summaryrefslogtreecommitdiff
path: root/src/arch/alpha/isa/decoder.isa
blob: 7014d4c221c7caa1efe49fe88a920074ff7ac1ef (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
// -*- mode:c++ -*-

// Copyright (c) 2003-2006 The Regents of The University of Michigan
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met: redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer;
// redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution;
// neither the name of the copyright holders nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Authors: Steve Reinhardt

////////////////////////////////////////////////////////////////////
//
// The actual decoder specification
//

decode OPCODE default Unknown::unknown() {

    format LoadAddress {
        0x08: lda({{ Ra = Rb + disp; }});
        0x09: ldah({{ Ra = Rb + (disp << 16); }});
    }

    format LoadOrNop {
        0x0a: ldbu({{ Ra.uq = Mem.ub; }});
        0x0c: ldwu({{ Ra.uq = Mem.uw; }});
        0x0b: ldq_u({{ Ra = Mem.uq; }}, ea_code = {{ EA = (Rb + disp) & ~7; }});
        0x23: ldt({{ Fa = Mem.df; }});
        0x2a: ldl_l({{ Ra.sl = Mem.sl; }}, mem_flags = LOCKED);
        0x2b: ldq_l({{ Ra.uq = Mem.uq; }}, mem_flags = LOCKED);
#ifdef USE_COPY
        0x20: MiscPrefetch::copy_load({{ EA = Ra; }},
                                      {{ fault = xc->copySrcTranslate(EA); }},
                                      inst_flags = [IsMemRef, IsLoad, IsCopy]);
#endif
    }

    format LoadOrPrefetch {
        0x28: ldl({{ Ra.sl = Mem.sl; }});
        0x29: ldq({{ Ra.uq = Mem.uq; }}, pf_flags = EVICT_NEXT);
        // IsFloating flag on lds gets the prefetch to disassemble
        // using f31 instead of r31... funcitonally it's unnecessary
        0x22: lds({{ Fa.uq = s_to_t(Mem.ul); }},
                  pf_flags = PF_EXCLUSIVE, inst_flags = IsFloating);
    }

    format Store {
        0x0e: stb({{ Mem.ub = Ra<7:0>; }});
        0x0d: stw({{ Mem.uw = Ra<15:0>; }});
        0x2c: stl({{ Mem.ul = Ra<31:0>; }});
        0x2d: stq({{ Mem.uq = Ra.uq; }});
        0x0f: stq_u({{ Mem.uq = Ra.uq; }}, {{ EA = (Rb + disp) & ~7; }});
        0x26: sts({{ Mem.ul = t_to_s(Fa.uq); }});
        0x27: stt({{ Mem.df = Fa; }});
#ifdef USE_COPY
        0x24: MiscPrefetch::copy_store({{ EA = Rb; }},
                                       {{ fault = xc->copy(EA); }},
                                       inst_flags = [IsMemRef, IsStore, IsCopy]);
#endif
    }

    format StoreCond {
        0x2e: stl_c({{ Mem.ul = Ra<31:0>; }},
                    {{
                        uint64_t tmp = write_result;
                        // see stq_c
                        Ra = (tmp == 0 || tmp == 1) ? tmp : Ra;
                    }}, mem_flags = LOCKED, inst_flags = IsStoreConditional);
        0x2f: stq_c({{ Mem.uq = Ra; }},
                    {{
                        uint64_t tmp = write_result;
                        // If the write operation returns 0 or 1, then
                        // this was a conventional store conditional,
                        // and the value indicates the success/failure
                        // of the operation.  If another value is
                        // returned, then this was a Turbolaser
                        // mailbox access, and we don't update the
                        // result register at all.
                        Ra = (tmp == 0 || tmp == 1) ? tmp : Ra;
                    }}, mem_flags = LOCKED, inst_flags = IsStoreConditional);
    }

    format IntegerOperate {

        0x10: decode INTFUNC {	// integer arithmetic operations

            0x00: addl({{ Rc.sl = Ra.sl + Rb_or_imm.sl; }});
            0x40: addlv({{
                uint32_t tmp  = Ra.sl + Rb_or_imm.sl;
                // signed overflow occurs when operands have same sign
                // and sign of result does not match.
                if (Ra.sl<31:> == Rb_or_imm.sl<31:> && tmp<31:> != Ra.sl<31:>)
                    fault = new IntegerOverflowFault;
                Rc.sl = tmp;
            }});
            0x02: s4addl({{ Rc.sl = (Ra.sl << 2) + Rb_or_imm.sl; }});
            0x12: s8addl({{ Rc.sl = (Ra.sl << 3) + Rb_or_imm.sl; }});

            0x20: addq({{ Rc = Ra + Rb_or_imm; }});
            0x60: addqv({{
                uint64_t tmp = Ra + Rb_or_imm;
                // signed overflow occurs when operands have same sign
                // and sign of result does not match.
                if (Ra<63:> == Rb_or_imm<63:> && tmp<63:> != Ra<63:>)
                    fault = new IntegerOverflowFault;
                Rc = tmp;
            }});
            0x22: s4addq({{ Rc = (Ra << 2) + Rb_or_imm; }});
            0x32: s8addq({{ Rc = (Ra << 3) + Rb_or_imm; }});

            0x09: subl({{ Rc.sl = Ra.sl - Rb_or_imm.sl; }});
            0x49: sublv({{
                uint32_t tmp  = Ra.sl - Rb_or_imm.sl;
                // signed overflow detection is same as for add,
                // except we need to look at the *complemented*
                // sign bit of the subtrahend (Rb), i.e., if the initial
                // signs are the *same* then no overflow can occur
                if (Ra.sl<31:> != Rb_or_imm.sl<31:> && tmp<31:> != Ra.sl<31:>)
                    fault = new IntegerOverflowFault;
                Rc.sl = tmp;
            }});
            0x0b: s4subl({{ Rc.sl = (Ra.sl << 2) - Rb_or_imm.sl; }});
            0x1b: s8subl({{ Rc.sl = (Ra.sl << 3) - Rb_or_imm.sl; }});

            0x29: subq({{ Rc = Ra - Rb_or_imm; }});
            0x69: subqv({{
                uint64_t tmp  = Ra - Rb_or_imm;
                // signed overflow detection is same as for add,
                // except we need to look at the *complemented*
                // sign bit of the subtrahend (Rb), i.e., if the initial
                // signs are the *same* then no overflow can occur
                if (Ra<63:> != Rb_or_imm<63:> && tmp<63:> != Ra<63:>)
                    fault = new IntegerOverflowFault;
                Rc = tmp;
            }});
            0x2b: s4subq({{ Rc = (Ra << 2) - Rb_or_imm; }});
            0x3b: s8subq({{ Rc = (Ra << 3) - Rb_or_imm; }});

            0x2d: cmpeq({{ Rc = (Ra == Rb_or_imm); }});
            0x6d: cmple({{ Rc = (Ra.sq <= Rb_or_imm.sq); }});
            0x4d: cmplt({{ Rc = (Ra.sq <  Rb_or_imm.sq); }});
            0x3d: cmpule({{ Rc = (Ra.uq <= Rb_or_imm.uq); }});
            0x1d: cmpult({{ Rc = (Ra.uq <  Rb_or_imm.uq); }});

            0x0f: cmpbge({{
                int hi = 7;
                int lo = 0;
                uint64_t tmp = 0;
                for (int i = 0; i < 8; ++i) {
                    tmp |= (Ra.uq<hi:lo> >= Rb_or_imm.uq<hi:lo>) << i;
                    hi += 8;
                    lo += 8;
                }
                Rc = tmp;
            }});
        }

        0x11: decode INTFUNC {	// integer logical operations

            0x00: and({{ Rc = Ra & Rb_or_imm; }});
            0x08: bic({{ Rc = Ra & ~Rb_or_imm; }});
            0x20: bis({{ Rc = Ra | Rb_or_imm; }});
            0x28: ornot({{ Rc = Ra | ~Rb_or_imm; }});
            0x40: xor({{ Rc = Ra ^ Rb_or_imm; }});
            0x48: eqv({{ Rc = Ra ^ ~Rb_or_imm; }});

            // conditional moves
            0x14: cmovlbs({{ Rc = ((Ra & 1) == 1) ? Rb_or_imm : Rc; }});
            0x16: cmovlbc({{ Rc = ((Ra & 1) == 0) ? Rb_or_imm : Rc; }});
            0x24: cmoveq({{ Rc = (Ra == 0) ? Rb_or_imm : Rc; }});
            0x26: cmovne({{ Rc = (Ra != 0) ? Rb_or_imm : Rc; }});
            0x44: cmovlt({{ Rc = (Ra.sq <  0) ? Rb_or_imm : Rc; }});
            0x46: cmovge({{ Rc = (Ra.sq >= 0) ? Rb_or_imm : Rc; }});
            0x64: cmovle({{ Rc = (Ra.sq <= 0) ? Rb_or_imm : Rc; }});
            0x66: cmovgt({{ Rc = (Ra.sq >  0) ? Rb_or_imm : Rc; }});

            // For AMASK, RA must be R31.
            0x61: decode RA {
                31: amask({{ Rc = Rb_or_imm & ~ULL(0x17); }});
            }

            // For IMPLVER, RA must be R31 and the B operand
            // must be the immediate value 1.
            0x6c: decode RA {
                31: decode IMM {
                    1: decode INTIMM {
                        // return EV5 for FULL_SYSTEM and EV6 otherwise
                        1: implver({{
#if FULL_SYSTEM
                             Rc = 1;
#else
                             Rc = 2;
#endif
                        }});
                    }
                }
            }

#if FULL_SYSTEM
            // The mysterious 11.25...
            0x25: WarnUnimpl::eleven25();
#endif
        }

        0x12: decode INTFUNC {
            0x39: sll({{ Rc = Ra << Rb_or_imm<5:0>; }});
            0x34: srl({{ Rc = Ra.uq >> Rb_or_imm<5:0>; }});
            0x3c: sra({{ Rc = Ra.sq >> Rb_or_imm<5:0>; }});

            0x02: mskbl({{ Rc = Ra & ~(mask( 8) << (Rb_or_imm<2:0> * 8)); }});
            0x12: mskwl({{ Rc = Ra & ~(mask(16) << (Rb_or_imm<2:0> * 8)); }});
            0x22: mskll({{ Rc = Ra & ~(mask(32) << (Rb_or_imm<2:0> * 8)); }});
            0x32: mskql({{ Rc = Ra & ~(mask(64) << (Rb_or_imm<2:0> * 8)); }});

            0x52: mskwh({{
                int bv = Rb_or_imm<2:0>;
                Rc =  bv ? (Ra & ~(mask(16) >> (64 - 8 * bv))) : Ra;
            }});
            0x62: msklh({{
                int bv = Rb_or_imm<2:0>;
                Rc =  bv ? (Ra & ~(mask(32) >> (64 - 8 * bv))) : Ra;
            }});
            0x72: mskqh({{
                int bv = Rb_or_imm<2:0>;
                Rc =  bv ? (Ra & ~(mask(64) >> (64 - 8 * bv))) : Ra;
            }});

            0x06: extbl({{ Rc = (Ra.uq >> (Rb_or_imm<2:0> * 8))< 7:0>; }});
            0x16: extwl({{ Rc = (Ra.uq >> (Rb_or_imm<2:0> * 8))<15:0>; }});
            0x26: extll({{ Rc = (Ra.uq >> (Rb_or_imm<2:0> * 8))<31:0>; }});
            0x36: extql({{ Rc = (Ra.uq >> (Rb_or_imm<2:0> * 8)); }});

            0x5a: extwh({{
                Rc = (Ra << (64 - (Rb_or_imm<2:0> * 8))<5:0>)<15:0>; }});
            0x6a: extlh({{
                Rc = (Ra << (64 - (Rb_or_imm<2:0> * 8))<5:0>)<31:0>; }});
            0x7a: extqh({{
                Rc = (Ra << (64 - (Rb_or_imm<2:0> * 8))<5:0>); }});

            0x0b: insbl({{ Rc = Ra< 7:0> << (Rb_or_imm<2:0> * 8); }});
            0x1b: inswl({{ Rc = Ra<15:0> << (Rb_or_imm<2:0> * 8); }});
            0x2b: insll({{ Rc = Ra<31:0> << (Rb_or_imm<2:0> * 8); }});
            0x3b: insql({{ Rc = Ra       << (Rb_or_imm<2:0> * 8); }});

            0x57: inswh({{
                int bv = Rb_or_imm<2:0>;
                Rc = bv ? (Ra.uq<15:0> >> (64 - 8 * bv)) : 0;
            }});
            0x67: inslh({{
                int bv = Rb_or_imm<2:0>;
                Rc = bv ? (Ra.uq<31:0> >> (64 - 8 * bv)) : 0;
            }});
            0x77: insqh({{
                int bv = Rb_or_imm<2:0>;
                Rc = bv ? (Ra.uq       >> (64 - 8 * bv)) : 0;
            }});

            0x30: zap({{
                uint64_t zapmask = 0;
                for (int i = 0; i < 8; ++i) {
                    if (Rb_or_imm<i:>)
                        zapmask |= (mask(8) << (i * 8));
                }
                Rc = Ra & ~zapmask;
            }});
            0x31: zapnot({{
                uint64_t zapmask = 0;
                for (int i = 0; i < 8; ++i) {
                    if (!Rb_or_imm<i:>)
                        zapmask |= (mask(8) << (i * 8));
                }
                Rc = Ra & ~zapmask;
            }});
        }

        0x13: decode INTFUNC {	// integer multiplies
            0x00: mull({{ Rc.sl = Ra.sl * Rb_or_imm.sl; }}, IntMultOp);
            0x20: mulq({{ Rc    = Ra    * Rb_or_imm;    }}, IntMultOp);
            0x30: umulh({{
                uint64_t hi, lo;
                mul128(Ra, Rb_or_imm, hi, lo);
                Rc = hi;
            }}, IntMultOp);
            0x40: mullv({{
                // 32-bit multiply with trap on overflow
                int64_t Rax = Ra.sl;	// sign extended version of Ra.sl
                int64_t Rbx = Rb_or_imm.sl;
                int64_t tmp = Rax * Rbx;
                // To avoid overflow, all the upper 32 bits must match
                // the sign bit of the lower 32.  We code this as
                // checking the upper 33 bits for all 0s or all 1s.
                uint64_t sign_bits = tmp<63:31>;
                if (sign_bits != 0 && sign_bits != mask(33))
                    fault = new IntegerOverflowFault;
                Rc.sl = tmp<31:0>;
            }}, IntMultOp);
            0x60: mulqv({{
                // 64-bit multiply with trap on overflow
                uint64_t hi, lo;
                mul128(Ra, Rb_or_imm, hi, lo);
                // all the upper 64 bits must match the sign bit of
                // the lower 64
                if (!((hi == 0 && lo<63:> == 0) ||
                      (hi == mask(64) && lo<63:> == 1)))
                    fault = new IntegerOverflowFault;
                Rc = lo;
            }}, IntMultOp);
        }

        0x1c: decode INTFUNC {
            0x00: decode RA { 31: sextb({{ Rc.sb = Rb_or_imm< 7:0>; }}); }
            0x01: decode RA { 31: sextw({{ Rc.sw = Rb_or_imm<15:0>; }}); }
            0x32: ctlz({{
                             uint64_t count = 0;
                             uint64_t temp = Rb;
                             if (temp<63:32>) temp >>= 32; else count += 32;
                             if (temp<31:16>) temp >>= 16; else count += 16;
                             if (temp<15:8>) temp >>= 8; else count += 8;
                             if (temp<7:4>) temp >>= 4; else count += 4;
                             if (temp<3:2>) temp >>= 2; else count += 2;
                             if (temp<1:1>) temp >>= 1; else count += 1;
                             if ((temp<0:0>) != 0x1) count += 1;
                             Rc = count;
                           }}, IntAluOp);

            0x33: cttz({{
                             uint64_t count = 0;
                             uint64_t temp = Rb;
                             if (!(temp<31:0>)) { temp >>= 32; count += 32; }
                             if (!(temp<15:0>)) { temp >>= 16; count += 16; }
                             if (!(temp<7:0>)) { temp >>= 8; count += 8; }
                             if (!(temp<3:0>)) { temp >>= 4; count += 4; }
                             if (!(temp<1:0>)) { temp >>= 2; count += 2; }
                             if (!(temp<0:0> & ULL(0x1))) count += 1;
                             Rc = count;
                           }}, IntAluOp);

            format FailUnimpl {
                0x30: ctpop();
                0x31: perr();
                0x34: unpkbw();
                0x35: unpkbl();
                0x36: pkwb();
                0x37: pklb();
                0x38: minsb8();
                0x39: minsw4();
                0x3a: minub8();
                0x3b: minuw4();
                0x3c: maxub8();
                0x3d: maxuw4();
                0x3e: maxsb8();
                0x3f: maxsw4();
            }

            format BasicOperateWithNopCheck {
                0x70: decode RB {
                    31: ftoit({{ Rc = Fa.uq; }}, FloatCvtOp);
                }
                0x78: decode RB {
                    31: ftois({{ Rc.sl = t_to_s(Fa.uq); }},
                              FloatCvtOp);
                }
            }
        }
    }

    // Conditional branches.
    format CondBranch {
        0x39: beq({{ cond = (Ra == 0); }});
        0x3d: bne({{ cond = (Ra != 0); }});
        0x3e: bge({{ cond = (Ra.sq >= 0); }});
        0x3f: bgt({{ cond = (Ra.sq >  0); }});
        0x3b: ble({{ cond = (Ra.sq <= 0); }});
        0x3a: blt({{ cond = (Ra.sq < 0); }});
        0x38: blbc({{ cond = ((Ra & 1) == 0); }});
        0x3c: blbs({{ cond = ((Ra & 1) == 1); }});

        0x31: fbeq({{ cond = (Fa == 0); }});
        0x35: fbne({{ cond = (Fa != 0); }});
        0x36: fbge({{ cond = (Fa >= 0); }});
        0x37: fbgt({{ cond = (Fa >  0); }});
        0x33: fble({{ cond = (Fa <= 0); }});
        0x32: fblt({{ cond = (Fa < 0); }});
    }

    // unconditional branches
    format UncondBranch {
        0x30: br();
        0x34: bsr(IsCall);
    }

    // indirect branches
    0x1a: decode JMPFUNC {
        format Jump {
            0: jmp();
            1: jsr(IsCall);
            2: ret(IsReturn);
            3: jsr_coroutine(IsCall, IsReturn);
        }
    }

    // Square root and integer-to-FP moves
    0x14: decode FP_SHORTFUNC {
        // Integer to FP register moves must have RB == 31
        0x4: decode RB {
            31: decode FP_FULLFUNC {
                format BasicOperateWithNopCheck {
                    0x004: itofs({{ Fc.uq = s_to_t(Ra.ul); }}, FloatCvtOp);
                    0x024: itoft({{ Fc.uq = Ra.uq; }}, FloatCvtOp);
                    0x014: FailUnimpl::itoff();	// VAX-format conversion
                }
            }
        }

        // Square root instructions must have FA == 31
        0xb: decode FA {
            31: decode FP_TYPEFUNC {
                format FloatingPointOperate {
#if SS_COMPATIBLE_FP
                    0x0b: sqrts({{
                        if (Fb < 0.0)
                            fault = new ArithmeticFault;
                        Fc = sqrt(Fb);
                    }}, FloatSqrtOp);
#else
                    0x0b: sqrts({{
                        if (Fb.sf < 0.0)
                            fault = new ArithmeticFault;
                        Fc.sf = sqrt(Fb.sf);
                    }}, FloatSqrtOp);
#endif
                    0x2b: sqrtt({{
                        if (Fb < 0.0)
                            fault = new ArithmeticFault;
                        Fc = sqrt(Fb);
                    }}, FloatSqrtOp);
                }
            }
        }

        // VAX-format sqrtf and sqrtg are not implemented
        0xa: FailUnimpl::sqrtfg();
    }

    // IEEE floating point
    0x16: decode FP_SHORTFUNC_TOP2 {
        // The top two bits of the short function code break this
        // space into four groups: binary ops, compares, reserved, and
        // conversions.  See Table 4-12 of AHB.  There are different
        // special cases in these different groups, so we decode on
        // these top two bits first just to select a decode strategy.
        // Most of these instructions may have various trapping and
        // rounding mode flags set; these are decoded in the
        // FloatingPointDecode template used by the
        // FloatingPointOperate format.

        // add/sub/mul/div: just decode on the short function code
        // and source type.  All valid trapping and rounding modes apply.
        0: decode FP_TRAPMODE {
            // check for valid trapping modes here
            0,1,5,7: decode FP_TYPEFUNC {
                   format FloatingPointOperate {
#if SS_COMPATIBLE_FP
                       0x00: adds({{ Fc = Fa + Fb; }});
                       0x01: subs({{ Fc = Fa - Fb; }});
                       0x02: muls({{ Fc = Fa * Fb; }}, FloatMultOp);
                       0x03: divs({{ Fc = Fa / Fb; }}, FloatDivOp);
#else
                       0x00: adds({{ Fc.sf = Fa.sf + Fb.sf; }});
                       0x01: subs({{ Fc.sf = Fa.sf - Fb.sf; }});
                       0x02: muls({{ Fc.sf = Fa.sf * Fb.sf; }}, FloatMultOp);
                       0x03: divs({{ Fc.sf = Fa.sf / Fb.sf; }}, FloatDivOp);
#endif

                       0x20: addt({{ Fc = Fa + Fb; }});
                       0x21: subt({{ Fc = Fa - Fb; }});
                       0x22: mult({{ Fc = Fa * Fb; }}, FloatMultOp);
                       0x23: divt({{ Fc = Fa / Fb; }}, FloatDivOp);
                   }
             }
        }

        // Floating-point compare instructions must have the default
        // rounding mode, and may use the default trapping mode or
        // /SU.  Both trapping modes are treated the same by M5; the
        // only difference on the real hardware (as far a I can tell)
        // is that without /SU you'd get an imprecise trap if you
        // tried to compare a NaN with something else (instead of an
        // "unordered" result).
        1: decode FP_FULLFUNC {
            format BasicOperateWithNopCheck {
                0x0a5, 0x5a5: cmpteq({{ Fc = (Fa == Fb) ? 2.0 : 0.0; }},
                                     FloatCmpOp);
                0x0a7, 0x5a7: cmptle({{ Fc = (Fa <= Fb) ? 2.0 : 0.0; }},
                                     FloatCmpOp);
                0x0a6, 0x5a6: cmptlt({{ Fc = (Fa <  Fb) ? 2.0 : 0.0; }},
                                     FloatCmpOp);
                0x0a4, 0x5a4: cmptun({{ // unordered
                    Fc = (!(Fa < Fb) && !(Fa == Fb) && !(Fa > Fb)) ? 2.0 : 0.0;
                }}, FloatCmpOp);
            }
        }

        // The FP-to-integer and integer-to-FP conversion insts
        // require that FA be 31.
        3: decode FA {
            31: decode FP_TYPEFUNC {
                format FloatingPointOperate {
                    0x2f: decode FP_ROUNDMODE {
                        format FPFixedRounding {
                            // "chopped" i.e. round toward zero
                            0: cvttq({{ Fc.sq = (int64_t)trunc(Fb); }},
                                     Chopped);
                            // round to minus infinity
                            1: cvttq({{ Fc.sq = (int64_t)floor(Fb); }},
                                     MinusInfinity);
                        }
                      default: cvttq({{ Fc.sq = (int64_t)nearbyint(Fb); }});
                    }

                    // The cvtts opcode is overloaded to be cvtst if the trap
                    // mode is 2 or 6 (which are not valid otherwise)
                    0x2c: decode FP_FULLFUNC {
                        format BasicOperateWithNopCheck {
                            // trap on denorm version "cvtst/s" is
                            // simulated same as cvtst
                            0x2ac, 0x6ac: cvtst({{ Fc = Fb.sf; }});
                        }
                      default: cvtts({{ Fc.sf = Fb; }});
                    }

                    // The trapping mode for integer-to-FP conversions
                    // must be /SUI or nothing; /U and /SU are not
                    // allowed.  The full set of rounding modes are
                    // supported though.
                    0x3c: decode FP_TRAPMODE {
                        0,7: cvtqs({{ Fc.sf = Fb.sq; }});
                    }
                    0x3e: decode FP_TRAPMODE {
                        0,7: cvtqt({{ Fc    = Fb.sq; }});
                    }
                }
            }
        }
    }

    // misc FP operate
    0x17: decode FP_FULLFUNC {
        format BasicOperateWithNopCheck {
            0x010: cvtlq({{
                Fc.sl = (Fb.uq<63:62> << 30) | Fb.uq<58:29>;
            }});
            0x030: cvtql({{
                Fc.uq = (Fb.uq<31:30> << 62) | (Fb.uq<29:0> << 29);
            }});

            // We treat the precise & imprecise trapping versions of
            // cvtql identically.
            0x130, 0x530: cvtqlv({{
                // To avoid overflow, all the upper 32 bits must match
                // the sign bit of the lower 32.  We code this as
                // checking the upper 33 bits for all 0s or all 1s.
                uint64_t sign_bits = Fb.uq<63:31>;
                if (sign_bits != 0 && sign_bits != mask(33))
                    fault = new IntegerOverflowFault;
                Fc.uq = (Fb.uq<31:30> << 62) | (Fb.uq<29:0> << 29);
            }});

            0x020: cpys({{  // copy sign
                Fc.uq = (Fa.uq<63:> << 63) | Fb.uq<62:0>;
            }});
            0x021: cpysn({{ // copy sign negated
                Fc.uq = (~Fa.uq<63:> << 63) | Fb.uq<62:0>;
            }});
            0x022: cpyse({{ // copy sign and exponent
                Fc.uq = (Fa.uq<63:52> << 52) | Fb.uq<51:0>;
            }});

            0x02a: fcmoveq({{ Fc = (Fa == 0) ? Fb : Fc; }});
            0x02b: fcmovne({{ Fc = (Fa != 0) ? Fb : Fc; }});
            0x02c: fcmovlt({{ Fc = (Fa <  0) ? Fb : Fc; }});
            0x02d: fcmovge({{ Fc = (Fa >= 0) ? Fb : Fc; }});
            0x02e: fcmovle({{ Fc = (Fa <= 0) ? Fb : Fc; }});
            0x02f: fcmovgt({{ Fc = (Fa >  0) ? Fb : Fc; }});

            0x024: mt_fpcr({{ FPCR = Fa.uq; }}, IsIprAccess);
            0x025: mf_fpcr({{ Fa.uq = FPCR; }}, IsIprAccess);
        }
    }

    // miscellaneous mem-format ops
    0x18: decode MEMFUNC {
        format WarnUnimpl {
            0x8000: fetch();
            0xa000: fetch_m();
            0xe800: ecb();
        }

        format MiscPrefetch {
            0xf800: wh64({{ EA = Rb & ~ULL(63); }},
                         {{ xc->writeHint(EA, 64, memAccessFlags); }},
                         mem_flags = NO_FAULT,
                         inst_flags = [IsMemRef, IsDataPrefetch,
                                       IsStore, MemWriteOp]);
        }

        format BasicOperate {
            0xc000: rpcc({{
#if FULL_SYSTEM
        /* Rb is a fake dependency so here is a fun way to get
         * the parser to understand that.
         */
                Ra = xc->readMiscRegWithEffect(AlphaISA::IPR_CC) + (Rb & 0);

#else
                Ra = curTick;
#endif
            }}, IsUnverifiable);

            // All of the barrier instructions below do nothing in
            // their execute() methods (hence the empty code blocks).
            // All of their functionality is hard-coded in the
            // pipeline based on the flags IsSerializing,
            // IsMemBarrier, and IsWriteBarrier.  In the current
            // detailed CPU model, the execute() function only gets
            // called at fetch, so there's no way to generate pipeline
            // behavior at any other stage.  Once we go to an
            // exec-in-exec CPU model we should be able to get rid of
            // these flags and implement this behavior via the
            // execute() methods.

            // trapb is just a barrier on integer traps, where excb is
            // a barrier on integer and FP traps.  "EXCB is thus a
            // superset of TRAPB." (Alpha ARM, Sec 4.11.4) We treat
            // them the same though.
            0x0000: trapb({{ }}, IsSerializing, IsSerializeBefore, No_OpClass);
            0x0400: excb({{ }}, IsSerializing, IsSerializeBefore, No_OpClass);
            0x4000: mb({{ }}, IsMemBarrier, MemReadOp);
            0x4400: wmb({{ }}, IsWriteBarrier, MemWriteOp);
        }

#if FULL_SYSTEM
        format BasicOperate {
            0xe000: rc({{
                Ra = IntrFlag;
                IntrFlag = 0;
            }}, IsNonSpeculative, IsUnverifiable);
            0xf000: rs({{
                Ra = IntrFlag;
                IntrFlag = 1;
            }}, IsNonSpeculative, IsUnverifiable);
        }
#else
        format FailUnimpl {
            0xe000: rc();
            0xf000: rs();
        }
#endif
    }

#if FULL_SYSTEM
    0x00: CallPal::call_pal({{
        if (!palValid ||
            (palPriv
             && xc->readMiscRegWithEffect(AlphaISA::IPR_ICM) != AlphaISA::mode_kernel)) {
            // invalid pal function code, or attempt to do privileged
            // PAL call in non-kernel mode
            fault = new UnimplementedOpcodeFault;
        }
        else {
            // check to see if simulator wants to do something special
            // on this PAL call (including maybe suppress it)
            bool dopal = xc->simPalCheck(palFunc);

            if (dopal) {
                xc->setMiscRegWithEffect(AlphaISA::IPR_EXC_ADDR, NPC);
                NPC = xc->readMiscRegWithEffect(AlphaISA::IPR_PAL_BASE) + palOffset;
            }
        }
    }}, IsNonSpeculative);
#else
    0x00: decode PALFUNC {
        format EmulatedCallPal {
            0x00: halt ({{
                exitSimLoop("halt instruction encountered");
            }}, IsNonSpeculative);
            0x83: callsys({{
                xc->syscall(R0);
            }}, IsSerializeAfter, IsNonSpeculative);
            // Read uniq reg into ABI return value register (r0)
            0x9e: rduniq({{ R0 = Runiq; }}, IsIprAccess);
            // Write uniq reg with value from ABI arg register (r16)
            0x9f: wruniq({{ Runiq = R16; }}, IsIprAccess);
        }
    }
#endif

#if FULL_SYSTEM
    0x1b: decode PALMODE {
        0: OpcdecFault::hw_st_quad();
        1: decode HW_LDST_QUAD {
            format HwLoad {
                0: hw_ld({{ EA = (Rb + disp) & ~3; }}, {{ Ra = Mem.ul; }}, L);
                1: hw_ld({{ EA = (Rb + disp) & ~7; }}, {{ Ra = Mem.uq; }}, Q);
            }
        }
    }

    0x1f: decode PALMODE {
        0: OpcdecFault::hw_st_cond();
        format HwStore {
            1: decode HW_LDST_COND {
                0: decode HW_LDST_QUAD {
                    0: hw_st({{ EA = (Rb + disp) & ~3; }},
                {{ Mem.ul = Ra<31:0>; }}, L);
                    1: hw_st({{ EA = (Rb + disp) & ~7; }},
                {{ Mem.uq = Ra.uq; }}, Q);
                }

                1: FailUnimpl::hw_st_cond();
            }
        }
    }

    0x19: decode PALMODE {
        0: OpcdecFault::hw_mfpr();
        format HwMoveIPR {
            1: hw_mfpr({{
                int miscRegIndex = (ipr_index < MaxInternalProcRegs) ?
                        IprToMiscRegIndex[ipr_index] : -1;
                if(miscRegIndex < 0 || !IprIsReadable(miscRegIndex) ||
                    miscRegIndex >= NumInternalProcRegs)
                        fault = new UnimplementedOpcodeFault;
                else
                    Ra = xc->readMiscRegWithEffect(miscRegIndex);
            }}, IsIprAccess);
        }
    }

    0x1d: decode PALMODE {
        0: OpcdecFault::hw_mtpr();
        format HwMoveIPR {
            1: hw_mtpr({{
                int miscRegIndex = (ipr_index < MaxInternalProcRegs) ?
                        IprToMiscRegIndex[ipr_index] : -1;
                if(miscRegIndex < 0 || !IprIsWritable(miscRegIndex) ||
                    miscRegIndex >= NumInternalProcRegs)
                        fault = new UnimplementedOpcodeFault;
                else
                    xc->setMiscRegWithEffect(miscRegIndex, Ra);
                if (traceData) { traceData->setData(Ra); }
            }}, IsIprAccess);
        }
    }

    format BasicOperate {
        0x1e: decode PALMODE {
            0: OpcdecFault::hw_rei();
            1:hw_rei({{ xc->hwrei(); }}, IsSerializing, IsSerializeBefore);
        }

        // M5 special opcodes use the reserved 0x01 opcode space
        0x01: decode M5FUNC {
            0x00: arm({{
                AlphaPseudo::arm(xc->tcBase());
            }}, IsNonSpeculative);
            0x01: quiesce({{
                AlphaPseudo::quiesce(xc->tcBase());
            }}, IsNonSpeculative, IsQuiesce);
            0x02: quiesceNs({{
                AlphaPseudo::quiesceNs(xc->tcBase(), R16);
            }}, IsNonSpeculative, IsQuiesce);
            0x03: quiesceCycles({{
                AlphaPseudo::quiesceCycles(xc->tcBase(), R16);
            }}, IsNonSpeculative, IsQuiesce, IsUnverifiable);
            0x04: quiesceTime({{
                R0 = AlphaPseudo::quiesceTime(xc->tcBase());
            }}, IsNonSpeculative, IsUnverifiable);
            0x20: m5exit_old({{
                AlphaPseudo::m5exit_old(xc->tcBase());
            }}, No_OpClass, IsNonSpeculative);
            0x21: m5exit({{
                AlphaPseudo::m5exit(xc->tcBase(), R16);
            }}, No_OpClass, IsNonSpeculative);
            0x31: loadsymbol({{
                AlphaPseudo::loadsymbol(xc->tcBase());
            }}, No_OpClass, IsNonSpeculative);
            0x30: initparam({{ Ra = xc->tcBase()->getCpuPtr()->system->init_param; }});
            0x40: resetstats({{
                AlphaPseudo::resetstats(xc->tcBase(), R16, R17);
            }}, IsNonSpeculative);
            0x41: dumpstats({{
                AlphaPseudo::dumpstats(xc->tcBase(), R16, R17);
            }}, IsNonSpeculative);
            0x42: dumpresetstats({{
                AlphaPseudo::dumpresetstats(xc->tcBase(), R16, R17);
            }}, IsNonSpeculative);
            0x43: m5checkpoint({{
                AlphaPseudo::m5checkpoint(xc->tcBase(), R16, R17);
            }}, IsNonSpeculative);
            0x50: m5readfile({{
                R0 = AlphaPseudo::readfile(xc->tcBase(), R16, R17, R18);
            }}, IsNonSpeculative);
            0x51: m5break({{
                AlphaPseudo::debugbreak(xc->tcBase());
            }}, IsNonSpeculative);
            0x52: m5switchcpu({{
                AlphaPseudo::switchcpu(xc->tcBase());
            }}, IsNonSpeculative);
            0x53: m5addsymbol({{
                AlphaPseudo::addsymbol(xc->tcBase(), R16, R17);
            }}, IsNonSpeculative);
            0x54: m5panic({{
                panic("M5 panic instruction called at pc=%#x.", xc->readPC());
            }}, IsNonSpeculative);
            0x55: m5anBegin({{
                AlphaPseudo::anBegin(xc->tcBase(), R16);
            }}, IsNonSpeculative);
            0x56: m5anWait({{
                AlphaPseudo::anWait(xc->tcBase(), R16, R17);
            }}, IsNonSpeculative);
        }
    }
#endif
}