1/*
2 * MIPS SIMD Architecture Module Instruction emulation helpers for QEMU.
3 *
4 * Copyright (c) 2014 Imagination Technologies
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
20#include "qemu/osdep.h"
21#include "cpu.h"
22#include "internal.h"
23#include "exec/exec-all.h"
24#include "exec/helper-proto.h"
25#include "fpu/softfloat.h"
26
27/* Data format min and max values */
28#define DF_BITS(df) (1 << ((df) + 3))
29
30#define DF_MAX_INT(df) (int64_t)((1LL << (DF_BITS(df) - 1)) - 1)
31#define M_MAX_INT(m) (int64_t)((1LL << ((m) - 1)) - 1)
32
33#define DF_MIN_INT(df) (int64_t)(-(1LL << (DF_BITS(df) - 1)))
34#define M_MIN_INT(m) (int64_t)(-(1LL << ((m) - 1)))
35
36#define DF_MAX_UINT(df) (uint64_t)(-1ULL >> (64 - DF_BITS(df)))
37#define M_MAX_UINT(m) (uint64_t)(-1ULL >> (64 - (m)))
38
39#define UNSIGNED(x, df) ((x) & DF_MAX_UINT(df))
40#define SIGNED(x, df) \
41 ((((int64_t)x) << (64 - DF_BITS(df))) >> (64 - DF_BITS(df)))
42
43/* Element-by-element access macros */
44#define DF_ELEMENTS(df) (MSA_WRLEN / DF_BITS(df))
45
46
47
48/*
49 * Bit Count
50 * ---------
51 *
52 * +---------------+----------------------------------------------------------+
53 * | NLOC.B | Vector Leading Ones Count (byte) |
54 * | NLOC.H | Vector Leading Ones Count (halfword) |
55 * | NLOC.W | Vector Leading Ones Count (word) |
56 * | NLOC.D | Vector Leading Ones Count (doubleword) |
57 * | NLZC.B | Vector Leading Zeros Count (byte) |
58 * | NLZC.H | Vector Leading Zeros Count (halfword) |
59 * | NLZC.W | Vector Leading Zeros Count (word) |
60 * | NLZC.D | Vector Leading Zeros Count (doubleword) |
61 * | PCNT.B | Vector Population Count (byte) |
62 * | PCNT.H | Vector Population Count (halfword) |
63 * | PCNT.W | Vector Population Count (word) |
64 * | PCNT.D | Vector Population Count (doubleword) |
65 * +---------------+----------------------------------------------------------+
66 */
67
68/* TODO: insert Bit Count group helpers here */
69
70
71/*
72 * Bit Move
73 * --------
74 *
75 * +---------------+----------------------------------------------------------+
76 * | BINSL.B | Vector Bit Insert Left (byte) |
77 * | BINSL.H | Vector Bit Insert Left (halfword) |
78 * | BINSL.W | Vector Bit Insert Left (word) |
79 * | BINSL.D | Vector Bit Insert Left (doubleword) |
80 * | BINSR.B | Vector Bit Insert Right (byte) |
81 * | BINSR.H | Vector Bit Insert Right (halfword) |
82 * | BINSR.W | Vector Bit Insert Right (word) |
83 * | BINSR.D | Vector Bit Insert Right (doubleword) |
84 * | BMNZ.V | Vector Bit Move If Not Zero |
85 * | BMZ.V | Vector Bit Move If Zero |
86 * | BSEL.V | Vector Bit Select |
87 * +---------------+----------------------------------------------------------+
88 */
89
90/* TODO: insert Bit Move group helpers here */
91
92
93/*
94 * Bit Set
95 * -------
96 *
97 * +---------------+----------------------------------------------------------+
98 * | BCLR.B | Vector Bit Clear (byte) |
99 * | BCLR.H | Vector Bit Clear (halfword) |
100 * | BCLR.W | Vector Bit Clear (word) |
101 * | BCLR.D | Vector Bit Clear (doubleword) |
102 * | BNEG.B | Vector Bit Negate (byte) |
103 * | BNEG.H | Vector Bit Negate (halfword) |
104 * | BNEG.W | Vector Bit Negate (word) |
105 * | BNEG.D | Vector Bit Negate (doubleword) |
106 * | BSET.B | Vector Bit Set (byte) |
107 * | BSET.H | Vector Bit Set (halfword) |
108 * | BSET.W | Vector Bit Set (word) |
109 * | BSET.D | Vector Bit Set (doubleword) |
110 * +---------------+----------------------------------------------------------+
111 */
112
113/* TODO: insert Bit Set group helpers here */
114
115
116/*
117 * Fixed Multiply
118 * --------------
119 *
120 * +---------------+----------------------------------------------------------+
121 * | MADD_Q.H | Vector Fixed-Point Multiply and Add (halfword) |
122 * | MADD_Q.W | Vector Fixed-Point Multiply and Add (word) |
123 * | MADDR_Q.H | Vector Fixed-Point Multiply and Add Rounded (halfword) |
124 * | MADDR_Q.W | Vector Fixed-Point Multiply and Add Rounded (word) |
125 * | MSUB_Q.H | Vector Fixed-Point Multiply and Subtr. (halfword) |
126 * | MSUB_Q.W | Vector Fixed-Point Multiply and Subtr. (word) |
127 * | MSUBR_Q.H | Vector Fixed-Point Multiply and Subtr. Rounded (halfword)|
128 * | MSUBR_Q.W | Vector Fixed-Point Multiply and Subtr. Rounded (word) |
129 * | MUL_Q.H | Vector Fixed-Point Multiply (halfword) |
130 * | MUL_Q.W | Vector Fixed-Point Multiply (word) |
131 * | MULR_Q.H | Vector Fixed-Point Multiply Rounded (halfword) |
132 * | MULR_Q.W | Vector Fixed-Point Multiply Rounded (word) |
133 * +---------------+----------------------------------------------------------+
134 */
135
136/* TODO: insert Fixed Multiply group helpers here */
137
138
139/*
140 * Float Max Min
141 * -------------
142 *
143 * +---------------+----------------------------------------------------------+
144 * | FMAX_A.W | Vector Floating-Point Maximum (Absolute) (word) |
145 * | FMAX_A.D | Vector Floating-Point Maximum (Absolute) (doubleword) |
146 * | FMAX.W | Vector Floating-Point Maximum (word) |
147 * | FMAX.D | Vector Floating-Point Maximum (doubleword) |
148 * | FMIN_A.W | Vector Floating-Point Minimum (Absolute) (word) |
149 * | FMIN_A.D | Vector Floating-Point Minimum (Absolute) (doubleword) |
150 * | FMIN.W | Vector Floating-Point Minimum (word) |
151 * | FMIN.D | Vector Floating-Point Minimum (doubleword) |
152 * +---------------+----------------------------------------------------------+
153 */
154
155/* TODO: insert Float Max Min group helpers here */
156
157
158/*
159 * Int Add
160 * -------
161 *
162 * +---------------+----------------------------------------------------------+
163 * | ADD_A.B | Vector Add Absolute Values (byte) |
164 * | ADD_A.H | Vector Add Absolute Values (halfword) |
165 * | ADD_A.W | Vector Add Absolute Values (word) |
166 * | ADD_A.D | Vector Add Absolute Values (doubleword) |
167 * | ADDS_A.B | Vector Signed Saturated Add (of Absolute) (byte) |
168 * | ADDS_A.H | Vector Signed Saturated Add (of Absolute) (halfword) |
169 * | ADDS_A.W | Vector Signed Saturated Add (of Absolute) (word) |
170 * | ADDS_A.D | Vector Signed Saturated Add (of Absolute) (doubleword) |
171 * | ADDS_S.B | Vector Signed Saturated Add (of Signed) (byte) |
172 * | ADDS_S.H | Vector Signed Saturated Add (of Signed) (halfword) |
173 * | ADDS_S.W | Vector Signed Saturated Add (of Signed) (word) |
174 * | ADDS_S.D | Vector Signed Saturated Add (of Signed) (doubleword) |
175 * | ADDS_U.B | Vector Unsigned Saturated Add (of Unsigned) (byte) |
176 * | ADDS_U.H | Vector Unsigned Saturated Add (of Unsigned) (halfword) |
177 * | ADDS_U.W | Vector Unsigned Saturated Add (of Unsigned) (word) |
178 * | ADDS_U.D | Vector Unsigned Saturated Add (of Unsigned) (doubleword) |
179 * | ADDV.B | Vector Add (byte) |
180 * | ADDV.H | Vector Add (halfword) |
181 * | ADDV.W | Vector Add (word) |
182 * | ADDV.D | Vector Add (doubleword) |
183 * | HADD_S.H | Vector Signed Horizontal Add (halfword) |
184 * | HADD_S.W | Vector Signed Horizontal Add (word) |
185 * | HADD_S.D | Vector Signed Horizontal Add (doubleword) |
186 * | HADD_U.H | Vector Unigned Horizontal Add (halfword) |
187 * | HADD_U.W | Vector Unigned Horizontal Add (word) |
188 * | HADD_U.D | Vector Unigned Horizontal Add (doubleword) |
189 * +---------------+----------------------------------------------------------+
190 */
191
192/* TODO: insert Int Add group helpers here */
193
194
195/*
196 * Int Average
197 * -----------
198 *
199 * +---------------+----------------------------------------------------------+
200 * | AVE_S.B | Vector Signed Average (byte) |
201 * | AVE_S.H | Vector Signed Average (halfword) |
202 * | AVE_S.W | Vector Signed Average (word) |
203 * | AVE_S.D | Vector Signed Average (doubleword) |
204 * | AVE_U.B | Vector Unsigned Average (byte) |
205 * | AVE_U.H | Vector Unsigned Average (halfword) |
206 * | AVE_U.W | Vector Unsigned Average (word) |
207 * | AVE_U.D | Vector Unsigned Average (doubleword) |
208 * | AVER_S.B | Vector Signed Average Rounded (byte) |
209 * | AVER_S.H | Vector Signed Average Rounded (halfword) |
210 * | AVER_S.W | Vector Signed Average Rounded (word) |
211 * | AVER_S.D | Vector Signed Average Rounded (doubleword) |
212 * | AVER_U.B | Vector Unsigned Average Rounded (byte) |
213 * | AVER_U.H | Vector Unsigned Average Rounded (halfword) |
214 * | AVER_U.W | Vector Unsigned Average Rounded (word) |
215 * | AVER_U.D | Vector Unsigned Average Rounded (doubleword) |
216 * +---------------+----------------------------------------------------------+
217 */
218
219/* TODO: insert Int Average group helpers here */
220
221
222/*
223 * Int Compare
224 * -----------
225 *
226 * +---------------+----------------------------------------------------------+
227 * | CEQ.B | Vector Compare Equal (byte) |
228 * | CEQ.H | Vector Compare Equal (halfword) |
229 * | CEQ.W | Vector Compare Equal (word) |
230 * | CEQ.D | Vector Compare Equal (doubleword) |
231 * | CLE_S.B | Vector Compare Signed Less Than or Equal (byte) |
232 * | CLE_S.H | Vector Compare Signed Less Than or Equal (halfword) |
233 * | CLE_S.W | Vector Compare Signed Less Than or Equal (word) |
234 * | CLE_S.D | Vector Compare Signed Less Than or Equal (doubleword) |
235 * | CLE_U.B | Vector Compare Unsigned Less Than or Equal (byte) |
236 * | CLE_U.H | Vector Compare Unsigned Less Than or Equal (halfword) |
237 * | CLE_U.W | Vector Compare Unsigned Less Than or Equal (word) |
238 * | CLE_U.D | Vector Compare Unsigned Less Than or Equal (doubleword) |
239 * | CLT_S.B | Vector Compare Signed Less Than (byte) |
240 * | CLT_S.H | Vector Compare Signed Less Than (halfword) |
241 * | CLT_S.W | Vector Compare Signed Less Than (word) |
242 * | CLT_S.D | Vector Compare Signed Less Than (doubleword) |
243 * | CLT_U.B | Vector Compare Unsigned Less Than (byte) |
244 * | CLT_U.H | Vector Compare Unsigned Less Than (halfword) |
245 * | CLT_U.W | Vector Compare Unsigned Less Than (word) |
246 * | CLT_U.D | Vector Compare Unsigned Less Than (doubleword) |
247 * +---------------+----------------------------------------------------------+
248 */
249
250/* TODO: insert Int Compare group helpers here */
251
252
253/*
254 * Int Divide
255 * ----------
256 *
257 * +---------------+----------------------------------------------------------+
258 * | DIV_S.B | Vector Signed Divide (byte) |
259 * | DIV_S.H | Vector Signed Divide (halfword) |
260 * | DIV_S.W | Vector Signed Divide (word) |
261 * | DIV_S.D | Vector Signed Divide (doubleword) |
262 * | DIV_U.B | Vector Unsigned Divide (byte) |
263 * | DIV_U.H | Vector Unsigned Divide (halfword) |
264 * | DIV_U.W | Vector Unsigned Divide (word) |
265 * | DIV_U.D | Vector Unsigned Divide (doubleword) |
266 * +---------------+----------------------------------------------------------+
267 */
268
269/* TODO: insert Int Divide group helpers here */
270
271
272/*
273 * Int Dot Product
274 * ---------------
275 *
276 * +---------------+----------------------------------------------------------+
277 * | DOTP_S.H | Vector Signed Dot Product (halfword) |
278 * | DOTP_S.W | Vector Signed Dot Product (word) |
279 * | DOTP_S.D | Vector Signed Dot Product (doubleword) |
280 * | DOTP_U.H | Vector Unsigned Dot Product (halfword) |
281 * | DOTP_U.W | Vector Unsigned Dot Product (word) |
282 * | DOTP_U.D | Vector Unsigned Dot Product (doubleword) |
283 * | DPADD_S.H | Vector Signed Dot Product (halfword) |
284 * | DPADD_S.W | Vector Signed Dot Product (word) |
285 * | DPADD_S.D | Vector Signed Dot Product (doubleword) |
286 * | DPADD_U.H | Vector Unsigned Dot Product (halfword) |
287 * | DPADD_U.W | Vector Unsigned Dot Product (word) |
288 * | DPADD_U.D | Vector Unsigned Dot Product (doubleword) |
289 * | DPSUB_S.H | Vector Signed Dot Product (halfword) |
290 * | DPSUB_S.W | Vector Signed Dot Product (word) |
291 * | DPSUB_S.D | Vector Signed Dot Product (doubleword) |
292 * | DPSUB_U.H | Vector Unsigned Dot Product (halfword) |
293 * | DPSUB_U.W | Vector Unsigned Dot Product (word) |
294 * | DPSUB_U.D | Vector Unsigned Dot Product (doubleword) |
295 * +---------------+----------------------------------------------------------+
296 */
297
298/* TODO: insert Int Dot Product group helpers here */
299
300
301/*
302 * Int Max Min
303 * -----------
304 *
305 * +---------------+----------------------------------------------------------+
306 * | MAX_A.B | Vector Maximum Based on Absolute Value (byte) |
307 * | MAX_A.H | Vector Maximum Based on Absolute Value (halfword) |
308 * | MAX_A.W | Vector Maximum Based on Absolute Value (word) |
309 * | MAX_A.D | Vector Maximum Based on Absolute Value (doubleword) |
310 * | MAX_S.B | Vector Signed Maximum (byte) |
311 * | MAX_S.H | Vector Signed Maximum (halfword) |
312 * | MAX_S.W | Vector Signed Maximum (word) |
313 * | MAX_S.D | Vector Signed Maximum (doubleword) |
314 * | MAX_U.B | Vector Unsigned Maximum (byte) |
315 * | MAX_U.H | Vector Unsigned Maximum (halfword) |
316 * | MAX_U.W | Vector Unsigned Maximum (word) |
317 * | MAX_U.D | Vector Unsigned Maximum (doubleword) |
318 * | MIN_A.B | Vector Minimum Based on Absolute Value (byte) |
319 * | MIN_A.H | Vector Minimum Based on Absolute Value (halfword) |
320 * | MIN_A.W | Vector Minimum Based on Absolute Value (word) |
321 * | MIN_A.D | Vector Minimum Based on Absolute Value (doubleword) |
322 * | MIN_S.B | Vector Signed Minimum (byte) |
323 * | MIN_S.H | Vector Signed Minimum (halfword) |
324 * | MIN_S.W | Vector Signed Minimum (word) |
325 * | MIN_S.D | Vector Signed Minimum (doubleword) |
326 * | MIN_U.B | Vector Unsigned Minimum (byte) |
327 * | MIN_U.H | Vector Unsigned Minimum (halfword) |
328 * | MIN_U.W | Vector Unsigned Minimum (word) |
329 * | MIN_U.D | Vector Unsigned Minimum (doubleword) |
330 * +---------------+----------------------------------------------------------+
331 */
332
333/* TODO: insert Int Max Min group helpers here */
334
335
336/*
337 * Int Modulo
338 * ----------
339 *
340 * +---------------+----------------------------------------------------------+
341 * | MOD_S.B | Vector Signed Modulo (byte) |
342 * | MOD_S.H | Vector Signed Modulo (halfword) |
343 * | MOD_S.W | Vector Signed Modulo (word) |
344 * | MOD_S.D | Vector Signed Modulo (doubleword) |
345 * | MOD_U.B | Vector Unsigned Modulo (byte) |
346 * | MOD_U.H | Vector Unsigned Modulo (halfword) |
347 * | MOD_U.W | Vector Unsigned Modulo (word) |
348 * | MOD_U.D | Vector Unsigned Modulo (doubleword) |
349 * +---------------+----------------------------------------------------------+
350 */
351
352/* TODO: insert Int Modulo group helpers here */
353
354
355/*
356 * Int Multiply
357 * ------------
358 *
359 * +---------------+----------------------------------------------------------+
360 * | MADDV.B | Vector Multiply and Add (byte) |
361 * | MADDV.H | Vector Multiply and Add (halfword) |
362 * | MADDV.W | Vector Multiply and Add (word) |
363 * | MADDV.D | Vector Multiply and Add (doubleword) |
364 * | MSUBV.B | Vector Multiply and Subtract (byte) |
365 * | MSUBV.H | Vector Multiply and Subtract (halfword) |
366 * | MSUBV.W | Vector Multiply and Subtract (word) |
367 * | MSUBV.D | Vector Multiply and Subtract (doubleword) |
368 * | MULV.B | Vector Multiply (byte) |
369 * | MULV.H | Vector Multiply (halfword) |
370 * | MULV.W | Vector Multiply (word) |
371 * | MULV.D | Vector Multiply (doubleword) |
372 * +---------------+----------------------------------------------------------+
373 */
374
375/* TODO: insert Int Multiply group helpers here */
376
377
378/*
379 * Int Subtract
380 * ------------
381 *
382 * +---------------+----------------------------------------------------------+
383 * | ASUB_S.B | Vector Absolute Values of Signed Subtract (byte) |
384 * | ASUB_S.H | Vector Absolute Values of Signed Subtract (halfword) |
385 * | ASUB_S.W | Vector Absolute Values of Signed Subtract (word) |
386 * | ASUB_S.D | Vector Absolute Values of Signed Subtract (doubleword) |
387 * | ASUB_U.B | Vector Absolute Values of Unsigned Subtract (byte) |
388 * | ASUB_U.H | Vector Absolute Values of Unsigned Subtract (halfword) |
389 * | ASUB_U.W | Vector Absolute Values of Unsigned Subtract (word) |
390 * | ASUB_U.D | Vector Absolute Values of Unsigned Subtract (doubleword) |
391 * | HSUB_S.H | Vector Signed Horizontal Subtract (halfword) |
392 * | HSUB_S.W | Vector Signed Horizontal Subtract (word) |
393 * | HSUB_S.D | Vector Signed Horizontal Subtract (doubleword) |
394 * | HSUB_U.H | Vector Unigned Horizontal Subtract (halfword) |
395 * | HSUB_U.W | Vector Unigned Horizontal Subtract (word) |
396 * | HSUB_U.D | Vector Unigned Horizontal Subtract (doubleword) |
397 * | SUBS_S.B | Vector Signed Saturated Subtract (of Signed) (byte) |
398 * | SUBS_S.H | Vector Signed Saturated Subtract (of Signed) (halfword) |
399 * | SUBS_S.W | Vector Signed Saturated Subtract (of Signed) (word) |
400 * | SUBS_S.D | Vector Signed Saturated Subtract (of Signed) (doubleword)|
401 * | SUBS_U.B | Vector Unsigned Saturated Subtract (of Uns.) (byte) |
402 * | SUBS_U.H | Vector Unsigned Saturated Subtract (of Uns.) (halfword) |
403 * | SUBS_U.W | Vector Unsigned Saturated Subtract (of Uns.) (word) |
404 * | SUBS_U.D | Vector Unsigned Saturated Subtract (of Uns.) (doubleword)|
405 * | SUBSUS_U.B | Vector Uns. Sat. Subtract (of S. from Uns.) (byte) |
406 * | SUBSUS_U.H | Vector Uns. Sat. Subtract (of S. from Uns.) (halfword) |
407 * | SUBSUS_U.W | Vector Uns. Sat. Subtract (of S. from Uns.) (word) |
408 * | SUBSUS_U.D | Vector Uns. Sat. Subtract (of S. from Uns.) (doubleword) |
409 * | SUBSUU_S.B | Vector Signed Saturated Subtract (of Uns.) (byte) |
410 * | SUBSUU_S.H | Vector Signed Saturated Subtract (of Uns.) (halfword) |
411 * | SUBSUU_S.W | Vector Signed Saturated Subtract (of Uns.) (word) |
412 * | SUBSUU_S.D | Vector Signed Saturated Subtract (of Uns.) (doubleword) |
413 * | SUBV.B | Vector Subtract (byte) |
414 * | SUBV.H | Vector Subtract (halfword) |
415 * | SUBV.W | Vector Subtract (word) |
416 * | SUBV.D | Vector Subtract (doubleword) |
417 * +---------------+----------------------------------------------------------+
418 */
419
420/* TODO: insert Int Subtract group helpers here */
421
422
423/*
424 * Interleave
425 * ----------
426 *
427 * +---------------+----------------------------------------------------------+
428 * | ILVEV.B | Vector Interleave Even (byte) |
429 * | ILVEV.H | Vector Interleave Even (halfword) |
430 * | ILVEV.W | Vector Interleave Even (word) |
431 * | ILVEV.D | Vector Interleave Even (doubleword) |
432 * | ILVOD.B | Vector Interleave Odd (byte) |
433 * | ILVOD.H | Vector Interleave Odd (halfword) |
434 * | ILVOD.W | Vector Interleave Odd (word) |
435 * | ILVOD.D | Vector Interleave Odd (doubleword) |
436 * | ILVL.B | Vector Interleave Left (byte) |
437 * | ILVL.H | Vector Interleave Left (halfword) |
438 * | ILVL.W | Vector Interleave Left (word) |
439 * | ILVL.D | Vector Interleave Left (doubleword) |
440 * | ILVR.B | Vector Interleave Right (byte) |
441 * | ILVR.H | Vector Interleave Right (halfword) |
442 * | ILVR.W | Vector Interleave Right (word) |
443 * | ILVR.D | Vector Interleave Right (doubleword) |
444 * +---------------+----------------------------------------------------------+
445 */
446
447/* TODO: insert Interleave group helpers here */
448
449
450/*
451 * Logic
452 * -----
453 *
454 * +---------------+----------------------------------------------------------+
455 * | AND.V | Vector Logical And |
456 * | NOR.V | Vector Logical Negated Or |
457 * | OR.V | Vector Logical Or |
458 * | XOR.V | Vector Logical Exclusive Or |
459 * +---------------+----------------------------------------------------------+
460 */
461
462/* TODO: insert Logic group helpers here */
463
464
465/*
466 * Move
467 * ----
468 *
469 * +---------------+----------------------------------------------------------+
470 * | MOVE.V | Vector Move |
471 * +---------------+----------------------------------------------------------+
472 */
473
474/* TODO: insert Move group helpers here */
475
476
477/*
478 * Pack
479 * ----
480 *
481 * +---------------+----------------------------------------------------------+
482 * | PCKEV.B | Vector Pack Even (byte) |
483 * | PCKEV.H | Vector Pack Even (halfword) |
484 * | PCKEV.W | Vector Pack Even (word) |
485 * | PCKEV.D | Vector Pack Even (doubleword) |
486 * | PCKOD.B | Vector Pack Odd (byte) |
487 * | PCKOD.H | Vector Pack Odd (halfword) |
488 * | PCKOD.W | Vector Pack Odd (word) |
489 * | PCKOD.D | Vector Pack Odd (doubleword) |
490 * | VSHF.B | Vector Data Preserving Shuffle (byte) |
491 * | VSHF.H | Vector Data Preserving Shuffle (halfword) |
492 * | VSHF.W | Vector Data Preserving Shuffle (word) |
493 * | VSHF.D | Vector Data Preserving Shuffle (doubleword) |
494 * +---------------+----------------------------------------------------------+
495 */
496
497/* TODO: insert Pack group helpers here */
498
499
500/*
501 * Shift
502 * -----
503 *
504 * +---------------+----------------------------------------------------------+
505 * | SLL.B | Vector Shift Left (byte) |
506 * | SLL.H | Vector Shift Left (halfword) |
507 * | SLL.W | Vector Shift Left (word) |
508 * | SLL.D | Vector Shift Left (doubleword) |
509 * | SRA.B | Vector Shift Right Arithmetic (byte) |
510 * | SRA.H | Vector Shift Right Arithmetic (halfword) |
511 * | SRA.W | Vector Shift Right Arithmetic (word) |
512 * | SRA.D | Vector Shift Right Arithmetic (doubleword) |
513 * | SRAR.B | Vector Shift Right Arithmetic Rounded (byte) |
514 * | SRAR.H | Vector Shift Right Arithmetic Rounded (halfword) |
515 * | SRAR.W | Vector Shift Right Arithmetic Rounded (word) |
516 * | SRAR.D | Vector Shift Right Arithmetic Rounded (doubleword) |
517 * | SRL.B | Vector Shift Right Logical (byte) |
518 * | SRL.H | Vector Shift Right Logical (halfword) |
519 * | SRL.W | Vector Shift Right Logical (word) |
520 * | SRL.D | Vector Shift Right Logical (doubleword) |
521 * | SRLR.B | Vector Shift Right Logical Rounded (byte) |
522 * | SRLR.H | Vector Shift Right Logical Rounded (halfword) |
523 * | SRLR.W | Vector Shift Right Logical Rounded (word) |
524 * | SRLR.D | Vector Shift Right Logical Rounded (doubleword) |
525 * +---------------+----------------------------------------------------------+
526 */
527
528/* TODO: insert Shift group helpers here */
529
530
531static inline void msa_move_v(wr_t *pwd, wr_t *pws)
532{
533 uint32_t i;
534
535 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
536 pwd->d[i] = pws->d[i];
537 }
538}
539
540#define MSA_FN_IMM8(FUNC, DEST, OPERATION) \
541void helper_msa_ ## FUNC(CPUMIPSState *env, uint32_t wd, uint32_t ws, \
542 uint32_t i8) \
543{ \
544 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
545 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
546 uint32_t i; \
547 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) { \
548 DEST = OPERATION; \
549 } \
550}
551
552MSA_FN_IMM8(andi_b, pwd->b[i], pws->b[i] & i8)
553MSA_FN_IMM8(ori_b, pwd->b[i], pws->b[i] | i8)
554MSA_FN_IMM8(nori_b, pwd->b[i], ~(pws->b[i] | i8))
555MSA_FN_IMM8(xori_b, pwd->b[i], pws->b[i] ^ i8)
556
557#define BIT_MOVE_IF_NOT_ZERO(dest, arg1, arg2, df) \
558 UNSIGNED(((dest & (~arg2)) | (arg1 & arg2)), df)
559MSA_FN_IMM8(bmnzi_b, pwd->b[i],
560 BIT_MOVE_IF_NOT_ZERO(pwd->b[i], pws->b[i], i8, DF_BYTE))
561
562#define BIT_MOVE_IF_ZERO(dest, arg1, arg2, df) \
563 UNSIGNED((dest & arg2) | (arg1 & (~arg2)), df)
564MSA_FN_IMM8(bmzi_b, pwd->b[i],
565 BIT_MOVE_IF_ZERO(pwd->b[i], pws->b[i], i8, DF_BYTE))
566
567#define BIT_SELECT(dest, arg1, arg2, df) \
568 UNSIGNED((arg1 & (~dest)) | (arg2 & dest), df)
569MSA_FN_IMM8(bseli_b, pwd->b[i],
570 BIT_SELECT(pwd->b[i], pws->b[i], i8, DF_BYTE))
571
572#undef MSA_FN_IMM8
573
574#define SHF_POS(i, imm) (((i) & 0xfc) + (((imm) >> (2 * ((i) & 0x03))) & 0x03))
575
576void helper_msa_shf_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
577 uint32_t ws, uint32_t imm)
578{
579 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
580 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
581 wr_t wx, *pwx = &wx;
582 uint32_t i;
583
584 switch (df) {
585 case DF_BYTE:
586 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) {
587 pwx->b[i] = pws->b[SHF_POS(i, imm)];
588 }
589 break;
590 case DF_HALF:
591 for (i = 0; i < DF_ELEMENTS(DF_HALF); i++) {
592 pwx->h[i] = pws->h[SHF_POS(i, imm)];
593 }
594 break;
595 case DF_WORD:
596 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
597 pwx->w[i] = pws->w[SHF_POS(i, imm)];
598 }
599 break;
600 default:
601 assert(0);
602 }
603 msa_move_v(pwd, pwx);
604}
605
606#define MSA_FN_VECTOR(FUNC, DEST, OPERATION) \
607void helper_msa_ ## FUNC(CPUMIPSState *env, uint32_t wd, uint32_t ws, \
608 uint32_t wt) \
609{ \
610 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
611 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
612 wr_t *pwt = &(env->active_fpu.fpr[wt].wr); \
613 uint32_t i; \
614 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) { \
615 DEST = OPERATION; \
616 } \
617}
618
619MSA_FN_VECTOR(bmnz_v, pwd->d[i],
620 BIT_MOVE_IF_NOT_ZERO(pwd->d[i], pws->d[i], pwt->d[i], DF_DOUBLE))
621MSA_FN_VECTOR(bmz_v, pwd->d[i],
622 BIT_MOVE_IF_ZERO(pwd->d[i], pws->d[i], pwt->d[i], DF_DOUBLE))
623MSA_FN_VECTOR(bsel_v, pwd->d[i],
624 BIT_SELECT(pwd->d[i], pws->d[i], pwt->d[i], DF_DOUBLE))
625#undef BIT_MOVE_IF_NOT_ZERO
626#undef BIT_MOVE_IF_ZERO
627#undef BIT_SELECT
628#undef MSA_FN_VECTOR
629
630void helper_msa_and_v(CPUMIPSState *env, uint32_t wd, uint32_t ws, uint32_t wt)
631{
632 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
633 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
634 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
635
636 pwd->d[0] = pws->d[0] & pwt->d[0];
637 pwd->d[1] = pws->d[1] & pwt->d[1];
638}
639
640void helper_msa_or_v(CPUMIPSState *env, uint32_t wd, uint32_t ws, uint32_t wt)
641{
642 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
643 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
644 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
645
646 pwd->d[0] = pws->d[0] | pwt->d[0];
647 pwd->d[1] = pws->d[1] | pwt->d[1];
648}
649
650void helper_msa_nor_v(CPUMIPSState *env, uint32_t wd, uint32_t ws, uint32_t wt)
651{
652 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
653 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
654 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
655
656 pwd->d[0] = ~(pws->d[0] | pwt->d[0]);
657 pwd->d[1] = ~(pws->d[1] | pwt->d[1]);
658}
659
660void helper_msa_xor_v(CPUMIPSState *env, uint32_t wd, uint32_t ws, uint32_t wt)
661{
662 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
663 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
664 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
665
666 pwd->d[0] = pws->d[0] ^ pwt->d[0];
667 pwd->d[1] = pws->d[1] ^ pwt->d[1];
668}
669
670static inline int64_t msa_addv_df(uint32_t df, int64_t arg1, int64_t arg2)
671{
672 return arg1 + arg2;
673}
674
675static inline int64_t msa_subv_df(uint32_t df, int64_t arg1, int64_t arg2)
676{
677 return arg1 - arg2;
678}
679
680static inline int64_t msa_ceq_df(uint32_t df, int64_t arg1, int64_t arg2)
681{
682 return arg1 == arg2 ? -1 : 0;
683}
684
685static inline int64_t msa_cle_s_df(uint32_t df, int64_t arg1, int64_t arg2)
686{
687 return arg1 <= arg2 ? -1 : 0;
688}
689
690static inline int64_t msa_cle_u_df(uint32_t df, int64_t arg1, int64_t arg2)
691{
692 uint64_t u_arg1 = UNSIGNED(arg1, df);
693 uint64_t u_arg2 = UNSIGNED(arg2, df);
694 return u_arg1 <= u_arg2 ? -1 : 0;
695}
696
697static inline int64_t msa_clt_s_df(uint32_t df, int64_t arg1, int64_t arg2)
698{
699 return arg1 < arg2 ? -1 : 0;
700}
701
702static inline int64_t msa_clt_u_df(uint32_t df, int64_t arg1, int64_t arg2)
703{
704 uint64_t u_arg1 = UNSIGNED(arg1, df);
705 uint64_t u_arg2 = UNSIGNED(arg2, df);
706 return u_arg1 < u_arg2 ? -1 : 0;
707}
708
709static inline int64_t msa_max_s_df(uint32_t df, int64_t arg1, int64_t arg2)
710{
711 return arg1 > arg2 ? arg1 : arg2;
712}
713
714static inline int64_t msa_max_u_df(uint32_t df, int64_t arg1, int64_t arg2)
715{
716 uint64_t u_arg1 = UNSIGNED(arg1, df);
717 uint64_t u_arg2 = UNSIGNED(arg2, df);
718 return u_arg1 > u_arg2 ? arg1 : arg2;
719}
720
721static inline int64_t msa_min_s_df(uint32_t df, int64_t arg1, int64_t arg2)
722{
723 return arg1 < arg2 ? arg1 : arg2;
724}
725
726static inline int64_t msa_min_u_df(uint32_t df, int64_t arg1, int64_t arg2)
727{
728 uint64_t u_arg1 = UNSIGNED(arg1, df);
729 uint64_t u_arg2 = UNSIGNED(arg2, df);
730 return u_arg1 < u_arg2 ? arg1 : arg2;
731}
732
733#define MSA_BINOP_IMM_DF(helper, func) \
734void helper_msa_ ## helper ## _df(CPUMIPSState *env, uint32_t df, \
735 uint32_t wd, uint32_t ws, int32_t u5) \
736{ \
737 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
738 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
739 uint32_t i; \
740 \
741 switch (df) { \
742 case DF_BYTE: \
743 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) { \
744 pwd->b[i] = msa_ ## func ## _df(df, pws->b[i], u5); \
745 } \
746 break; \
747 case DF_HALF: \
748 for (i = 0; i < DF_ELEMENTS(DF_HALF); i++) { \
749 pwd->h[i] = msa_ ## func ## _df(df, pws->h[i], u5); \
750 } \
751 break; \
752 case DF_WORD: \
753 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) { \
754 pwd->w[i] = msa_ ## func ## _df(df, pws->w[i], u5); \
755 } \
756 break; \
757 case DF_DOUBLE: \
758 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) { \
759 pwd->d[i] = msa_ ## func ## _df(df, pws->d[i], u5); \
760 } \
761 break; \
762 default: \
763 assert(0); \
764 } \
765}
766
767MSA_BINOP_IMM_DF(addvi, addv)
768MSA_BINOP_IMM_DF(subvi, subv)
769MSA_BINOP_IMM_DF(ceqi, ceq)
770MSA_BINOP_IMM_DF(clei_s, cle_s)
771MSA_BINOP_IMM_DF(clei_u, cle_u)
772MSA_BINOP_IMM_DF(clti_s, clt_s)
773MSA_BINOP_IMM_DF(clti_u, clt_u)
774MSA_BINOP_IMM_DF(maxi_s, max_s)
775MSA_BINOP_IMM_DF(maxi_u, max_u)
776MSA_BINOP_IMM_DF(mini_s, min_s)
777MSA_BINOP_IMM_DF(mini_u, min_u)
778#undef MSA_BINOP_IMM_DF
779
780void helper_msa_ldi_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
781 int32_t s10)
782{
783 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
784 uint32_t i;
785
786 switch (df) {
787 case DF_BYTE:
788 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) {
789 pwd->b[i] = (int8_t)s10;
790 }
791 break;
792 case DF_HALF:
793 for (i = 0; i < DF_ELEMENTS(DF_HALF); i++) {
794 pwd->h[i] = (int16_t)s10;
795 }
796 break;
797 case DF_WORD:
798 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
799 pwd->w[i] = (int32_t)s10;
800 }
801 break;
802 case DF_DOUBLE:
803 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
804 pwd->d[i] = (int64_t)s10;
805 }
806 break;
807 default:
808 assert(0);
809 }
810}
811
812/* Data format bit position and unsigned values */
813#define BIT_POSITION(x, df) ((uint64_t)(x) % DF_BITS(df))
814
815static inline int64_t msa_sll_df(uint32_t df, int64_t arg1, int64_t arg2)
816{
817 int32_t b_arg2 = BIT_POSITION(arg2, df);
818 return arg1 << b_arg2;
819}
820
821static inline int64_t msa_sra_df(uint32_t df, int64_t arg1, int64_t arg2)
822{
823 int32_t b_arg2 = BIT_POSITION(arg2, df);
824 return arg1 >> b_arg2;
825}
826
827static inline int64_t msa_srl_df(uint32_t df, int64_t arg1, int64_t arg2)
828{
829 uint64_t u_arg1 = UNSIGNED(arg1, df);
830 int32_t b_arg2 = BIT_POSITION(arg2, df);
831 return u_arg1 >> b_arg2;
832}
833
834static inline int64_t msa_bclr_df(uint32_t df, int64_t arg1, int64_t arg2)
835{
836 int32_t b_arg2 = BIT_POSITION(arg2, df);
837 return UNSIGNED(arg1 & (~(1LL << b_arg2)), df);
838}
839
840static inline int64_t msa_bset_df(uint32_t df, int64_t arg1,
841 int64_t arg2)
842{
843 int32_t b_arg2 = BIT_POSITION(arg2, df);
844 return UNSIGNED(arg1 | (1LL << b_arg2), df);
845}
846
847static inline int64_t msa_bneg_df(uint32_t df, int64_t arg1, int64_t arg2)
848{
849 int32_t b_arg2 = BIT_POSITION(arg2, df);
850 return UNSIGNED(arg1 ^ (1LL << b_arg2), df);
851}
852
853static inline int64_t msa_binsl_df(uint32_t df, int64_t dest, int64_t arg1,
854 int64_t arg2)
855{
856 uint64_t u_arg1 = UNSIGNED(arg1, df);
857 uint64_t u_dest = UNSIGNED(dest, df);
858 int32_t sh_d = BIT_POSITION(arg2, df) + 1;
859 int32_t sh_a = DF_BITS(df) - sh_d;
860 if (sh_d == DF_BITS(df)) {
861 return u_arg1;
862 } else {
863 return UNSIGNED(UNSIGNED(u_dest << sh_d, df) >> sh_d, df) |
864 UNSIGNED(UNSIGNED(u_arg1 >> sh_a, df) << sh_a, df);
865 }
866}
867
868static inline int64_t msa_binsr_df(uint32_t df, int64_t dest, int64_t arg1,
869 int64_t arg2)
870{
871 uint64_t u_arg1 = UNSIGNED(arg1, df);
872 uint64_t u_dest = UNSIGNED(dest, df);
873 int32_t sh_d = BIT_POSITION(arg2, df) + 1;
874 int32_t sh_a = DF_BITS(df) - sh_d;
875 if (sh_d == DF_BITS(df)) {
876 return u_arg1;
877 } else {
878 return UNSIGNED(UNSIGNED(u_dest >> sh_d, df) << sh_d, df) |
879 UNSIGNED(UNSIGNED(u_arg1 << sh_a, df) >> sh_a, df);
880 }
881}
882
883static inline int64_t msa_sat_s_df(uint32_t df, int64_t arg, uint32_t m)
884{
885 return arg < M_MIN_INT(m + 1) ? M_MIN_INT(m + 1) :
886 arg > M_MAX_INT(m + 1) ? M_MAX_INT(m + 1) :
887 arg;
888}
889
890static inline int64_t msa_sat_u_df(uint32_t df, int64_t arg, uint32_t m)
891{
892 uint64_t u_arg = UNSIGNED(arg, df);
893 return u_arg < M_MAX_UINT(m + 1) ? u_arg :
894 M_MAX_UINT(m + 1);
895}
896
897static inline int64_t msa_srar_df(uint32_t df, int64_t arg1, int64_t arg2)
898{
899 int32_t b_arg2 = BIT_POSITION(arg2, df);
900 if (b_arg2 == 0) {
901 return arg1;
902 } else {
903 int64_t r_bit = (arg1 >> (b_arg2 - 1)) & 1;
904 return (arg1 >> b_arg2) + r_bit;
905 }
906}
907
908static inline int64_t msa_srlr_df(uint32_t df, int64_t arg1, int64_t arg2)
909{
910 uint64_t u_arg1 = UNSIGNED(arg1, df);
911 int32_t b_arg2 = BIT_POSITION(arg2, df);
912 if (b_arg2 == 0) {
913 return u_arg1;
914 } else {
915 uint64_t r_bit = (u_arg1 >> (b_arg2 - 1)) & 1;
916 return (u_arg1 >> b_arg2) + r_bit;
917 }
918}
919
920#define MSA_BINOP_IMMU_DF(helper, func) \
921void helper_msa_ ## helper ## _df(CPUMIPSState *env, uint32_t df, uint32_t wd, \
922 uint32_t ws, uint32_t u5) \
923{ \
924 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
925 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
926 uint32_t i; \
927 \
928 switch (df) { \
929 case DF_BYTE: \
930 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) { \
931 pwd->b[i] = msa_ ## func ## _df(df, pws->b[i], u5); \
932 } \
933 break; \
934 case DF_HALF: \
935 for (i = 0; i < DF_ELEMENTS(DF_HALF); i++) { \
936 pwd->h[i] = msa_ ## func ## _df(df, pws->h[i], u5); \
937 } \
938 break; \
939 case DF_WORD: \
940 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) { \
941 pwd->w[i] = msa_ ## func ## _df(df, pws->w[i], u5); \
942 } \
943 break; \
944 case DF_DOUBLE: \
945 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) { \
946 pwd->d[i] = msa_ ## func ## _df(df, pws->d[i], u5); \
947 } \
948 break; \
949 default: \
950 assert(0); \
951 } \
952}
953
954MSA_BINOP_IMMU_DF(slli, sll)
955MSA_BINOP_IMMU_DF(srai, sra)
956MSA_BINOP_IMMU_DF(srli, srl)
957MSA_BINOP_IMMU_DF(bclri, bclr)
958MSA_BINOP_IMMU_DF(bseti, bset)
959MSA_BINOP_IMMU_DF(bnegi, bneg)
960MSA_BINOP_IMMU_DF(sat_s, sat_s)
961MSA_BINOP_IMMU_DF(sat_u, sat_u)
962MSA_BINOP_IMMU_DF(srari, srar)
963MSA_BINOP_IMMU_DF(srlri, srlr)
964#undef MSA_BINOP_IMMU_DF
965
966#define MSA_TEROP_IMMU_DF(helper, func) \
967void helper_msa_ ## helper ## _df(CPUMIPSState *env, uint32_t df, \
968 uint32_t wd, uint32_t ws, uint32_t u5) \
969{ \
970 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
971 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
972 uint32_t i; \
973 \
974 switch (df) { \
975 case DF_BYTE: \
976 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) { \
977 pwd->b[i] = msa_ ## func ## _df(df, pwd->b[i], pws->b[i], \
978 u5); \
979 } \
980 break; \
981 case DF_HALF: \
982 for (i = 0; i < DF_ELEMENTS(DF_HALF); i++) { \
983 pwd->h[i] = msa_ ## func ## _df(df, pwd->h[i], pws->h[i], \
984 u5); \
985 } \
986 break; \
987 case DF_WORD: \
988 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) { \
989 pwd->w[i] = msa_ ## func ## _df(df, pwd->w[i], pws->w[i], \
990 u5); \
991 } \
992 break; \
993 case DF_DOUBLE: \
994 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) { \
995 pwd->d[i] = msa_ ## func ## _df(df, pwd->d[i], pws->d[i], \
996 u5); \
997 } \
998 break; \
999 default: \
1000 assert(0); \
1001 } \
1002}
1003
1004MSA_TEROP_IMMU_DF(binsli, binsl)
1005MSA_TEROP_IMMU_DF(binsri, binsr)
1006#undef MSA_TEROP_IMMU_DF
1007
1008static inline int64_t msa_max_a_df(uint32_t df, int64_t arg1, int64_t arg2)
1009{
1010 uint64_t abs_arg1 = arg1 >= 0 ? arg1 : -arg1;
1011 uint64_t abs_arg2 = arg2 >= 0 ? arg2 : -arg2;
1012 return abs_arg1 > abs_arg2 ? arg1 : arg2;
1013}
1014
1015static inline int64_t msa_min_a_df(uint32_t df, int64_t arg1, int64_t arg2)
1016{
1017 uint64_t abs_arg1 = arg1 >= 0 ? arg1 : -arg1;
1018 uint64_t abs_arg2 = arg2 >= 0 ? arg2 : -arg2;
1019 return abs_arg1 < abs_arg2 ? arg1 : arg2;
1020}
1021
1022static inline int64_t msa_add_a_df(uint32_t df, int64_t arg1, int64_t arg2)
1023{
1024 uint64_t abs_arg1 = arg1 >= 0 ? arg1 : -arg1;
1025 uint64_t abs_arg2 = arg2 >= 0 ? arg2 : -arg2;
1026 return abs_arg1 + abs_arg2;
1027}
1028
1029static inline int64_t msa_adds_a_df(uint32_t df, int64_t arg1, int64_t arg2)
1030{
1031 uint64_t max_int = (uint64_t)DF_MAX_INT(df);
1032 uint64_t abs_arg1 = arg1 >= 0 ? arg1 : -arg1;
1033 uint64_t abs_arg2 = arg2 >= 0 ? arg2 : -arg2;
1034 if (abs_arg1 > max_int || abs_arg2 > max_int) {
1035 return (int64_t)max_int;
1036 } else {
1037 return (abs_arg1 < max_int - abs_arg2) ? abs_arg1 + abs_arg2 : max_int;
1038 }
1039}
1040
1041static inline int64_t msa_adds_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1042{
1043 int64_t max_int = DF_MAX_INT(df);
1044 int64_t min_int = DF_MIN_INT(df);
1045 if (arg1 < 0) {
1046 return (min_int - arg1 < arg2) ? arg1 + arg2 : min_int;
1047 } else {
1048 return (arg2 < max_int - arg1) ? arg1 + arg2 : max_int;
1049 }
1050}
1051
1052static inline uint64_t msa_adds_u_df(uint32_t df, uint64_t arg1, uint64_t arg2)
1053{
1054 uint64_t max_uint = DF_MAX_UINT(df);
1055 uint64_t u_arg1 = UNSIGNED(arg1, df);
1056 uint64_t u_arg2 = UNSIGNED(arg2, df);
1057 return (u_arg1 < max_uint - u_arg2) ? u_arg1 + u_arg2 : max_uint;
1058}
1059
1060static inline int64_t msa_ave_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1061{
1062 /* signed shift */
1063 return (arg1 >> 1) + (arg2 >> 1) + (arg1 & arg2 & 1);
1064}
1065
1066static inline uint64_t msa_ave_u_df(uint32_t df, uint64_t arg1, uint64_t arg2)
1067{
1068 uint64_t u_arg1 = UNSIGNED(arg1, df);
1069 uint64_t u_arg2 = UNSIGNED(arg2, df);
1070 /* unsigned shift */
1071 return (u_arg1 >> 1) + (u_arg2 >> 1) + (u_arg1 & u_arg2 & 1);
1072}
1073
1074static inline int64_t msa_aver_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1075{
1076 /* signed shift */
1077 return (arg1 >> 1) + (arg2 >> 1) + ((arg1 | arg2) & 1);
1078}
1079
1080static inline uint64_t msa_aver_u_df(uint32_t df, uint64_t arg1, uint64_t arg2)
1081{
1082 uint64_t u_arg1 = UNSIGNED(arg1, df);
1083 uint64_t u_arg2 = UNSIGNED(arg2, df);
1084 /* unsigned shift */
1085 return (u_arg1 >> 1) + (u_arg2 >> 1) + ((u_arg1 | u_arg2) & 1);
1086}
1087
1088static inline int64_t msa_subs_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1089{
1090 int64_t max_int = DF_MAX_INT(df);
1091 int64_t min_int = DF_MIN_INT(df);
1092 if (arg2 > 0) {
1093 return (min_int + arg2 < arg1) ? arg1 - arg2 : min_int;
1094 } else {
1095 return (arg1 < max_int + arg2) ? arg1 - arg2 : max_int;
1096 }
1097}
1098
1099static inline int64_t msa_subs_u_df(uint32_t df, int64_t arg1, int64_t arg2)
1100{
1101 uint64_t u_arg1 = UNSIGNED(arg1, df);
1102 uint64_t u_arg2 = UNSIGNED(arg2, df);
1103 return (u_arg1 > u_arg2) ? u_arg1 - u_arg2 : 0;
1104}
1105
1106static inline int64_t msa_subsus_u_df(uint32_t df, int64_t arg1, int64_t arg2)
1107{
1108 uint64_t u_arg1 = UNSIGNED(arg1, df);
1109 uint64_t max_uint = DF_MAX_UINT(df);
1110 if (arg2 >= 0) {
1111 uint64_t u_arg2 = (uint64_t)arg2;
1112 return (u_arg1 > u_arg2) ?
1113 (int64_t)(u_arg1 - u_arg2) :
1114 0;
1115 } else {
1116 uint64_t u_arg2 = (uint64_t)(-arg2);
1117 return (u_arg1 < max_uint - u_arg2) ?
1118 (int64_t)(u_arg1 + u_arg2) :
1119 (int64_t)max_uint;
1120 }
1121}
1122
1123static inline int64_t msa_subsuu_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1124{
1125 uint64_t u_arg1 = UNSIGNED(arg1, df);
1126 uint64_t u_arg2 = UNSIGNED(arg2, df);
1127 int64_t max_int = DF_MAX_INT(df);
1128 int64_t min_int = DF_MIN_INT(df);
1129 if (u_arg1 > u_arg2) {
1130 return u_arg1 - u_arg2 < (uint64_t)max_int ?
1131 (int64_t)(u_arg1 - u_arg2) :
1132 max_int;
1133 } else {
1134 return u_arg2 - u_arg1 < (uint64_t)(-min_int) ?
1135 (int64_t)(u_arg1 - u_arg2) :
1136 min_int;
1137 }
1138}
1139
1140static inline int64_t msa_asub_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1141{
1142 /* signed compare */
1143 return (arg1 < arg2) ?
1144 (uint64_t)(arg2 - arg1) : (uint64_t)(arg1 - arg2);
1145}
1146
1147static inline uint64_t msa_asub_u_df(uint32_t df, uint64_t arg1, uint64_t arg2)
1148{
1149 uint64_t u_arg1 = UNSIGNED(arg1, df);
1150 uint64_t u_arg2 = UNSIGNED(arg2, df);
1151 /* unsigned compare */
1152 return (u_arg1 < u_arg2) ?
1153 (uint64_t)(u_arg2 - u_arg1) : (uint64_t)(u_arg1 - u_arg2);
1154}
1155
1156static inline int64_t msa_mulv_df(uint32_t df, int64_t arg1, int64_t arg2)
1157{
1158 return arg1 * arg2;
1159}
1160
1161static inline int64_t msa_div_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1162{
1163 if (arg1 == DF_MIN_INT(df) && arg2 == -1) {
1164 return DF_MIN_INT(df);
1165 }
1166 return arg2 ? arg1 / arg2
1167 : arg1 >= 0 ? -1 : 1;
1168}
1169
1170static inline int64_t msa_div_u_df(uint32_t df, int64_t arg1, int64_t arg2)
1171{
1172 uint64_t u_arg1 = UNSIGNED(arg1, df);
1173 uint64_t u_arg2 = UNSIGNED(arg2, df);
1174 return arg2 ? u_arg1 / u_arg2 : -1;
1175}
1176
1177static inline int64_t msa_mod_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1178{
1179 if (arg1 == DF_MIN_INT(df) && arg2 == -1) {
1180 return 0;
1181 }
1182 return arg2 ? arg1 % arg2 : arg1;
1183}
1184
1185static inline int64_t msa_mod_u_df(uint32_t df, int64_t arg1, int64_t arg2)
1186{
1187 uint64_t u_arg1 = UNSIGNED(arg1, df);
1188 uint64_t u_arg2 = UNSIGNED(arg2, df);
1189 return u_arg2 ? u_arg1 % u_arg2 : u_arg1;
1190}
1191
1192#define SIGNED_EVEN(a, df) \
1193 ((((int64_t)(a)) << (64 - DF_BITS(df) / 2)) >> (64 - DF_BITS(df) / 2))
1194
1195#define UNSIGNED_EVEN(a, df) \
1196 ((((uint64_t)(a)) << (64 - DF_BITS(df) / 2)) >> (64 - DF_BITS(df) / 2))
1197
1198#define SIGNED_ODD(a, df) \
1199 ((((int64_t)(a)) << (64 - DF_BITS(df))) >> (64 - DF_BITS(df) / 2))
1200
1201#define UNSIGNED_ODD(a, df) \
1202 ((((uint64_t)(a)) << (64 - DF_BITS(df))) >> (64 - DF_BITS(df) / 2))
1203
1204#define SIGNED_EXTRACT(e, o, a, df) \
1205 do { \
1206 e = SIGNED_EVEN(a, df); \
1207 o = SIGNED_ODD(a, df); \
1208 } while (0)
1209
1210#define UNSIGNED_EXTRACT(e, o, a, df) \
1211 do { \
1212 e = UNSIGNED_EVEN(a, df); \
1213 o = UNSIGNED_ODD(a, df); \
1214 } while (0)
1215
1216static inline int64_t msa_dotp_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1217{
1218 int64_t even_arg1;
1219 int64_t even_arg2;
1220 int64_t odd_arg1;
1221 int64_t odd_arg2;
1222 SIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
1223 SIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
1224 return (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
1225}
1226
1227static inline int64_t msa_dotp_u_df(uint32_t df, int64_t arg1, int64_t arg2)
1228{
1229 int64_t even_arg1;
1230 int64_t even_arg2;
1231 int64_t odd_arg1;
1232 int64_t odd_arg2;
1233 UNSIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
1234 UNSIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
1235 return (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
1236}
1237
1238#define CONCATENATE_AND_SLIDE(s, k) \
1239 do { \
1240 for (i = 0; i < s; i++) { \
1241 v[i] = pws->b[s * k + i]; \
1242 v[i + s] = pwd->b[s * k + i]; \
1243 } \
1244 for (i = 0; i < s; i++) { \
1245 pwd->b[s * k + i] = v[i + n]; \
1246 } \
1247 } while (0)
1248
1249static inline void msa_sld_df(uint32_t df, wr_t *pwd,
1250 wr_t *pws, target_ulong rt)
1251{
1252 uint32_t n = rt % DF_ELEMENTS(df);
1253 uint8_t v[64];
1254 uint32_t i, k;
1255
1256 switch (df) {
1257 case DF_BYTE:
1258 CONCATENATE_AND_SLIDE(DF_ELEMENTS(DF_BYTE), 0);
1259 break;
1260 case DF_HALF:
1261 for (k = 0; k < 2; k++) {
1262 CONCATENATE_AND_SLIDE(DF_ELEMENTS(DF_HALF), k);
1263 }
1264 break;
1265 case DF_WORD:
1266 for (k = 0; k < 4; k++) {
1267 CONCATENATE_AND_SLIDE(DF_ELEMENTS(DF_WORD), k);
1268 }
1269 break;
1270 case DF_DOUBLE:
1271 for (k = 0; k < 8; k++) {
1272 CONCATENATE_AND_SLIDE(DF_ELEMENTS(DF_DOUBLE), k);
1273 }
1274 break;
1275 default:
1276 assert(0);
1277 }
1278}
1279
1280static inline int64_t msa_hadd_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1281{
1282 return SIGNED_ODD(arg1, df) + SIGNED_EVEN(arg2, df);
1283}
1284
1285static inline int64_t msa_hadd_u_df(uint32_t df, int64_t arg1, int64_t arg2)
1286{
1287 return UNSIGNED_ODD(arg1, df) + UNSIGNED_EVEN(arg2, df);
1288}
1289
1290static inline int64_t msa_hsub_s_df(uint32_t df, int64_t arg1, int64_t arg2)
1291{
1292 return SIGNED_ODD(arg1, df) - SIGNED_EVEN(arg2, df);
1293}
1294
1295static inline int64_t msa_hsub_u_df(uint32_t df, int64_t arg1, int64_t arg2)
1296{
1297 return UNSIGNED_ODD(arg1, df) - UNSIGNED_EVEN(arg2, df);
1298}
1299
1300static inline int64_t msa_mul_q_df(uint32_t df, int64_t arg1, int64_t arg2)
1301{
1302 int64_t q_min = DF_MIN_INT(df);
1303 int64_t q_max = DF_MAX_INT(df);
1304
1305 if (arg1 == q_min && arg2 == q_min) {
1306 return q_max;
1307 }
1308 return (arg1 * arg2) >> (DF_BITS(df) - 1);
1309}
1310
1311static inline int64_t msa_mulr_q_df(uint32_t df, int64_t arg1, int64_t arg2)
1312{
1313 int64_t q_min = DF_MIN_INT(df);
1314 int64_t q_max = DF_MAX_INT(df);
1315 int64_t r_bit = 1 << (DF_BITS(df) - 2);
1316
1317 if (arg1 == q_min && arg2 == q_min) {
1318 return q_max;
1319 }
1320 return (arg1 * arg2 + r_bit) >> (DF_BITS(df) - 1);
1321}
1322
1323#define MSA_BINOP_DF(func) \
1324void helper_msa_ ## func ## _df(CPUMIPSState *env, uint32_t df, \
1325 uint32_t wd, uint32_t ws, uint32_t wt) \
1326{ \
1327 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
1328 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
1329 wr_t *pwt = &(env->active_fpu.fpr[wt].wr); \
1330 \
1331 switch (df) { \
1332 case DF_BYTE: \
1333 pwd->b[0] = msa_ ## func ## _df(df, pws->b[0], pwt->b[0]); \
1334 pwd->b[1] = msa_ ## func ## _df(df, pws->b[1], pwt->b[1]); \
1335 pwd->b[2] = msa_ ## func ## _df(df, pws->b[2], pwt->b[2]); \
1336 pwd->b[3] = msa_ ## func ## _df(df, pws->b[3], pwt->b[3]); \
1337 pwd->b[4] = msa_ ## func ## _df(df, pws->b[4], pwt->b[4]); \
1338 pwd->b[5] = msa_ ## func ## _df(df, pws->b[5], pwt->b[5]); \
1339 pwd->b[6] = msa_ ## func ## _df(df, pws->b[6], pwt->b[6]); \
1340 pwd->b[7] = msa_ ## func ## _df(df, pws->b[7], pwt->b[7]); \
1341 pwd->b[8] = msa_ ## func ## _df(df, pws->b[8], pwt->b[8]); \
1342 pwd->b[9] = msa_ ## func ## _df(df, pws->b[9], pwt->b[9]); \
1343 pwd->b[10] = msa_ ## func ## _df(df, pws->b[10], pwt->b[10]); \
1344 pwd->b[11] = msa_ ## func ## _df(df, pws->b[11], pwt->b[11]); \
1345 pwd->b[12] = msa_ ## func ## _df(df, pws->b[12], pwt->b[12]); \
1346 pwd->b[13] = msa_ ## func ## _df(df, pws->b[13], pwt->b[13]); \
1347 pwd->b[14] = msa_ ## func ## _df(df, pws->b[14], pwt->b[14]); \
1348 pwd->b[15] = msa_ ## func ## _df(df, pws->b[15], pwt->b[15]); \
1349 break; \
1350 case DF_HALF: \
1351 pwd->h[0] = msa_ ## func ## _df(df, pws->h[0], pwt->h[0]); \
1352 pwd->h[1] = msa_ ## func ## _df(df, pws->h[1], pwt->h[1]); \
1353 pwd->h[2] = msa_ ## func ## _df(df, pws->h[2], pwt->h[2]); \
1354 pwd->h[3] = msa_ ## func ## _df(df, pws->h[3], pwt->h[3]); \
1355 pwd->h[4] = msa_ ## func ## _df(df, pws->h[4], pwt->h[4]); \
1356 pwd->h[5] = msa_ ## func ## _df(df, pws->h[5], pwt->h[5]); \
1357 pwd->h[6] = msa_ ## func ## _df(df, pws->h[6], pwt->h[6]); \
1358 pwd->h[7] = msa_ ## func ## _df(df, pws->h[7], pwt->h[7]); \
1359 break; \
1360 case DF_WORD: \
1361 pwd->w[0] = msa_ ## func ## _df(df, pws->w[0], pwt->w[0]); \
1362 pwd->w[1] = msa_ ## func ## _df(df, pws->w[1], pwt->w[1]); \
1363 pwd->w[2] = msa_ ## func ## _df(df, pws->w[2], pwt->w[2]); \
1364 pwd->w[3] = msa_ ## func ## _df(df, pws->w[3], pwt->w[3]); \
1365 break; \
1366 case DF_DOUBLE: \
1367 pwd->d[0] = msa_ ## func ## _df(df, pws->d[0], pwt->d[0]); \
1368 pwd->d[1] = msa_ ## func ## _df(df, pws->d[1], pwt->d[1]); \
1369 break; \
1370 default: \
1371 assert(0); \
1372 } \
1373}
1374
1375MSA_BINOP_DF(sll)
1376MSA_BINOP_DF(sra)
1377MSA_BINOP_DF(srl)
1378MSA_BINOP_DF(bclr)
1379MSA_BINOP_DF(bset)
1380MSA_BINOP_DF(bneg)
1381MSA_BINOP_DF(addv)
1382MSA_BINOP_DF(subv)
1383MSA_BINOP_DF(max_s)
1384MSA_BINOP_DF(max_u)
1385MSA_BINOP_DF(min_s)
1386MSA_BINOP_DF(min_u)
1387MSA_BINOP_DF(max_a)
1388MSA_BINOP_DF(min_a)
1389MSA_BINOP_DF(ceq)
1390MSA_BINOP_DF(clt_s)
1391MSA_BINOP_DF(clt_u)
1392MSA_BINOP_DF(cle_s)
1393MSA_BINOP_DF(cle_u)
1394MSA_BINOP_DF(add_a)
1395MSA_BINOP_DF(adds_a)
1396MSA_BINOP_DF(adds_s)
1397MSA_BINOP_DF(adds_u)
1398MSA_BINOP_DF(ave_s)
1399MSA_BINOP_DF(ave_u)
1400MSA_BINOP_DF(aver_s)
1401MSA_BINOP_DF(aver_u)
1402MSA_BINOP_DF(subs_s)
1403MSA_BINOP_DF(subs_u)
1404MSA_BINOP_DF(subsus_u)
1405MSA_BINOP_DF(subsuu_s)
1406MSA_BINOP_DF(asub_s)
1407MSA_BINOP_DF(asub_u)
1408MSA_BINOP_DF(mulv)
1409MSA_BINOP_DF(div_s)
1410MSA_BINOP_DF(div_u)
1411MSA_BINOP_DF(mod_s)
1412MSA_BINOP_DF(mod_u)
1413MSA_BINOP_DF(dotp_s)
1414MSA_BINOP_DF(dotp_u)
1415MSA_BINOP_DF(srar)
1416MSA_BINOP_DF(srlr)
1417MSA_BINOP_DF(hadd_s)
1418MSA_BINOP_DF(hadd_u)
1419MSA_BINOP_DF(hsub_s)
1420MSA_BINOP_DF(hsub_u)
1421
1422MSA_BINOP_DF(mul_q)
1423MSA_BINOP_DF(mulr_q)
1424#undef MSA_BINOP_DF
1425
1426void helper_msa_sld_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
1427 uint32_t ws, uint32_t rt)
1428{
1429 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
1430 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
1431
1432 msa_sld_df(df, pwd, pws, env->active_tc.gpr[rt]);
1433}
1434
1435static inline int64_t msa_maddv_df(uint32_t df, int64_t dest, int64_t arg1,
1436 int64_t arg2)
1437{
1438 return dest + arg1 * arg2;
1439}
1440
1441static inline int64_t msa_msubv_df(uint32_t df, int64_t dest, int64_t arg1,
1442 int64_t arg2)
1443{
1444 return dest - arg1 * arg2;
1445}
1446
1447static inline int64_t msa_dpadd_s_df(uint32_t df, int64_t dest, int64_t arg1,
1448 int64_t arg2)
1449{
1450 int64_t even_arg1;
1451 int64_t even_arg2;
1452 int64_t odd_arg1;
1453 int64_t odd_arg2;
1454 SIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
1455 SIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
1456 return dest + (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
1457}
1458
1459static inline int64_t msa_dpadd_u_df(uint32_t df, int64_t dest, int64_t arg1,
1460 int64_t arg2)
1461{
1462 int64_t even_arg1;
1463 int64_t even_arg2;
1464 int64_t odd_arg1;
1465 int64_t odd_arg2;
1466 UNSIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
1467 UNSIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
1468 return dest + (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
1469}
1470
1471static inline int64_t msa_dpsub_s_df(uint32_t df, int64_t dest, int64_t arg1,
1472 int64_t arg2)
1473{
1474 int64_t even_arg1;
1475 int64_t even_arg2;
1476 int64_t odd_arg1;
1477 int64_t odd_arg2;
1478 SIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
1479 SIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
1480 return dest - ((even_arg1 * even_arg2) + (odd_arg1 * odd_arg2));
1481}
1482
1483static inline int64_t msa_dpsub_u_df(uint32_t df, int64_t dest, int64_t arg1,
1484 int64_t arg2)
1485{
1486 int64_t even_arg1;
1487 int64_t even_arg2;
1488 int64_t odd_arg1;
1489 int64_t odd_arg2;
1490 UNSIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
1491 UNSIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
1492 return dest - ((even_arg1 * even_arg2) + (odd_arg1 * odd_arg2));
1493}
1494
1495static inline int64_t msa_madd_q_df(uint32_t df, int64_t dest, int64_t arg1,
1496 int64_t arg2)
1497{
1498 int64_t q_prod, q_ret;
1499
1500 int64_t q_max = DF_MAX_INT(df);
1501 int64_t q_min = DF_MIN_INT(df);
1502
1503 q_prod = arg1 * arg2;
1504 q_ret = ((dest << (DF_BITS(df) - 1)) + q_prod) >> (DF_BITS(df) - 1);
1505
1506 return (q_ret < q_min) ? q_min : (q_max < q_ret) ? q_max : q_ret;
1507}
1508
1509static inline int64_t msa_msub_q_df(uint32_t df, int64_t dest, int64_t arg1,
1510 int64_t arg2)
1511{
1512 int64_t q_prod, q_ret;
1513
1514 int64_t q_max = DF_MAX_INT(df);
1515 int64_t q_min = DF_MIN_INT(df);
1516
1517 q_prod = arg1 * arg2;
1518 q_ret = ((dest << (DF_BITS(df) - 1)) - q_prod) >> (DF_BITS(df) - 1);
1519
1520 return (q_ret < q_min) ? q_min : (q_max < q_ret) ? q_max : q_ret;
1521}
1522
1523static inline int64_t msa_maddr_q_df(uint32_t df, int64_t dest, int64_t arg1,
1524 int64_t arg2)
1525{
1526 int64_t q_prod, q_ret;
1527
1528 int64_t q_max = DF_MAX_INT(df);
1529 int64_t q_min = DF_MIN_INT(df);
1530 int64_t r_bit = 1 << (DF_BITS(df) - 2);
1531
1532 q_prod = arg1 * arg2;
1533 q_ret = ((dest << (DF_BITS(df) - 1)) + q_prod + r_bit) >> (DF_BITS(df) - 1);
1534
1535 return (q_ret < q_min) ? q_min : (q_max < q_ret) ? q_max : q_ret;
1536}
1537
1538static inline int64_t msa_msubr_q_df(uint32_t df, int64_t dest, int64_t arg1,
1539 int64_t arg2)
1540{
1541 int64_t q_prod, q_ret;
1542
1543 int64_t q_max = DF_MAX_INT(df);
1544 int64_t q_min = DF_MIN_INT(df);
1545 int64_t r_bit = 1 << (DF_BITS(df) - 2);
1546
1547 q_prod = arg1 * arg2;
1548 q_ret = ((dest << (DF_BITS(df) - 1)) - q_prod + r_bit) >> (DF_BITS(df) - 1);
1549
1550 return (q_ret < q_min) ? q_min : (q_max < q_ret) ? q_max : q_ret;
1551}
1552
1553#define MSA_TEROP_DF(func) \
1554void helper_msa_ ## func ## _df(CPUMIPSState *env, uint32_t df, uint32_t wd, \
1555 uint32_t ws, uint32_t wt) \
1556{ \
1557 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
1558 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
1559 wr_t *pwt = &(env->active_fpu.fpr[wt].wr); \
1560 \
1561 switch (df) { \
1562 case DF_BYTE: \
1563 pwd->b[0] = msa_ ## func ## _df(df, pwd->b[0], pws->b[0], \
1564 pwt->b[0]); \
1565 pwd->b[1] = msa_ ## func ## _df(df, pwd->b[1], pws->b[1], \
1566 pwt->b[1]); \
1567 pwd->b[2] = msa_ ## func ## _df(df, pwd->b[2], pws->b[2], \
1568 pwt->b[2]); \
1569 pwd->b[3] = msa_ ## func ## _df(df, pwd->b[3], pws->b[3], \
1570 pwt->b[3]); \
1571 pwd->b[4] = msa_ ## func ## _df(df, pwd->b[4], pws->b[4], \
1572 pwt->b[4]); \
1573 pwd->b[5] = msa_ ## func ## _df(df, pwd->b[5], pws->b[5], \
1574 pwt->b[5]); \
1575 pwd->b[6] = msa_ ## func ## _df(df, pwd->b[6], pws->b[6], \
1576 pwt->b[6]); \
1577 pwd->b[7] = msa_ ## func ## _df(df, pwd->b[7], pws->b[7], \
1578 pwt->b[7]); \
1579 pwd->b[8] = msa_ ## func ## _df(df, pwd->b[8], pws->b[8], \
1580 pwt->b[8]); \
1581 pwd->b[9] = msa_ ## func ## _df(df, pwd->b[9], pws->b[9], \
1582 pwt->b[9]); \
1583 pwd->b[10] = msa_ ## func ## _df(df, pwd->b[10], pws->b[10], \
1584 pwt->b[10]); \
1585 pwd->b[11] = msa_ ## func ## _df(df, pwd->b[11], pws->b[11], \
1586 pwt->b[11]); \
1587 pwd->b[12] = msa_ ## func ## _df(df, pwd->b[12], pws->b[12], \
1588 pwt->b[12]); \
1589 pwd->b[13] = msa_ ## func ## _df(df, pwd->b[13], pws->b[13], \
1590 pwt->b[13]); \
1591 pwd->b[14] = msa_ ## func ## _df(df, pwd->b[14], pws->b[14], \
1592 pwt->b[14]); \
1593 pwd->b[15] = msa_ ## func ## _df(df, pwd->b[15], pws->b[15], \
1594 pwt->b[15]); \
1595 break; \
1596 case DF_HALF: \
1597 pwd->h[0] = msa_ ## func ## _df(df, pwd->h[0], pws->h[0], pwt->h[0]); \
1598 pwd->h[1] = msa_ ## func ## _df(df, pwd->h[1], pws->h[1], pwt->h[1]); \
1599 pwd->h[2] = msa_ ## func ## _df(df, pwd->h[2], pws->h[2], pwt->h[2]); \
1600 pwd->h[3] = msa_ ## func ## _df(df, pwd->h[3], pws->h[3], pwt->h[3]); \
1601 pwd->h[4] = msa_ ## func ## _df(df, pwd->h[4], pws->h[4], pwt->h[4]); \
1602 pwd->h[5] = msa_ ## func ## _df(df, pwd->h[5], pws->h[5], pwt->h[5]); \
1603 pwd->h[6] = msa_ ## func ## _df(df, pwd->h[6], pws->h[6], pwt->h[6]); \
1604 pwd->h[7] = msa_ ## func ## _df(df, pwd->h[7], pws->h[7], pwt->h[7]); \
1605 break; \
1606 case DF_WORD: \
1607 pwd->w[0] = msa_ ## func ## _df(df, pwd->w[0], pws->w[0], pwt->w[0]); \
1608 pwd->w[1] = msa_ ## func ## _df(df, pwd->w[1], pws->w[1], pwt->w[1]); \
1609 pwd->w[2] = msa_ ## func ## _df(df, pwd->w[2], pws->w[2], pwt->w[2]); \
1610 pwd->w[3] = msa_ ## func ## _df(df, pwd->w[3], pws->w[3], pwt->w[3]); \
1611 break; \
1612 case DF_DOUBLE: \
1613 pwd->d[0] = msa_ ## func ## _df(df, pwd->d[0], pws->d[0], pwt->d[0]); \
1614 pwd->d[1] = msa_ ## func ## _df(df, pwd->d[1], pws->d[1], pwt->d[1]); \
1615 break; \
1616 default: \
1617 assert(0); \
1618 } \
1619}
1620
1621MSA_TEROP_DF(maddv)
1622MSA_TEROP_DF(msubv)
1623MSA_TEROP_DF(dpadd_s)
1624MSA_TEROP_DF(dpadd_u)
1625MSA_TEROP_DF(dpsub_s)
1626MSA_TEROP_DF(dpsub_u)
1627MSA_TEROP_DF(binsl)
1628MSA_TEROP_DF(binsr)
1629MSA_TEROP_DF(madd_q)
1630MSA_TEROP_DF(msub_q)
1631MSA_TEROP_DF(maddr_q)
1632MSA_TEROP_DF(msubr_q)
1633#undef MSA_TEROP_DF
1634
1635static inline void msa_splat_df(uint32_t df, wr_t *pwd,
1636 wr_t *pws, target_ulong rt)
1637{
1638 uint32_t n = rt % DF_ELEMENTS(df);
1639 uint32_t i;
1640
1641 switch (df) {
1642 case DF_BYTE:
1643 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) {
1644 pwd->b[i] = pws->b[n];
1645 }
1646 break;
1647 case DF_HALF:
1648 for (i = 0; i < DF_ELEMENTS(DF_HALF); i++) {
1649 pwd->h[i] = pws->h[n];
1650 }
1651 break;
1652 case DF_WORD:
1653 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
1654 pwd->w[i] = pws->w[n];
1655 }
1656 break;
1657 case DF_DOUBLE:
1658 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
1659 pwd->d[i] = pws->d[n];
1660 }
1661 break;
1662 default:
1663 assert(0);
1664 }
1665}
1666
1667void helper_msa_splat_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
1668 uint32_t ws, uint32_t rt)
1669{
1670 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
1671 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
1672
1673 msa_splat_df(df, pwd, pws, env->active_tc.gpr[rt]);
1674}
1675
1676#define MSA_DO_B MSA_DO(b)
1677#define MSA_DO_H MSA_DO(h)
1678#define MSA_DO_W MSA_DO(w)
1679#define MSA_DO_D MSA_DO(d)
1680
1681#define MSA_LOOP_B MSA_LOOP(B)
1682#define MSA_LOOP_H MSA_LOOP(H)
1683#define MSA_LOOP_W MSA_LOOP(W)
1684#define MSA_LOOP_D MSA_LOOP(D)
1685
1686#define MSA_LOOP_COND_B MSA_LOOP_COND(DF_BYTE)
1687#define MSA_LOOP_COND_H MSA_LOOP_COND(DF_HALF)
1688#define MSA_LOOP_COND_W MSA_LOOP_COND(DF_WORD)
1689#define MSA_LOOP_COND_D MSA_LOOP_COND(DF_DOUBLE)
1690
1691#define MSA_LOOP(DF) \
1692 do { \
1693 for (i = 0; i < (MSA_LOOP_COND_ ## DF) ; i++) { \
1694 MSA_DO_ ## DF; \
1695 } \
1696 } while (0)
1697
1698#define MSA_FN_DF(FUNC) \
1699void helper_msa_##FUNC(CPUMIPSState *env, uint32_t df, uint32_t wd, \
1700 uint32_t ws, uint32_t wt) \
1701{ \
1702 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
1703 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
1704 wr_t *pwt = &(env->active_fpu.fpr[wt].wr); \
1705 wr_t wx, *pwx = &wx; \
1706 uint32_t i; \
1707 switch (df) { \
1708 case DF_BYTE: \
1709 MSA_LOOP_B; \
1710 break; \
1711 case DF_HALF: \
1712 MSA_LOOP_H; \
1713 break; \
1714 case DF_WORD: \
1715 MSA_LOOP_W; \
1716 break; \
1717 case DF_DOUBLE: \
1718 MSA_LOOP_D; \
1719 break; \
1720 default: \
1721 assert(0); \
1722 } \
1723 msa_move_v(pwd, pwx); \
1724}
1725
1726#define MSA_LOOP_COND(DF) \
1727 (DF_ELEMENTS(DF) / 2)
1728
1729#define Rb(pwr, i) (pwr->b[i])
1730#define Lb(pwr, i) (pwr->b[i + DF_ELEMENTS(DF_BYTE) / 2])
1731#define Rh(pwr, i) (pwr->h[i])
1732#define Lh(pwr, i) (pwr->h[i + DF_ELEMENTS(DF_HALF) / 2])
1733#define Rw(pwr, i) (pwr->w[i])
1734#define Lw(pwr, i) (pwr->w[i + DF_ELEMENTS(DF_WORD) / 2])
1735#define Rd(pwr, i) (pwr->d[i])
1736#define Ld(pwr, i) (pwr->d[i + DF_ELEMENTS(DF_DOUBLE) / 2])
1737
1738#undef MSA_LOOP_COND
1739
1740#define MSA_LOOP_COND(DF) \
1741 (DF_ELEMENTS(DF))
1742
1743#define MSA_DO(DF) \
1744 do { \
1745 uint32_t n = DF_ELEMENTS(df); \
1746 uint32_t k = (pwd->DF[i] & 0x3f) % (2 * n); \
1747 pwx->DF[i] = \
1748 (pwd->DF[i] & 0xc0) ? 0 : k < n ? pwt->DF[k] : pws->DF[k - n]; \
1749 } while (0)
1750MSA_FN_DF(vshf_df)
1751#undef MSA_DO
1752#undef MSA_LOOP_COND
1753#undef MSA_FN_DF
1754
1755
1756void helper_msa_ilvev_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
1757 uint32_t ws, uint32_t wt)
1758{
1759 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
1760 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
1761 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
1762
1763 switch (df) {
1764 case DF_BYTE:
1765#if defined(HOST_WORDS_BIGENDIAN)
1766 pwd->b[8] = pws->b[9];
1767 pwd->b[9] = pwt->b[9];
1768 pwd->b[10] = pws->b[11];
1769 pwd->b[11] = pwt->b[11];
1770 pwd->b[12] = pws->b[13];
1771 pwd->b[13] = pwt->b[13];
1772 pwd->b[14] = pws->b[15];
1773 pwd->b[15] = pwt->b[15];
1774 pwd->b[0] = pws->b[1];
1775 pwd->b[1] = pwt->b[1];
1776 pwd->b[2] = pws->b[3];
1777 pwd->b[3] = pwt->b[3];
1778 pwd->b[4] = pws->b[5];
1779 pwd->b[5] = pwt->b[5];
1780 pwd->b[6] = pws->b[7];
1781 pwd->b[7] = pwt->b[7];
1782#else
1783 pwd->b[15] = pws->b[14];
1784 pwd->b[14] = pwt->b[14];
1785 pwd->b[13] = pws->b[12];
1786 pwd->b[12] = pwt->b[12];
1787 pwd->b[11] = pws->b[10];
1788 pwd->b[10] = pwt->b[10];
1789 pwd->b[9] = pws->b[8];
1790 pwd->b[8] = pwt->b[8];
1791 pwd->b[7] = pws->b[6];
1792 pwd->b[6] = pwt->b[6];
1793 pwd->b[5] = pws->b[4];
1794 pwd->b[4] = pwt->b[4];
1795 pwd->b[3] = pws->b[2];
1796 pwd->b[2] = pwt->b[2];
1797 pwd->b[1] = pws->b[0];
1798 pwd->b[0] = pwt->b[0];
1799#endif
1800 break;
1801 case DF_HALF:
1802#if defined(HOST_WORDS_BIGENDIAN)
1803 pwd->h[4] = pws->h[5];
1804 pwd->h[5] = pwt->h[5];
1805 pwd->h[6] = pws->h[7];
1806 pwd->h[7] = pwt->h[7];
1807 pwd->h[0] = pws->h[1];
1808 pwd->h[1] = pwt->h[1];
1809 pwd->h[2] = pws->h[3];
1810 pwd->h[3] = pwt->h[3];
1811#else
1812 pwd->h[7] = pws->h[6];
1813 pwd->h[6] = pwt->h[6];
1814 pwd->h[5] = pws->h[4];
1815 pwd->h[4] = pwt->h[4];
1816 pwd->h[3] = pws->h[2];
1817 pwd->h[2] = pwt->h[2];
1818 pwd->h[1] = pws->h[0];
1819 pwd->h[0] = pwt->h[0];
1820#endif
1821 break;
1822 case DF_WORD:
1823#if defined(HOST_WORDS_BIGENDIAN)
1824 pwd->w[2] = pws->w[3];
1825 pwd->w[3] = pwt->w[3];
1826 pwd->w[0] = pws->w[1];
1827 pwd->w[1] = pwt->w[1];
1828#else
1829 pwd->w[3] = pws->w[2];
1830 pwd->w[2] = pwt->w[2];
1831 pwd->w[1] = pws->w[0];
1832 pwd->w[0] = pwt->w[0];
1833#endif
1834 break;
1835 case DF_DOUBLE:
1836 pwd->d[1] = pws->d[0];
1837 pwd->d[0] = pwt->d[0];
1838 break;
1839 default:
1840 assert(0);
1841 }
1842}
1843
1844void helper_msa_ilvod_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
1845 uint32_t ws, uint32_t wt)
1846{
1847 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
1848 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
1849 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
1850
1851 switch (df) {
1852 case DF_BYTE:
1853#if defined(HOST_WORDS_BIGENDIAN)
1854 pwd->b[7] = pwt->b[6];
1855 pwd->b[6] = pws->b[6];
1856 pwd->b[5] = pwt->b[4];
1857 pwd->b[4] = pws->b[4];
1858 pwd->b[3] = pwt->b[2];
1859 pwd->b[2] = pws->b[2];
1860 pwd->b[1] = pwt->b[0];
1861 pwd->b[0] = pws->b[0];
1862 pwd->b[15] = pwt->b[14];
1863 pwd->b[14] = pws->b[14];
1864 pwd->b[13] = pwt->b[12];
1865 pwd->b[12] = pws->b[12];
1866 pwd->b[11] = pwt->b[10];
1867 pwd->b[10] = pws->b[10];
1868 pwd->b[9] = pwt->b[8];
1869 pwd->b[8] = pws->b[8];
1870#else
1871 pwd->b[0] = pwt->b[1];
1872 pwd->b[1] = pws->b[1];
1873 pwd->b[2] = pwt->b[3];
1874 pwd->b[3] = pws->b[3];
1875 pwd->b[4] = pwt->b[5];
1876 pwd->b[5] = pws->b[5];
1877 pwd->b[6] = pwt->b[7];
1878 pwd->b[7] = pws->b[7];
1879 pwd->b[8] = pwt->b[9];
1880 pwd->b[9] = pws->b[9];
1881 pwd->b[10] = pwt->b[11];
1882 pwd->b[11] = pws->b[11];
1883 pwd->b[12] = pwt->b[13];
1884 pwd->b[13] = pws->b[13];
1885 pwd->b[14] = pwt->b[15];
1886 pwd->b[15] = pws->b[15];
1887#endif
1888 break;
1889 case DF_HALF:
1890#if defined(HOST_WORDS_BIGENDIAN)
1891 pwd->h[3] = pwt->h[2];
1892 pwd->h[2] = pws->h[2];
1893 pwd->h[1] = pwt->h[0];
1894 pwd->h[0] = pws->h[0];
1895 pwd->h[7] = pwt->h[6];
1896 pwd->h[6] = pws->h[6];
1897 pwd->h[5] = pwt->h[4];
1898 pwd->h[4] = pws->h[4];
1899#else
1900 pwd->h[0] = pwt->h[1];
1901 pwd->h[1] = pws->h[1];
1902 pwd->h[2] = pwt->h[3];
1903 pwd->h[3] = pws->h[3];
1904 pwd->h[4] = pwt->h[5];
1905 pwd->h[5] = pws->h[5];
1906 pwd->h[6] = pwt->h[7];
1907 pwd->h[7] = pws->h[7];
1908#endif
1909 break;
1910 case DF_WORD:
1911#if defined(HOST_WORDS_BIGENDIAN)
1912 pwd->w[1] = pwt->w[0];
1913 pwd->w[0] = pws->w[0];
1914 pwd->w[3] = pwt->w[2];
1915 pwd->w[2] = pws->w[2];
1916#else
1917 pwd->w[0] = pwt->w[1];
1918 pwd->w[1] = pws->w[1];
1919 pwd->w[2] = pwt->w[3];
1920 pwd->w[3] = pws->w[3];
1921#endif
1922 break;
1923 case DF_DOUBLE:
1924 pwd->d[0] = pwt->d[1];
1925 pwd->d[1] = pws->d[1];
1926 break;
1927 default:
1928 assert(0);
1929 }
1930}
1931
1932void helper_msa_ilvl_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
1933 uint32_t ws, uint32_t wt)
1934{
1935 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
1936 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
1937 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
1938
1939 switch (df) {
1940 case DF_BYTE:
1941#if defined(HOST_WORDS_BIGENDIAN)
1942 pwd->b[7] = pwt->b[15];
1943 pwd->b[6] = pws->b[15];
1944 pwd->b[5] = pwt->b[14];
1945 pwd->b[4] = pws->b[14];
1946 pwd->b[3] = pwt->b[13];
1947 pwd->b[2] = pws->b[13];
1948 pwd->b[1] = pwt->b[12];
1949 pwd->b[0] = pws->b[12];
1950 pwd->b[15] = pwt->b[11];
1951 pwd->b[14] = pws->b[11];
1952 pwd->b[13] = pwt->b[10];
1953 pwd->b[12] = pws->b[10];
1954 pwd->b[11] = pwt->b[9];
1955 pwd->b[10] = pws->b[9];
1956 pwd->b[9] = pwt->b[8];
1957 pwd->b[8] = pws->b[8];
1958#else
1959 pwd->b[0] = pwt->b[8];
1960 pwd->b[1] = pws->b[8];
1961 pwd->b[2] = pwt->b[9];
1962 pwd->b[3] = pws->b[9];
1963 pwd->b[4] = pwt->b[10];
1964 pwd->b[5] = pws->b[10];
1965 pwd->b[6] = pwt->b[11];
1966 pwd->b[7] = pws->b[11];
1967 pwd->b[8] = pwt->b[12];
1968 pwd->b[9] = pws->b[12];
1969 pwd->b[10] = pwt->b[13];
1970 pwd->b[11] = pws->b[13];
1971 pwd->b[12] = pwt->b[14];
1972 pwd->b[13] = pws->b[14];
1973 pwd->b[14] = pwt->b[15];
1974 pwd->b[15] = pws->b[15];
1975#endif
1976 break;
1977 case DF_HALF:
1978#if defined(HOST_WORDS_BIGENDIAN)
1979 pwd->h[3] = pwt->h[7];
1980 pwd->h[2] = pws->h[7];
1981 pwd->h[1] = pwt->h[6];
1982 pwd->h[0] = pws->h[6];
1983 pwd->h[7] = pwt->h[5];
1984 pwd->h[6] = pws->h[5];
1985 pwd->h[5] = pwt->h[4];
1986 pwd->h[4] = pws->h[4];
1987#else
1988 pwd->h[0] = pwt->h[4];
1989 pwd->h[1] = pws->h[4];
1990 pwd->h[2] = pwt->h[5];
1991 pwd->h[3] = pws->h[5];
1992 pwd->h[4] = pwt->h[6];
1993 pwd->h[5] = pws->h[6];
1994 pwd->h[6] = pwt->h[7];
1995 pwd->h[7] = pws->h[7];
1996#endif
1997 break;
1998 case DF_WORD:
1999#if defined(HOST_WORDS_BIGENDIAN)
2000 pwd->w[1] = pwt->w[3];
2001 pwd->w[0] = pws->w[3];
2002 pwd->w[3] = pwt->w[2];
2003 pwd->w[2] = pws->w[2];
2004#else
2005 pwd->w[0] = pwt->w[2];
2006 pwd->w[1] = pws->w[2];
2007 pwd->w[2] = pwt->w[3];
2008 pwd->w[3] = pws->w[3];
2009#endif
2010 break;
2011 case DF_DOUBLE:
2012 pwd->d[0] = pwt->d[1];
2013 pwd->d[1] = pws->d[1];
2014 break;
2015 default:
2016 assert(0);
2017 }
2018}
2019
2020void helper_msa_ilvr_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
2021 uint32_t ws, uint32_t wt)
2022{
2023 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2024 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
2025 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
2026
2027 switch (df) {
2028 case DF_BYTE:
2029#if defined(HOST_WORDS_BIGENDIAN)
2030 pwd->b[8] = pws->b[0];
2031 pwd->b[9] = pwt->b[0];
2032 pwd->b[10] = pws->b[1];
2033 pwd->b[11] = pwt->b[1];
2034 pwd->b[12] = pws->b[2];
2035 pwd->b[13] = pwt->b[2];
2036 pwd->b[14] = pws->b[3];
2037 pwd->b[15] = pwt->b[3];
2038 pwd->b[0] = pws->b[4];
2039 pwd->b[1] = pwt->b[4];
2040 pwd->b[2] = pws->b[5];
2041 pwd->b[3] = pwt->b[5];
2042 pwd->b[4] = pws->b[6];
2043 pwd->b[5] = pwt->b[6];
2044 pwd->b[6] = pws->b[7];
2045 pwd->b[7] = pwt->b[7];
2046#else
2047 pwd->b[15] = pws->b[7];
2048 pwd->b[14] = pwt->b[7];
2049 pwd->b[13] = pws->b[6];
2050 pwd->b[12] = pwt->b[6];
2051 pwd->b[11] = pws->b[5];
2052 pwd->b[10] = pwt->b[5];
2053 pwd->b[9] = pws->b[4];
2054 pwd->b[8] = pwt->b[4];
2055 pwd->b[7] = pws->b[3];
2056 pwd->b[6] = pwt->b[3];
2057 pwd->b[5] = pws->b[2];
2058 pwd->b[4] = pwt->b[2];
2059 pwd->b[3] = pws->b[1];
2060 pwd->b[2] = pwt->b[1];
2061 pwd->b[1] = pws->b[0];
2062 pwd->b[0] = pwt->b[0];
2063#endif
2064 break;
2065 case DF_HALF:
2066#if defined(HOST_WORDS_BIGENDIAN)
2067 pwd->h[4] = pws->h[0];
2068 pwd->h[5] = pwt->h[0];
2069 pwd->h[6] = pws->h[1];
2070 pwd->h[7] = pwt->h[1];
2071 pwd->h[0] = pws->h[2];
2072 pwd->h[1] = pwt->h[2];
2073 pwd->h[2] = pws->h[3];
2074 pwd->h[3] = pwt->h[3];
2075#else
2076 pwd->h[7] = pws->h[3];
2077 pwd->h[6] = pwt->h[3];
2078 pwd->h[5] = pws->h[2];
2079 pwd->h[4] = pwt->h[2];
2080 pwd->h[3] = pws->h[1];
2081 pwd->h[2] = pwt->h[1];
2082 pwd->h[1] = pws->h[0];
2083 pwd->h[0] = pwt->h[0];
2084#endif
2085 break;
2086 case DF_WORD:
2087#if defined(HOST_WORDS_BIGENDIAN)
2088 pwd->w[2] = pws->w[0];
2089 pwd->w[3] = pwt->w[0];
2090 pwd->w[0] = pws->w[1];
2091 pwd->w[1] = pwt->w[1];
2092#else
2093 pwd->w[3] = pws->w[1];
2094 pwd->w[2] = pwt->w[1];
2095 pwd->w[1] = pws->w[0];
2096 pwd->w[0] = pwt->w[0];
2097#endif
2098 break;
2099 case DF_DOUBLE:
2100 pwd->d[1] = pws->d[0];
2101 pwd->d[0] = pwt->d[0];
2102 break;
2103 default:
2104 assert(0);
2105 }
2106}
2107
2108void helper_msa_pckev_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
2109 uint32_t ws, uint32_t wt)
2110{
2111 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2112 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
2113 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
2114
2115 switch (df) {
2116 case DF_BYTE:
2117#if defined(HOST_WORDS_BIGENDIAN)
2118 pwd->b[8] = pws->b[9];
2119 pwd->b[10] = pws->b[13];
2120 pwd->b[12] = pws->b[1];
2121 pwd->b[14] = pws->b[5];
2122 pwd->b[0] = pwt->b[9];
2123 pwd->b[2] = pwt->b[13];
2124 pwd->b[4] = pwt->b[1];
2125 pwd->b[6] = pwt->b[5];
2126 pwd->b[9] = pws->b[11];
2127 pwd->b[13] = pws->b[3];
2128 pwd->b[1] = pwt->b[11];
2129 pwd->b[5] = pwt->b[3];
2130 pwd->b[11] = pws->b[15];
2131 pwd->b[3] = pwt->b[15];
2132 pwd->b[15] = pws->b[7];
2133 pwd->b[7] = pwt->b[7];
2134#else
2135 pwd->b[15] = pws->b[14];
2136 pwd->b[13] = pws->b[10];
2137 pwd->b[11] = pws->b[6];
2138 pwd->b[9] = pws->b[2];
2139 pwd->b[7] = pwt->b[14];
2140 pwd->b[5] = pwt->b[10];
2141 pwd->b[3] = pwt->b[6];
2142 pwd->b[1] = pwt->b[2];
2143 pwd->b[14] = pws->b[12];
2144 pwd->b[10] = pws->b[4];
2145 pwd->b[6] = pwt->b[12];
2146 pwd->b[2] = pwt->b[4];
2147 pwd->b[12] = pws->b[8];
2148 pwd->b[4] = pwt->b[8];
2149 pwd->b[8] = pws->b[0];
2150 pwd->b[0] = pwt->b[0];
2151#endif
2152 break;
2153 case DF_HALF:
2154#if defined(HOST_WORDS_BIGENDIAN)
2155 pwd->h[4] = pws->h[5];
2156 pwd->h[6] = pws->h[1];
2157 pwd->h[0] = pwt->h[5];
2158 pwd->h[2] = pwt->h[1];
2159 pwd->h[5] = pws->h[7];
2160 pwd->h[1] = pwt->h[7];
2161 pwd->h[7] = pws->h[3];
2162 pwd->h[3] = pwt->h[3];
2163#else
2164 pwd->h[7] = pws->h[6];
2165 pwd->h[5] = pws->h[2];
2166 pwd->h[3] = pwt->h[6];
2167 pwd->h[1] = pwt->h[2];
2168 pwd->h[6] = pws->h[4];
2169 pwd->h[2] = pwt->h[4];
2170 pwd->h[4] = pws->h[0];
2171 pwd->h[0] = pwt->h[0];
2172#endif
2173 break;
2174 case DF_WORD:
2175#if defined(HOST_WORDS_BIGENDIAN)
2176 pwd->w[2] = pws->w[3];
2177 pwd->w[0] = pwt->w[3];
2178 pwd->w[3] = pws->w[1];
2179 pwd->w[1] = pwt->w[1];
2180#else
2181 pwd->w[3] = pws->w[2];
2182 pwd->w[1] = pwt->w[2];
2183 pwd->w[2] = pws->w[0];
2184 pwd->w[0] = pwt->w[0];
2185#endif
2186 break;
2187 case DF_DOUBLE:
2188 pwd->d[1] = pws->d[0];
2189 pwd->d[0] = pwt->d[0];
2190 break;
2191 default:
2192 assert(0);
2193 }
2194}
2195
2196void helper_msa_pckod_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
2197 uint32_t ws, uint32_t wt)
2198{
2199 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2200 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
2201 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
2202
2203 switch (df) {
2204 case DF_BYTE:
2205#if defined(HOST_WORDS_BIGENDIAN)
2206 pwd->b[7] = pwt->b[6];
2207 pwd->b[5] = pwt->b[2];
2208 pwd->b[3] = pwt->b[14];
2209 pwd->b[1] = pwt->b[10];
2210 pwd->b[15] = pws->b[6];
2211 pwd->b[13] = pws->b[2];
2212 pwd->b[11] = pws->b[14];
2213 pwd->b[9] = pws->b[10];
2214 pwd->b[6] = pwt->b[4];
2215 pwd->b[2] = pwt->b[12];
2216 pwd->b[14] = pws->b[4];
2217 pwd->b[10] = pws->b[12];
2218 pwd->b[4] = pwt->b[0];
2219 pwd->b[12] = pws->b[0];
2220 pwd->b[0] = pwt->b[8];
2221 pwd->b[8] = pws->b[8];
2222#else
2223 pwd->b[0] = pwt->b[1];
2224 pwd->b[2] = pwt->b[5];
2225 pwd->b[4] = pwt->b[9];
2226 pwd->b[6] = pwt->b[13];
2227 pwd->b[8] = pws->b[1];
2228 pwd->b[10] = pws->b[5];
2229 pwd->b[12] = pws->b[9];
2230 pwd->b[14] = pws->b[13];
2231 pwd->b[1] = pwt->b[3];
2232 pwd->b[5] = pwt->b[11];
2233 pwd->b[9] = pws->b[3];
2234 pwd->b[13] = pws->b[11];
2235 pwd->b[3] = pwt->b[7];
2236 pwd->b[11] = pws->b[7];
2237 pwd->b[7] = pwt->b[15];
2238 pwd->b[15] = pws->b[15];
2239#endif
2240 break;
2241 case DF_HALF:
2242#if defined(HOST_WORDS_BIGENDIAN)
2243 pwd->h[3] = pwt->h[2];
2244 pwd->h[1] = pwt->h[6];
2245 pwd->h[7] = pws->h[2];
2246 pwd->h[5] = pws->h[6];
2247 pwd->h[2] = pwt->h[0];
2248 pwd->h[6] = pws->h[0];
2249 pwd->h[0] = pwt->h[4];
2250 pwd->h[4] = pws->h[4];
2251#else
2252 pwd->h[0] = pwt->h[1];
2253 pwd->h[2] = pwt->h[5];
2254 pwd->h[4] = pws->h[1];
2255 pwd->h[6] = pws->h[5];
2256 pwd->h[1] = pwt->h[3];
2257 pwd->h[5] = pws->h[3];
2258 pwd->h[3] = pwt->h[7];
2259 pwd->h[7] = pws->h[7];
2260#endif
2261 break;
2262 case DF_WORD:
2263#if defined(HOST_WORDS_BIGENDIAN)
2264 pwd->w[1] = pwt->w[0];
2265 pwd->w[3] = pws->w[0];
2266 pwd->w[0] = pwt->w[2];
2267 pwd->w[2] = pws->w[2];
2268#else
2269 pwd->w[0] = pwt->w[1];
2270 pwd->w[2] = pws->w[1];
2271 pwd->w[1] = pwt->w[3];
2272 pwd->w[3] = pws->w[3];
2273#endif
2274 break;
2275 case DF_DOUBLE:
2276 pwd->d[0] = pwt->d[1];
2277 pwd->d[1] = pws->d[1];
2278 break;
2279 default:
2280 assert(0);
2281 }
2282}
2283
2284
2285void helper_msa_sldi_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
2286 uint32_t ws, uint32_t n)
2287{
2288 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2289 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
2290
2291 msa_sld_df(df, pwd, pws, n);
2292}
2293
2294void helper_msa_splati_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
2295 uint32_t ws, uint32_t n)
2296{
2297 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2298 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
2299
2300 msa_splat_df(df, pwd, pws, n);
2301}
2302
2303void helper_msa_copy_s_b(CPUMIPSState *env, uint32_t rd,
2304 uint32_t ws, uint32_t n)
2305{
2306 n %= 16;
2307#if defined(HOST_WORDS_BIGENDIAN)
2308 if (n < 8) {
2309 n = 8 - n - 1;
2310 } else {
2311 n = 24 - n - 1;
2312 }
2313#endif
2314 env->active_tc.gpr[rd] = (int8_t)env->active_fpu.fpr[ws].wr.b[n];
2315}
2316
2317void helper_msa_copy_s_h(CPUMIPSState *env, uint32_t rd,
2318 uint32_t ws, uint32_t n)
2319{
2320 n %= 8;
2321#if defined(HOST_WORDS_BIGENDIAN)
2322 if (n < 4) {
2323 n = 4 - n - 1;
2324 } else {
2325 n = 12 - n - 1;
2326 }
2327#endif
2328 env->active_tc.gpr[rd] = (int16_t)env->active_fpu.fpr[ws].wr.h[n];
2329}
2330
2331void helper_msa_copy_s_w(CPUMIPSState *env, uint32_t rd,
2332 uint32_t ws, uint32_t n)
2333{
2334 n %= 4;
2335#if defined(HOST_WORDS_BIGENDIAN)
2336 if (n < 2) {
2337 n = 2 - n - 1;
2338 } else {
2339 n = 6 - n - 1;
2340 }
2341#endif
2342 env->active_tc.gpr[rd] = (int32_t)env->active_fpu.fpr[ws].wr.w[n];
2343}
2344
2345void helper_msa_copy_s_d(CPUMIPSState *env, uint32_t rd,
2346 uint32_t ws, uint32_t n)
2347{
2348 n %= 2;
2349 env->active_tc.gpr[rd] = (int64_t)env->active_fpu.fpr[ws].wr.d[n];
2350}
2351
2352void helper_msa_copy_u_b(CPUMIPSState *env, uint32_t rd,
2353 uint32_t ws, uint32_t n)
2354{
2355 n %= 16;
2356#if defined(HOST_WORDS_BIGENDIAN)
2357 if (n < 8) {
2358 n = 8 - n - 1;
2359 } else {
2360 n = 24 - n - 1;
2361 }
2362#endif
2363 env->active_tc.gpr[rd] = (uint8_t)env->active_fpu.fpr[ws].wr.b[n];
2364}
2365
2366void helper_msa_copy_u_h(CPUMIPSState *env, uint32_t rd,
2367 uint32_t ws, uint32_t n)
2368{
2369 n %= 8;
2370#if defined(HOST_WORDS_BIGENDIAN)
2371 if (n < 4) {
2372 n = 4 - n - 1;
2373 } else {
2374 n = 12 - n - 1;
2375 }
2376#endif
2377 env->active_tc.gpr[rd] = (uint16_t)env->active_fpu.fpr[ws].wr.h[n];
2378}
2379
2380void helper_msa_copy_u_w(CPUMIPSState *env, uint32_t rd,
2381 uint32_t ws, uint32_t n)
2382{
2383 n %= 4;
2384#if defined(HOST_WORDS_BIGENDIAN)
2385 if (n < 2) {
2386 n = 2 - n - 1;
2387 } else {
2388 n = 6 - n - 1;
2389 }
2390#endif
2391 env->active_tc.gpr[rd] = (uint32_t)env->active_fpu.fpr[ws].wr.w[n];
2392}
2393
2394void helper_msa_insert_b(CPUMIPSState *env, uint32_t wd,
2395 uint32_t rs_num, uint32_t n)
2396{
2397 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2398 target_ulong rs = env->active_tc.gpr[rs_num];
2399 n %= 16;
2400#if defined(HOST_WORDS_BIGENDIAN)
2401 if (n < 8) {
2402 n = 8 - n - 1;
2403 } else {
2404 n = 24 - n - 1;
2405 }
2406#endif
2407 pwd->b[n] = (int8_t)rs;
2408}
2409
2410void helper_msa_insert_h(CPUMIPSState *env, uint32_t wd,
2411 uint32_t rs_num, uint32_t n)
2412{
2413 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2414 target_ulong rs = env->active_tc.gpr[rs_num];
2415 n %= 8;
2416#if defined(HOST_WORDS_BIGENDIAN)
2417 if (n < 4) {
2418 n = 4 - n - 1;
2419 } else {
2420 n = 12 - n - 1;
2421 }
2422#endif
2423 pwd->h[n] = (int16_t)rs;
2424}
2425
2426void helper_msa_insert_w(CPUMIPSState *env, uint32_t wd,
2427 uint32_t rs_num, uint32_t n)
2428{
2429 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2430 target_ulong rs = env->active_tc.gpr[rs_num];
2431 n %= 4;
2432#if defined(HOST_WORDS_BIGENDIAN)
2433 if (n < 2) {
2434 n = 2 - n - 1;
2435 } else {
2436 n = 6 - n - 1;
2437 }
2438#endif
2439 pwd->w[n] = (int32_t)rs;
2440}
2441
2442void helper_msa_insert_d(CPUMIPSState *env, uint32_t wd,
2443 uint32_t rs_num, uint32_t n)
2444{
2445 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2446 target_ulong rs = env->active_tc.gpr[rs_num];
2447 n %= 2;
2448 pwd->d[n] = (int64_t)rs;
2449}
2450
2451void helper_msa_insve_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
2452 uint32_t ws, uint32_t n)
2453{
2454 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2455 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
2456
2457 switch (df) {
2458 case DF_BYTE:
2459 pwd->b[n] = (int8_t)pws->b[0];
2460 break;
2461 case DF_HALF:
2462 pwd->h[n] = (int16_t)pws->h[0];
2463 break;
2464 case DF_WORD:
2465 pwd->w[n] = (int32_t)pws->w[0];
2466 break;
2467 case DF_DOUBLE:
2468 pwd->d[n] = (int64_t)pws->d[0];
2469 break;
2470 default:
2471 assert(0);
2472 }
2473}
2474
2475void helper_msa_ctcmsa(CPUMIPSState *env, target_ulong elm, uint32_t cd)
2476{
2477 switch (cd) {
2478 case 0:
2479 break;
2480 case 1:
2481 env->active_tc.msacsr = (int32_t)elm & MSACSR_MASK;
2482 restore_msa_fp_status(env);
2483 /* check exception */
2484 if ((GET_FP_ENABLE(env->active_tc.msacsr) | FP_UNIMPLEMENTED)
2485 & GET_FP_CAUSE(env->active_tc.msacsr)) {
2486 do_raise_exception(env, EXCP_MSAFPE, GETPC());
2487 }
2488 break;
2489 }
2490}
2491
2492target_ulong helper_msa_cfcmsa(CPUMIPSState *env, uint32_t cs)
2493{
2494 switch (cs) {
2495 case 0:
2496 return env->msair;
2497 case 1:
2498 return env->active_tc.msacsr & MSACSR_MASK;
2499 }
2500 return 0;
2501}
2502
2503void helper_msa_move_v(CPUMIPSState *env, uint32_t wd, uint32_t ws)
2504{
2505 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2506 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
2507
2508 msa_move_v(pwd, pws);
2509}
2510
2511static inline int64_t msa_pcnt_df(uint32_t df, int64_t arg)
2512{
2513 uint64_t x;
2514
2515 x = UNSIGNED(arg, df);
2516
2517 x = (x & 0x5555555555555555ULL) + ((x >> 1) & 0x5555555555555555ULL);
2518 x = (x & 0x3333333333333333ULL) + ((x >> 2) & 0x3333333333333333ULL);
2519 x = (x & 0x0F0F0F0F0F0F0F0FULL) + ((x >> 4) & 0x0F0F0F0F0F0F0F0FULL);
2520 x = (x & 0x00FF00FF00FF00FFULL) + ((x >> 8) & 0x00FF00FF00FF00FFULL);
2521 x = (x & 0x0000FFFF0000FFFFULL) + ((x >> 16) & 0x0000FFFF0000FFFFULL);
2522 x = (x & 0x00000000FFFFFFFFULL) + ((x >> 32));
2523
2524 return x;
2525}
2526
2527static inline int64_t msa_nlzc_df(uint32_t df, int64_t arg)
2528{
2529 uint64_t x, y;
2530 int n, c;
2531
2532 x = UNSIGNED(arg, df);
2533 n = DF_BITS(df);
2534 c = DF_BITS(df) / 2;
2535
2536 do {
2537 y = x >> c;
2538 if (y != 0) {
2539 n = n - c;
2540 x = y;
2541 }
2542 c = c >> 1;
2543 } while (c != 0);
2544
2545 return n - x;
2546}
2547
2548static inline int64_t msa_nloc_df(uint32_t df, int64_t arg)
2549{
2550 return msa_nlzc_df(df, UNSIGNED((~arg), df));
2551}
2552
2553void helper_msa_fill_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
2554 uint32_t rs)
2555{
2556 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
2557 uint32_t i;
2558
2559 switch (df) {
2560 case DF_BYTE:
2561 for (i = 0; i < DF_ELEMENTS(DF_BYTE); i++) {
2562 pwd->b[i] = (int8_t)env->active_tc.gpr[rs];
2563 }
2564 break;
2565 case DF_HALF:
2566 for (i = 0; i < DF_ELEMENTS(DF_HALF); i++) {
2567 pwd->h[i] = (int16_t)env->active_tc.gpr[rs];
2568 }
2569 break;
2570 case DF_WORD:
2571 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
2572 pwd->w[i] = (int32_t)env->active_tc.gpr[rs];
2573 }
2574 break;
2575 case DF_DOUBLE:
2576 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
2577 pwd->d[i] = (int64_t)env->active_tc.gpr[rs];
2578 }
2579 break;
2580 default:
2581 assert(0);
2582 }
2583}
2584
2585#define MSA_UNOP_DF(func) \
2586void helper_msa_ ## func ## _df(CPUMIPSState *env, uint32_t df, \
2587 uint32_t wd, uint32_t ws) \
2588{ \
2589 wr_t *pwd = &(env->active_fpu.fpr[wd].wr); \
2590 wr_t *pws = &(env->active_fpu.fpr[ws].wr); \
2591 \
2592 switch (df) { \
2593 case DF_BYTE: \
2594 pwd->b[0] = msa_ ## func ## _df(df, pws->b[0]); \
2595 pwd->b[1] = msa_ ## func ## _df(df, pws->b[1]); \
2596 pwd->b[2] = msa_ ## func ## _df(df, pws->b[2]); \
2597 pwd->b[3] = msa_ ## func ## _df(df, pws->b[3]); \
2598 pwd->b[4] = msa_ ## func ## _df(df, pws->b[4]); \
2599 pwd->b[5] = msa_ ## func ## _df(df, pws->b[5]); \
2600 pwd->b[6] = msa_ ## func ## _df(df, pws->b[6]); \
2601 pwd->b[7] = msa_ ## func ## _df(df, pws->b[7]); \
2602 pwd->b[8] = msa_ ## func ## _df(df, pws->b[8]); \
2603 pwd->b[9] = msa_ ## func ## _df(df, pws->b[9]); \
2604 pwd->b[10] = msa_ ## func ## _df(df, pws->b[10]); \
2605 pwd->b[11] = msa_ ## func ## _df(df, pws->b[11]); \
2606 pwd->b[12] = msa_ ## func ## _df(df, pws->b[12]); \
2607 pwd->b[13] = msa_ ## func ## _df(df, pws->b[13]); \
2608 pwd->b[14] = msa_ ## func ## _df(df, pws->b[14]); \
2609 pwd->b[15] = msa_ ## func ## _df(df, pws->b[15]); \
2610 break; \
2611 case DF_HALF: \
2612 pwd->h[0] = msa_ ## func ## _df(df, pws->h[0]); \
2613 pwd->h[1] = msa_ ## func ## _df(df, pws->h[1]); \
2614 pwd->h[2] = msa_ ## func ## _df(df, pws->h[2]); \
2615 pwd->h[3] = msa_ ## func ## _df(df, pws->h[3]); \
2616 pwd->h[4] = msa_ ## func ## _df(df, pws->h[4]); \
2617 pwd->h[5] = msa_ ## func ## _df(df, pws->h[5]); \
2618 pwd->h[6] = msa_ ## func ## _df(df, pws->h[6]); \
2619 pwd->h[7] = msa_ ## func ## _df(df, pws->h[7]); \
2620 break; \
2621 case DF_WORD: \
2622 pwd->w[0] = msa_ ## func ## _df(df, pws->w[0]); \
2623 pwd->w[1] = msa_ ## func ## _df(df, pws->w[1]); \
2624 pwd->w[2] = msa_ ## func ## _df(df, pws->w[2]); \
2625 pwd->w[3] = msa_ ## func ## _df(df, pws->w[3]); \
2626 break; \
2627 case DF_DOUBLE: \
2628 pwd->d[0] = msa_ ## func ## _df(df, pws->d[0]); \
2629 pwd->d[1] = msa_ ## func ## _df(df, pws->d[1]); \
2630 break; \
2631 default: \
2632 assert(0); \
2633 } \
2634}
2635
2636MSA_UNOP_DF(nlzc)
2637MSA_UNOP_DF(nloc)
2638MSA_UNOP_DF(pcnt)
2639#undef MSA_UNOP_DF
2640
2641#define FLOAT_ONE32 make_float32(0x3f8 << 20)
2642#define FLOAT_ONE64 make_float64(0x3ffULL << 52)
2643
2644#define FLOAT_SNAN16(s) (float16_default_nan(s) ^ 0x0220)
2645 /* 0x7c20 */
2646#define FLOAT_SNAN32(s) (float32_default_nan(s) ^ 0x00400020)
2647 /* 0x7f800020 */
2648#define FLOAT_SNAN64(s) (float64_default_nan(s) ^ 0x0008000000000020ULL)
2649 /* 0x7ff0000000000020 */
2650
2651static inline void clear_msacsr_cause(CPUMIPSState *env)
2652{
2653 SET_FP_CAUSE(env->active_tc.msacsr, 0);
2654}
2655
2656static inline void check_msacsr_cause(CPUMIPSState *env, uintptr_t retaddr)
2657{
2658 if ((GET_FP_CAUSE(env->active_tc.msacsr) &
2659 (GET_FP_ENABLE(env->active_tc.msacsr) | FP_UNIMPLEMENTED)) == 0) {
2660 UPDATE_FP_FLAGS(env->active_tc.msacsr,
2661 GET_FP_CAUSE(env->active_tc.msacsr));
2662 } else {
2663 do_raise_exception(env, EXCP_MSAFPE, retaddr);
2664 }
2665}
2666
2667/* Flush-to-zero use cases for update_msacsr() */
2668#define CLEAR_FS_UNDERFLOW 1
2669#define CLEAR_IS_INEXACT 2
2670#define RECIPROCAL_INEXACT 4
2671
2672static inline int update_msacsr(CPUMIPSState *env, int action, int denormal)
2673{
2674 int ieee_ex;
2675
2676 int c;
2677 int cause;
2678 int enable;
2679
2680 ieee_ex = get_float_exception_flags(&env->active_tc.msa_fp_status);
2681
2682 /* QEMU softfloat does not signal all underflow cases */
2683 if (denormal) {
2684 ieee_ex |= float_flag_underflow;
2685 }
2686
2687 c = ieee_ex_to_mips(ieee_ex);
2688 enable = GET_FP_ENABLE(env->active_tc.msacsr) | FP_UNIMPLEMENTED;
2689
2690 /* Set Inexact (I) when flushing inputs to zero */
2691 if ((ieee_ex & float_flag_input_denormal) &&
2692 (env->active_tc.msacsr & MSACSR_FS_MASK) != 0) {
2693 if (action & CLEAR_IS_INEXACT) {
2694 c &= ~FP_INEXACT;
2695 } else {
2696 c |= FP_INEXACT;
2697 }
2698 }
2699
2700 /* Set Inexact (I) and Underflow (U) when flushing outputs to zero */
2701 if ((ieee_ex & float_flag_output_denormal) &&
2702 (env->active_tc.msacsr & MSACSR_FS_MASK) != 0) {
2703 c |= FP_INEXACT;
2704 if (action & CLEAR_FS_UNDERFLOW) {
2705 c &= ~FP_UNDERFLOW;
2706 } else {
2707 c |= FP_UNDERFLOW;
2708 }
2709 }
2710
2711 /* Set Inexact (I) when Overflow (O) is not enabled */
2712 if ((c & FP_OVERFLOW) != 0 && (enable & FP_OVERFLOW) == 0) {
2713 c |= FP_INEXACT;
2714 }
2715
2716 /* Clear Exact Underflow when Underflow (U) is not enabled */
2717 if ((c & FP_UNDERFLOW) != 0 && (enable & FP_UNDERFLOW) == 0 &&
2718 (c & FP_INEXACT) == 0) {
2719 c &= ~FP_UNDERFLOW;
2720 }
2721
2722 /*
2723 * Reciprocal operations set only Inexact when valid and not
2724 * divide by zero
2725 */
2726 if ((action & RECIPROCAL_INEXACT) &&
2727 (c & (FP_INVALID | FP_DIV0)) == 0) {
2728 c = FP_INEXACT;
2729 }
2730
2731 cause = c & enable; /* all current enabled exceptions */
2732
2733 if (cause == 0) {
2734 /*
2735 * No enabled exception, update the MSACSR Cause
2736 * with all current exceptions
2737 */
2738 SET_FP_CAUSE(env->active_tc.msacsr,
2739 (GET_FP_CAUSE(env->active_tc.msacsr) | c));
2740 } else {
2741 /* Current exceptions are enabled */
2742 if ((env->active_tc.msacsr & MSACSR_NX_MASK) == 0) {
2743 /*
2744 * Exception(s) will trap, update MSACSR Cause
2745 * with all enabled exceptions
2746 */
2747 SET_FP_CAUSE(env->active_tc.msacsr,
2748 (GET_FP_CAUSE(env->active_tc.msacsr) | c));
2749 }
2750 }
2751
2752 return c;
2753}
2754
2755static inline int get_enabled_exceptions(const CPUMIPSState *env, int c)
2756{
2757 int enable = GET_FP_ENABLE(env->active_tc.msacsr) | FP_UNIMPLEMENTED;
2758 return c & enable;
2759}
2760
2761static inline float16 float16_from_float32(int32_t a, flag ieee,
2762 float_status *status)
2763{
2764 float16 f_val;
2765
2766 f_val = float32_to_float16((float32)a, ieee, status);
2767
2768 return a < 0 ? (f_val | (1 << 15)) : f_val;
2769}
2770
2771static inline float32 float32_from_float64(int64_t a, float_status *status)
2772{
2773 float32 f_val;
2774
2775 f_val = float64_to_float32((float64)a, status);
2776
2777 return a < 0 ? (f_val | (1 << 31)) : f_val;
2778}
2779
2780static inline float32 float32_from_float16(int16_t a, flag ieee,
2781 float_status *status)
2782{
2783 float32 f_val;
2784
2785 f_val = float16_to_float32((float16)a, ieee, status);
2786
2787 return a < 0 ? (f_val | (1 << 31)) : f_val;
2788}
2789
2790static inline float64 float64_from_float32(int32_t a, float_status *status)
2791{
2792 float64 f_val;
2793
2794 f_val = float32_to_float64((float64)a, status);
2795
2796 return a < 0 ? (f_val | (1ULL << 63)) : f_val;
2797}
2798
2799static inline float32 float32_from_q16(int16_t a, float_status *status)
2800{
2801 float32 f_val;
2802
2803 /* conversion as integer and scaling */
2804 f_val = int32_to_float32(a, status);
2805 f_val = float32_scalbn(f_val, -15, status);
2806
2807 return f_val;
2808}
2809
2810static inline float64 float64_from_q32(int32_t a, float_status *status)
2811{
2812 float64 f_val;
2813
2814 /* conversion as integer and scaling */
2815 f_val = int32_to_float64(a, status);
2816 f_val = float64_scalbn(f_val, -31, status);
2817
2818 return f_val;
2819}
2820
2821static inline int16_t float32_to_q16(float32 a, float_status *status)
2822{
2823 int32_t q_val;
2824 int32_t q_min = 0xffff8000;
2825 int32_t q_max = 0x00007fff;
2826
2827 int ieee_ex;
2828
2829 if (float32_is_any_nan(a)) {
2830 float_raise(float_flag_invalid, status);
2831 return 0;
2832 }
2833
2834 /* scaling */
2835 a = float32_scalbn(a, 15, status);
2836
2837 ieee_ex = get_float_exception_flags(status);
2838 set_float_exception_flags(ieee_ex & (~float_flag_underflow)
2839 , status);
2840
2841 if (ieee_ex & float_flag_overflow) {
2842 float_raise(float_flag_inexact, status);
2843 return (int32_t)a < 0 ? q_min : q_max;
2844 }
2845
2846 /* conversion to int */
2847 q_val = float32_to_int32(a, status);
2848
2849 ieee_ex = get_float_exception_flags(status);
2850 set_float_exception_flags(ieee_ex & (~float_flag_underflow)
2851 , status);
2852
2853 if (ieee_ex & float_flag_invalid) {
2854 set_float_exception_flags(ieee_ex & (~float_flag_invalid)
2855 , status);
2856 float_raise(float_flag_overflow | float_flag_inexact, status);
2857 return (int32_t)a < 0 ? q_min : q_max;
2858 }
2859
2860 if (q_val < q_min) {
2861 float_raise(float_flag_overflow | float_flag_inexact, status);
2862 return (int16_t)q_min;
2863 }
2864
2865 if (q_max < q_val) {
2866 float_raise(float_flag_overflow | float_flag_inexact, status);
2867 return (int16_t)q_max;
2868 }
2869
2870 return (int16_t)q_val;
2871}
2872
2873static inline int32_t float64_to_q32(float64 a, float_status *status)
2874{
2875 int64_t q_val;
2876 int64_t q_min = 0xffffffff80000000LL;
2877 int64_t q_max = 0x000000007fffffffLL;
2878
2879 int ieee_ex;
2880
2881 if (float64_is_any_nan(a)) {
2882 float_raise(float_flag_invalid, status);
2883 return 0;
2884 }
2885
2886 /* scaling */
2887 a = float64_scalbn(a, 31, status);
2888
2889 ieee_ex = get_float_exception_flags(status);
2890 set_float_exception_flags(ieee_ex & (~float_flag_underflow)
2891 , status);
2892
2893 if (ieee_ex & float_flag_overflow) {
2894 float_raise(float_flag_inexact, status);
2895 return (int64_t)a < 0 ? q_min : q_max;
2896 }
2897
2898 /* conversion to integer */
2899 q_val = float64_to_int64(a, status);
2900
2901 ieee_ex = get_float_exception_flags(status);
2902 set_float_exception_flags(ieee_ex & (~float_flag_underflow)
2903 , status);
2904
2905 if (ieee_ex & float_flag_invalid) {
2906 set_float_exception_flags(ieee_ex & (~float_flag_invalid)
2907 , status);
2908 float_raise(float_flag_overflow | float_flag_inexact, status);
2909 return (int64_t)a < 0 ? q_min : q_max;
2910 }
2911
2912 if (q_val < q_min) {
2913 float_raise(float_flag_overflow | float_flag_inexact, status);
2914 return (int32_t)q_min;
2915 }
2916
2917 if (q_max < q_val) {
2918 float_raise(float_flag_overflow | float_flag_inexact, status);
2919 return (int32_t)q_max;
2920 }
2921
2922 return (int32_t)q_val;
2923}
2924
2925#define MSA_FLOAT_COND(DEST, OP, ARG1, ARG2, BITS, QUIET) \
2926 do { \
2927 float_status *status = &env->active_tc.msa_fp_status; \
2928 int c; \
2929 int64_t cond; \
2930 set_float_exception_flags(0, status); \
2931 if (!QUIET) { \
2932 cond = float ## BITS ## _ ## OP(ARG1, ARG2, status); \
2933 } else { \
2934 cond = float ## BITS ## _ ## OP ## _quiet(ARG1, ARG2, status); \
2935 } \
2936 DEST = cond ? M_MAX_UINT(BITS) : 0; \
2937 c = update_msacsr(env, CLEAR_IS_INEXACT, 0); \
2938 \
2939 if (get_enabled_exceptions(env, c)) { \
2940 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
2941 } \
2942 } while (0)
2943
2944#define MSA_FLOAT_AF(DEST, ARG1, ARG2, BITS, QUIET) \
2945 do { \
2946 MSA_FLOAT_COND(DEST, eq, ARG1, ARG2, BITS, QUIET); \
2947 if ((DEST & M_MAX_UINT(BITS)) == M_MAX_UINT(BITS)) { \
2948 DEST = 0; \
2949 } \
2950 } while (0)
2951
2952#define MSA_FLOAT_UEQ(DEST, ARG1, ARG2, BITS, QUIET) \
2953 do { \
2954 MSA_FLOAT_COND(DEST, unordered, ARG1, ARG2, BITS, QUIET); \
2955 if (DEST == 0) { \
2956 MSA_FLOAT_COND(DEST, eq, ARG1, ARG2, BITS, QUIET); \
2957 } \
2958 } while (0)
2959
2960#define MSA_FLOAT_NE(DEST, ARG1, ARG2, BITS, QUIET) \
2961 do { \
2962 MSA_FLOAT_COND(DEST, lt, ARG1, ARG2, BITS, QUIET); \
2963 if (DEST == 0) { \
2964 MSA_FLOAT_COND(DEST, lt, ARG2, ARG1, BITS, QUIET); \
2965 } \
2966 } while (0)
2967
2968#define MSA_FLOAT_UNE(DEST, ARG1, ARG2, BITS, QUIET) \
2969 do { \
2970 MSA_FLOAT_COND(DEST, unordered, ARG1, ARG2, BITS, QUIET); \
2971 if (DEST == 0) { \
2972 MSA_FLOAT_COND(DEST, lt, ARG1, ARG2, BITS, QUIET); \
2973 if (DEST == 0) { \
2974 MSA_FLOAT_COND(DEST, lt, ARG2, ARG1, BITS, QUIET); \
2975 } \
2976 } \
2977 } while (0)
2978
2979#define MSA_FLOAT_ULE(DEST, ARG1, ARG2, BITS, QUIET) \
2980 do { \
2981 MSA_FLOAT_COND(DEST, unordered, ARG1, ARG2, BITS, QUIET); \
2982 if (DEST == 0) { \
2983 MSA_FLOAT_COND(DEST, le, ARG1, ARG2, BITS, QUIET); \
2984 } \
2985 } while (0)
2986
2987#define MSA_FLOAT_ULT(DEST, ARG1, ARG2, BITS, QUIET) \
2988 do { \
2989 MSA_FLOAT_COND(DEST, unordered, ARG1, ARG2, BITS, QUIET); \
2990 if (DEST == 0) { \
2991 MSA_FLOAT_COND(DEST, lt, ARG1, ARG2, BITS, QUIET); \
2992 } \
2993 } while (0)
2994
2995#define MSA_FLOAT_OR(DEST, ARG1, ARG2, BITS, QUIET) \
2996 do { \
2997 MSA_FLOAT_COND(DEST, le, ARG1, ARG2, BITS, QUIET); \
2998 if (DEST == 0) { \
2999 MSA_FLOAT_COND(DEST, le, ARG2, ARG1, BITS, QUIET); \
3000 } \
3001 } while (0)
3002
3003static inline void compare_af(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3004 wr_t *pwt, uint32_t df, int quiet,
3005 uintptr_t retaddr)
3006{
3007 wr_t wx, *pwx = &wx;
3008 uint32_t i;
3009
3010 clear_msacsr_cause(env);
3011
3012 switch (df) {
3013 case DF_WORD:
3014 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3015 MSA_FLOAT_AF(pwx->w[i], pws->w[i], pwt->w[i], 32, quiet);
3016 }
3017 break;
3018 case DF_DOUBLE:
3019 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3020 MSA_FLOAT_AF(pwx->d[i], pws->d[i], pwt->d[i], 64, quiet);
3021 }
3022 break;
3023 default:
3024 assert(0);
3025 }
3026
3027 check_msacsr_cause(env, retaddr);
3028
3029 msa_move_v(pwd, pwx);
3030}
3031
3032static inline void compare_un(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3033 wr_t *pwt, uint32_t df, int quiet,
3034 uintptr_t retaddr)
3035{
3036 wr_t wx, *pwx = &wx;
3037 uint32_t i;
3038
3039 clear_msacsr_cause(env);
3040
3041 switch (df) {
3042 case DF_WORD:
3043 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3044 MSA_FLOAT_COND(pwx->w[i], unordered, pws->w[i], pwt->w[i], 32,
3045 quiet);
3046 }
3047 break;
3048 case DF_DOUBLE:
3049 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3050 MSA_FLOAT_COND(pwx->d[i], unordered, pws->d[i], pwt->d[i], 64,
3051 quiet);
3052 }
3053 break;
3054 default:
3055 assert(0);
3056 }
3057
3058 check_msacsr_cause(env, retaddr);
3059
3060 msa_move_v(pwd, pwx);
3061}
3062
3063static inline void compare_eq(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3064 wr_t *pwt, uint32_t df, int quiet,
3065 uintptr_t retaddr)
3066{
3067 wr_t wx, *pwx = &wx;
3068 uint32_t i;
3069
3070 clear_msacsr_cause(env);
3071
3072 switch (df) {
3073 case DF_WORD:
3074 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3075 MSA_FLOAT_COND(pwx->w[i], eq, pws->w[i], pwt->w[i], 32, quiet);
3076 }
3077 break;
3078 case DF_DOUBLE:
3079 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3080 MSA_FLOAT_COND(pwx->d[i], eq, pws->d[i], pwt->d[i], 64, quiet);
3081 }
3082 break;
3083 default:
3084 assert(0);
3085 }
3086
3087 check_msacsr_cause(env, retaddr);
3088
3089 msa_move_v(pwd, pwx);
3090}
3091
3092static inline void compare_ueq(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3093 wr_t *pwt, uint32_t df, int quiet,
3094 uintptr_t retaddr)
3095{
3096 wr_t wx, *pwx = &wx;
3097 uint32_t i;
3098
3099 clear_msacsr_cause(env);
3100
3101 switch (df) {
3102 case DF_WORD:
3103 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3104 MSA_FLOAT_UEQ(pwx->w[i], pws->w[i], pwt->w[i], 32, quiet);
3105 }
3106 break;
3107 case DF_DOUBLE:
3108 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3109 MSA_FLOAT_UEQ(pwx->d[i], pws->d[i], pwt->d[i], 64, quiet);
3110 }
3111 break;
3112 default:
3113 assert(0);
3114 }
3115
3116 check_msacsr_cause(env, retaddr);
3117
3118 msa_move_v(pwd, pwx);
3119}
3120
3121static inline void compare_lt(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3122 wr_t *pwt, uint32_t df, int quiet,
3123 uintptr_t retaddr)
3124{
3125 wr_t wx, *pwx = &wx;
3126 uint32_t i;
3127
3128 clear_msacsr_cause(env);
3129
3130 switch (df) {
3131 case DF_WORD:
3132 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3133 MSA_FLOAT_COND(pwx->w[i], lt, pws->w[i], pwt->w[i], 32, quiet);
3134 }
3135 break;
3136 case DF_DOUBLE:
3137 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3138 MSA_FLOAT_COND(pwx->d[i], lt, pws->d[i], pwt->d[i], 64, quiet);
3139 }
3140 break;
3141 default:
3142 assert(0);
3143 }
3144
3145 check_msacsr_cause(env, retaddr);
3146
3147 msa_move_v(pwd, pwx);
3148}
3149
3150static inline void compare_ult(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3151 wr_t *pwt, uint32_t df, int quiet,
3152 uintptr_t retaddr)
3153{
3154 wr_t wx, *pwx = &wx;
3155 uint32_t i;
3156
3157 clear_msacsr_cause(env);
3158
3159 switch (df) {
3160 case DF_WORD:
3161 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3162 MSA_FLOAT_ULT(pwx->w[i], pws->w[i], pwt->w[i], 32, quiet);
3163 }
3164 break;
3165 case DF_DOUBLE:
3166 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3167 MSA_FLOAT_ULT(pwx->d[i], pws->d[i], pwt->d[i], 64, quiet);
3168 }
3169 break;
3170 default:
3171 assert(0);
3172 }
3173
3174 check_msacsr_cause(env, retaddr);
3175
3176 msa_move_v(pwd, pwx);
3177}
3178
3179static inline void compare_le(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3180 wr_t *pwt, uint32_t df, int quiet,
3181 uintptr_t retaddr)
3182{
3183 wr_t wx, *pwx = &wx;
3184 uint32_t i;
3185
3186 clear_msacsr_cause(env);
3187
3188 switch (df) {
3189 case DF_WORD:
3190 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3191 MSA_FLOAT_COND(pwx->w[i], le, pws->w[i], pwt->w[i], 32, quiet);
3192 }
3193 break;
3194 case DF_DOUBLE:
3195 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3196 MSA_FLOAT_COND(pwx->d[i], le, pws->d[i], pwt->d[i], 64, quiet);
3197 }
3198 break;
3199 default:
3200 assert(0);
3201 }
3202
3203 check_msacsr_cause(env, retaddr);
3204
3205 msa_move_v(pwd, pwx);
3206}
3207
3208static inline void compare_ule(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3209 wr_t *pwt, uint32_t df, int quiet,
3210 uintptr_t retaddr)
3211{
3212 wr_t wx, *pwx = &wx;
3213 uint32_t i;
3214
3215 clear_msacsr_cause(env);
3216
3217 switch (df) {
3218 case DF_WORD:
3219 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3220 MSA_FLOAT_ULE(pwx->w[i], pws->w[i], pwt->w[i], 32, quiet);
3221 }
3222 break;
3223 case DF_DOUBLE:
3224 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3225 MSA_FLOAT_ULE(pwx->d[i], pws->d[i], pwt->d[i], 64, quiet);
3226 }
3227 break;
3228 default:
3229 assert(0);
3230 }
3231
3232 check_msacsr_cause(env, retaddr);
3233
3234 msa_move_v(pwd, pwx);
3235}
3236
3237static inline void compare_or(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3238 wr_t *pwt, uint32_t df, int quiet,
3239 uintptr_t retaddr)
3240{
3241 wr_t wx, *pwx = &wx;
3242 uint32_t i;
3243
3244 clear_msacsr_cause(env);
3245
3246 switch (df) {
3247 case DF_WORD:
3248 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3249 MSA_FLOAT_OR(pwx->w[i], pws->w[i], pwt->w[i], 32, quiet);
3250 }
3251 break;
3252 case DF_DOUBLE:
3253 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3254 MSA_FLOAT_OR(pwx->d[i], pws->d[i], pwt->d[i], 64, quiet);
3255 }
3256 break;
3257 default:
3258 assert(0);
3259 }
3260
3261 check_msacsr_cause(env, retaddr);
3262
3263 msa_move_v(pwd, pwx);
3264}
3265
3266static inline void compare_une(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3267 wr_t *pwt, uint32_t df, int quiet,
3268 uintptr_t retaddr)
3269{
3270 wr_t wx, *pwx = &wx;
3271 uint32_t i;
3272
3273 clear_msacsr_cause(env);
3274
3275 switch (df) {
3276 case DF_WORD:
3277 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3278 MSA_FLOAT_UNE(pwx->w[i], pws->w[i], pwt->w[i], 32, quiet);
3279 }
3280 break;
3281 case DF_DOUBLE:
3282 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3283 MSA_FLOAT_UNE(pwx->d[i], pws->d[i], pwt->d[i], 64, quiet);
3284 }
3285 break;
3286 default:
3287 assert(0);
3288 }
3289
3290 check_msacsr_cause(env, retaddr);
3291
3292 msa_move_v(pwd, pwx);
3293}
3294
3295static inline void compare_ne(CPUMIPSState *env, wr_t *pwd, wr_t *pws,
3296 wr_t *pwt, uint32_t df, int quiet,
3297 uintptr_t retaddr)
3298{
3299 wr_t wx, *pwx = &wx;
3300 uint32_t i;
3301
3302 clear_msacsr_cause(env);
3303
3304 switch (df) {
3305 case DF_WORD:
3306 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3307 MSA_FLOAT_NE(pwx->w[i], pws->w[i], pwt->w[i], 32, quiet);
3308 }
3309 break;
3310 case DF_DOUBLE:
3311 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3312 MSA_FLOAT_NE(pwx->d[i], pws->d[i], pwt->d[i], 64, quiet);
3313 }
3314 break;
3315 default:
3316 assert(0);
3317 }
3318
3319 check_msacsr_cause(env, retaddr);
3320
3321 msa_move_v(pwd, pwx);
3322}
3323
3324void helper_msa_fcaf_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3325 uint32_t ws, uint32_t wt)
3326{
3327 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3328 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3329 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3330 compare_af(env, pwd, pws, pwt, df, 1, GETPC());
3331}
3332
3333void helper_msa_fcun_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3334 uint32_t ws, uint32_t wt)
3335{
3336 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3337 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3338 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3339 compare_un(env, pwd, pws, pwt, df, 1, GETPC());
3340}
3341
3342void helper_msa_fceq_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3343 uint32_t ws, uint32_t wt)
3344{
3345 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3346 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3347 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3348 compare_eq(env, pwd, pws, pwt, df, 1, GETPC());
3349}
3350
3351void helper_msa_fcueq_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3352 uint32_t ws, uint32_t wt)
3353{
3354 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3355 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3356 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3357 compare_ueq(env, pwd, pws, pwt, df, 1, GETPC());
3358}
3359
3360void helper_msa_fclt_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3361 uint32_t ws, uint32_t wt)
3362{
3363 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3364 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3365 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3366 compare_lt(env, pwd, pws, pwt, df, 1, GETPC());
3367}
3368
3369void helper_msa_fcult_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3370 uint32_t ws, uint32_t wt)
3371{
3372 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3373 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3374 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3375 compare_ult(env, pwd, pws, pwt, df, 1, GETPC());
3376}
3377
3378void helper_msa_fcle_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3379 uint32_t ws, uint32_t wt)
3380{
3381 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3382 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3383 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3384 compare_le(env, pwd, pws, pwt, df, 1, GETPC());
3385}
3386
3387void helper_msa_fcule_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3388 uint32_t ws, uint32_t wt)
3389{
3390 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3391 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3392 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3393 compare_ule(env, pwd, pws, pwt, df, 1, GETPC());
3394}
3395
3396void helper_msa_fsaf_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3397 uint32_t ws, uint32_t wt)
3398{
3399 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3400 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3401 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3402 compare_af(env, pwd, pws, pwt, df, 0, GETPC());
3403}
3404
3405void helper_msa_fsun_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3406 uint32_t ws, uint32_t wt)
3407{
3408 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3409 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3410 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3411 compare_un(env, pwd, pws, pwt, df, 0, GETPC());
3412}
3413
3414void helper_msa_fseq_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3415 uint32_t ws, uint32_t wt)
3416{
3417 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3418 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3419 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3420 compare_eq(env, pwd, pws, pwt, df, 0, GETPC());
3421}
3422
3423void helper_msa_fsueq_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3424 uint32_t ws, uint32_t wt)
3425{
3426 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3427 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3428 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3429 compare_ueq(env, pwd, pws, pwt, df, 0, GETPC());
3430}
3431
3432void helper_msa_fslt_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3433 uint32_t ws, uint32_t wt)
3434{
3435 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3436 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3437 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3438 compare_lt(env, pwd, pws, pwt, df, 0, GETPC());
3439}
3440
3441void helper_msa_fsult_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3442 uint32_t ws, uint32_t wt)
3443{
3444 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3445 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3446 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3447 compare_ult(env, pwd, pws, pwt, df, 0, GETPC());
3448}
3449
3450void helper_msa_fsle_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3451 uint32_t ws, uint32_t wt)
3452{
3453 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3454 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3455 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3456 compare_le(env, pwd, pws, pwt, df, 0, GETPC());
3457}
3458
3459void helper_msa_fsule_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3460 uint32_t ws, uint32_t wt)
3461{
3462 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3463 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3464 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3465 compare_ule(env, pwd, pws, pwt, df, 0, GETPC());
3466}
3467
3468void helper_msa_fcor_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3469 uint32_t ws, uint32_t wt)
3470{
3471 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3472 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3473 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3474 compare_or(env, pwd, pws, pwt, df, 1, GETPC());
3475}
3476
3477void helper_msa_fcune_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3478 uint32_t ws, uint32_t wt)
3479{
3480 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3481 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3482 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3483 compare_une(env, pwd, pws, pwt, df, 1, GETPC());
3484}
3485
3486void helper_msa_fcne_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3487 uint32_t ws, uint32_t wt)
3488{
3489 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3490 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3491 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3492 compare_ne(env, pwd, pws, pwt, df, 1, GETPC());
3493}
3494
3495void helper_msa_fsor_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3496 uint32_t ws, uint32_t wt)
3497{
3498 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3499 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3500 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3501 compare_or(env, pwd, pws, pwt, df, 0, GETPC());
3502}
3503
3504void helper_msa_fsune_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3505 uint32_t ws, uint32_t wt)
3506{
3507 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3508 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3509 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3510 compare_une(env, pwd, pws, pwt, df, 0, GETPC());
3511}
3512
3513void helper_msa_fsne_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3514 uint32_t ws, uint32_t wt)
3515{
3516 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3517 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3518 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3519 compare_ne(env, pwd, pws, pwt, df, 0, GETPC());
3520}
3521
3522#define float16_is_zero(ARG) 0
3523#define float16_is_zero_or_denormal(ARG) 0
3524
3525#define IS_DENORMAL(ARG, BITS) \
3526 (!float ## BITS ## _is_zero(ARG) \
3527 && float ## BITS ## _is_zero_or_denormal(ARG))
3528
3529#define MSA_FLOAT_BINOP(DEST, OP, ARG1, ARG2, BITS) \
3530 do { \
3531 float_status *status = &env->active_tc.msa_fp_status; \
3532 int c; \
3533 \
3534 set_float_exception_flags(0, status); \
3535 DEST = float ## BITS ## _ ## OP(ARG1, ARG2, status); \
3536 c = update_msacsr(env, 0, IS_DENORMAL(DEST, BITS)); \
3537 \
3538 if (get_enabled_exceptions(env, c)) { \
3539 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
3540 } \
3541 } while (0)
3542
3543void helper_msa_fadd_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3544 uint32_t ws, uint32_t wt)
3545{
3546 wr_t wx, *pwx = &wx;
3547 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3548 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3549 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3550 uint32_t i;
3551
3552 clear_msacsr_cause(env);
3553
3554 switch (df) {
3555 case DF_WORD:
3556 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3557 MSA_FLOAT_BINOP(pwx->w[i], add, pws->w[i], pwt->w[i], 32);
3558 }
3559 break;
3560 case DF_DOUBLE:
3561 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3562 MSA_FLOAT_BINOP(pwx->d[i], add, pws->d[i], pwt->d[i], 64);
3563 }
3564 break;
3565 default:
3566 assert(0);
3567 }
3568
3569 check_msacsr_cause(env, GETPC());
3570 msa_move_v(pwd, pwx);
3571}
3572
3573void helper_msa_fsub_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3574 uint32_t ws, uint32_t wt)
3575{
3576 wr_t wx, *pwx = &wx;
3577 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3578 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3579 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3580 uint32_t i;
3581
3582 clear_msacsr_cause(env);
3583
3584 switch (df) {
3585 case DF_WORD:
3586 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3587 MSA_FLOAT_BINOP(pwx->w[i], sub, pws->w[i], pwt->w[i], 32);
3588 }
3589 break;
3590 case DF_DOUBLE:
3591 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3592 MSA_FLOAT_BINOP(pwx->d[i], sub, pws->d[i], pwt->d[i], 64);
3593 }
3594 break;
3595 default:
3596 assert(0);
3597 }
3598
3599 check_msacsr_cause(env, GETPC());
3600 msa_move_v(pwd, pwx);
3601}
3602
3603void helper_msa_fmul_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3604 uint32_t ws, uint32_t wt)
3605{
3606 wr_t wx, *pwx = &wx;
3607 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3608 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3609 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3610 uint32_t i;
3611
3612 clear_msacsr_cause(env);
3613
3614 switch (df) {
3615 case DF_WORD:
3616 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3617 MSA_FLOAT_BINOP(pwx->w[i], mul, pws->w[i], pwt->w[i], 32);
3618 }
3619 break;
3620 case DF_DOUBLE:
3621 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3622 MSA_FLOAT_BINOP(pwx->d[i], mul, pws->d[i], pwt->d[i], 64);
3623 }
3624 break;
3625 default:
3626 assert(0);
3627 }
3628
3629 check_msacsr_cause(env, GETPC());
3630
3631 msa_move_v(pwd, pwx);
3632}
3633
3634void helper_msa_fdiv_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3635 uint32_t ws, uint32_t wt)
3636{
3637 wr_t wx, *pwx = &wx;
3638 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3639 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3640 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3641 uint32_t i;
3642
3643 clear_msacsr_cause(env);
3644
3645 switch (df) {
3646 case DF_WORD:
3647 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3648 MSA_FLOAT_BINOP(pwx->w[i], div, pws->w[i], pwt->w[i], 32);
3649 }
3650 break;
3651 case DF_DOUBLE:
3652 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3653 MSA_FLOAT_BINOP(pwx->d[i], div, pws->d[i], pwt->d[i], 64);
3654 }
3655 break;
3656 default:
3657 assert(0);
3658 }
3659
3660 check_msacsr_cause(env, GETPC());
3661
3662 msa_move_v(pwd, pwx);
3663}
3664
3665#define MSA_FLOAT_MULADD(DEST, ARG1, ARG2, ARG3, NEGATE, BITS) \
3666 do { \
3667 float_status *status = &env->active_tc.msa_fp_status; \
3668 int c; \
3669 \
3670 set_float_exception_flags(0, status); \
3671 DEST = float ## BITS ## _muladd(ARG2, ARG3, ARG1, NEGATE, status); \
3672 c = update_msacsr(env, 0, IS_DENORMAL(DEST, BITS)); \
3673 \
3674 if (get_enabled_exceptions(env, c)) { \
3675 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
3676 } \
3677 } while (0)
3678
3679void helper_msa_fmadd_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3680 uint32_t ws, uint32_t wt)
3681{
3682 wr_t wx, *pwx = &wx;
3683 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3684 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3685 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3686 uint32_t i;
3687
3688 clear_msacsr_cause(env);
3689
3690 switch (df) {
3691 case DF_WORD:
3692 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3693 MSA_FLOAT_MULADD(pwx->w[i], pwd->w[i],
3694 pws->w[i], pwt->w[i], 0, 32);
3695 }
3696 break;
3697 case DF_DOUBLE:
3698 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3699 MSA_FLOAT_MULADD(pwx->d[i], pwd->d[i],
3700 pws->d[i], pwt->d[i], 0, 64);
3701 }
3702 break;
3703 default:
3704 assert(0);
3705 }
3706
3707 check_msacsr_cause(env, GETPC());
3708
3709 msa_move_v(pwd, pwx);
3710}
3711
3712void helper_msa_fmsub_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3713 uint32_t ws, uint32_t wt)
3714{
3715 wr_t wx, *pwx = &wx;
3716 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3717 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3718 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3719 uint32_t i;
3720
3721 clear_msacsr_cause(env);
3722
3723 switch (df) {
3724 case DF_WORD:
3725 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3726 MSA_FLOAT_MULADD(pwx->w[i], pwd->w[i],
3727 pws->w[i], pwt->w[i],
3728 float_muladd_negate_product, 32);
3729 }
3730 break;
3731 case DF_DOUBLE:
3732 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3733 MSA_FLOAT_MULADD(pwx->d[i], pwd->d[i],
3734 pws->d[i], pwt->d[i],
3735 float_muladd_negate_product, 64);
3736 }
3737 break;
3738 default:
3739 assert(0);
3740 }
3741
3742 check_msacsr_cause(env, GETPC());
3743
3744 msa_move_v(pwd, pwx);
3745}
3746
3747void helper_msa_fexp2_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3748 uint32_t ws, uint32_t wt)
3749{
3750 wr_t wx, *pwx = &wx;
3751 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3752 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3753 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3754 uint32_t i;
3755
3756 clear_msacsr_cause(env);
3757
3758 switch (df) {
3759 case DF_WORD:
3760 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3761 MSA_FLOAT_BINOP(pwx->w[i], scalbn, pws->w[i],
3762 pwt->w[i] > 0x200 ? 0x200 :
3763 pwt->w[i] < -0x200 ? -0x200 : pwt->w[i],
3764 32);
3765 }
3766 break;
3767 case DF_DOUBLE:
3768 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3769 MSA_FLOAT_BINOP(pwx->d[i], scalbn, pws->d[i],
3770 pwt->d[i] > 0x1000 ? 0x1000 :
3771 pwt->d[i] < -0x1000 ? -0x1000 : pwt->d[i],
3772 64);
3773 }
3774 break;
3775 default:
3776 assert(0);
3777 }
3778
3779 check_msacsr_cause(env, GETPC());
3780
3781 msa_move_v(pwd, pwx);
3782}
3783
3784#define MSA_FLOAT_UNOP(DEST, OP, ARG, BITS) \
3785 do { \
3786 float_status *status = &env->active_tc.msa_fp_status; \
3787 int c; \
3788 \
3789 set_float_exception_flags(0, status); \
3790 DEST = float ## BITS ## _ ## OP(ARG, status); \
3791 c = update_msacsr(env, 0, IS_DENORMAL(DEST, BITS)); \
3792 \
3793 if (get_enabled_exceptions(env, c)) { \
3794 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
3795 } \
3796 } while (0)
3797
3798void helper_msa_fexdo_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3799 uint32_t ws, uint32_t wt)
3800{
3801 wr_t wx, *pwx = &wx;
3802 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3803 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3804 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3805 uint32_t i;
3806
3807 clear_msacsr_cause(env);
3808
3809 switch (df) {
3810 case DF_WORD:
3811 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3812 /*
3813 * Half precision floats come in two formats: standard
3814 * IEEE and "ARM" format. The latter gains extra exponent
3815 * range by omitting the NaN/Inf encodings.
3816 */
3817 flag ieee = 1;
3818
3819 MSA_FLOAT_BINOP(Lh(pwx, i), from_float32, pws->w[i], ieee, 16);
3820 MSA_FLOAT_BINOP(Rh(pwx, i), from_float32, pwt->w[i], ieee, 16);
3821 }
3822 break;
3823 case DF_DOUBLE:
3824 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3825 MSA_FLOAT_UNOP(Lw(pwx, i), from_float64, pws->d[i], 32);
3826 MSA_FLOAT_UNOP(Rw(pwx, i), from_float64, pwt->d[i], 32);
3827 }
3828 break;
3829 default:
3830 assert(0);
3831 }
3832
3833 check_msacsr_cause(env, GETPC());
3834 msa_move_v(pwd, pwx);
3835}
3836
3837#define MSA_FLOAT_UNOP_XD(DEST, OP, ARG, BITS, XBITS) \
3838 do { \
3839 float_status *status = &env->active_tc.msa_fp_status; \
3840 int c; \
3841 \
3842 set_float_exception_flags(0, status); \
3843 DEST = float ## BITS ## _ ## OP(ARG, status); \
3844 c = update_msacsr(env, CLEAR_FS_UNDERFLOW, 0); \
3845 \
3846 if (get_enabled_exceptions(env, c)) { \
3847 DEST = ((FLOAT_SNAN ## XBITS(status) >> 6) << 6) | c; \
3848 } \
3849 } while (0)
3850
3851void helper_msa_ftq_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3852 uint32_t ws, uint32_t wt)
3853{
3854 wr_t wx, *pwx = &wx;
3855 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3856 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3857 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3858 uint32_t i;
3859
3860 clear_msacsr_cause(env);
3861
3862 switch (df) {
3863 case DF_WORD:
3864 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
3865 MSA_FLOAT_UNOP_XD(Lh(pwx, i), to_q16, pws->w[i], 32, 16);
3866 MSA_FLOAT_UNOP_XD(Rh(pwx, i), to_q16, pwt->w[i], 32, 16);
3867 }
3868 break;
3869 case DF_DOUBLE:
3870 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
3871 MSA_FLOAT_UNOP_XD(Lw(pwx, i), to_q32, pws->d[i], 64, 32);
3872 MSA_FLOAT_UNOP_XD(Rw(pwx, i), to_q32, pwt->d[i], 64, 32);
3873 }
3874 break;
3875 default:
3876 assert(0);
3877 }
3878
3879 check_msacsr_cause(env, GETPC());
3880
3881 msa_move_v(pwd, pwx);
3882}
3883
3884#define NUMBER_QNAN_PAIR(ARG1, ARG2, BITS, STATUS) \
3885 !float ## BITS ## _is_any_nan(ARG1) \
3886 && float ## BITS ## _is_quiet_nan(ARG2, STATUS)
3887
3888#define MSA_FLOAT_MAXOP(DEST, OP, ARG1, ARG2, BITS) \
3889 do { \
3890 float_status *status = &env->active_tc.msa_fp_status; \
3891 int c; \
3892 \
3893 set_float_exception_flags(0, status); \
3894 DEST = float ## BITS ## _ ## OP(ARG1, ARG2, status); \
3895 c = update_msacsr(env, 0, 0); \
3896 \
3897 if (get_enabled_exceptions(env, c)) { \
3898 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
3899 } \
3900 } while (0)
3901
3902#define FMAXMIN_A(F, G, X, _S, _T, BITS, STATUS) \
3903 do { \
3904 uint## BITS ##_t S = _S, T = _T; \
3905 uint## BITS ##_t as, at, xs, xt, xd; \
3906 if (NUMBER_QNAN_PAIR(S, T, BITS, STATUS)) { \
3907 T = S; \
3908 } \
3909 else if (NUMBER_QNAN_PAIR(T, S, BITS, STATUS)) { \
3910 S = T; \
3911 } \
3912 as = float## BITS ##_abs(S); \
3913 at = float## BITS ##_abs(T); \
3914 MSA_FLOAT_MAXOP(xs, F, S, T, BITS); \
3915 MSA_FLOAT_MAXOP(xt, G, S, T, BITS); \
3916 MSA_FLOAT_MAXOP(xd, F, as, at, BITS); \
3917 X = (as == at || xd == float## BITS ##_abs(xs)) ? xs : xt; \
3918 } while (0)
3919
3920void helper_msa_fmin_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3921 uint32_t ws, uint32_t wt)
3922{
3923 float_status *status = &env->active_tc.msa_fp_status;
3924 wr_t wx, *pwx = &wx;
3925 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
3926 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
3927 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
3928
3929 clear_msacsr_cause(env);
3930
3931 if (df == DF_WORD) {
3932
3933 if (NUMBER_QNAN_PAIR(pws->w[0], pwt->w[0], 32, status)) {
3934 MSA_FLOAT_MAXOP(pwx->w[0], min, pws->w[0], pws->w[0], 32);
3935 } else if (NUMBER_QNAN_PAIR(pwt->w[0], pws->w[0], 32, status)) {
3936 MSA_FLOAT_MAXOP(pwx->w[0], min, pwt->w[0], pwt->w[0], 32);
3937 } else {
3938 MSA_FLOAT_MAXOP(pwx->w[0], min, pws->w[0], pwt->w[0], 32);
3939 }
3940
3941 if (NUMBER_QNAN_PAIR(pws->w[1], pwt->w[1], 32, status)) {
3942 MSA_FLOAT_MAXOP(pwx->w[1], min, pws->w[1], pws->w[1], 32);
3943 } else if (NUMBER_QNAN_PAIR(pwt->w[1], pws->w[1], 32, status)) {
3944 MSA_FLOAT_MAXOP(pwx->w[1], min, pwt->w[1], pwt->w[1], 32);
3945 } else {
3946 MSA_FLOAT_MAXOP(pwx->w[1], min, pws->w[1], pwt->w[1], 32);
3947 }
3948
3949 if (NUMBER_QNAN_PAIR(pws->w[2], pwt->w[2], 32, status)) {
3950 MSA_FLOAT_MAXOP(pwx->w[2], min, pws->w[2], pws->w[2], 32);
3951 } else if (NUMBER_QNAN_PAIR(pwt->w[2], pws->w[2], 32, status)) {
3952 MSA_FLOAT_MAXOP(pwx->w[2], min, pwt->w[2], pwt->w[2], 32);
3953 } else {
3954 MSA_FLOAT_MAXOP(pwx->w[2], min, pws->w[2], pwt->w[2], 32);
3955 }
3956
3957 if (NUMBER_QNAN_PAIR(pws->w[3], pwt->w[3], 32, status)) {
3958 MSA_FLOAT_MAXOP(pwx->w[3], min, pws->w[3], pws->w[3], 32);
3959 } else if (NUMBER_QNAN_PAIR(pwt->w[3], pws->w[3], 32, status)) {
3960 MSA_FLOAT_MAXOP(pwx->w[3], min, pwt->w[3], pwt->w[3], 32);
3961 } else {
3962 MSA_FLOAT_MAXOP(pwx->w[3], min, pws->w[3], pwt->w[3], 32);
3963 }
3964
3965 } else if (df == DF_DOUBLE) {
3966
3967 if (NUMBER_QNAN_PAIR(pws->d[0], pwt->d[0], 64, status)) {
3968 MSA_FLOAT_MAXOP(pwx->d[0], min, pws->d[0], pws->d[0], 64);
3969 } else if (NUMBER_QNAN_PAIR(pwt->d[0], pws->d[0], 64, status)) {
3970 MSA_FLOAT_MAXOP(pwx->d[0], min, pwt->d[0], pwt->d[0], 64);
3971 } else {
3972 MSA_FLOAT_MAXOP(pwx->d[0], min, pws->d[0], pwt->d[0], 64);
3973 }
3974
3975 if (NUMBER_QNAN_PAIR(pws->d[1], pwt->d[1], 64, status)) {
3976 MSA_FLOAT_MAXOP(pwx->d[1], min, pws->d[1], pws->d[1], 64);
3977 } else if (NUMBER_QNAN_PAIR(pwt->d[1], pws->d[1], 64, status)) {
3978 MSA_FLOAT_MAXOP(pwx->d[1], min, pwt->d[1], pwt->d[1], 64);
3979 } else {
3980 MSA_FLOAT_MAXOP(pwx->d[1], min, pws->d[1], pwt->d[1], 64);
3981 }
3982
3983 } else {
3984
3985 assert(0);
3986
3987 }
3988
3989 check_msacsr_cause(env, GETPC());
3990
3991 msa_move_v(pwd, pwx);
3992}
3993
3994void helper_msa_fmin_a_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
3995 uint32_t ws, uint32_t wt)
3996{
3997 float_status *status = &env->active_tc.msa_fp_status;
3998 wr_t wx, *pwx = &wx;
3999 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4000 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4001 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
4002
4003 clear_msacsr_cause(env);
4004
4005 if (df == DF_WORD) {
4006 FMAXMIN_A(min, max, pwx->w[0], pws->w[0], pwt->w[0], 32, status);
4007 FMAXMIN_A(min, max, pwx->w[1], pws->w[1], pwt->w[1], 32, status);
4008 FMAXMIN_A(min, max, pwx->w[2], pws->w[2], pwt->w[2], 32, status);
4009 FMAXMIN_A(min, max, pwx->w[3], pws->w[3], pwt->w[3], 32, status);
4010 } else if (df == DF_DOUBLE) {
4011 FMAXMIN_A(min, max, pwx->d[0], pws->d[0], pwt->d[0], 64, status);
4012 FMAXMIN_A(min, max, pwx->d[1], pws->d[1], pwt->d[1], 64, status);
4013 } else {
4014 assert(0);
4015 }
4016
4017 check_msacsr_cause(env, GETPC());
4018
4019 msa_move_v(pwd, pwx);
4020}
4021
4022void helper_msa_fmax_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4023 uint32_t ws, uint32_t wt)
4024{
4025 float_status *status = &env->active_tc.msa_fp_status;
4026 wr_t wx, *pwx = &wx;
4027 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4028 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4029 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
4030
4031 clear_msacsr_cause(env);
4032
4033 if (df == DF_WORD) {
4034
4035 if (NUMBER_QNAN_PAIR(pws->w[0], pwt->w[0], 32, status)) {
4036 MSA_FLOAT_MAXOP(pwx->w[0], max, pws->w[0], pws->w[0], 32);
4037 } else if (NUMBER_QNAN_PAIR(pwt->w[0], pws->w[0], 32, status)) {
4038 MSA_FLOAT_MAXOP(pwx->w[0], max, pwt->w[0], pwt->w[0], 32);
4039 } else {
4040 MSA_FLOAT_MAXOP(pwx->w[0], max, pws->w[0], pwt->w[0], 32);
4041 }
4042
4043 if (NUMBER_QNAN_PAIR(pws->w[1], pwt->w[1], 32, status)) {
4044 MSA_FLOAT_MAXOP(pwx->w[1], max, pws->w[1], pws->w[1], 32);
4045 } else if (NUMBER_QNAN_PAIR(pwt->w[1], pws->w[1], 32, status)) {
4046 MSA_FLOAT_MAXOP(pwx->w[1], max, pwt->w[1], pwt->w[1], 32);
4047 } else {
4048 MSA_FLOAT_MAXOP(pwx->w[1], max, pws->w[1], pwt->w[1], 32);
4049 }
4050
4051 if (NUMBER_QNAN_PAIR(pws->w[2], pwt->w[2], 32, status)) {
4052 MSA_FLOAT_MAXOP(pwx->w[2], max, pws->w[2], pws->w[2], 32);
4053 } else if (NUMBER_QNAN_PAIR(pwt->w[2], pws->w[2], 32, status)) {
4054 MSA_FLOAT_MAXOP(pwx->w[2], max, pwt->w[2], pwt->w[2], 32);
4055 } else {
4056 MSA_FLOAT_MAXOP(pwx->w[2], max, pws->w[2], pwt->w[2], 32);
4057 }
4058
4059 if (NUMBER_QNAN_PAIR(pws->w[3], pwt->w[3], 32, status)) {
4060 MSA_FLOAT_MAXOP(pwx->w[3], max, pws->w[3], pws->w[3], 32);
4061 } else if (NUMBER_QNAN_PAIR(pwt->w[3], pws->w[3], 32, status)) {
4062 MSA_FLOAT_MAXOP(pwx->w[3], max, pwt->w[3], pwt->w[3], 32);
4063 } else {
4064 MSA_FLOAT_MAXOP(pwx->w[3], max, pws->w[3], pwt->w[3], 32);
4065 }
4066
4067 } else if (df == DF_DOUBLE) {
4068
4069 if (NUMBER_QNAN_PAIR(pws->d[0], pwt->d[0], 64, status)) {
4070 MSA_FLOAT_MAXOP(pwx->d[0], max, pws->d[0], pws->d[0], 64);
4071 } else if (NUMBER_QNAN_PAIR(pwt->d[0], pws->d[0], 64, status)) {
4072 MSA_FLOAT_MAXOP(pwx->d[0], max, pwt->d[0], pwt->d[0], 64);
4073 } else {
4074 MSA_FLOAT_MAXOP(pwx->d[0], max, pws->d[0], pwt->d[0], 64);
4075 }
4076
4077 if (NUMBER_QNAN_PAIR(pws->d[1], pwt->d[1], 64, status)) {
4078 MSA_FLOAT_MAXOP(pwx->d[1], max, pws->d[1], pws->d[1], 64);
4079 } else if (NUMBER_QNAN_PAIR(pwt->d[1], pws->d[1], 64, status)) {
4080 MSA_FLOAT_MAXOP(pwx->d[1], max, pwt->d[1], pwt->d[1], 64);
4081 } else {
4082 MSA_FLOAT_MAXOP(pwx->d[1], max, pws->d[1], pwt->d[1], 64);
4083 }
4084
4085 } else {
4086
4087 assert(0);
4088
4089 }
4090
4091 check_msacsr_cause(env, GETPC());
4092
4093 msa_move_v(pwd, pwx);
4094}
4095
4096void helper_msa_fmax_a_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4097 uint32_t ws, uint32_t wt)
4098{
4099 float_status *status = &env->active_tc.msa_fp_status;
4100 wr_t wx, *pwx = &wx;
4101 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4102 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4103 wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
4104
4105 clear_msacsr_cause(env);
4106
4107 if (df == DF_WORD) {
4108 FMAXMIN_A(max, min, pwx->w[0], pws->w[0], pwt->w[0], 32, status);
4109 FMAXMIN_A(max, min, pwx->w[1], pws->w[1], pwt->w[1], 32, status);
4110 FMAXMIN_A(max, min, pwx->w[2], pws->w[2], pwt->w[2], 32, status);
4111 FMAXMIN_A(max, min, pwx->w[3], pws->w[3], pwt->w[3], 32, status);
4112 } else if (df == DF_DOUBLE) {
4113 FMAXMIN_A(max, min, pwx->d[0], pws->d[0], pwt->d[0], 64, status);
4114 FMAXMIN_A(max, min, pwx->d[1], pws->d[1], pwt->d[1], 64, status);
4115 } else {
4116 assert(0);
4117 }
4118
4119 check_msacsr_cause(env, GETPC());
4120
4121 msa_move_v(pwd, pwx);
4122}
4123
4124void helper_msa_fclass_df(CPUMIPSState *env, uint32_t df,
4125 uint32_t wd, uint32_t ws)
4126{
4127 float_status *status = &env->active_tc.msa_fp_status;
4128
4129 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4130 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4131 if (df == DF_WORD) {
4132 pwd->w[0] = float_class_s(pws->w[0], status);
4133 pwd->w[1] = float_class_s(pws->w[1], status);
4134 pwd->w[2] = float_class_s(pws->w[2], status);
4135 pwd->w[3] = float_class_s(pws->w[3], status);
4136 } else if (df == DF_DOUBLE) {
4137 pwd->d[0] = float_class_d(pws->d[0], status);
4138 pwd->d[1] = float_class_d(pws->d[1], status);
4139 } else {
4140 assert(0);
4141 }
4142}
4143
4144#define MSA_FLOAT_UNOP0(DEST, OP, ARG, BITS) \
4145 do { \
4146 float_status *status = &env->active_tc.msa_fp_status; \
4147 int c; \
4148 \
4149 set_float_exception_flags(0, status); \
4150 DEST = float ## BITS ## _ ## OP(ARG, status); \
4151 c = update_msacsr(env, CLEAR_FS_UNDERFLOW, 0); \
4152 \
4153 if (get_enabled_exceptions(env, c)) { \
4154 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
4155 } else if (float ## BITS ## _is_any_nan(ARG)) { \
4156 DEST = 0; \
4157 } \
4158 } while (0)
4159
4160void helper_msa_ftrunc_s_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4161 uint32_t ws)
4162{
4163 wr_t wx, *pwx = &wx;
4164 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4165 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4166 uint32_t i;
4167
4168 clear_msacsr_cause(env);
4169
4170 switch (df) {
4171 case DF_WORD:
4172 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4173 MSA_FLOAT_UNOP0(pwx->w[i], to_int32_round_to_zero, pws->w[i], 32);
4174 }
4175 break;
4176 case DF_DOUBLE:
4177 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4178 MSA_FLOAT_UNOP0(pwx->d[i], to_int64_round_to_zero, pws->d[i], 64);
4179 }
4180 break;
4181 default:
4182 assert(0);
4183 }
4184
4185 check_msacsr_cause(env, GETPC());
4186
4187 msa_move_v(pwd, pwx);
4188}
4189
4190void helper_msa_ftrunc_u_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4191 uint32_t ws)
4192{
4193 wr_t wx, *pwx = &wx;
4194 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4195 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4196 uint32_t i;
4197
4198 clear_msacsr_cause(env);
4199
4200 switch (df) {
4201 case DF_WORD:
4202 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4203 MSA_FLOAT_UNOP0(pwx->w[i], to_uint32_round_to_zero, pws->w[i], 32);
4204 }
4205 break;
4206 case DF_DOUBLE:
4207 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4208 MSA_FLOAT_UNOP0(pwx->d[i], to_uint64_round_to_zero, pws->d[i], 64);
4209 }
4210 break;
4211 default:
4212 assert(0);
4213 }
4214
4215 check_msacsr_cause(env, GETPC());
4216
4217 msa_move_v(pwd, pwx);
4218}
4219
4220void helper_msa_fsqrt_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4221 uint32_t ws)
4222{
4223 wr_t wx, *pwx = &wx;
4224 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4225 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4226 uint32_t i;
4227
4228 clear_msacsr_cause(env);
4229
4230 switch (df) {
4231 case DF_WORD:
4232 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4233 MSA_FLOAT_UNOP(pwx->w[i], sqrt, pws->w[i], 32);
4234 }
4235 break;
4236 case DF_DOUBLE:
4237 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4238 MSA_FLOAT_UNOP(pwx->d[i], sqrt, pws->d[i], 64);
4239 }
4240 break;
4241 default:
4242 assert(0);
4243 }
4244
4245 check_msacsr_cause(env, GETPC());
4246
4247 msa_move_v(pwd, pwx);
4248}
4249
4250#define MSA_FLOAT_RECIPROCAL(DEST, ARG, BITS) \
4251 do { \
4252 float_status *status = &env->active_tc.msa_fp_status; \
4253 int c; \
4254 \
4255 set_float_exception_flags(0, status); \
4256 DEST = float ## BITS ## _ ## div(FLOAT_ONE ## BITS, ARG, status); \
4257 c = update_msacsr(env, float ## BITS ## _is_infinity(ARG) || \
4258 float ## BITS ## _is_quiet_nan(DEST, status) ? \
4259 0 : RECIPROCAL_INEXACT, \
4260 IS_DENORMAL(DEST, BITS)); \
4261 \
4262 if (get_enabled_exceptions(env, c)) { \
4263 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
4264 } \
4265 } while (0)
4266
4267void helper_msa_frsqrt_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4268 uint32_t ws)
4269{
4270 wr_t wx, *pwx = &wx;
4271 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4272 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4273 uint32_t i;
4274
4275 clear_msacsr_cause(env);
4276
4277 switch (df) {
4278 case DF_WORD:
4279 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4280 MSA_FLOAT_RECIPROCAL(pwx->w[i], float32_sqrt(pws->w[i],
4281 &env->active_tc.msa_fp_status), 32);
4282 }
4283 break;
4284 case DF_DOUBLE:
4285 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4286 MSA_FLOAT_RECIPROCAL(pwx->d[i], float64_sqrt(pws->d[i],
4287 &env->active_tc.msa_fp_status), 64);
4288 }
4289 break;
4290 default:
4291 assert(0);
4292 }
4293
4294 check_msacsr_cause(env, GETPC());
4295
4296 msa_move_v(pwd, pwx);
4297}
4298
4299void helper_msa_frcp_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4300 uint32_t ws)
4301{
4302 wr_t wx, *pwx = &wx;
4303 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4304 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4305 uint32_t i;
4306
4307 clear_msacsr_cause(env);
4308
4309 switch (df) {
4310 case DF_WORD:
4311 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4312 MSA_FLOAT_RECIPROCAL(pwx->w[i], pws->w[i], 32);
4313 }
4314 break;
4315 case DF_DOUBLE:
4316 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4317 MSA_FLOAT_RECIPROCAL(pwx->d[i], pws->d[i], 64);
4318 }
4319 break;
4320 default:
4321 assert(0);
4322 }
4323
4324 check_msacsr_cause(env, GETPC());
4325
4326 msa_move_v(pwd, pwx);
4327}
4328
4329void helper_msa_frint_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4330 uint32_t ws)
4331{
4332 wr_t wx, *pwx = &wx;
4333 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4334 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4335 uint32_t i;
4336
4337 clear_msacsr_cause(env);
4338
4339 switch (df) {
4340 case DF_WORD:
4341 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4342 MSA_FLOAT_UNOP(pwx->w[i], round_to_int, pws->w[i], 32);
4343 }
4344 break;
4345 case DF_DOUBLE:
4346 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4347 MSA_FLOAT_UNOP(pwx->d[i], round_to_int, pws->d[i], 64);
4348 }
4349 break;
4350 default:
4351 assert(0);
4352 }
4353
4354 check_msacsr_cause(env, GETPC());
4355
4356 msa_move_v(pwd, pwx);
4357}
4358
4359#define MSA_FLOAT_LOGB(DEST, ARG, BITS) \
4360 do { \
4361 float_status *status = &env->active_tc.msa_fp_status; \
4362 int c; \
4363 \
4364 set_float_exception_flags(0, status); \
4365 set_float_rounding_mode(float_round_down, status); \
4366 DEST = float ## BITS ## _ ## log2(ARG, status); \
4367 DEST = float ## BITS ## _ ## round_to_int(DEST, status); \
4368 set_float_rounding_mode(ieee_rm[(env->active_tc.msacsr & \
4369 MSACSR_RM_MASK) >> MSACSR_RM], \
4370 status); \
4371 \
4372 set_float_exception_flags(get_float_exception_flags(status) & \
4373 (~float_flag_inexact), \
4374 status); \
4375 \
4376 c = update_msacsr(env, 0, IS_DENORMAL(DEST, BITS)); \
4377 \
4378 if (get_enabled_exceptions(env, c)) { \
4379 DEST = ((FLOAT_SNAN ## BITS(status) >> 6) << 6) | c; \
4380 } \
4381 } while (0)
4382
4383void helper_msa_flog2_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4384 uint32_t ws)
4385{
4386 wr_t wx, *pwx = &wx;
4387 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4388 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4389 uint32_t i;
4390
4391 clear_msacsr_cause(env);
4392
4393 switch (df) {
4394 case DF_WORD:
4395 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4396 MSA_FLOAT_LOGB(pwx->w[i], pws->w[i], 32);
4397 }
4398 break;
4399 case DF_DOUBLE:
4400 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4401 MSA_FLOAT_LOGB(pwx->d[i], pws->d[i], 64);
4402 }
4403 break;
4404 default:
4405 assert(0);
4406 }
4407
4408 check_msacsr_cause(env, GETPC());
4409
4410 msa_move_v(pwd, pwx);
4411}
4412
4413void helper_msa_fexupl_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4414 uint32_t ws)
4415{
4416 wr_t wx, *pwx = &wx;
4417 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4418 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4419 uint32_t i;
4420
4421 clear_msacsr_cause(env);
4422
4423 switch (df) {
4424 case DF_WORD:
4425 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4426 /*
4427 * Half precision floats come in two formats: standard
4428 * IEEE and "ARM" format. The latter gains extra exponent
4429 * range by omitting the NaN/Inf encodings.
4430 */
4431 flag ieee = 1;
4432
4433 MSA_FLOAT_BINOP(pwx->w[i], from_float16, Lh(pws, i), ieee, 32);
4434 }
4435 break;
4436 case DF_DOUBLE:
4437 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4438 MSA_FLOAT_UNOP(pwx->d[i], from_float32, Lw(pws, i), 64);
4439 }
4440 break;
4441 default:
4442 assert(0);
4443 }
4444
4445 check_msacsr_cause(env, GETPC());
4446 msa_move_v(pwd, pwx);
4447}
4448
4449void helper_msa_fexupr_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4450 uint32_t ws)
4451{
4452 wr_t wx, *pwx = &wx;
4453 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4454 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4455 uint32_t i;
4456
4457 clear_msacsr_cause(env);
4458
4459 switch (df) {
4460 case DF_WORD:
4461 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4462 /*
4463 * Half precision floats come in two formats: standard
4464 * IEEE and "ARM" format. The latter gains extra exponent
4465 * range by omitting the NaN/Inf encodings.
4466 */
4467 flag ieee = 1;
4468
4469 MSA_FLOAT_BINOP(pwx->w[i], from_float16, Rh(pws, i), ieee, 32);
4470 }
4471 break;
4472 case DF_DOUBLE:
4473 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4474 MSA_FLOAT_UNOP(pwx->d[i], from_float32, Rw(pws, i), 64);
4475 }
4476 break;
4477 default:
4478 assert(0);
4479 }
4480
4481 check_msacsr_cause(env, GETPC());
4482 msa_move_v(pwd, pwx);
4483}
4484
4485void helper_msa_ffql_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4486 uint32_t ws)
4487{
4488 wr_t wx, *pwx = &wx;
4489 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4490 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4491 uint32_t i;
4492
4493 switch (df) {
4494 case DF_WORD:
4495 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4496 MSA_FLOAT_UNOP(pwx->w[i], from_q16, Lh(pws, i), 32);
4497 }
4498 break;
4499 case DF_DOUBLE:
4500 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4501 MSA_FLOAT_UNOP(pwx->d[i], from_q32, Lw(pws, i), 64);
4502 }
4503 break;
4504 default:
4505 assert(0);
4506 }
4507
4508 msa_move_v(pwd, pwx);
4509}
4510
4511void helper_msa_ffqr_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4512 uint32_t ws)
4513{
4514 wr_t wx, *pwx = &wx;
4515 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4516 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4517 uint32_t i;
4518
4519 switch (df) {
4520 case DF_WORD:
4521 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4522 MSA_FLOAT_UNOP(pwx->w[i], from_q16, Rh(pws, i), 32);
4523 }
4524 break;
4525 case DF_DOUBLE:
4526 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4527 MSA_FLOAT_UNOP(pwx->d[i], from_q32, Rw(pws, i), 64);
4528 }
4529 break;
4530 default:
4531 assert(0);
4532 }
4533
4534 msa_move_v(pwd, pwx);
4535}
4536
4537void helper_msa_ftint_s_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4538 uint32_t ws)
4539{
4540 wr_t wx, *pwx = &wx;
4541 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4542 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4543 uint32_t i;
4544
4545 clear_msacsr_cause(env);
4546
4547 switch (df) {
4548 case DF_WORD:
4549 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4550 MSA_FLOAT_UNOP0(pwx->w[i], to_int32, pws->w[i], 32);
4551 }
4552 break;
4553 case DF_DOUBLE:
4554 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4555 MSA_FLOAT_UNOP0(pwx->d[i], to_int64, pws->d[i], 64);
4556 }
4557 break;
4558 default:
4559 assert(0);
4560 }
4561
4562 check_msacsr_cause(env, GETPC());
4563
4564 msa_move_v(pwd, pwx);
4565}
4566
4567void helper_msa_ftint_u_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4568 uint32_t ws)
4569{
4570 wr_t wx, *pwx = &wx;
4571 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4572 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4573 uint32_t i;
4574
4575 clear_msacsr_cause(env);
4576
4577 switch (df) {
4578 case DF_WORD:
4579 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4580 MSA_FLOAT_UNOP0(pwx->w[i], to_uint32, pws->w[i], 32);
4581 }
4582 break;
4583 case DF_DOUBLE:
4584 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4585 MSA_FLOAT_UNOP0(pwx->d[i], to_uint64, pws->d[i], 64);
4586 }
4587 break;
4588 default:
4589 assert(0);
4590 }
4591
4592 check_msacsr_cause(env, GETPC());
4593
4594 msa_move_v(pwd, pwx);
4595}
4596
4597#define float32_from_int32 int32_to_float32
4598#define float32_from_uint32 uint32_to_float32
4599
4600#define float64_from_int64 int64_to_float64
4601#define float64_from_uint64 uint64_to_float64
4602
4603void helper_msa_ffint_s_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4604 uint32_t ws)
4605{
4606 wr_t wx, *pwx = &wx;
4607 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4608 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4609 uint32_t i;
4610
4611 clear_msacsr_cause(env);
4612
4613 switch (df) {
4614 case DF_WORD:
4615 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4616 MSA_FLOAT_UNOP(pwx->w[i], from_int32, pws->w[i], 32);
4617 }
4618 break;
4619 case DF_DOUBLE:
4620 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4621 MSA_FLOAT_UNOP(pwx->d[i], from_int64, pws->d[i], 64);
4622 }
4623 break;
4624 default:
4625 assert(0);
4626 }
4627
4628 check_msacsr_cause(env, GETPC());
4629
4630 msa_move_v(pwd, pwx);
4631}
4632
4633void helper_msa_ffint_u_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
4634 uint32_t ws)
4635{
4636 wr_t wx, *pwx = &wx;
4637 wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
4638 wr_t *pws = &(env->active_fpu.fpr[ws].wr);
4639 uint32_t i;
4640
4641 clear_msacsr_cause(env);
4642
4643 switch (df) {
4644 case DF_WORD:
4645 for (i = 0; i < DF_ELEMENTS(DF_WORD); i++) {
4646 MSA_FLOAT_UNOP(pwx->w[i], from_uint32, pws->w[i], 32);
4647 }
4648 break;
4649 case DF_DOUBLE:
4650 for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {
4651 MSA_FLOAT_UNOP(pwx->d[i], from_uint64, pws->d[i], 64);
4652 }
4653 break;
4654 default:
4655 assert(0);
4656 }
4657
4658 check_msacsr_cause(env, GETPC());
4659
4660 msa_move_v(pwd, pwx);
4661}
4662