forked from llvm/llvm-project
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathSparcISelLowering.cpp
3441 lines (2966 loc) · 134 KB
/
SparcISelLowering.cpp
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
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
//===-- SparcISelLowering.cpp - Sparc DAG Lowering Implementation ---------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file implements the interfaces that Sparc uses to lower LLVM code into a
// selection DAG.
//
//===----------------------------------------------------------------------===//
#include "SparcISelLowering.h"
#include "MCTargetDesc/SparcMCExpr.h"
#include "SparcMachineFunctionInfo.h"
#include "SparcRegisterInfo.h"
#include "SparcTargetMachine.h"
#include "SparcTargetObjectFile.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/ADT/StringSwitch.h"
#include "llvm/CodeGen/CallingConvLower.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/Module.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/KnownBits.h"
using namespace llvm;
//===----------------------------------------------------------------------===//
// Calling Convention Implementation
//===----------------------------------------------------------------------===//
static bool CC_Sparc_Assign_SRet(unsigned &ValNo, MVT &ValVT,
MVT &LocVT, CCValAssign::LocInfo &LocInfo,
ISD::ArgFlagsTy &ArgFlags, CCState &State)
{
assert (ArgFlags.isSRet());
// Assign SRet argument.
State.addLoc(CCValAssign::getCustomMem(ValNo, ValVT,
0,
LocVT, LocInfo));
return true;
}
static bool CC_Sparc_Assign_Split_64(unsigned &ValNo, MVT &ValVT,
MVT &LocVT, CCValAssign::LocInfo &LocInfo,
ISD::ArgFlagsTy &ArgFlags, CCState &State)
{
static const MCPhysReg RegList[] = {
SP::I0, SP::I1, SP::I2, SP::I3, SP::I4, SP::I5
};
// Try to get first reg.
if (Register Reg = State.AllocateReg(RegList)) {
State.addLoc(CCValAssign::getCustomReg(ValNo, ValVT, Reg, LocVT, LocInfo));
} else {
// Assign whole thing in stack.
State.addLoc(CCValAssign::getCustomMem(
ValNo, ValVT, State.AllocateStack(8, Align(4)), LocVT, LocInfo));
return true;
}
// Try to get second reg.
if (Register Reg = State.AllocateReg(RegList))
State.addLoc(CCValAssign::getCustomReg(ValNo, ValVT, Reg, LocVT, LocInfo));
else
State.addLoc(CCValAssign::getCustomMem(
ValNo, ValVT, State.AllocateStack(4, Align(4)), LocVT, LocInfo));
return true;
}
static bool CC_Sparc_Assign_Ret_Split_64(unsigned &ValNo, MVT &ValVT,
MVT &LocVT, CCValAssign::LocInfo &LocInfo,
ISD::ArgFlagsTy &ArgFlags, CCState &State)
{
static const MCPhysReg RegList[] = {
SP::I0, SP::I1, SP::I2, SP::I3, SP::I4, SP::I5
};
// Try to get first reg.
if (Register Reg = State.AllocateReg(RegList))
State.addLoc(CCValAssign::getCustomReg(ValNo, ValVT, Reg, LocVT, LocInfo));
else
return false;
// Try to get second reg.
if (Register Reg = State.AllocateReg(RegList))
State.addLoc(CCValAssign::getCustomReg(ValNo, ValVT, Reg, LocVT, LocInfo));
else
return false;
return true;
}
// Allocate a full-sized argument for the 64-bit ABI.
static bool CC_Sparc64_Full(unsigned &ValNo, MVT &ValVT,
MVT &LocVT, CCValAssign::LocInfo &LocInfo,
ISD::ArgFlagsTy &ArgFlags, CCState &State) {
assert((LocVT == MVT::f32 || LocVT == MVT::f128
|| LocVT.getSizeInBits() == 64) &&
"Can't handle non-64 bits locations");
// Stack space is allocated for all arguments starting from [%fp+BIAS+128].
unsigned size = (LocVT == MVT::f128) ? 16 : 8;
Align alignment = (LocVT == MVT::f128) ? Align(16) : Align(8);
unsigned Offset = State.AllocateStack(size, alignment);
unsigned Reg = 0;
if (LocVT == MVT::i64 && Offset < 6*8)
// Promote integers to %i0-%i5.
Reg = SP::I0 + Offset/8;
else if (LocVT == MVT::f64 && Offset < 16*8)
// Promote doubles to %d0-%d30. (Which LLVM calls D0-D15).
Reg = SP::D0 + Offset/8;
else if (LocVT == MVT::f32 && Offset < 16*8)
// Promote floats to %f1, %f3, ...
Reg = SP::F1 + Offset/4;
else if (LocVT == MVT::f128 && Offset < 16*8)
// Promote long doubles to %q0-%q28. (Which LLVM calls Q0-Q7).
Reg = SP::Q0 + Offset/16;
// Promote to register when possible, otherwise use the stack slot.
if (Reg) {
State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
return true;
}
// This argument goes on the stack in an 8-byte slot.
// When passing floats, LocVT is smaller than 8 bytes. Adjust the offset to
// the right-aligned float. The first 4 bytes of the stack slot are undefined.
if (LocVT == MVT::f32)
Offset += 4;
State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo));
return true;
}
// Allocate a half-sized argument for the 64-bit ABI.
//
// This is used when passing { float, int } structs by value in registers.
static bool CC_Sparc64_Half(unsigned &ValNo, MVT &ValVT,
MVT &LocVT, CCValAssign::LocInfo &LocInfo,
ISD::ArgFlagsTy &ArgFlags, CCState &State) {
assert(LocVT.getSizeInBits() == 32 && "Can't handle non-32 bits locations");
unsigned Offset = State.AllocateStack(4, Align(4));
if (LocVT == MVT::f32 && Offset < 16*8) {
// Promote floats to %f0-%f31.
State.addLoc(CCValAssign::getReg(ValNo, ValVT, SP::F0 + Offset/4,
LocVT, LocInfo));
return true;
}
if (LocVT == MVT::i32 && Offset < 6*8) {
// Promote integers to %i0-%i5, using half the register.
unsigned Reg = SP::I0 + Offset/8;
LocVT = MVT::i64;
LocInfo = CCValAssign::AExt;
// Set the Custom bit if this i32 goes in the high bits of a register.
if (Offset % 8 == 0)
State.addLoc(CCValAssign::getCustomReg(ValNo, ValVT, Reg,
LocVT, LocInfo));
else
State.addLoc(CCValAssign::getReg(ValNo, ValVT, Reg, LocVT, LocInfo));
return true;
}
State.addLoc(CCValAssign::getMem(ValNo, ValVT, Offset, LocVT, LocInfo));
return true;
}
#include "SparcGenCallingConv.inc"
// The calling conventions in SparcCallingConv.td are described in terms of the
// callee's register window. This function translates registers to the
// corresponding caller window %o register.
static unsigned toCallerWindow(unsigned Reg) {
static_assert(SP::I0 + 7 == SP::I7 && SP::O0 + 7 == SP::O7,
"Unexpected enum");
if (Reg >= SP::I0 && Reg <= SP::I7)
return Reg - SP::I0 + SP::O0;
return Reg;
}
SDValue
SparcTargetLowering::LowerReturn(SDValue Chain, CallingConv::ID CallConv,
bool IsVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,
const SDLoc &DL, SelectionDAG &DAG) const {
if (Subtarget->is64Bit())
return LowerReturn_64(Chain, CallConv, IsVarArg, Outs, OutVals, DL, DAG);
return LowerReturn_32(Chain, CallConv, IsVarArg, Outs, OutVals, DL, DAG);
}
SDValue
SparcTargetLowering::LowerReturn_32(SDValue Chain, CallingConv::ID CallConv,
bool IsVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,
const SDLoc &DL, SelectionDAG &DAG) const {
MachineFunction &MF = DAG.getMachineFunction();
// CCValAssign - represent the assignment of the return value to locations.
SmallVector<CCValAssign, 16> RVLocs;
// CCState - Info about the registers and stack slot.
CCState CCInfo(CallConv, IsVarArg, DAG.getMachineFunction(), RVLocs,
*DAG.getContext());
// Analyze return values.
CCInfo.AnalyzeReturn(Outs, RetCC_Sparc32);
SDValue Flag;
SmallVector<SDValue, 4> RetOps(1, Chain);
// Make room for the return address offset.
RetOps.push_back(SDValue());
// Copy the result values into the output registers.
for (unsigned i = 0, realRVLocIdx = 0;
i != RVLocs.size();
++i, ++realRVLocIdx) {
CCValAssign &VA = RVLocs[i];
assert(VA.isRegLoc() && "Can only return in registers!");
SDValue Arg = OutVals[realRVLocIdx];
if (VA.needsCustom()) {
assert(VA.getLocVT() == MVT::v2i32);
// Legalize ret v2i32 -> ret 2 x i32 (Basically: do what would
// happen by default if this wasn't a legal type)
SDValue Part0 = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, MVT::i32,
Arg,
DAG.getConstant(0, DL, getVectorIdxTy(DAG.getDataLayout())));
SDValue Part1 = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, MVT::i32,
Arg,
DAG.getConstant(1, DL, getVectorIdxTy(DAG.getDataLayout())));
Chain = DAG.getCopyToReg(Chain, DL, VA.getLocReg(), Part0, Flag);
Flag = Chain.getValue(1);
RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
VA = RVLocs[++i]; // skip ahead to next loc
Chain = DAG.getCopyToReg(Chain, DL, VA.getLocReg(), Part1,
Flag);
} else
Chain = DAG.getCopyToReg(Chain, DL, VA.getLocReg(), Arg, Flag);
// Guarantee that all emitted copies are stuck together with flags.
Flag = Chain.getValue(1);
RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
}
unsigned RetAddrOffset = 8; // Call Inst + Delay Slot
// If the function returns a struct, copy the SRetReturnReg to I0
if (MF.getFunction().hasStructRetAttr()) {
SparcMachineFunctionInfo *SFI = MF.getInfo<SparcMachineFunctionInfo>();
Register Reg = SFI->getSRetReturnReg();
if (!Reg)
llvm_unreachable("sret virtual register not created in the entry block");
auto PtrVT = getPointerTy(DAG.getDataLayout());
SDValue Val = DAG.getCopyFromReg(Chain, DL, Reg, PtrVT);
Chain = DAG.getCopyToReg(Chain, DL, SP::I0, Val, Flag);
Flag = Chain.getValue(1);
RetOps.push_back(DAG.getRegister(SP::I0, PtrVT));
RetAddrOffset = 12; // CallInst + Delay Slot + Unimp
}
RetOps[0] = Chain; // Update chain.
RetOps[1] = DAG.getConstant(RetAddrOffset, DL, MVT::i32);
// Add the flag if we have it.
if (Flag.getNode())
RetOps.push_back(Flag);
return DAG.getNode(SPISD::RET_FLAG, DL, MVT::Other, RetOps);
}
// Lower return values for the 64-bit ABI.
// Return values are passed the exactly the same way as function arguments.
SDValue
SparcTargetLowering::LowerReturn_64(SDValue Chain, CallingConv::ID CallConv,
bool IsVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,
const SDLoc &DL, SelectionDAG &DAG) const {
// CCValAssign - represent the assignment of the return value to locations.
SmallVector<CCValAssign, 16> RVLocs;
// CCState - Info about the registers and stack slot.
CCState CCInfo(CallConv, IsVarArg, DAG.getMachineFunction(), RVLocs,
*DAG.getContext());
// Analyze return values.
CCInfo.AnalyzeReturn(Outs, RetCC_Sparc64);
SDValue Flag;
SmallVector<SDValue, 4> RetOps(1, Chain);
// The second operand on the return instruction is the return address offset.
// The return address is always %i7+8 with the 64-bit ABI.
RetOps.push_back(DAG.getConstant(8, DL, MVT::i32));
// Copy the result values into the output registers.
for (unsigned i = 0; i != RVLocs.size(); ++i) {
CCValAssign &VA = RVLocs[i];
assert(VA.isRegLoc() && "Can only return in registers!");
SDValue OutVal = OutVals[i];
// Integer return values must be sign or zero extended by the callee.
switch (VA.getLocInfo()) {
case CCValAssign::Full: break;
case CCValAssign::SExt:
OutVal = DAG.getNode(ISD::SIGN_EXTEND, DL, VA.getLocVT(), OutVal);
break;
case CCValAssign::ZExt:
OutVal = DAG.getNode(ISD::ZERO_EXTEND, DL, VA.getLocVT(), OutVal);
break;
case CCValAssign::AExt:
OutVal = DAG.getNode(ISD::ANY_EXTEND, DL, VA.getLocVT(), OutVal);
break;
default:
llvm_unreachable("Unknown loc info!");
}
// The custom bit on an i32 return value indicates that it should be passed
// in the high bits of the register.
if (VA.getValVT() == MVT::i32 && VA.needsCustom()) {
OutVal = DAG.getNode(ISD::SHL, DL, MVT::i64, OutVal,
DAG.getConstant(32, DL, MVT::i32));
// The next value may go in the low bits of the same register.
// Handle both at once.
if (i+1 < RVLocs.size() && RVLocs[i+1].getLocReg() == VA.getLocReg()) {
SDValue NV = DAG.getNode(ISD::ZERO_EXTEND, DL, MVT::i64, OutVals[i+1]);
OutVal = DAG.getNode(ISD::OR, DL, MVT::i64, OutVal, NV);
// Skip the next value, it's already done.
++i;
}
}
Chain = DAG.getCopyToReg(Chain, DL, VA.getLocReg(), OutVal, Flag);
// Guarantee that all emitted copies are stuck together with flags.
Flag = Chain.getValue(1);
RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
}
RetOps[0] = Chain; // Update chain.
// Add the flag if we have it.
if (Flag.getNode())
RetOps.push_back(Flag);
return DAG.getNode(SPISD::RET_FLAG, DL, MVT::Other, RetOps);
}
SDValue SparcTargetLowering::LowerFormalArguments(
SDValue Chain, CallingConv::ID CallConv, bool IsVarArg,
const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &DL,
SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const {
if (Subtarget->is64Bit())
return LowerFormalArguments_64(Chain, CallConv, IsVarArg, Ins,
DL, DAG, InVals);
return LowerFormalArguments_32(Chain, CallConv, IsVarArg, Ins,
DL, DAG, InVals);
}
/// LowerFormalArguments32 - V8 uses a very simple ABI, where all values are
/// passed in either one or two GPRs, including FP values. TODO: we should
/// pass FP values in FP registers for fastcc functions.
SDValue SparcTargetLowering::LowerFormalArguments_32(
SDValue Chain, CallingConv::ID CallConv, bool isVarArg,
const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl,
SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const {
MachineFunction &MF = DAG.getMachineFunction();
MachineRegisterInfo &RegInfo = MF.getRegInfo();
SparcMachineFunctionInfo *FuncInfo = MF.getInfo<SparcMachineFunctionInfo>();
// Assign locations to all of the incoming arguments.
SmallVector<CCValAssign, 16> ArgLocs;
CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs,
*DAG.getContext());
CCInfo.AnalyzeFormalArguments(Ins, CC_Sparc32);
const unsigned StackOffset = 92;
bool IsLittleEndian = DAG.getDataLayout().isLittleEndian();
unsigned InIdx = 0;
for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i, ++InIdx) {
CCValAssign &VA = ArgLocs[i];
if (Ins[InIdx].Flags.isSRet()) {
if (InIdx != 0)
report_fatal_error("sparc only supports sret on the first parameter");
// Get SRet from [%fp+64].
int FrameIdx = MF.getFrameInfo().CreateFixedObject(4, 64, true);
SDValue FIPtr = DAG.getFrameIndex(FrameIdx, MVT::i32);
SDValue Arg =
DAG.getLoad(MVT::i32, dl, Chain, FIPtr, MachinePointerInfo());
InVals.push_back(Arg);
continue;
}
if (VA.isRegLoc()) {
if (VA.needsCustom()) {
assert(VA.getLocVT() == MVT::f64 || VA.getLocVT() == MVT::v2i32);
Register VRegHi = RegInfo.createVirtualRegister(&SP::IntRegsRegClass);
MF.getRegInfo().addLiveIn(VA.getLocReg(), VRegHi);
SDValue HiVal = DAG.getCopyFromReg(Chain, dl, VRegHi, MVT::i32);
assert(i+1 < e);
CCValAssign &NextVA = ArgLocs[++i];
SDValue LoVal;
if (NextVA.isMemLoc()) {
int FrameIdx = MF.getFrameInfo().
CreateFixedObject(4, StackOffset+NextVA.getLocMemOffset(),true);
SDValue FIPtr = DAG.getFrameIndex(FrameIdx, MVT::i32);
LoVal = DAG.getLoad(MVT::i32, dl, Chain, FIPtr, MachinePointerInfo());
} else {
Register loReg = MF.addLiveIn(NextVA.getLocReg(),
&SP::IntRegsRegClass);
LoVal = DAG.getCopyFromReg(Chain, dl, loReg, MVT::i32);
}
if (IsLittleEndian)
std::swap(LoVal, HiVal);
SDValue WholeValue =
DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, LoVal, HiVal);
WholeValue = DAG.getNode(ISD::BITCAST, dl, VA.getLocVT(), WholeValue);
InVals.push_back(WholeValue);
continue;
}
Register VReg = RegInfo.createVirtualRegister(&SP::IntRegsRegClass);
MF.getRegInfo().addLiveIn(VA.getLocReg(), VReg);
SDValue Arg = DAG.getCopyFromReg(Chain, dl, VReg, MVT::i32);
if (VA.getLocVT() == MVT::f32)
Arg = DAG.getNode(ISD::BITCAST, dl, MVT::f32, Arg);
else if (VA.getLocVT() != MVT::i32) {
Arg = DAG.getNode(ISD::AssertSext, dl, MVT::i32, Arg,
DAG.getValueType(VA.getLocVT()));
Arg = DAG.getNode(ISD::TRUNCATE, dl, VA.getLocVT(), Arg);
}
InVals.push_back(Arg);
continue;
}
assert(VA.isMemLoc());
unsigned Offset = VA.getLocMemOffset()+StackOffset;
auto PtrVT = getPointerTy(DAG.getDataLayout());
if (VA.needsCustom()) {
assert(VA.getValVT() == MVT::f64 || VA.getValVT() == MVT::v2i32);
// If it is double-word aligned, just load.
if (Offset % 8 == 0) {
int FI = MF.getFrameInfo().CreateFixedObject(8,
Offset,
true);
SDValue FIPtr = DAG.getFrameIndex(FI, PtrVT);
SDValue Load =
DAG.getLoad(VA.getValVT(), dl, Chain, FIPtr, MachinePointerInfo());
InVals.push_back(Load);
continue;
}
int FI = MF.getFrameInfo().CreateFixedObject(4,
Offset,
true);
SDValue FIPtr = DAG.getFrameIndex(FI, PtrVT);
SDValue HiVal =
DAG.getLoad(MVT::i32, dl, Chain, FIPtr, MachinePointerInfo());
int FI2 = MF.getFrameInfo().CreateFixedObject(4,
Offset+4,
true);
SDValue FIPtr2 = DAG.getFrameIndex(FI2, PtrVT);
SDValue LoVal =
DAG.getLoad(MVT::i32, dl, Chain, FIPtr2, MachinePointerInfo());
if (IsLittleEndian)
std::swap(LoVal, HiVal);
SDValue WholeValue =
DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, LoVal, HiVal);
WholeValue = DAG.getNode(ISD::BITCAST, dl, VA.getValVT(), WholeValue);
InVals.push_back(WholeValue);
continue;
}
int FI = MF.getFrameInfo().CreateFixedObject(4,
Offset,
true);
SDValue FIPtr = DAG.getFrameIndex(FI, PtrVT);
SDValue Load ;
if (VA.getValVT() == MVT::i32 || VA.getValVT() == MVT::f32) {
Load = DAG.getLoad(VA.getValVT(), dl, Chain, FIPtr, MachinePointerInfo());
} else if (VA.getValVT() == MVT::f128) {
report_fatal_error("SPARCv8 does not handle f128 in calls; "
"pass indirectly");
} else {
// We shouldn't see any other value types here.
llvm_unreachable("Unexpected ValVT encountered in frame lowering.");
}
InVals.push_back(Load);
}
if (MF.getFunction().hasStructRetAttr()) {
// Copy the SRet Argument to SRetReturnReg.
SparcMachineFunctionInfo *SFI = MF.getInfo<SparcMachineFunctionInfo>();
Register Reg = SFI->getSRetReturnReg();
if (!Reg) {
Reg = MF.getRegInfo().createVirtualRegister(&SP::IntRegsRegClass);
SFI->setSRetReturnReg(Reg);
}
SDValue Copy = DAG.getCopyToReg(DAG.getEntryNode(), dl, Reg, InVals[0]);
Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Copy, Chain);
}
// Store remaining ArgRegs to the stack if this is a varargs function.
if (isVarArg) {
static const MCPhysReg ArgRegs[] = {
SP::I0, SP::I1, SP::I2, SP::I3, SP::I4, SP::I5
};
unsigned NumAllocated = CCInfo.getFirstUnallocated(ArgRegs);
const MCPhysReg *CurArgReg = ArgRegs+NumAllocated, *ArgRegEnd = ArgRegs+6;
unsigned ArgOffset = CCInfo.getNextStackOffset();
if (NumAllocated == 6)
ArgOffset += StackOffset;
else {
assert(!ArgOffset);
ArgOffset = 68+4*NumAllocated;
}
// Remember the vararg offset for the va_start implementation.
FuncInfo->setVarArgsFrameOffset(ArgOffset);
std::vector<SDValue> OutChains;
for (; CurArgReg != ArgRegEnd; ++CurArgReg) {
Register VReg = RegInfo.createVirtualRegister(&SP::IntRegsRegClass);
MF.getRegInfo().addLiveIn(*CurArgReg, VReg);
SDValue Arg = DAG.getCopyFromReg(DAG.getRoot(), dl, VReg, MVT::i32);
int FrameIdx = MF.getFrameInfo().CreateFixedObject(4, ArgOffset,
true);
SDValue FIPtr = DAG.getFrameIndex(FrameIdx, MVT::i32);
OutChains.push_back(
DAG.getStore(DAG.getRoot(), dl, Arg, FIPtr, MachinePointerInfo()));
ArgOffset += 4;
}
if (!OutChains.empty()) {
OutChains.push_back(Chain);
Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains);
}
}
return Chain;
}
// Lower formal arguments for the 64 bit ABI.
SDValue SparcTargetLowering::LowerFormalArguments_64(
SDValue Chain, CallingConv::ID CallConv, bool IsVarArg,
const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &DL,
SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals) const {
MachineFunction &MF = DAG.getMachineFunction();
// Analyze arguments according to CC_Sparc64.
SmallVector<CCValAssign, 16> ArgLocs;
CCState CCInfo(CallConv, IsVarArg, DAG.getMachineFunction(), ArgLocs,
*DAG.getContext());
CCInfo.AnalyzeFormalArguments(Ins, CC_Sparc64);
// The argument array begins at %fp+BIAS+128, after the register save area.
const unsigned ArgArea = 128;
for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
CCValAssign &VA = ArgLocs[i];
if (VA.isRegLoc()) {
// This argument is passed in a register.
// All integer register arguments are promoted by the caller to i64.
// Create a virtual register for the promoted live-in value.
Register VReg = MF.addLiveIn(VA.getLocReg(),
getRegClassFor(VA.getLocVT()));
SDValue Arg = DAG.getCopyFromReg(Chain, DL, VReg, VA.getLocVT());
// Get the high bits for i32 struct elements.
if (VA.getValVT() == MVT::i32 && VA.needsCustom())
Arg = DAG.getNode(ISD::SRL, DL, VA.getLocVT(), Arg,
DAG.getConstant(32, DL, MVT::i32));
// The caller promoted the argument, so insert an Assert?ext SDNode so we
// won't promote the value again in this function.
switch (VA.getLocInfo()) {
case CCValAssign::SExt:
Arg = DAG.getNode(ISD::AssertSext, DL, VA.getLocVT(), Arg,
DAG.getValueType(VA.getValVT()));
break;
case CCValAssign::ZExt:
Arg = DAG.getNode(ISD::AssertZext, DL, VA.getLocVT(), Arg,
DAG.getValueType(VA.getValVT()));
break;
default:
break;
}
// Truncate the register down to the argument type.
if (VA.isExtInLoc())
Arg = DAG.getNode(ISD::TRUNCATE, DL, VA.getValVT(), Arg);
InVals.push_back(Arg);
continue;
}
// The registers are exhausted. This argument was passed on the stack.
assert(VA.isMemLoc());
// The CC_Sparc64_Full/Half functions compute stack offsets relative to the
// beginning of the arguments area at %fp+BIAS+128.
unsigned Offset = VA.getLocMemOffset() + ArgArea;
unsigned ValSize = VA.getValVT().getSizeInBits() / 8;
// Adjust offset for extended arguments, SPARC is big-endian.
// The caller will have written the full slot with extended bytes, but we
// prefer our own extending loads.
if (VA.isExtInLoc())
Offset += 8 - ValSize;
int FI = MF.getFrameInfo().CreateFixedObject(ValSize, Offset, true);
InVals.push_back(
DAG.getLoad(VA.getValVT(), DL, Chain,
DAG.getFrameIndex(FI, getPointerTy(MF.getDataLayout())),
MachinePointerInfo::getFixedStack(MF, FI)));
}
if (!IsVarArg)
return Chain;
// This function takes variable arguments, some of which may have been passed
// in registers %i0-%i5. Variable floating point arguments are never passed
// in floating point registers. They go on %i0-%i5 or on the stack like
// integer arguments.
//
// The va_start intrinsic needs to know the offset to the first variable
// argument.
unsigned ArgOffset = CCInfo.getNextStackOffset();
SparcMachineFunctionInfo *FuncInfo = MF.getInfo<SparcMachineFunctionInfo>();
// Skip the 128 bytes of register save area.
FuncInfo->setVarArgsFrameOffset(ArgOffset + ArgArea +
Subtarget->getStackPointerBias());
// Save the variable arguments that were passed in registers.
// The caller is required to reserve stack space for 6 arguments regardless
// of how many arguments were actually passed.
SmallVector<SDValue, 8> OutChains;
for (; ArgOffset < 6*8; ArgOffset += 8) {
Register VReg = MF.addLiveIn(SP::I0 + ArgOffset/8, &SP::I64RegsRegClass);
SDValue VArg = DAG.getCopyFromReg(Chain, DL, VReg, MVT::i64);
int FI = MF.getFrameInfo().CreateFixedObject(8, ArgOffset + ArgArea, true);
auto PtrVT = getPointerTy(MF.getDataLayout());
OutChains.push_back(
DAG.getStore(Chain, DL, VArg, DAG.getFrameIndex(FI, PtrVT),
MachinePointerInfo::getFixedStack(MF, FI)));
}
if (!OutChains.empty())
Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, OutChains);
return Chain;
}
SDValue
SparcTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
SmallVectorImpl<SDValue> &InVals) const {
if (Subtarget->is64Bit())
return LowerCall_64(CLI, InVals);
return LowerCall_32(CLI, InVals);
}
static bool hasReturnsTwiceAttr(SelectionDAG &DAG, SDValue Callee,
const CallBase *Call) {
if (Call)
return Call->hasFnAttr(Attribute::ReturnsTwice);
const Function *CalleeFn = nullptr;
if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee)) {
CalleeFn = dyn_cast<Function>(G->getGlobal());
} else if (ExternalSymbolSDNode *E =
dyn_cast<ExternalSymbolSDNode>(Callee)) {
const Function &Fn = DAG.getMachineFunction().getFunction();
const Module *M = Fn.getParent();
const char *CalleeName = E->getSymbol();
CalleeFn = M->getFunction(CalleeName);
}
if (!CalleeFn)
return false;
return CalleeFn->hasFnAttribute(Attribute::ReturnsTwice);
}
// Lower a call for the 32-bit ABI.
SDValue
SparcTargetLowering::LowerCall_32(TargetLowering::CallLoweringInfo &CLI,
SmallVectorImpl<SDValue> &InVals) const {
SelectionDAG &DAG = CLI.DAG;
SDLoc &dl = CLI.DL;
SmallVectorImpl<ISD::OutputArg> &Outs = CLI.Outs;
SmallVectorImpl<SDValue> &OutVals = CLI.OutVals;
SmallVectorImpl<ISD::InputArg> &Ins = CLI.Ins;
SDValue Chain = CLI.Chain;
SDValue Callee = CLI.Callee;
bool &isTailCall = CLI.IsTailCall;
CallingConv::ID CallConv = CLI.CallConv;
bool isVarArg = CLI.IsVarArg;
// Sparc target does not yet support tail call optimization.
isTailCall = false;
// Analyze operands of the call, assigning locations to each operand.
SmallVector<CCValAssign, 16> ArgLocs;
CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs,
*DAG.getContext());
CCInfo.AnalyzeCallOperands(Outs, CC_Sparc32);
// Get the size of the outgoing arguments stack space requirement.
unsigned ArgsSize = CCInfo.getNextStackOffset();
// Keep stack frames 8-byte aligned.
ArgsSize = (ArgsSize+7) & ~7;
MachineFrameInfo &MFI = DAG.getMachineFunction().getFrameInfo();
// Create local copies for byval args.
SmallVector<SDValue, 8> ByValArgs;
for (unsigned i = 0, e = Outs.size(); i != e; ++i) {
ISD::ArgFlagsTy Flags = Outs[i].Flags;
if (!Flags.isByVal())
continue;
SDValue Arg = OutVals[i];
unsigned Size = Flags.getByValSize();
Align Alignment = Flags.getNonZeroByValAlign();
if (Size > 0U) {
int FI = MFI.CreateStackObject(Size, Alignment, false);
SDValue FIPtr = DAG.getFrameIndex(FI, getPointerTy(DAG.getDataLayout()));
SDValue SizeNode = DAG.getConstant(Size, dl, MVT::i32);
Chain = DAG.getMemcpy(Chain, dl, FIPtr, Arg, SizeNode, Alignment,
false, // isVolatile,
(Size <= 32), // AlwaysInline if size <= 32,
false, // isTailCall
MachinePointerInfo(), MachinePointerInfo());
ByValArgs.push_back(FIPtr);
}
else {
SDValue nullVal;
ByValArgs.push_back(nullVal);
}
}
Chain = DAG.getCALLSEQ_START(Chain, ArgsSize, 0, dl);
SmallVector<std::pair<unsigned, SDValue>, 8> RegsToPass;
SmallVector<SDValue, 8> MemOpChains;
const unsigned StackOffset = 92;
bool hasStructRetAttr = false;
unsigned SRetArgSize = 0;
// Walk the register/memloc assignments, inserting copies/loads.
for (unsigned i = 0, realArgIdx = 0, byvalArgIdx = 0, e = ArgLocs.size();
i != e;
++i, ++realArgIdx) {
CCValAssign &VA = ArgLocs[i];
SDValue Arg = OutVals[realArgIdx];
ISD::ArgFlagsTy Flags = Outs[realArgIdx].Flags;
// Use local copy if it is a byval arg.
if (Flags.isByVal()) {
Arg = ByValArgs[byvalArgIdx++];
if (!Arg) {
continue;
}
}
// Promote the value if needed.
switch (VA.getLocInfo()) {
default: llvm_unreachable("Unknown loc info!");
case CCValAssign::Full: break;
case CCValAssign::SExt:
Arg = DAG.getNode(ISD::SIGN_EXTEND, dl, VA.getLocVT(), Arg);
break;
case CCValAssign::ZExt:
Arg = DAG.getNode(ISD::ZERO_EXTEND, dl, VA.getLocVT(), Arg);
break;
case CCValAssign::AExt:
Arg = DAG.getNode(ISD::ANY_EXTEND, dl, VA.getLocVT(), Arg);
break;
case CCValAssign::BCvt:
Arg = DAG.getNode(ISD::BITCAST, dl, VA.getLocVT(), Arg);
break;
}
if (Flags.isSRet()) {
assert(VA.needsCustom());
// store SRet argument in %sp+64
SDValue StackPtr = DAG.getRegister(SP::O6, MVT::i32);
SDValue PtrOff = DAG.getIntPtrConstant(64, dl);
PtrOff = DAG.getNode(ISD::ADD, dl, MVT::i32, StackPtr, PtrOff);
MemOpChains.push_back(
DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo()));
hasStructRetAttr = true;
// sret only allowed on first argument
assert(Outs[realArgIdx].OrigArgIndex == 0);
PointerType *Ty = cast<PointerType>(CLI.getArgs()[0].Ty);
Type *ElementTy = Ty->getPointerElementType();
SRetArgSize = DAG.getDataLayout().getTypeAllocSize(ElementTy);
continue;
}
if (VA.needsCustom()) {
assert(VA.getLocVT() == MVT::f64 || VA.getLocVT() == MVT::v2i32);
if (VA.isMemLoc()) {
unsigned Offset = VA.getLocMemOffset() + StackOffset;
// if it is double-word aligned, just store.
if (Offset % 8 == 0) {
SDValue StackPtr = DAG.getRegister(SP::O6, MVT::i32);
SDValue PtrOff = DAG.getIntPtrConstant(Offset, dl);
PtrOff = DAG.getNode(ISD::ADD, dl, MVT::i32, StackPtr, PtrOff);
MemOpChains.push_back(
DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo()));
continue;
}
}
if (VA.getLocVT() == MVT::f64) {
// Move from the float value from float registers into the
// integer registers.
if (ConstantFPSDNode *C = dyn_cast<ConstantFPSDNode>(Arg))
Arg = bitcastConstantFPToInt(C, dl, DAG);
else
Arg = DAG.getNode(ISD::BITCAST, dl, MVT::v2i32, Arg);
}
SDValue Part0 = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i32,
Arg,
DAG.getConstant(0, dl, getVectorIdxTy(DAG.getDataLayout())));
SDValue Part1 = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i32,
Arg,
DAG.getConstant(1, dl, getVectorIdxTy(DAG.getDataLayout())));
if (VA.isRegLoc()) {
RegsToPass.push_back(std::make_pair(VA.getLocReg(), Part0));
assert(i+1 != e);
CCValAssign &NextVA = ArgLocs[++i];
if (NextVA.isRegLoc()) {
RegsToPass.push_back(std::make_pair(NextVA.getLocReg(), Part1));
} else {
// Store the second part in stack.
unsigned Offset = NextVA.getLocMemOffset() + StackOffset;
SDValue StackPtr = DAG.getRegister(SP::O6, MVT::i32);
SDValue PtrOff = DAG.getIntPtrConstant(Offset, dl);
PtrOff = DAG.getNode(ISD::ADD, dl, MVT::i32, StackPtr, PtrOff);
MemOpChains.push_back(
DAG.getStore(Chain, dl, Part1, PtrOff, MachinePointerInfo()));
}
} else {
unsigned Offset = VA.getLocMemOffset() + StackOffset;
// Store the first part.
SDValue StackPtr = DAG.getRegister(SP::O6, MVT::i32);
SDValue PtrOff = DAG.getIntPtrConstant(Offset, dl);
PtrOff = DAG.getNode(ISD::ADD, dl, MVT::i32, StackPtr, PtrOff);
MemOpChains.push_back(
DAG.getStore(Chain, dl, Part0, PtrOff, MachinePointerInfo()));
// Store the second part.
PtrOff = DAG.getIntPtrConstant(Offset + 4, dl);
PtrOff = DAG.getNode(ISD::ADD, dl, MVT::i32, StackPtr, PtrOff);
MemOpChains.push_back(
DAG.getStore(Chain, dl, Part1, PtrOff, MachinePointerInfo()));
}
continue;
}
// Arguments that can be passed on register must be kept at
// RegsToPass vector
if (VA.isRegLoc()) {
if (VA.getLocVT() != MVT::f32) {
RegsToPass.push_back(std::make_pair(VA.getLocReg(), Arg));
continue;
}
Arg = DAG.getNode(ISD::BITCAST, dl, MVT::i32, Arg);
RegsToPass.push_back(std::make_pair(VA.getLocReg(), Arg));
continue;
}
assert(VA.isMemLoc());
// Create a store off the stack pointer for this argument.
SDValue StackPtr = DAG.getRegister(SP::O6, MVT::i32);
SDValue PtrOff = DAG.getIntPtrConstant(VA.getLocMemOffset() + StackOffset,
dl);
PtrOff = DAG.getNode(ISD::ADD, dl, MVT::i32, StackPtr, PtrOff);
MemOpChains.push_back(
DAG.getStore(Chain, dl, Arg, PtrOff, MachinePointerInfo()));
}
// Emit all stores, make sure the occur before any copies into physregs.
if (!MemOpChains.empty())
Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains);
// Build a sequence of copy-to-reg nodes chained together with token
// chain and flag operands which copy the outgoing args into registers.
// The InFlag in necessary since all emitted instructions must be
// stuck together.
SDValue InFlag;
for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) {
Register Reg = toCallerWindow(RegsToPass[i].first);
Chain = DAG.getCopyToReg(Chain, dl, Reg, RegsToPass[i].second, InFlag);
InFlag = Chain.getValue(1);
}
bool hasReturnsTwice = hasReturnsTwiceAttr(DAG, Callee, CLI.CB);
// If the callee is a GlobalAddress node (quite common, every direct call is)
// turn it into a TargetGlobalAddress node so that legalize doesn't hack it.
// Likewise ExternalSymbol -> TargetExternalSymbol.
unsigned TF = isPositionIndependent() ? SparcMCExpr::VK_Sparc_WPLT30
: SparcMCExpr::VK_Sparc_WDISP30;
if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee))
Callee = DAG.getTargetGlobalAddress(G->getGlobal(), dl, MVT::i32, 0, TF);
else if (ExternalSymbolSDNode *E = dyn_cast<ExternalSymbolSDNode>(Callee))
Callee = DAG.getTargetExternalSymbol(E->getSymbol(), MVT::i32, TF);
// Returns a chain & a flag for retval copy to use
SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
SmallVector<SDValue, 8> Ops;
Ops.push_back(Chain);
Ops.push_back(Callee);
if (hasStructRetAttr)
Ops.push_back(DAG.getTargetConstant(SRetArgSize, dl, MVT::i32));
for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i)
Ops.push_back(DAG.getRegister(toCallerWindow(RegsToPass[i].first),
RegsToPass[i].second.getValueType()));
// Add a register mask operand representing the call-preserved registers.
const SparcRegisterInfo *TRI = Subtarget->getRegisterInfo();
const uint32_t *Mask =
((hasReturnsTwice)
? TRI->getRTCallPreservedMask(CallConv)
: TRI->getCallPreservedMask(DAG.getMachineFunction(), CallConv));
assert(Mask && "Missing call preserved mask for calling convention");
Ops.push_back(DAG.getRegisterMask(Mask));
if (InFlag.getNode())
Ops.push_back(InFlag);
Chain = DAG.getNode(SPISD::CALL, dl, NodeTys, Ops);
InFlag = Chain.getValue(1);
Chain = DAG.getCALLSEQ_END(Chain, DAG.getIntPtrConstant(ArgsSize, dl, true),
DAG.getIntPtrConstant(0, dl, true), InFlag, dl);
InFlag = Chain.getValue(1);
// Assign locations to each value returned by this call.
SmallVector<CCValAssign, 16> RVLocs;
CCState RVInfo(CallConv, isVarArg, DAG.getMachineFunction(), RVLocs,
*DAG.getContext());
RVInfo.AnalyzeCallResult(Ins, RetCC_Sparc32);
// Copy all of the result registers out of their specified physreg.
for (unsigned i = 0; i != RVLocs.size(); ++i) {
if (RVLocs[i].getLocVT() == MVT::v2i32) {
SDValue Vec = DAG.getNode(ISD::UNDEF, dl, MVT::v2i32);
SDValue Lo = DAG.getCopyFromReg(
Chain, dl, toCallerWindow(RVLocs[i++].getLocReg()), MVT::i32, InFlag);
Chain = Lo.getValue(1);
InFlag = Lo.getValue(2);
Vec = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, MVT::v2i32, Vec, Lo,
DAG.getConstant(0, dl, MVT::i32));
SDValue Hi = DAG.getCopyFromReg(
Chain, dl, toCallerWindow(RVLocs[i].getLocReg()), MVT::i32, InFlag);
Chain = Hi.getValue(1);
InFlag = Hi.getValue(2);
Vec = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, MVT::v2i32, Vec, Hi,