forked from rust-lang/rust
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmod.rs
1274 lines (1097 loc) · 48.6 KB
/
mod.rs
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
//! Codegen of intrinsics. This includes `extern "rust-intrinsic"`, `extern "platform-intrinsic"`
//! and LLVM intrinsics that have symbol names starting with `llvm.`.
macro_rules! intrinsic_args {
($fx:expr, $args:expr => ($($arg:tt),*); $intrinsic:expr) => {
#[allow(unused_parens)]
let ($($arg),*) = if let [$($arg),*] = $args {
($(codegen_operand($fx, &($arg).node)),*)
} else {
$crate::intrinsics::bug_on_incorrect_arg_count($intrinsic);
};
}
}
mod llvm;
mod llvm_aarch64;
mod llvm_x86;
mod simd;
use cranelift_codegen::ir::AtomicRmwOp;
use rustc_middle::ty;
use rustc_middle::ty::layout::{HasParamEnv, ValidityRequirement};
use rustc_middle::ty::print::{with_no_trimmed_paths, with_no_visible_paths};
use rustc_middle::ty::GenericArgsRef;
use rustc_span::source_map::Spanned;
use rustc_span::symbol::{sym, Symbol};
pub(crate) use self::llvm::codegen_llvm_intrinsic_call;
use crate::prelude::*;
fn bug_on_incorrect_arg_count(intrinsic: impl std::fmt::Display) -> ! {
bug!("wrong number of args for intrinsic {}", intrinsic);
}
fn report_atomic_type_validation_error<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
intrinsic: Symbol,
span: Span,
ty: Ty<'tcx>,
) {
fx.tcx.dcx().span_err(
span,
format!(
"`{}` intrinsic: expected basic integer or raw pointer type, found `{:?}`",
intrinsic, ty
),
);
// Prevent verifier error
fx.bcx.ins().trap(TrapCode::UnreachableCodeReached);
}
pub(crate) fn clif_vector_type<'tcx>(tcx: TyCtxt<'tcx>, layout: TyAndLayout<'tcx>) -> Type {
let (element, count) = match layout.abi {
Abi::Vector { element, count } => (element, count),
_ => unreachable!(),
};
scalar_to_clif_type(tcx, element).by(u32::try_from(count).unwrap()).unwrap()
}
fn simd_for_each_lane<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
val: CValue<'tcx>,
ret: CPlace<'tcx>,
f: &dyn Fn(&mut FunctionCx<'_, '_, 'tcx>, Ty<'tcx>, Ty<'tcx>, Value) -> Value,
) {
let layout = val.layout();
let (lane_count, lane_ty) = layout.ty.simd_size_and_type(fx.tcx);
let lane_layout = fx.layout_of(lane_ty);
let (ret_lane_count, ret_lane_ty) = ret.layout().ty.simd_size_and_type(fx.tcx);
let ret_lane_layout = fx.layout_of(ret_lane_ty);
assert_eq!(lane_count, ret_lane_count);
for lane_idx in 0..lane_count {
let lane = val.value_lane(fx, lane_idx).load_scalar(fx);
let res_lane = f(fx, lane_layout.ty, ret_lane_layout.ty, lane);
let res_lane = CValue::by_val(res_lane, ret_lane_layout);
ret.place_lane(fx, lane_idx).write_cvalue(fx, res_lane);
}
}
fn simd_pair_for_each_lane_typed<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
x: CValue<'tcx>,
y: CValue<'tcx>,
ret: CPlace<'tcx>,
f: &dyn Fn(&mut FunctionCx<'_, '_, 'tcx>, CValue<'tcx>, CValue<'tcx>) -> CValue<'tcx>,
) {
assert_eq!(x.layout(), y.layout());
let layout = x.layout();
let (lane_count, _lane_ty) = layout.ty.simd_size_and_type(fx.tcx);
let (ret_lane_count, _ret_lane_ty) = ret.layout().ty.simd_size_and_type(fx.tcx);
assert_eq!(lane_count, ret_lane_count);
for lane_idx in 0..lane_count {
let x_lane = x.value_lane(fx, lane_idx);
let y_lane = y.value_lane(fx, lane_idx);
let res_lane = f(fx, x_lane, y_lane);
ret.place_lane(fx, lane_idx).write_cvalue(fx, res_lane);
}
}
fn simd_pair_for_each_lane<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
x: CValue<'tcx>,
y: CValue<'tcx>,
ret: CPlace<'tcx>,
f: &dyn Fn(&mut FunctionCx<'_, '_, 'tcx>, Ty<'tcx>, Ty<'tcx>, Value, Value) -> Value,
) {
assert_eq!(x.layout(), y.layout());
let layout = x.layout();
let (lane_count, lane_ty) = layout.ty.simd_size_and_type(fx.tcx);
let lane_layout = fx.layout_of(lane_ty);
let (ret_lane_count, ret_lane_ty) = ret.layout().ty.simd_size_and_type(fx.tcx);
let ret_lane_layout = fx.layout_of(ret_lane_ty);
assert_eq!(lane_count, ret_lane_count);
for lane_idx in 0..lane_count {
let x_lane = x.value_lane(fx, lane_idx).load_scalar(fx);
let y_lane = y.value_lane(fx, lane_idx).load_scalar(fx);
let res_lane = f(fx, lane_layout.ty, ret_lane_layout.ty, x_lane, y_lane);
let res_lane = CValue::by_val(res_lane, ret_lane_layout);
ret.place_lane(fx, lane_idx).write_cvalue(fx, res_lane);
}
}
fn simd_horizontal_pair_for_each_lane<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
x: CValue<'tcx>,
y: CValue<'tcx>,
ret: CPlace<'tcx>,
f: &dyn Fn(&mut FunctionCx<'_, '_, 'tcx>, Ty<'tcx>, Ty<'tcx>, Value, Value) -> Value,
) {
assert_eq!(x.layout(), y.layout());
let layout = x.layout();
let (lane_count, lane_ty) = layout.ty.simd_size_and_type(fx.tcx);
let lane_layout = fx.layout_of(lane_ty);
let (ret_lane_count, ret_lane_ty) = ret.layout().ty.simd_size_and_type(fx.tcx);
let ret_lane_layout = fx.layout_of(ret_lane_ty);
assert_eq!(lane_count, ret_lane_count);
for lane_idx in 0..lane_count {
let src = if lane_idx < (lane_count / 2) { x } else { y };
let src_idx = lane_idx % (lane_count / 2);
let lhs_lane = src.value_lane(fx, src_idx * 2).load_scalar(fx);
let rhs_lane = src.value_lane(fx, src_idx * 2 + 1).load_scalar(fx);
let res_lane = f(fx, lane_layout.ty, ret_lane_layout.ty, lhs_lane, rhs_lane);
let res_lane = CValue::by_val(res_lane, ret_lane_layout);
ret.place_lane(fx, lane_idx).write_cvalue(fx, res_lane);
}
}
fn simd_trio_for_each_lane<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
x: CValue<'tcx>,
y: CValue<'tcx>,
z: CValue<'tcx>,
ret: CPlace<'tcx>,
f: &dyn Fn(&mut FunctionCx<'_, '_, 'tcx>, Ty<'tcx>, Ty<'tcx>, Value, Value, Value) -> Value,
) {
assert_eq!(x.layout(), y.layout());
let layout = x.layout();
let (lane_count, lane_ty) = layout.ty.simd_size_and_type(fx.tcx);
let lane_layout = fx.layout_of(lane_ty);
let (ret_lane_count, ret_lane_ty) = ret.layout().ty.simd_size_and_type(fx.tcx);
let ret_lane_layout = fx.layout_of(ret_lane_ty);
assert_eq!(lane_count, ret_lane_count);
for lane_idx in 0..lane_count {
let x_lane = x.value_lane(fx, lane_idx).load_scalar(fx);
let y_lane = y.value_lane(fx, lane_idx).load_scalar(fx);
let z_lane = z.value_lane(fx, lane_idx).load_scalar(fx);
let res_lane = f(fx, lane_layout.ty, ret_lane_layout.ty, x_lane, y_lane, z_lane);
let res_lane = CValue::by_val(res_lane, ret_lane_layout);
ret.place_lane(fx, lane_idx).write_cvalue(fx, res_lane);
}
}
fn simd_reduce<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
val: CValue<'tcx>,
acc: Option<Value>,
ret: CPlace<'tcx>,
f: &dyn Fn(&mut FunctionCx<'_, '_, 'tcx>, Ty<'tcx>, Value, Value) -> Value,
) {
let (lane_count, lane_ty) = val.layout().ty.simd_size_and_type(fx.tcx);
let lane_layout = fx.layout_of(lane_ty);
assert_eq!(lane_layout, ret.layout());
let (mut res_val, start_lane) =
if let Some(acc) = acc { (acc, 0) } else { (val.value_lane(fx, 0).load_scalar(fx), 1) };
for lane_idx in start_lane..lane_count {
let lane = val.value_lane(fx, lane_idx).load_scalar(fx);
res_val = f(fx, lane_layout.ty, res_val, lane);
}
let res = CValue::by_val(res_val, lane_layout);
ret.write_cvalue(fx, res);
}
// FIXME move all uses to `simd_reduce`
fn simd_reduce_bool<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
val: CValue<'tcx>,
ret: CPlace<'tcx>,
f: &dyn Fn(&mut FunctionCx<'_, '_, 'tcx>, Value, Value) -> Value,
) {
let (lane_count, _lane_ty) = val.layout().ty.simd_size_and_type(fx.tcx);
assert!(ret.layout().ty.is_bool());
let res_val = val.value_lane(fx, 0).load_scalar(fx);
let mut res_val = fx.bcx.ins().band_imm(res_val, 1); // mask to boolean
for lane_idx in 1..lane_count {
let lane = val.value_lane(fx, lane_idx).load_scalar(fx);
let lane = fx.bcx.ins().band_imm(lane, 1); // mask to boolean
res_val = f(fx, res_val, lane);
}
let res_val = if fx.bcx.func.dfg.value_type(res_val) != types::I8 {
fx.bcx.ins().ireduce(types::I8, res_val)
} else {
res_val
};
let res = CValue::by_val(res_val, ret.layout());
ret.write_cvalue(fx, res);
}
fn bool_to_zero_or_max_uint<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
ty: Ty<'tcx>,
val: Value,
) -> Value {
let ty = fx.clif_type(ty).unwrap();
let int_ty = match ty {
types::F32 => types::I32,
types::F64 => types::I64,
ty => ty,
};
let mut res = fx.bcx.ins().bmask(int_ty, val);
if ty.is_float() {
res = codegen_bitcast(fx, ty, res);
}
res
}
pub(crate) fn codegen_intrinsic_call<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
instance: Instance<'tcx>,
args: &[Spanned<mir::Operand<'tcx>>],
destination: CPlace<'tcx>,
target: Option<BasicBlock>,
source_info: mir::SourceInfo,
) -> Result<(), Instance<'tcx>> {
let intrinsic = fx.tcx.item_name(instance.def_id());
let instance_args = instance.args;
if intrinsic.as_str().starts_with("simd_") {
self::simd::codegen_simd_intrinsic_call(
fx,
intrinsic,
instance_args,
args,
destination,
target.expect("target for simd intrinsic"),
source_info.span,
);
} else if codegen_float_intrinsic_call(fx, intrinsic, args, destination) {
let ret_block = fx.get_block(target.expect("target for float intrinsic"));
fx.bcx.ins().jump(ret_block, &[]);
} else {
codegen_regular_intrinsic_call(
fx,
instance,
intrinsic,
instance_args,
args,
destination,
target,
source_info,
)?;
}
Ok(())
}
fn codegen_float_intrinsic_call<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
intrinsic: Symbol,
args: &[Spanned<mir::Operand<'tcx>>],
ret: CPlace<'tcx>,
) -> bool {
let (name, arg_count, ty, clif_ty) = match intrinsic {
sym::expf32 => ("expf", 1, fx.tcx.types.f32, types::F32),
sym::expf64 => ("exp", 1, fx.tcx.types.f64, types::F64),
sym::exp2f32 => ("exp2f", 1, fx.tcx.types.f32, types::F32),
sym::exp2f64 => ("exp2", 1, fx.tcx.types.f64, types::F64),
sym::sqrtf32 => ("sqrtf", 1, fx.tcx.types.f32, types::F32),
sym::sqrtf64 => ("sqrt", 1, fx.tcx.types.f64, types::F64),
sym::powif32 => ("__powisf2", 2, fx.tcx.types.f32, types::F32), // compiler-builtins
sym::powif64 => ("__powidf2", 2, fx.tcx.types.f64, types::F64), // compiler-builtins
sym::powf32 => ("powf", 2, fx.tcx.types.f32, types::F32),
sym::powf64 => ("pow", 2, fx.tcx.types.f64, types::F64),
sym::logf32 => ("logf", 1, fx.tcx.types.f32, types::F32),
sym::logf64 => ("log", 1, fx.tcx.types.f64, types::F64),
sym::log2f32 => ("log2f", 1, fx.tcx.types.f32, types::F32),
sym::log2f64 => ("log2", 1, fx.tcx.types.f64, types::F64),
sym::log10f32 => ("log10f", 1, fx.tcx.types.f32, types::F32),
sym::log10f64 => ("log10", 1, fx.tcx.types.f64, types::F64),
sym::fabsf32 => ("fabsf", 1, fx.tcx.types.f32, types::F32),
sym::fabsf64 => ("fabs", 1, fx.tcx.types.f64, types::F64),
sym::fmaf32 => ("fmaf", 3, fx.tcx.types.f32, types::F32),
sym::fmaf64 => ("fma", 3, fx.tcx.types.f64, types::F64),
sym::copysignf32 => ("copysignf", 2, fx.tcx.types.f32, types::F32),
sym::copysignf64 => ("copysign", 2, fx.tcx.types.f64, types::F64),
sym::floorf32 => ("floorf", 1, fx.tcx.types.f32, types::F32),
sym::floorf64 => ("floor", 1, fx.tcx.types.f64, types::F64),
sym::ceilf32 => ("ceilf", 1, fx.tcx.types.f32, types::F32),
sym::ceilf64 => ("ceil", 1, fx.tcx.types.f64, types::F64),
sym::truncf32 => ("truncf", 1, fx.tcx.types.f32, types::F32),
sym::truncf64 => ("trunc", 1, fx.tcx.types.f64, types::F64),
sym::rintf32 => ("rintf", 1, fx.tcx.types.f32, types::F32),
sym::rintf64 => ("rint", 1, fx.tcx.types.f64, types::F64),
sym::roundf32 => ("roundf", 1, fx.tcx.types.f32, types::F32),
sym::roundf64 => ("round", 1, fx.tcx.types.f64, types::F64),
sym::roundevenf32 => ("roundevenf", 1, fx.tcx.types.f32, types::F32),
sym::roundevenf64 => ("roundeven", 1, fx.tcx.types.f64, types::F64),
sym::sinf32 => ("sinf", 1, fx.tcx.types.f32, types::F32),
sym::sinf64 => ("sin", 1, fx.tcx.types.f64, types::F64),
sym::cosf32 => ("cosf", 1, fx.tcx.types.f32, types::F32),
sym::cosf64 => ("cos", 1, fx.tcx.types.f64, types::F64),
_ => return false,
};
if args.len() != arg_count {
bug!("wrong number of args for intrinsic {:?}", intrinsic);
}
let (a, b, c);
let args = match args {
[x] => {
a = [codegen_operand(fx, &x.node).load_scalar(fx)];
&a as &[_]
}
[x, y] => {
b = [
codegen_operand(fx, &x.node).load_scalar(fx),
codegen_operand(fx, &y.node).load_scalar(fx),
];
&b
}
[x, y, z] => {
c = [
codegen_operand(fx, &x.node).load_scalar(fx),
codegen_operand(fx, &y.node).load_scalar(fx),
codegen_operand(fx, &z.node).load_scalar(fx),
];
&c
}
_ => unreachable!(),
};
let layout = fx.layout_of(ty);
let res = match intrinsic {
sym::fmaf32 | sym::fmaf64 => {
CValue::by_val(fx.bcx.ins().fma(args[0], args[1], args[2]), layout)
}
sym::copysignf32 | sym::copysignf64 => {
CValue::by_val(fx.bcx.ins().fcopysign(args[0], args[1]), layout)
}
sym::fabsf32
| sym::fabsf64
| sym::floorf32
| sym::floorf64
| sym::ceilf32
| sym::ceilf64
| sym::truncf32
| sym::truncf64 => {
let val = match intrinsic {
sym::fabsf32 | sym::fabsf64 => fx.bcx.ins().fabs(args[0]),
sym::floorf32 | sym::floorf64 => fx.bcx.ins().floor(args[0]),
sym::ceilf32 | sym::ceilf64 => fx.bcx.ins().ceil(args[0]),
sym::truncf32 | sym::truncf64 => fx.bcx.ins().trunc(args[0]),
_ => unreachable!(),
};
CValue::by_val(val, layout)
}
// These intrinsics aren't supported natively by Cranelift.
// Lower them to a libcall.
sym::powif32 | sym::powif64 => {
let input_tys: Vec<_> = vec![AbiParam::new(clif_ty), AbiParam::new(types::I32)];
let ret_val = fx.lib_call(name, input_tys, vec![AbiParam::new(clif_ty)], &args)[0];
CValue::by_val(ret_val, fx.layout_of(ty))
}
_ => {
let input_tys: Vec<_> = args.iter().map(|_| AbiParam::new(clif_ty)).collect();
let ret_val = fx.lib_call(name, input_tys, vec![AbiParam::new(clif_ty)], &args)[0];
CValue::by_val(ret_val, fx.layout_of(ty))
}
};
ret.write_cvalue(fx, res);
true
}
fn codegen_regular_intrinsic_call<'tcx>(
fx: &mut FunctionCx<'_, '_, 'tcx>,
instance: Instance<'tcx>,
intrinsic: Symbol,
generic_args: GenericArgsRef<'tcx>,
args: &[Spanned<mir::Operand<'tcx>>],
ret: CPlace<'tcx>,
destination: Option<BasicBlock>,
source_info: mir::SourceInfo,
) -> Result<(), Instance<'tcx>> {
assert_eq!(generic_args, instance.args);
let usize_layout = fx.layout_of(fx.tcx.types.usize);
match intrinsic {
sym::abort => {
fx.bcx.ins().trap(TrapCode::User(0));
return Ok(());
}
sym::likely | sym::unlikely => {
intrinsic_args!(fx, args => (a); intrinsic);
ret.write_cvalue(fx, a);
}
sym::breakpoint => {
intrinsic_args!(fx, args => (); intrinsic);
fx.bcx.ins().debugtrap();
}
sym::copy => {
intrinsic_args!(fx, args => (src, dst, count); intrinsic);
let src = src.load_scalar(fx);
let dst = dst.load_scalar(fx);
let count = count.load_scalar(fx);
let elem_ty = generic_args.type_at(0);
let elem_size: u64 = fx.layout_of(elem_ty).size.bytes();
assert_eq!(args.len(), 3);
let byte_amount =
if elem_size != 1 { fx.bcx.ins().imul_imm(count, elem_size as i64) } else { count };
// FIXME emit_small_memmove
fx.bcx.call_memmove(fx.target_config, dst, src, byte_amount);
}
sym::volatile_copy_memory | sym::volatile_copy_nonoverlapping_memory => {
// NOTE: the volatile variants have src and dst swapped
intrinsic_args!(fx, args => (dst, src, count); intrinsic);
let dst = dst.load_scalar(fx);
let src = src.load_scalar(fx);
let count = count.load_scalar(fx);
let elem_ty = generic_args.type_at(0);
let elem_size: u64 = fx.layout_of(elem_ty).size.bytes();
assert_eq!(args.len(), 3);
let byte_amount =
if elem_size != 1 { fx.bcx.ins().imul_imm(count, elem_size as i64) } else { count };
// FIXME make the copy actually volatile when using emit_small_mem{cpy,move}
if intrinsic == sym::volatile_copy_nonoverlapping_memory {
// FIXME emit_small_memcpy
fx.bcx.call_memcpy(fx.target_config, dst, src, byte_amount);
} else {
// FIXME emit_small_memmove
fx.bcx.call_memmove(fx.target_config, dst, src, byte_amount);
}
}
sym::size_of_val => {
intrinsic_args!(fx, args => (ptr); intrinsic);
let layout = fx.layout_of(generic_args.type_at(0));
// Note: Can't use is_unsized here as truly unsized types need to take the fixed size
// branch
let meta = if let Abi::ScalarPair(_, _) = ptr.layout().abi {
Some(ptr.load_scalar_pair(fx).1)
} else {
None
};
let (size, _align) = crate::unsize::size_and_align_of(fx, layout, meta);
ret.write_cvalue(fx, CValue::by_val(size, usize_layout));
}
sym::min_align_of_val => {
intrinsic_args!(fx, args => (ptr); intrinsic);
let layout = fx.layout_of(generic_args.type_at(0));
// Note: Can't use is_unsized here as truly unsized types need to take the fixed size
// branch
let meta = if let Abi::ScalarPair(_, _) = ptr.layout().abi {
Some(ptr.load_scalar_pair(fx).1)
} else {
None
};
let (_size, align) = crate::unsize::size_and_align_of(fx, layout, meta);
ret.write_cvalue(fx, CValue::by_val(align, usize_layout));
}
sym::vtable_size => {
intrinsic_args!(fx, args => (vtable); intrinsic);
let vtable = vtable.load_scalar(fx);
let size = crate::vtable::size_of_obj(fx, vtable);
ret.write_cvalue(fx, CValue::by_val(size, usize_layout));
}
sym::vtable_align => {
intrinsic_args!(fx, args => (vtable); intrinsic);
let vtable = vtable.load_scalar(fx);
let align = crate::vtable::min_align_of_obj(fx, vtable);
ret.write_cvalue(fx, CValue::by_val(align, usize_layout));
}
sym::exact_div => {
intrinsic_args!(fx, args => (x, y); intrinsic);
// FIXME trap on inexact
let res = crate::num::codegen_int_binop(fx, BinOp::Div, x, y);
ret.write_cvalue(fx, res);
}
sym::saturating_add | sym::saturating_sub => {
intrinsic_args!(fx, args => (lhs, rhs); intrinsic);
assert_eq!(lhs.layout().ty, rhs.layout().ty);
let bin_op = match intrinsic {
sym::saturating_add => BinOp::Add,
sym::saturating_sub => BinOp::Sub,
_ => unreachable!(),
};
let res = crate::num::codegen_saturating_int_binop(fx, bin_op, lhs, rhs);
ret.write_cvalue(fx, res);
}
sym::rotate_left => {
intrinsic_args!(fx, args => (x, y); intrinsic);
let y = y.load_scalar(fx);
let layout = x.layout();
let x = x.load_scalar(fx);
let res = fx.bcx.ins().rotl(x, y);
ret.write_cvalue(fx, CValue::by_val(res, layout));
}
sym::rotate_right => {
intrinsic_args!(fx, args => (x, y); intrinsic);
let y = y.load_scalar(fx);
let layout = x.layout();
let x = x.load_scalar(fx);
let res = fx.bcx.ins().rotr(x, y);
ret.write_cvalue(fx, CValue::by_val(res, layout));
}
// The only difference between offset and arith_offset is regarding UB. Because Cranelift
// doesn't have UB both are codegen'ed the same way
sym::arith_offset => {
intrinsic_args!(fx, args => (base, offset); intrinsic);
let offset = offset.load_scalar(fx);
let pointee_ty = base.layout().ty.builtin_deref(true).unwrap().ty;
let pointee_size = fx.layout_of(pointee_ty).size.bytes();
let ptr_diff = if pointee_size != 1 {
fx.bcx.ins().imul_imm(offset, pointee_size as i64)
} else {
offset
};
let base_val = base.load_scalar(fx);
let res = fx.bcx.ins().iadd(base_val, ptr_diff);
ret.write_cvalue(fx, CValue::by_val(res, base.layout()));
}
sym::ptr_mask => {
intrinsic_args!(fx, args => (ptr, mask); intrinsic);
let ptr = ptr.load_scalar(fx);
let mask = mask.load_scalar(fx);
fx.bcx.ins().band(ptr, mask);
}
sym::write_bytes | sym::volatile_set_memory => {
intrinsic_args!(fx, args => (dst, val, count); intrinsic);
let val = val.load_scalar(fx);
let count = count.load_scalar(fx);
let pointee_ty = dst.layout().ty.builtin_deref(true).unwrap().ty;
let pointee_size = fx.layout_of(pointee_ty).size.bytes();
let count = if pointee_size != 1 {
fx.bcx.ins().imul_imm(count, pointee_size as i64)
} else {
count
};
let dst_ptr = dst.load_scalar(fx);
// FIXME make the memset actually volatile when switching to emit_small_memset
// FIXME use emit_small_memset
fx.bcx.call_memset(fx.target_config, dst_ptr, val, count);
}
sym::ctlz | sym::ctlz_nonzero => {
intrinsic_args!(fx, args => (arg); intrinsic);
let val = arg.load_scalar(fx);
// FIXME trap on `ctlz_nonzero` with zero arg.
let res = fx.bcx.ins().clz(val);
let res = CValue::by_val(res, arg.layout());
ret.write_cvalue(fx, res);
}
sym::cttz | sym::cttz_nonzero => {
intrinsic_args!(fx, args => (arg); intrinsic);
let val = arg.load_scalar(fx);
// FIXME trap on `cttz_nonzero` with zero arg.
let res = fx.bcx.ins().ctz(val);
let res = CValue::by_val(res, arg.layout());
ret.write_cvalue(fx, res);
}
sym::ctpop => {
intrinsic_args!(fx, args => (arg); intrinsic);
let val = arg.load_scalar(fx);
let res = fx.bcx.ins().popcnt(val);
let res = CValue::by_val(res, arg.layout());
ret.write_cvalue(fx, res);
}
sym::bitreverse => {
intrinsic_args!(fx, args => (arg); intrinsic);
let val = arg.load_scalar(fx);
let res = fx.bcx.ins().bitrev(val);
let res = CValue::by_val(res, arg.layout());
ret.write_cvalue(fx, res);
}
sym::bswap => {
intrinsic_args!(fx, args => (arg); intrinsic);
let val = arg.load_scalar(fx);
let res = if fx.bcx.func.dfg.value_type(val) == types::I8 {
val
} else {
fx.bcx.ins().bswap(val)
};
let res = CValue::by_val(res, arg.layout());
ret.write_cvalue(fx, res);
}
sym::assert_inhabited | sym::assert_zero_valid | sym::assert_mem_uninitialized_valid => {
intrinsic_args!(fx, args => (); intrinsic);
let ty = generic_args.type_at(0);
let requirement = ValidityRequirement::from_intrinsic(intrinsic);
if let Some(requirement) = requirement {
let do_panic = !fx
.tcx
.check_validity_requirement((requirement, fx.param_env().and(ty)))
.expect("expect to have layout during codegen");
if do_panic {
let layout = fx.layout_of(ty);
let msg_str = with_no_visible_paths!({
with_no_trimmed_paths!({
if layout.abi.is_uninhabited() {
// Use this error even for the other intrinsics as it is more precise.
format!("attempted to instantiate uninhabited type `{}`", ty)
} else if intrinsic == sym::assert_zero_valid {
format!(
"attempted to zero-initialize type `{}`, which is invalid",
ty
)
} else {
format!(
"attempted to leave type `{}` uninitialized, which is invalid",
ty
)
}
})
});
crate::base::codegen_panic_nounwind(fx, &msg_str, Some(source_info.span));
return Ok(());
}
}
}
sym::volatile_load | sym::unaligned_volatile_load => {
intrinsic_args!(fx, args => (ptr); intrinsic);
// Cranelift treats loads as volatile by default
// FIXME correctly handle unaligned_volatile_load
let inner_layout = fx.layout_of(ptr.layout().ty.builtin_deref(true).unwrap().ty);
let val = CValue::by_ref(Pointer::new(ptr.load_scalar(fx)), inner_layout);
ret.write_cvalue(fx, val);
}
sym::volatile_store | sym::unaligned_volatile_store | sym::nontemporal_store => {
intrinsic_args!(fx, args => (ptr, val); intrinsic);
let ptr = ptr.load_scalar(fx);
// Cranelift treats stores as volatile by default
// FIXME correctly handle unaligned_volatile_store
// FIXME actually do nontemporal stores if requested
let dest = CPlace::for_ptr(Pointer::new(ptr), val.layout());
dest.write_cvalue(fx, val);
}
sym::pref_align_of
| sym::needs_drop
| sym::type_id
| sym::type_name
| sym::variant_count => {
intrinsic_args!(fx, args => (); intrinsic);
let const_val =
fx.tcx.const_eval_instance(ParamEnv::reveal_all(), instance, None).unwrap();
let val = crate::constant::codegen_const_value(fx, const_val, ret.layout().ty);
ret.write_cvalue(fx, val);
}
sym::ptr_offset_from | sym::ptr_offset_from_unsigned => {
intrinsic_args!(fx, args => (ptr, base); intrinsic);
let ptr = ptr.load_scalar(fx);
let base = base.load_scalar(fx);
let ty = generic_args.type_at(0);
let pointee_size: u64 = fx.layout_of(ty).size.bytes();
let diff_bytes = fx.bcx.ins().isub(ptr, base);
// FIXME this can be an exact division.
let val = if intrinsic == sym::ptr_offset_from_unsigned {
let usize_layout = fx.layout_of(fx.tcx.types.usize);
// Because diff_bytes ULE isize::MAX, this would be fine as signed,
// but unsigned is slightly easier to codegen, so might as well.
CValue::by_val(fx.bcx.ins().udiv_imm(diff_bytes, pointee_size as i64), usize_layout)
} else {
let isize_layout = fx.layout_of(fx.tcx.types.isize);
CValue::by_val(fx.bcx.ins().sdiv_imm(diff_bytes, pointee_size as i64), isize_layout)
};
ret.write_cvalue(fx, val);
}
sym::ptr_guaranteed_cmp => {
intrinsic_args!(fx, args => (a, b); intrinsic);
let val = crate::num::codegen_ptr_binop(fx, BinOp::Eq, a, b).load_scalar(fx);
ret.write_cvalue(fx, CValue::by_val(val, fx.layout_of(fx.tcx.types.u8)));
}
sym::caller_location => {
intrinsic_args!(fx, args => (); intrinsic);
let caller_location = fx.get_caller_location(source_info);
ret.write_cvalue(fx, caller_location);
}
_ if intrinsic.as_str().starts_with("atomic_fence") => {
intrinsic_args!(fx, args => (); intrinsic);
fx.bcx.ins().fence();
}
_ if intrinsic.as_str().starts_with("atomic_singlethreadfence") => {
intrinsic_args!(fx, args => (); intrinsic);
// FIXME use a compiler fence once Cranelift supports it
fx.bcx.ins().fence();
}
_ if intrinsic.as_str().starts_with("atomic_load") => {
intrinsic_args!(fx, args => (ptr); intrinsic);
let ptr = ptr.load_scalar(fx);
let ty = generic_args.type_at(0);
match ty.kind() {
ty::Uint(UintTy::U128) | ty::Int(IntTy::I128) => {
// FIXME implement 128bit atomics
if fx.tcx.is_compiler_builtins(LOCAL_CRATE) {
// special case for compiler-builtins to avoid having to patch it
crate::trap::trap_unimplemented(fx, "128bit atomics not yet supported");
return Ok(());
} else {
fx.tcx
.dcx()
.span_fatal(source_info.span, "128bit atomics not yet supported");
}
}
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, ty);
return Ok(());
}
}
let clif_ty = fx.clif_type(ty).unwrap();
let val = fx.bcx.ins().atomic_load(clif_ty, MemFlags::trusted(), ptr);
let val = CValue::by_val(val, fx.layout_of(ty));
ret.write_cvalue(fx, val);
}
_ if intrinsic.as_str().starts_with("atomic_store") => {
intrinsic_args!(fx, args => (ptr, val); intrinsic);
let ptr = ptr.load_scalar(fx);
let ty = generic_args.type_at(0);
match ty.kind() {
ty::Uint(UintTy::U128) | ty::Int(IntTy::I128) => {
// FIXME implement 128bit atomics
if fx.tcx.is_compiler_builtins(LOCAL_CRATE) {
// special case for compiler-builtins to avoid having to patch it
crate::trap::trap_unimplemented(fx, "128bit atomics not yet supported");
return Ok(());
} else {
fx.tcx
.dcx()
.span_fatal(source_info.span, "128bit atomics not yet supported");
}
}
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, ty);
return Ok(());
}
}
let val = val.load_scalar(fx);
fx.bcx.ins().atomic_store(MemFlags::trusted(), val, ptr);
}
_ if intrinsic.as_str().starts_with("atomic_xchg") => {
intrinsic_args!(fx, args => (ptr, new); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = new.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);
return Ok(());
}
}
let ty = fx.clif_type(layout.ty).unwrap();
let new = new.load_scalar(fx);
let old = fx.bcx.ins().atomic_rmw(ty, MemFlags::trusted(), AtomicRmwOp::Xchg, ptr, new);
let old = CValue::by_val(old, layout);
ret.write_cvalue(fx, old);
}
_ if intrinsic.as_str().starts_with("atomic_cxchg") => {
// both atomic_cxchg_* and atomic_cxchgweak_*
intrinsic_args!(fx, args => (ptr, test_old, new); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = new.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);
return Ok(());
}
}
let test_old = test_old.load_scalar(fx);
let new = new.load_scalar(fx);
let old = fx.bcx.ins().atomic_cas(MemFlags::trusted(), ptr, test_old, new);
let is_eq = fx.bcx.ins().icmp(IntCC::Equal, old, test_old);
let ret_val = CValue::by_val_pair(old, is_eq, ret.layout());
ret.write_cvalue(fx, ret_val)
}
_ if intrinsic.as_str().starts_with("atomic_xadd") => {
intrinsic_args!(fx, args => (ptr, amount); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = amount.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);
return Ok(());
}
}
let ty = fx.clif_type(layout.ty).unwrap();
let amount = amount.load_scalar(fx);
let old =
fx.bcx.ins().atomic_rmw(ty, MemFlags::trusted(), AtomicRmwOp::Add, ptr, amount);
let old = CValue::by_val(old, layout);
ret.write_cvalue(fx, old);
}
_ if intrinsic.as_str().starts_with("atomic_xsub") => {
intrinsic_args!(fx, args => (ptr, amount); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = amount.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);
return Ok(());
}
}
let ty = fx.clif_type(layout.ty).unwrap();
let amount = amount.load_scalar(fx);
let old =
fx.bcx.ins().atomic_rmw(ty, MemFlags::trusted(), AtomicRmwOp::Sub, ptr, amount);
let old = CValue::by_val(old, layout);
ret.write_cvalue(fx, old);
}
_ if intrinsic.as_str().starts_with("atomic_and") => {
intrinsic_args!(fx, args => (ptr, src); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = src.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);
return Ok(());
}
}
let ty = fx.clif_type(layout.ty).unwrap();
let src = src.load_scalar(fx);
let old = fx.bcx.ins().atomic_rmw(ty, MemFlags::trusted(), AtomicRmwOp::And, ptr, src);
let old = CValue::by_val(old, layout);
ret.write_cvalue(fx, old);
}
_ if intrinsic.as_str().starts_with("atomic_or") => {
intrinsic_args!(fx, args => (ptr, src); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = src.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);
return Ok(());
}
}
let ty = fx.clif_type(layout.ty).unwrap();
let src = src.load_scalar(fx);
let old = fx.bcx.ins().atomic_rmw(ty, MemFlags::trusted(), AtomicRmwOp::Or, ptr, src);
let old = CValue::by_val(old, layout);
ret.write_cvalue(fx, old);
}
_ if intrinsic.as_str().starts_with("atomic_xor") => {
intrinsic_args!(fx, args => (ptr, src); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = src.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);
return Ok(());
}
}
let ty = fx.clif_type(layout.ty).unwrap();
let src = src.load_scalar(fx);
let old = fx.bcx.ins().atomic_rmw(ty, MemFlags::trusted(), AtomicRmwOp::Xor, ptr, src);
let old = CValue::by_val(old, layout);
ret.write_cvalue(fx, old);
}
_ if intrinsic.as_str().starts_with("atomic_nand") => {
intrinsic_args!(fx, args => (ptr, src); intrinsic);
let ptr = ptr.load_scalar(fx);
let layout = src.layout();
match layout.ty.kind() {
ty::Uint(_) | ty::Int(_) | ty::RawPtr(..) => {}
_ => {
report_atomic_type_validation_error(fx, intrinsic, source_info.span, layout.ty);