1, 0) | |
1869 | VSX_ADD_SUB(xssubsp, sub, 1, float64, VsrD(0), 1, 1) | |
1870 | VSX_ADD_SUB(xvsubdp, sub, 2, float64, VsrD(i), 0, 0) | |
1871 | VSX_ADD_SUB(xvsubsp, sub, 4, float32, VsrW(i), 0, 0) | |
5e591d88 TM |
1872 | |
1873 | /* VSX_MUL - VSX floating point multiply | |
1874 | * op - instruction mnemonic | |
1875 | * nels - number of elements (1, 2 or 4) | |
1876 | * tp - type (float32 or float64) | |
bcb7652e | 1877 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
5e591d88 TM |
1878 | * sfprf - set FPRF |
1879 | */ | |
ab9408a2 | 1880 | #define VSX_MUL(op, nels, tp, fld, sfprf, r2sp) \ |
5e591d88 TM |
1881 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
1882 | { \ | |
1883 | ppc_vsr_t xt, xa, xb; \ | |
1884 | int i; \ | |
1885 | \ | |
1886 | getVSR(xA(opcode), &xa, env); \ | |
1887 | getVSR(xB(opcode), &xb, env); \ | |
1888 | getVSR(xT(opcode), &xt, env); \ | |
1889 | helper_reset_fpstatus(env); \ | |
1890 | \ | |
1891 | for (i = 0; i < nels; i++) { \ | |
1892 | float_status tstat = env->fp_status; \ | |
1893 | set_float_exception_flags(0, &tstat); \ | |
bcb7652e | 1894 | xt.fld = tp##_mul(xa.fld, xb.fld, &tstat); \ |
5e591d88 TM |
1895 | env->fp_status.float_exception_flags |= tstat.float_exception_flags; \ |
1896 | \ | |
1897 | if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \ | |
bcb7652e TM |
1898 | if ((tp##_is_infinity(xa.fld) && tp##_is_zero(xb.fld)) || \ |
1899 | (tp##_is_infinity(xb.fld) && tp##_is_zero(xa.fld))) { \ | |
5e591d88 | 1900 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIMZ, sfprf); \ |
bcb7652e TM |
1901 | } else if (tp##_is_signaling_nan(xa.fld) || \ |
1902 | tp##_is_signaling_nan(xb.fld)) { \ | |
5e591d88 TM |
1903 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \ |
1904 | } \ | |
1905 | } \ | |
1906 | \ | |
ab9408a2 | 1907 | if (r2sp) { \ |
bcb7652e | 1908 | xt.fld = helper_frsp(env, xt.fld); \ |
ab9408a2 TM |
1909 | } \ |
1910 | \ | |
5e591d88 | 1911 | if (sfprf) { \ |
bcb7652e | 1912 | helper_compute_fprf(env, xt.fld, sfprf); \ |
5e591d88 TM |
1913 | } \ |
1914 | } \ | |
1915 | \ | |
1916 | putVSR(xT(opcode), &xt, env); \ | |
1917 | helper_float_check_status(env); \ | |
1918 | } | |
1919 | ||
bcb7652e TM |
1920 | VSX_MUL(xsmuldp, 1, float64, VsrD(0), 1, 0) |
1921 | VSX_MUL(xsmulsp, 1, float64, VsrD(0), 1, 1) | |
1922 | VSX_MUL(xvmuldp, 2, float64, VsrD(i), 0, 0) | |
1923 | VSX_MUL(xvmulsp, 4, float32, VsrW(i), 0, 0) | |
4b98eeef TM |
1924 | |
1925 | /* VSX_DIV - VSX floating point divide | |
1926 | * op - instruction mnemonic | |
1927 | * nels - number of elements (1, 2 or 4) | |
1928 | * tp - type (float32 or float64) | |
bcb7652e | 1929 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
4b98eeef TM |
1930 | * sfprf - set FPRF |
1931 | */ | |
b24d0b47 | 1932 | #define VSX_DIV(op, nels, tp, fld, sfprf, r2sp) \ |
4b98eeef TM |
1933 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
1934 | { \ | |
1935 | ppc_vsr_t xt, xa, xb; \ | |
1936 | int i; \ | |
1937 | \ | |
1938 | getVSR(xA(opcode), &xa, env); \ | |
1939 | getVSR(xB(opcode), &xb, env); \ | |
1940 | getVSR(xT(opcode), &xt, env); \ | |
1941 | helper_reset_fpstatus(env); \ | |
1942 | \ | |
1943 | for (i = 0; i < nels; i++) { \ | |
1944 | float_status tstat = env->fp_status; \ | |
1945 | set_float_exception_flags(0, &tstat); \ | |
bcb7652e | 1946 | xt.fld = tp##_div(xa.fld, xb.fld, &tstat); \ |
4b98eeef TM |
1947 | env->fp_status.float_exception_flags |= tstat.float_exception_flags; \ |
1948 | \ | |
1949 | if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \ | |
bcb7652e | 1950 | if (tp##_is_infinity(xa.fld) && tp##_is_infinity(xb.fld)) { \ |
4b98eeef | 1951 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXIDI, sfprf); \ |
bcb7652e TM |
1952 | } else if (tp##_is_zero(xa.fld) && \ |
1953 | tp##_is_zero(xb.fld)) { \ | |
4b98eeef | 1954 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXZDZ, sfprf); \ |
bcb7652e TM |
1955 | } else if (tp##_is_signaling_nan(xa.fld) || \ |
1956 | tp##_is_signaling_nan(xb.fld)) { \ | |
4b98eeef TM |
1957 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \ |
1958 | } \ | |
1959 | } \ | |
1960 | \ | |
b24d0b47 | 1961 | if (r2sp) { \ |
bcb7652e | 1962 | xt.fld = helper_frsp(env, xt.fld); \ |
b24d0b47 TM |
1963 | } \ |
1964 | \ | |
4b98eeef | 1965 | if (sfprf) { \ |
bcb7652e | 1966 | helper_compute_fprf(env, xt.fld, sfprf); \ |
4b98eeef TM |
1967 | } \ |
1968 | } \ | |
1969 | \ | |
1970 | putVSR(xT(opcode), &xt, env); \ | |
1971 | helper_float_check_status(env); \ | |
1972 | } | |
1973 | ||
bcb7652e TM |
1974 | VSX_DIV(xsdivdp, 1, float64, VsrD(0), 1, 0) |
1975 | VSX_DIV(xsdivsp, 1, float64, VsrD(0), 1, 1) | |
1976 | VSX_DIV(xvdivdp, 2, float64, VsrD(i), 0, 0) | |
1977 | VSX_DIV(xvdivsp, 4, float32, VsrW(i), 0, 0) | |
2009227f TM |
1978 | |
1979 | /* VSX_RE - VSX floating point reciprocal estimate | |
1980 | * op - instruction mnemonic | |
1981 | * nels - number of elements (1, 2 or 4) | |
1982 | * tp - type (float32 or float64) | |
bcb7652e | 1983 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
2009227f TM |
1984 | * sfprf - set FPRF |
1985 | */ | |
2c0c52ae | 1986 | #define VSX_RE(op, nels, tp, fld, sfprf, r2sp) \ |
2009227f TM |
1987 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
1988 | { \ | |
1989 | ppc_vsr_t xt, xb; \ | |
1990 | int i; \ | |
1991 | \ | |
1992 | getVSR(xB(opcode), &xb, env); \ | |
1993 | getVSR(xT(opcode), &xt, env); \ | |
1994 | helper_reset_fpstatus(env); \ | |
1995 | \ | |
1996 | for (i = 0; i < nels; i++) { \ | |
bcb7652e | 1997 | if (unlikely(tp##_is_signaling_nan(xb.fld))) { \ |
2009227f TM |
1998 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \ |
1999 | } \ | |
bcb7652e | 2000 | xt.fld = tp##_div(tp##_one, xb.fld, &env->fp_status); \ |
2c0c52ae TM |
2001 | \ |
2002 | if (r2sp) { \ | |
bcb7652e | 2003 | xt.fld = helper_frsp(env, xt.fld); \ |
2c0c52ae TM |
2004 | } \ |
2005 | \ | |
2009227f | 2006 | if (sfprf) { \ |
bcb7652e | 2007 | helper_compute_fprf(env, xt.fld, sfprf); \ |
2009227f TM |
2008 | } \ |
2009 | } \ | |
2010 | \ | |
2011 | putVSR(xT(opcode), &xt, env); \ | |
2012 | helper_float_check_status(env); \ | |
2013 | } | |
2014 | ||
bcb7652e TM |
2015 | VSX_RE(xsredp, 1, float64, VsrD(0), 1, 0) |
2016 | VSX_RE(xsresp, 1, float64, VsrD(0), 1, 1) | |
2017 | VSX_RE(xvredp, 2, float64, VsrD(i), 0, 0) | |
2018 | VSX_RE(xvresp, 4, float32, VsrW(i), 0, 0) | |
d32404fe TM |
2019 | |
2020 | /* VSX_SQRT - VSX floating point square root | |
2021 | * op - instruction mnemonic | |
2022 | * nels - number of elements (1, 2 or 4) | |
2023 | * tp - type (float32 or float64) | |
bcb7652e | 2024 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
d32404fe TM |
2025 | * sfprf - set FPRF |
2026 | */ | |
cea4e574 | 2027 | #define VSX_SQRT(op, nels, tp, fld, sfprf, r2sp) \ |
d32404fe TM |
2028 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
2029 | { \ | |
2030 | ppc_vsr_t xt, xb; \ | |
2031 | int i; \ | |
2032 | \ | |
2033 | getVSR(xB(opcode), &xb, env); \ | |
2034 | getVSR(xT(opcode), &xt, env); \ | |
2035 | helper_reset_fpstatus(env); \ | |
2036 | \ | |
2037 | for (i = 0; i < nels; i++) { \ | |
2038 | float_status tstat = env->fp_status; \ | |
2039 | set_float_exception_flags(0, &tstat); \ | |
bcb7652e | 2040 | xt.fld = tp##_sqrt(xb.fld, &tstat); \ |
d32404fe TM |
2041 | env->fp_status.float_exception_flags |= tstat.float_exception_flags; \ |
2042 | \ | |
2043 | if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \ | |
bcb7652e | 2044 | if (tp##_is_neg(xb.fld) && !tp##_is_zero(xb.fld)) { \ |
d32404fe | 2045 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, sfprf); \ |
bcb7652e | 2046 | } else if (tp##_is_signaling_nan(xb.fld)) { \ |
d32404fe TM |
2047 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \ |
2048 | } \ | |
2049 | } \ | |
2050 | \ | |
cea4e574 | 2051 | if (r2sp) { \ |
bcb7652e | 2052 | xt.fld = helper_frsp(env, xt.fld); \ |
cea4e574 TM |
2053 | } \ |
2054 | \ | |
d32404fe | 2055 | if (sfprf) { \ |
bcb7652e | 2056 | helper_compute_fprf(env, xt.fld, sfprf); \ |
d32404fe TM |
2057 | } \ |
2058 | } \ | |
2059 | \ | |
2060 | putVSR(xT(opcode), &xt, env); \ | |
2061 | helper_float_check_status(env); \ | |
2062 | } | |
2063 | ||
bcb7652e TM |
2064 | VSX_SQRT(xssqrtdp, 1, float64, VsrD(0), 1, 0) |
2065 | VSX_SQRT(xssqrtsp, 1, float64, VsrD(0), 1, 1) | |
2066 | VSX_SQRT(xvsqrtdp, 2, float64, VsrD(i), 0, 0) | |
2067 | VSX_SQRT(xvsqrtsp, 4, float32, VsrW(i), 0, 0) | |
d3f9df8f TM |
2068 | |
2069 | /* VSX_RSQRTE - VSX floating point reciprocal square root estimate | |
2070 | * op - instruction mnemonic | |
2071 | * nels - number of elements (1, 2 or 4) | |
2072 | * tp - type (float32 or float64) | |
bcb7652e | 2073 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
d3f9df8f TM |
2074 | * sfprf - set FPRF |
2075 | */ | |
968e76bc | 2076 | #define VSX_RSQRTE(op, nels, tp, fld, sfprf, r2sp) \ |
d3f9df8f TM |
2077 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
2078 | { \ | |
2079 | ppc_vsr_t xt, xb; \ | |
2080 | int i; \ | |
2081 | \ | |
2082 | getVSR(xB(opcode), &xb, env); \ | |
2083 | getVSR(xT(opcode), &xt, env); \ | |
2084 | helper_reset_fpstatus(env); \ | |
2085 | \ | |
2086 | for (i = 0; i < nels; i++) { \ | |
2087 | float_status tstat = env->fp_status; \ | |
2088 | set_float_exception_flags(0, &tstat); \ | |
bcb7652e TM |
2089 | xt.fld = tp##_sqrt(xb.fld, &tstat); \ |
2090 | xt.fld = tp##_div(tp##_one, xt.fld, &tstat); \ | |
d3f9df8f TM |
2091 | env->fp_status.float_exception_flags |= tstat.float_exception_flags; \ |
2092 | \ | |
2093 | if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \ | |
bcb7652e | 2094 | if (tp##_is_neg(xb.fld) && !tp##_is_zero(xb.fld)) { \ |
d3f9df8f | 2095 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSQRT, sfprf); \ |
bcb7652e | 2096 | } else if (tp##_is_signaling_nan(xb.fld)) { \ |
d3f9df8f TM |
2097 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \ |
2098 | } \ | |
2099 | } \ | |
2100 | \ | |
968e76bc | 2101 | if (r2sp) { \ |
bcb7652e | 2102 | xt.fld = helper_frsp(env, xt.fld); \ |
968e76bc TM |
2103 | } \ |
2104 | \ | |
d3f9df8f | 2105 | if (sfprf) { \ |
bcb7652e | 2106 | helper_compute_fprf(env, xt.fld, sfprf); \ |
d3f9df8f TM |
2107 | } \ |
2108 | } \ | |
2109 | \ | |
2110 | putVSR(xT(opcode), &xt, env); \ | |
2111 | helper_float_check_status(env); \ | |
2112 | } | |
2113 | ||
bcb7652e TM |
2114 | VSX_RSQRTE(xsrsqrtedp, 1, float64, VsrD(0), 1, 0) |
2115 | VSX_RSQRTE(xsrsqrtesp, 1, float64, VsrD(0), 1, 1) | |
2116 | VSX_RSQRTE(xvrsqrtedp, 2, float64, VsrD(i), 0, 0) | |
2117 | VSX_RSQRTE(xvrsqrtesp, 4, float32, VsrW(i), 0, 0) | |
bc80838f | 2118 | |
bc80838f TM |
2119 | /* VSX_TDIV - VSX floating point test for divide |
2120 | * op - instruction mnemonic | |
2121 | * nels - number of elements (1, 2 or 4) | |
2122 | * tp - type (float32 or float64) | |
bcb7652e | 2123 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
bc80838f TM |
2124 | * emin - minimum unbiased exponent |
2125 | * emax - maximum unbiased exponent | |
2126 | * nbits - number of fraction bits | |
2127 | */ | |
2128 | #define VSX_TDIV(op, nels, tp, fld, emin, emax, nbits) \ | |
2129 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ | |
2130 | { \ | |
2131 | ppc_vsr_t xa, xb; \ | |
2132 | int i; \ | |
2133 | int fe_flag = 0; \ | |
2134 | int fg_flag = 0; \ | |
2135 | \ | |
2136 | getVSR(xA(opcode), &xa, env); \ | |
2137 | getVSR(xB(opcode), &xb, env); \ | |
2138 | \ | |
2139 | for (i = 0; i < nels; i++) { \ | |
bcb7652e TM |
2140 | if (unlikely(tp##_is_infinity(xa.fld) || \ |
2141 | tp##_is_infinity(xb.fld) || \ | |
2142 | tp##_is_zero(xb.fld))) { \ | |
bc80838f TM |
2143 | fe_flag = 1; \ |
2144 | fg_flag = 1; \ | |
2145 | } else { \ | |
bcb7652e TM |
2146 | int e_a = ppc_##tp##_get_unbiased_exp(xa.fld); \ |
2147 | int e_b = ppc_##tp##_get_unbiased_exp(xb.fld); \ | |
bc80838f | 2148 | \ |
bcb7652e TM |
2149 | if (unlikely(tp##_is_any_nan(xa.fld) || \ |
2150 | tp##_is_any_nan(xb.fld))) { \ | |
bc80838f TM |
2151 | fe_flag = 1; \ |
2152 | } else if ((e_b <= emin) || (e_b >= (emax-2))) { \ | |
2153 | fe_flag = 1; \ | |
bcb7652e | 2154 | } else if (!tp##_is_zero(xa.fld) && \ |
bc80838f TM |
2155 | (((e_a - e_b) >= emax) || \ |
2156 | ((e_a - e_b) <= (emin+1)) || \ | |
2157 | (e_a <= (emin+nbits)))) { \ | |
2158 | fe_flag = 1; \ | |
2159 | } \ | |
2160 | \ | |
bcb7652e | 2161 | if (unlikely(tp##_is_zero_or_denormal(xb.fld))) { \ |
bc80838f TM |
2162 | /* XB is not zero because of the above check and */ \ |
2163 | /* so must be denormalized. */ \ | |
2164 | fg_flag = 1; \ | |
2165 | } \ | |
2166 | } \ | |
2167 | } \ | |
2168 | \ | |
2169 | env->crf[BF(opcode)] = 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); \ | |
2170 | } | |
2171 | ||
bcb7652e TM |
2172 | VSX_TDIV(xstdivdp, 1, float64, VsrD(0), -1022, 1023, 52) |
2173 | VSX_TDIV(xvtdivdp, 2, float64, VsrD(i), -1022, 1023, 52) | |
2174 | VSX_TDIV(xvtdivsp, 4, float32, VsrW(i), -126, 127, 23) | |
5cb151ac TM |
2175 | |
2176 | /* VSX_TSQRT - VSX floating point test for square root | |
2177 | * op - instruction mnemonic | |
2178 | * nels - number of elements (1, 2 or 4) | |
2179 | * tp - type (float32 or float64) | |
bcb7652e | 2180 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
5cb151ac TM |
2181 | * emin - minimum unbiased exponent |
2182 | * emax - maximum unbiased exponent | |
2183 | * nbits - number of fraction bits | |
2184 | */ | |
2185 | #define VSX_TSQRT(op, nels, tp, fld, emin, nbits) \ | |
2186 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ | |
2187 | { \ | |
2188 | ppc_vsr_t xa, xb; \ | |
2189 | int i; \ | |
2190 | int fe_flag = 0; \ | |
2191 | int fg_flag = 0; \ | |
2192 | \ | |
2193 | getVSR(xA(opcode), &xa, env); \ | |
2194 | getVSR(xB(opcode), &xb, env); \ | |
2195 | \ | |
2196 | for (i = 0; i < nels; i++) { \ | |
bcb7652e TM |
2197 | if (unlikely(tp##_is_infinity(xb.fld) || \ |
2198 | tp##_is_zero(xb.fld))) { \ | |
5cb151ac TM |
2199 | fe_flag = 1; \ |
2200 | fg_flag = 1; \ | |
2201 | } else { \ | |
bcb7652e | 2202 | int e_b = ppc_##tp##_get_unbiased_exp(xb.fld); \ |
5cb151ac | 2203 | \ |
bcb7652e | 2204 | if (unlikely(tp##_is_any_nan(xb.fld))) { \ |
5cb151ac | 2205 | fe_flag = 1; \ |
bcb7652e | 2206 | } else if (unlikely(tp##_is_zero(xb.fld))) { \ |
5cb151ac | 2207 | fe_flag = 1; \ |
bcb7652e | 2208 | } else if (unlikely(tp##_is_neg(xb.fld))) { \ |
5cb151ac | 2209 | fe_flag = 1; \ |
bcb7652e | 2210 | } else if (!tp##_is_zero(xb.fld) && \ |
5cb151ac TM |
2211 | (e_b <= (emin+nbits))) { \ |
2212 | fe_flag = 1; \ | |
2213 | } \ | |
2214 | \ | |
bcb7652e | 2215 | if (unlikely(tp##_is_zero_or_denormal(xb.fld))) { \ |
5cb151ac TM |
2216 | /* XB is not zero because of the above check and */ \ |
2217 | /* therefore must be denormalized. */ \ | |
2218 | fg_flag = 1; \ | |
2219 | } \ | |
2220 | } \ | |
2221 | } \ | |
2222 | \ | |
2223 | env->crf[BF(opcode)] = 0x8 | (fg_flag ? 4 : 0) | (fe_flag ? 2 : 0); \ | |
2224 | } | |
2225 | ||
bcb7652e TM |
2226 | VSX_TSQRT(xstsqrtdp, 1, float64, VsrD(0), -1022, 52) |
2227 | VSX_TSQRT(xvtsqrtdp, 2, float64, VsrD(i), -1022, 52) | |
2228 | VSX_TSQRT(xvtsqrtsp, 4, float32, VsrW(i), -126, 23) | |
595c6eef TM |
2229 | |
2230 | /* VSX_MADD - VSX floating point muliply/add variations | |
2231 | * op - instruction mnemonic | |
2232 | * nels - number of elements (1, 2 or 4) | |
2233 | * tp - type (float32 or float64) | |
bcb7652e | 2234 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
595c6eef TM |
2235 | * maddflgs - flags for the float*muladd routine that control the |
2236 | * various forms (madd, msub, nmadd, nmsub) | |
2237 | * afrm - A form (1=A, 0=M) | |
2238 | * sfprf - set FPRF | |
2239 | */ | |
f53f81e0 | 2240 | #define VSX_MADD(op, nels, tp, fld, maddflgs, afrm, sfprf, r2sp) \ |
595c6eef TM |
2241 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
2242 | { \ | |
2243 | ppc_vsr_t xt_in, xa, xb, xt_out; \ | |
2244 | ppc_vsr_t *b, *c; \ | |
2245 | int i; \ | |
2246 | \ | |
2247 | if (afrm) { /* AxB + T */ \ | |
2248 | b = &xb; \ | |
2249 | c = &xt_in; \ | |
2250 | } else { /* AxT + B */ \ | |
2251 | b = &xt_in; \ | |
2252 | c = &xb; \ | |
2253 | } \ | |
2254 | \ | |
2255 | getVSR(xA(opcode), &xa, env); \ | |
2256 | getVSR(xB(opcode), &xb, env); \ | |
2257 | getVSR(xT(opcode), &xt_in, env); \ | |
2258 | \ | |
2259 | xt_out = xt_in; \ | |
2260 | \ | |
2261 | helper_reset_fpstatus(env); \ | |
2262 | \ | |
2263 | for (i = 0; i < nels; i++) { \ | |
2264 | float_status tstat = env->fp_status; \ | |
2265 | set_float_exception_flags(0, &tstat); \ | |
f53f81e0 TM |
2266 | if (r2sp && (tstat.float_rounding_mode == float_round_nearest_even)) {\ |
2267 | /* Avoid double rounding errors by rounding the intermediate */ \ | |
2268 | /* result to odd. */ \ | |
2269 | set_float_rounding_mode(float_round_to_zero, &tstat); \ | |
bcb7652e | 2270 | xt_out.fld = tp##_muladd(xa.fld, b->fld, c->fld, \ |
f53f81e0 | 2271 | maddflgs, &tstat); \ |
bcb7652e | 2272 | xt_out.fld |= (get_float_exception_flags(&tstat) & \ |
f53f81e0 TM |
2273 | float_flag_inexact) != 0; \ |
2274 | } else { \ | |
bcb7652e | 2275 | xt_out.fld = tp##_muladd(xa.fld, b->fld, c->fld, \ |
f53f81e0 TM |
2276 | maddflgs, &tstat); \ |
2277 | } \ | |
595c6eef TM |
2278 | env->fp_status.float_exception_flags |= tstat.float_exception_flags; \ |
2279 | \ | |
2280 | if (unlikely(tstat.float_exception_flags & float_flag_invalid)) { \ | |
bcb7652e TM |
2281 | if (tp##_is_signaling_nan(xa.fld) || \ |
2282 | tp##_is_signaling_nan(b->fld) || \ | |
2283 | tp##_is_signaling_nan(c->fld)) { \ | |
595c6eef TM |
2284 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, sfprf); \ |
2285 | tstat.float_exception_flags &= ~float_flag_invalid; \ | |
2286 | } \ | |
bcb7652e TM |
2287 | if ((tp##_is_infinity(xa.fld) && tp##_is_zero(b->fld)) || \ |
2288 | (tp##_is_zero(xa.fld) && tp##_is_infinity(b->fld))) { \ | |
2289 | xt_out.fld = float64_to_##tp(fload_invalid_op_excp(env, \ | |
595c6eef TM |
2290 | POWERPC_EXCP_FP_VXIMZ, sfprf), &env->fp_status); \ |
2291 | tstat.float_exception_flags &= ~float_flag_invalid; \ | |
2292 | } \ | |
2293 | if ((tstat.float_exception_flags & float_flag_invalid) && \ | |
bcb7652e TM |
2294 | ((tp##_is_infinity(xa.fld) || \ |
2295 | tp##_is_infinity(b->fld)) && \ | |
2296 | tp##_is_infinity(c->fld))) { \ | |
595c6eef TM |
2297 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, sfprf); \ |
2298 | } \ | |
2299 | } \ | |
f53f81e0 TM |
2300 | \ |
2301 | if (r2sp) { \ | |
bcb7652e | 2302 | xt_out.fld = helper_frsp(env, xt_out.fld); \ |
f53f81e0 TM |
2303 | } \ |
2304 | \ | |
595c6eef | 2305 | if (sfprf) { \ |
bcb7652e | 2306 | helper_compute_fprf(env, xt_out.fld, sfprf); \ |
595c6eef TM |
2307 | } \ |
2308 | } \ | |
2309 | putVSR(xT(opcode), &xt_out, env); \ | |
2310 | helper_float_check_status(env); \ | |
2311 | } | |
2312 | ||
2313 | #define MADD_FLGS 0 | |
2314 | #define MSUB_FLGS float_muladd_negate_c | |
2315 | #define NMADD_FLGS float_muladd_negate_result | |
2316 | #define NMSUB_FLGS (float_muladd_negate_c | float_muladd_negate_result) | |
2317 | ||
bcb7652e TM |
2318 | VSX_MADD(xsmaddadp, 1, float64, VsrD(0), MADD_FLGS, 1, 1, 0) |
2319 | VSX_MADD(xsmaddmdp, 1, float64, VsrD(0), MADD_FLGS, 0, 1, 0) | |
2320 | VSX_MADD(xsmsubadp, 1, float64, VsrD(0), MSUB_FLGS, 1, 1, 0) | |
2321 | VSX_MADD(xsmsubmdp, 1, float64, VsrD(0), MSUB_FLGS, 0, 1, 0) | |
2322 | VSX_MADD(xsnmaddadp, 1, float64, VsrD(0), NMADD_FLGS, 1, 1, 0) | |
2323 | VSX_MADD(xsnmaddmdp, 1, float64, VsrD(0), NMADD_FLGS, 0, 1, 0) | |
2324 | VSX_MADD(xsnmsubadp, 1, float64, VsrD(0), NMSUB_FLGS, 1, 1, 0) | |
2325 | VSX_MADD(xsnmsubmdp, 1, float64, VsrD(0), NMSUB_FLGS, 0, 1, 0) | |
2326 | ||
2327 | VSX_MADD(xsmaddasp, 1, float64, VsrD(0), MADD_FLGS, 1, 1, 1) | |
2328 | VSX_MADD(xsmaddmsp, 1, float64, VsrD(0), MADD_FLGS, 0, 1, 1) | |
2329 | VSX_MADD(xsmsubasp, 1, float64, VsrD(0), MSUB_FLGS, 1, 1, 1) | |
2330 | VSX_MADD(xsmsubmsp, 1, float64, VsrD(0), MSUB_FLGS, 0, 1, 1) | |
2331 | VSX_MADD(xsnmaddasp, 1, float64, VsrD(0), NMADD_FLGS, 1, 1, 1) | |
2332 | VSX_MADD(xsnmaddmsp, 1, float64, VsrD(0), NMADD_FLGS, 0, 1, 1) | |
2333 | VSX_MADD(xsnmsubasp, 1, float64, VsrD(0), NMSUB_FLGS, 1, 1, 1) | |
2334 | VSX_MADD(xsnmsubmsp, 1, float64, VsrD(0), NMSUB_FLGS, 0, 1, 1) | |
2335 | ||
2336 | VSX_MADD(xvmaddadp, 2, float64, VsrD(i), MADD_FLGS, 1, 0, 0) | |
2337 | VSX_MADD(xvmaddmdp, 2, float64, VsrD(i), MADD_FLGS, 0, 0, 0) | |
2338 | VSX_MADD(xvmsubadp, 2, float64, VsrD(i), MSUB_FLGS, 1, 0, 0) | |
2339 | VSX_MADD(xvmsubmdp, 2, float64, VsrD(i), MSUB_FLGS, 0, 0, 0) | |
2340 | VSX_MADD(xvnmaddadp, 2, float64, VsrD(i), NMADD_FLGS, 1, 0, 0) | |
2341 | VSX_MADD(xvnmaddmdp, 2, float64, VsrD(i), NMADD_FLGS, 0, 0, 0) | |
2342 | VSX_MADD(xvnmsubadp, 2, float64, VsrD(i), NMSUB_FLGS, 1, 0, 0) | |
2343 | VSX_MADD(xvnmsubmdp, 2, float64, VsrD(i), NMSUB_FLGS, 0, 0, 0) | |
2344 | ||
2345 | VSX_MADD(xvmaddasp, 4, float32, VsrW(i), MADD_FLGS, 1, 0, 0) | |
2346 | VSX_MADD(xvmaddmsp, 4, float32, VsrW(i), MADD_FLGS, 0, 0, 0) | |
2347 | VSX_MADD(xvmsubasp, 4, float32, VsrW(i), MSUB_FLGS, 1, 0, 0) | |
2348 | VSX_MADD(xvmsubmsp, 4, float32, VsrW(i), MSUB_FLGS, 0, 0, 0) | |
2349 | VSX_MADD(xvnmaddasp, 4, float32, VsrW(i), NMADD_FLGS, 1, 0, 0) | |
2350 | VSX_MADD(xvnmaddmsp, 4, float32, VsrW(i), NMADD_FLGS, 0, 0, 0) | |
2351 | VSX_MADD(xvnmsubasp, 4, float32, VsrW(i), NMSUB_FLGS, 1, 0, 0) | |
2352 | VSX_MADD(xvnmsubmsp, 4, float32, VsrW(i), NMSUB_FLGS, 0, 0, 0) | |
4f17e9c7 TM |
2353 | |
2354 | #define VSX_SCALAR_CMP(op, ordered) \ | |
2355 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ | |
2356 | { \ | |
2357 | ppc_vsr_t xa, xb; \ | |
2358 | uint32_t cc = 0; \ | |
2359 | \ | |
2360 | getVSR(xA(opcode), &xa, env); \ | |
2361 | getVSR(xB(opcode), &xb, env); \ | |
2362 | \ | |
50fc89e7 TM |
2363 | if (unlikely(float64_is_any_nan(xa.VsrD(0)) || \ |
2364 | float64_is_any_nan(xb.VsrD(0)))) { \ | |
2365 | if (float64_is_signaling_nan(xa.VsrD(0)) || \ | |
2366 | float64_is_signaling_nan(xb.VsrD(0))) { \ | |
4f17e9c7 TM |
2367 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \ |
2368 | } \ | |
2369 | if (ordered) { \ | |
2370 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXVC, 0); \ | |
2371 | } \ | |
2372 | cc = 1; \ | |
2373 | } else { \ | |
50fc89e7 | 2374 | if (float64_lt(xa.VsrD(0), xb.VsrD(0), &env->fp_status)) { \ |
4f17e9c7 | 2375 | cc = 8; \ |
50fc89e7 TM |
2376 | } else if (!float64_le(xa.VsrD(0), xb.VsrD(0), \ |
2377 | &env->fp_status)) { \ | |
4f17e9c7 TM |
2378 | cc = 4; \ |
2379 | } else { \ | |
2380 | cc = 2; \ | |
2381 | } \ | |
2382 | } \ | |
2383 | \ | |
2384 | env->fpscr &= ~(0x0F << FPSCR_FPRF); \ | |
2385 | env->fpscr |= cc << FPSCR_FPRF; \ | |
2386 | env->crf[BF(opcode)] = cc; \ | |
2387 | \ | |
2388 | helper_float_check_status(env); \ | |
2389 | } | |
2390 | ||
2391 | VSX_SCALAR_CMP(xscmpodp, 1) | |
2392 | VSX_SCALAR_CMP(xscmpudp, 0) | |
959e9c9d | 2393 | |
7dff9abe | 2394 | #define float64_snan_to_qnan(x) ((x) | 0x0008000000000000ULL) |
959e9c9d TM |
2395 | #define float32_snan_to_qnan(x) ((x) | 0x00400000) |
2396 | ||
2397 | /* VSX_MAX_MIN - VSX floating point maximum/minimum | |
2398 | * name - instruction mnemonic | |
2399 | * op - operation (max or min) | |
2400 | * nels - number of elements (1, 2 or 4) | |
2401 | * tp - type (float32 or float64) | |
bcb7652e | 2402 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
959e9c9d TM |
2403 | */ |
2404 | #define VSX_MAX_MIN(name, op, nels, tp, fld) \ | |
2405 | void helper_##name(CPUPPCState *env, uint32_t opcode) \ | |
2406 | { \ | |
2407 | ppc_vsr_t xt, xa, xb; \ | |
2408 | int i; \ | |
2409 | \ | |
2410 | getVSR(xA(opcode), &xa, env); \ | |
2411 | getVSR(xB(opcode), &xb, env); \ | |
2412 | getVSR(xT(opcode), &xt, env); \ | |
2413 | \ | |
2414 | for (i = 0; i < nels; i++) { \ | |
bcb7652e TM |
2415 | xt.fld = tp##_##op(xa.fld, xb.fld, &env->fp_status); \ |
2416 | if (unlikely(tp##_is_signaling_nan(xa.fld) || \ | |
2417 | tp##_is_signaling_nan(xb.fld))) { \ | |
959e9c9d TM |
2418 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \ |
2419 | } \ | |
2420 | } \ | |
2421 | \ | |
2422 | putVSR(xT(opcode), &xt, env); \ | |
2423 | helper_float_check_status(env); \ | |
2424 | } | |
2425 | ||
bcb7652e TM |
2426 | VSX_MAX_MIN(xsmaxdp, maxnum, 1, float64, VsrD(0)) |
2427 | VSX_MAX_MIN(xvmaxdp, maxnum, 2, float64, VsrD(i)) | |
2428 | VSX_MAX_MIN(xvmaxsp, maxnum, 4, float32, VsrW(i)) | |
2429 | VSX_MAX_MIN(xsmindp, minnum, 1, float64, VsrD(0)) | |
2430 | VSX_MAX_MIN(xvmindp, minnum, 2, float64, VsrD(i)) | |
2431 | VSX_MAX_MIN(xvminsp, minnum, 4, float32, VsrW(i)) | |
354a6dec TM |
2432 | |
2433 | /* VSX_CMP - VSX floating point compare | |
2434 | * op - instruction mnemonic | |
2435 | * nels - number of elements (1, 2 or 4) | |
2436 | * tp - type (float32 or float64) | |
bcb7652e | 2437 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
354a6dec TM |
2438 | * cmp - comparison operation |
2439 | * svxvc - set VXVC bit | |
2440 | */ | |
2441 | #define VSX_CMP(op, nels, tp, fld, cmp, svxvc) \ | |
2442 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ | |
2443 | { \ | |
2444 | ppc_vsr_t xt, xa, xb; \ | |
2445 | int i; \ | |
2446 | int all_true = 1; \ | |
2447 | int all_false = 1; \ | |
2448 | \ | |
2449 | getVSR(xA(opcode), &xa, env); \ | |
2450 | getVSR(xB(opcode), &xb, env); \ | |
2451 | getVSR(xT(opcode), &xt, env); \ | |
2452 | \ | |
2453 | for (i = 0; i < nels; i++) { \ | |
bcb7652e TM |
2454 | if (unlikely(tp##_is_any_nan(xa.fld) || \ |
2455 | tp##_is_any_nan(xb.fld))) { \ | |
2456 | if (tp##_is_signaling_nan(xa.fld) || \ | |
2457 | tp##_is_signaling_nan(xb.fld)) { \ | |
354a6dec TM |
2458 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \ |
2459 | } \ | |
2460 | if (svxvc) { \ | |
2461 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXVC, 0); \ | |
2462 | } \ | |
bcb7652e | 2463 | xt.fld = 0; \ |
354a6dec TM |
2464 | all_true = 0; \ |
2465 | } else { \ | |
bcb7652e TM |
2466 | if (tp##_##cmp(xb.fld, xa.fld, &env->fp_status) == 1) { \ |
2467 | xt.fld = -1; \ | |
354a6dec TM |
2468 | all_false = 0; \ |
2469 | } else { \ | |
bcb7652e | 2470 | xt.fld = 0; \ |
354a6dec TM |
2471 | all_true = 0; \ |
2472 | } \ | |
2473 | } \ | |
2474 | } \ | |
2475 | \ | |
2476 | putVSR(xT(opcode), &xt, env); \ | |
2477 | if ((opcode >> (31-21)) & 1) { \ | |
2478 | env->crf[6] = (all_true ? 0x8 : 0) | (all_false ? 0x2 : 0); \ | |
2479 | } \ | |
2480 | helper_float_check_status(env); \ | |
2481 | } | |
2482 | ||
bcb7652e TM |
2483 | VSX_CMP(xvcmpeqdp, 2, float64, VsrD(i), eq, 0) |
2484 | VSX_CMP(xvcmpgedp, 2, float64, VsrD(i), le, 1) | |
2485 | VSX_CMP(xvcmpgtdp, 2, float64, VsrD(i), lt, 1) | |
2486 | VSX_CMP(xvcmpeqsp, 4, float32, VsrW(i), eq, 0) | |
2487 | VSX_CMP(xvcmpgesp, 4, float32, VsrW(i), le, 1) | |
2488 | VSX_CMP(xvcmpgtsp, 4, float32, VsrW(i), lt, 1) | |
ed8ac568 | 2489 | |
ed8ac568 TM |
2490 | /* VSX_CVT_FP_TO_FP - VSX floating point/floating point conversion |
2491 | * op - instruction mnemonic | |
2492 | * nels - number of elements (1, 2 or 4) | |
2493 | * stp - source type (float32 or float64) | |
2494 | * ttp - target type (float32 or float64) | |
2495 | * sfld - source vsr_t field | |
2496 | * tfld - target vsr_t field (f32 or f64) | |
2497 | * sfprf - set FPRF | |
2498 | */ | |
2499 | #define VSX_CVT_FP_TO_FP(op, nels, stp, ttp, sfld, tfld, sfprf) \ | |
2500 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ | |
2501 | { \ | |
2502 | ppc_vsr_t xt, xb; \ | |
2503 | int i; \ | |
2504 | \ | |
2505 | getVSR(xB(opcode), &xb, env); \ | |
2506 | getVSR(xT(opcode), &xt, env); \ | |
2507 | \ | |
2508 | for (i = 0; i < nels; i++) { \ | |
ed8ac568 TM |
2509 | xt.tfld = stp##_to_##ttp(xb.sfld, &env->fp_status); \ |
2510 | if (unlikely(stp##_is_signaling_nan(xb.sfld))) { \ | |
2511 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \ | |
2512 | xt.tfld = ttp##_snan_to_qnan(xt.tfld); \ | |
2513 | } \ | |
2514 | if (sfprf) { \ | |
2515 | helper_compute_fprf(env, ttp##_to_float64(xt.tfld, \ | |
2516 | &env->fp_status), sfprf); \ | |
2517 | } \ | |
2518 | } \ | |
2519 | \ | |
2520 | putVSR(xT(opcode), &xt, env); \ | |
2521 | helper_float_check_status(env); \ | |
2522 | } | |
2523 | ||
6bbad7a9 TM |
2524 | VSX_CVT_FP_TO_FP(xscvdpsp, 1, float64, float32, VsrD(0), VsrW(0), 1) |
2525 | VSX_CVT_FP_TO_FP(xscvspdp, 1, float32, float64, VsrW(0), VsrD(0), 1) | |
2526 | VSX_CVT_FP_TO_FP(xvcvdpsp, 2, float64, float32, VsrD(i), VsrW(2*i), 0) | |
2527 | VSX_CVT_FP_TO_FP(xvcvspdp, 2, float32, float64, VsrW(2*i), VsrD(i), 0) | |
5177d2ca | 2528 | |
7ee19fb9 TM |
2529 | uint64_t helper_xscvdpspn(CPUPPCState *env, uint64_t xb) |
2530 | { | |
2531 | float_status tstat = env->fp_status; | |
2532 | set_float_exception_flags(0, &tstat); | |
2533 | ||
2534 | return (uint64_t)float64_to_float32(xb, &tstat) << 32; | |
2535 | } | |
2536 | ||
2537 | uint64_t helper_xscvspdpn(CPUPPCState *env, uint64_t xb) | |
2538 | { | |
2539 | float_status tstat = env->fp_status; | |
2540 | set_float_exception_flags(0, &tstat); | |
2541 | ||
2542 | return float32_to_float64(xb >> 32, &tstat); | |
2543 | } | |
2544 | ||
5177d2ca TM |
2545 | /* VSX_CVT_FP_TO_INT - VSX floating point to integer conversion |
2546 | * op - instruction mnemonic | |
2547 | * nels - number of elements (1, 2 or 4) | |
2548 | * stp - source type (float32 or float64) | |
2549 | * ttp - target type (int32, uint32, int64 or uint64) | |
2550 | * sfld - source vsr_t field | |
2551 | * tfld - target vsr_t field | |
5177d2ca TM |
2552 | * rnan - resulting NaN |
2553 | */ | |
d1dec5ef | 2554 | #define VSX_CVT_FP_TO_INT(op, nels, stp, ttp, sfld, tfld, rnan) \ |
5177d2ca TM |
2555 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
2556 | { \ | |
2557 | ppc_vsr_t xt, xb; \ | |
2558 | int i; \ | |
2559 | \ | |
2560 | getVSR(xB(opcode), &xb, env); \ | |
2561 | getVSR(xT(opcode), &xt, env); \ | |
2562 | \ | |
2563 | for (i = 0; i < nels; i++) { \ | |
5177d2ca TM |
2564 | if (unlikely(stp##_is_any_nan(xb.sfld))) { \ |
2565 | if (stp##_is_signaling_nan(xb.sfld)) { \ | |
2566 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \ | |
2567 | } \ | |
2568 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXCVI, 0); \ | |
2569 | xt.tfld = rnan; \ | |
2570 | } else { \ | |
0453099b TM |
2571 | xt.tfld = stp##_to_##ttp##_round_to_zero(xb.sfld, \ |
2572 | &env->fp_status); \ | |
5177d2ca TM |
2573 | if (env->fp_status.float_exception_flags & float_flag_invalid) { \ |
2574 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXCVI, 0); \ | |
2575 | } \ | |
2576 | } \ | |
2577 | } \ | |
2578 | \ | |
2579 | putVSR(xT(opcode), &xt, env); \ | |
2580 | helper_float_check_status(env); \ | |
2581 | } | |
2582 | ||
d1dec5ef | 2583 | VSX_CVT_FP_TO_INT(xscvdpsxds, 1, float64, int64, VsrD(0), VsrD(0), \ |
7dff9abe | 2584 | 0x8000000000000000ULL) |
d1dec5ef TM |
2585 | VSX_CVT_FP_TO_INT(xscvdpsxws, 1, float64, int32, VsrD(0), VsrW(1), \ |
2586 | 0x80000000U) | |
2587 | VSX_CVT_FP_TO_INT(xscvdpuxds, 1, float64, uint64, VsrD(0), VsrD(0), 0ULL) | |
2588 | VSX_CVT_FP_TO_INT(xscvdpuxws, 1, float64, uint32, VsrD(0), VsrW(1), 0U) | |
2589 | VSX_CVT_FP_TO_INT(xvcvdpsxds, 2, float64, int64, VsrD(i), VsrD(i), \ | |
7dff9abe | 2590 | 0x8000000000000000ULL) |
d1dec5ef | 2591 | VSX_CVT_FP_TO_INT(xvcvdpsxws, 2, float64, int32, VsrD(i), VsrW(2*i), \ |
7dff9abe | 2592 | 0x80000000U) |
d1dec5ef TM |
2593 | VSX_CVT_FP_TO_INT(xvcvdpuxds, 2, float64, uint64, VsrD(i), VsrD(i), 0ULL) |
2594 | VSX_CVT_FP_TO_INT(xvcvdpuxws, 2, float64, uint32, VsrD(i), VsrW(2*i), 0U) | |
2595 | VSX_CVT_FP_TO_INT(xvcvspsxds, 2, float32, int64, VsrW(2*i), VsrD(i), \ | |
2596 | 0x8000000000000000ULL) | |
2597 | VSX_CVT_FP_TO_INT(xvcvspsxws, 4, float32, int32, VsrW(i), VsrW(i), 0x80000000U) | |
2598 | VSX_CVT_FP_TO_INT(xvcvspuxds, 2, float32, uint64, VsrW(2*i), VsrD(i), 0ULL) | |
2599 | VSX_CVT_FP_TO_INT(xvcvspuxws, 4, float32, uint32, VsrW(i), VsrW(i), 0U) | |
5177d2ca TM |
2600 | |
2601 | /* VSX_CVT_INT_TO_FP - VSX integer to floating point conversion | |
2602 | * op - instruction mnemonic | |
2603 | * nels - number of elements (1, 2 or 4) | |
2604 | * stp - source type (int32, uint32, int64 or uint64) | |
2605 | * ttp - target type (float32 or float64) | |
2606 | * sfld - source vsr_t field | |
2607 | * tfld - target vsr_t field | |
2608 | * jdef - definition of the j index (i or 2*i) | |
2609 | * sfprf - set FPRF | |
2610 | */ | |
6cd7db3d | 2611 | #define VSX_CVT_INT_TO_FP(op, nels, stp, ttp, sfld, tfld, sfprf, r2sp) \ |
5177d2ca TM |
2612 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ |
2613 | { \ | |
2614 | ppc_vsr_t xt, xb; \ | |
2615 | int i; \ | |
2616 | \ | |
2617 | getVSR(xB(opcode), &xb, env); \ | |
2618 | getVSR(xT(opcode), &xt, env); \ | |
2619 | \ | |
2620 | for (i = 0; i < nels; i++) { \ | |
5177d2ca | 2621 | xt.tfld = stp##_to_##ttp(xb.sfld, &env->fp_status); \ |
74698350 TM |
2622 | if (r2sp) { \ |
2623 | xt.tfld = helper_frsp(env, xt.tfld); \ | |
2624 | } \ | |
5177d2ca TM |
2625 | if (sfprf) { \ |
2626 | helper_compute_fprf(env, xt.tfld, sfprf); \ | |
2627 | } \ | |
2628 | } \ | |
2629 | \ | |
2630 | putVSR(xT(opcode), &xt, env); \ | |
2631 | helper_float_check_status(env); \ | |
2632 | } | |
2633 | ||
6cd7db3d TM |
2634 | VSX_CVT_INT_TO_FP(xscvsxddp, 1, int64, float64, VsrD(0), VsrD(0), 1, 0) |
2635 | VSX_CVT_INT_TO_FP(xscvuxddp, 1, uint64, float64, VsrD(0), VsrD(0), 1, 0) | |
2636 | VSX_CVT_INT_TO_FP(xscvsxdsp, 1, int64, float64, VsrD(0), VsrD(0), 1, 1) | |
2637 | VSX_CVT_INT_TO_FP(xscvuxdsp, 1, uint64, float64, VsrD(0), VsrD(0), 1, 1) | |
2638 | VSX_CVT_INT_TO_FP(xvcvsxddp, 2, int64, float64, VsrD(i), VsrD(i), 0, 0) | |
2639 | VSX_CVT_INT_TO_FP(xvcvuxddp, 2, uint64, float64, VsrD(i), VsrD(i), 0, 0) | |
2640 | VSX_CVT_INT_TO_FP(xvcvsxwdp, 2, int32, float64, VsrW(2*i), VsrD(i), 0, 0) | |
2641 | VSX_CVT_INT_TO_FP(xvcvuxwdp, 2, uint64, float64, VsrW(2*i), VsrD(i), 0, 0) | |
2642 | VSX_CVT_INT_TO_FP(xvcvsxdsp, 2, int64, float32, VsrD(i), VsrW(2*i), 0, 0) | |
2643 | VSX_CVT_INT_TO_FP(xvcvuxdsp, 2, uint64, float32, VsrD(i), VsrW(2*i), 0, 0) | |
2644 | VSX_CVT_INT_TO_FP(xvcvsxwsp, 4, int32, float32, VsrW(i), VsrW(i), 0, 0) | |
2645 | VSX_CVT_INT_TO_FP(xvcvuxwsp, 4, uint32, float32, VsrW(i), VsrW(i), 0, 0) | |
88e33d08 TM |
2646 | |
2647 | /* For "use current rounding mode", define a value that will not be one of | |
2648 | * the existing rounding model enums. | |
2649 | */ | |
2650 | #define FLOAT_ROUND_CURRENT (float_round_nearest_even + float_round_down + \ | |
2651 | float_round_up + float_round_to_zero) | |
2652 | ||
2653 | /* VSX_ROUND - VSX floating point round | |
2654 | * op - instruction mnemonic | |
2655 | * nels - number of elements (1, 2 or 4) | |
2656 | * tp - type (float32 or float64) | |
bcb7652e | 2657 | * fld - vsr_t field (VsrD(*) or VsrW(*)) |
88e33d08 TM |
2658 | * rmode - rounding mode |
2659 | * sfprf - set FPRF | |
2660 | */ | |
2661 | #define VSX_ROUND(op, nels, tp, fld, rmode, sfprf) \ | |
2662 | void helper_##op(CPUPPCState *env, uint32_t opcode) \ | |
2663 | { \ | |
2664 | ppc_vsr_t xt, xb; \ | |
2665 | int i; \ | |
2666 | getVSR(xB(opcode), &xb, env); \ | |
2667 | getVSR(xT(opcode), &xt, env); \ | |
2668 | \ | |
2669 | if (rmode != FLOAT_ROUND_CURRENT) { \ | |
2670 | set_float_rounding_mode(rmode, &env->fp_status); \ | |
2671 | } \ | |
2672 | \ | |
2673 | for (i = 0; i < nels; i++) { \ | |
bcb7652e | 2674 | if (unlikely(tp##_is_signaling_nan(xb.fld))) { \ |
88e33d08 | 2675 | fload_invalid_op_excp(env, POWERPC_EXCP_FP_VXSNAN, 0); \ |
bcb7652e | 2676 | xt.fld = tp##_snan_to_qnan(xb.fld); \ |
88e33d08 | 2677 | } else { \ |
bcb7652e | 2678 | xt.fld = tp##_round_to_int(xb.fld, &env->fp_status); \ |
88e33d08 TM |
2679 | } \ |
2680 | if (sfprf) { \ | |
bcb7652e | 2681 | helper_compute_fprf(env, xt.fld, sfprf); \ |
88e33d08 TM |
2682 | } \ |
2683 | } \ | |
2684 | \ | |
2685 | /* If this is not a "use current rounding mode" instruction, \ | |
2686 | * then inhibit setting of the XX bit and restore rounding \ | |
2687 | * mode from FPSCR */ \ | |
2688 | if (rmode != FLOAT_ROUND_CURRENT) { \ | |
2689 | fpscr_set_rounding_mode(env); \ | |
2690 | env->fp_status.float_exception_flags &= ~float_flag_inexact; \ | |
2691 | } \ | |
2692 | \ | |
2693 | putVSR(xT(opcode), &xt, env); \ | |
2694 | helper_float_check_status(env); \ | |
2695 | } | |
2696 | ||
bcb7652e TM |
2697 | VSX_ROUND(xsrdpi, 1, float64, VsrD(0), float_round_nearest_even, 1) |
2698 | VSX_ROUND(xsrdpic, 1, float64, VsrD(0), FLOAT_ROUND_CURRENT, 1) | |
2699 | VSX_ROUND(xsrdpim, 1, float64, VsrD(0), float_round_down, 1) | |
2700 | VSX_ROUND(xsrdpip, 1, float64, VsrD(0), float_round_up, 1) | |
2701 | VSX_ROUND(xsrdpiz, 1, float64, VsrD(0), float_round_to_zero, 1) | |
88e33d08 | 2702 | |
bcb7652e TM |
2703 | VSX_ROUND(xvrdpi, 2, float64, VsrD(i), float_round_nearest_even, 0) |
2704 | VSX_ROUND(xvrdpic, 2, float64, VsrD(i), FLOAT_ROUND_CURRENT, 0) | |
2705 | VSX_ROUND(xvrdpim, 2, float64, VsrD(i), float_round_down, 0) | |
2706 | VSX_ROUND(xvrdpip, 2, float64, VsrD(i), float_round_up, 0) | |
2707 | VSX_ROUND(xvrdpiz, 2, float64, VsrD(i), float_round_to_zero, 0) | |
88e33d08 | 2708 | |
bcb7652e TM |
2709 | VSX_ROUND(xvrspi, 4, float32, VsrW(i), float_round_nearest_even, 0) |
2710 | VSX_ROUND(xvrspic, 4, float32, VsrW(i), FLOAT_ROUND_CURRENT, 0) | |
2711 | VSX_ROUND(xvrspim, 4, float32, VsrW(i), float_round_down, 0) | |
2712 | VSX_ROUND(xvrspip, 4, float32, VsrW(i), float_round_up, 0) | |
2713 | VSX_ROUND(xvrspiz, 4, float32, VsrW(i), float_round_to_zero, 0) | |
3d1140bf TM |
2714 | |
2715 | uint64_t helper_xsrsp(CPUPPCState *env, uint64_t xb) | |
2716 | { | |
2717 | helper_reset_fpstatus(env); | |
2718 | ||
2719 | uint64_t xt = helper_frsp(env, xb); | |
2720 | ||
2721 | helper_compute_fprf(env, xt, 1); | |
2722 | helper_float_check_status(env); | |
2723 | return xt; | |
2724 | } |