rec_arm64.cpp 31 KB
Newer Older
Flyinghead's avatar
Flyinghead committed
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
/*
	Copyright 2019 flyinghead

	This file is part of reicast.

    reicast is free software: you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation, either version 2 of the License, or
    (at your option) any later version.

    reicast is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with reicast.  If not, see <https://www.gnu.org/licenses/>.
 */

#include "types.h"

#if FEAT_SHREC == DYNAREC_JIT

#include <unistd.h>
#include <sys/mman.h>
#include <map>

#include "deps/vixl/aarch64/macro-assembler-aarch64.h"
using namespace vixl::aarch64;

#include "hw/sh4/sh4_opcode_list.h"

#include "hw/sh4/sh4_mmr.h"
#include "hw/sh4/sh4_interrupts.h"
#include "hw/sh4/sh4_core.h"
#include "hw/sh4/dyna/ngen.h"
#include "hw/sh4/sh4_mem.h"
#include "hw/sh4/sh4_rom.h"
#include "arm64_regalloc.h"

#undef do_sqw_nommu

struct DynaRBI : RuntimeBlockInfo
{
	virtual u32 Relink() {
		return 0;
	}

	virtual void Relocate(void* dst) {
		verify(false);
	}
};

// Code borrowed from Dolphin https://github.com/dolphin-emu/dolphin
static void CacheFlush(void* start, void* end)
{
	if (start == end)
		return;

#if HOST_OS == OS_DARWIN
	// Header file says this is equivalent to: sys_icache_invalidate(start, end - start);
	sys_cache_control(kCacheFunctionPrepareForExecution, start, end - start);
#else
	// Don't rely on GCC's __clear_cache implementation, as it caches
	// icache/dcache cache line sizes, that can vary between cores on
	// big.LITTLE architectures.
	u64 addr, ctr_el0;
	static size_t icache_line_size = 0xffff, dcache_line_size = 0xffff;
	size_t isize, dsize;

	__asm__ volatile("mrs %0, ctr_el0" : "=r"(ctr_el0));
	isize = 4 << ((ctr_el0 >> 0) & 0xf);
	dsize = 4 << ((ctr_el0 >> 16) & 0xf);

	// use the global minimum cache line size
	icache_line_size = isize = icache_line_size < isize ? icache_line_size : isize;
	dcache_line_size = dsize = dcache_line_size < dsize ? dcache_line_size : dsize;

	addr = (u64)start & ~(u64)(dsize - 1);
	for (; addr < (u64)end; addr += dsize)
		// use "civac" instead of "cvau", as this is the suggested workaround for
		// Cortex-A53 errata 819472, 826319, 827319 and 824069.
		__asm__ volatile("dc civac, %0" : : "r"(addr) : "memory");
	__asm__ volatile("dsb ish" : : : "memory");

	addr = (u64)start & ~(u64)(isize - 1);
	for (; addr < (u64)end; addr += isize)
		__asm__ volatile("ic ivau, %0" : : "r"(addr) : "memory");

	__asm__ volatile("dsb ish" : : : "memory");
	__asm__ volatile("isb" : : : "memory");
#endif
}

static void ngen_FailedToFindBlock_internal() {
	rdv_FailedToFindBlock(Sh4cntx.pc);
}

void(*ngen_FailedToFindBlock)() = &ngen_FailedToFindBlock_internal;

void ngen_mainloop(void* v_cntx)
{
	Sh4RCB* ctx = (Sh4RCB*)((u8*)v_cntx - sizeof(Sh4RCB));

	__asm__ volatile
	(
		"stp x19, x20, [sp, #-144]!	\n\t"
		"stp x21, x22, [sp, #16]	\n\t"
		"stp x23, x24, [sp, #32]	\n\t"
		"stp x25, x26, [sp, #48]	\n\t"
		"stp x27, x28, [sp, #64]	\n\t"
		"stp s8, s9, [sp, #80]		\n\t"
		"stp s10, s11, [sp, #96]	\n\t"
		"stp s12, s13, [sp, #112]	\n\t"
		"stp s14, s15, [sp, #128]	\n\t"
		// Use x28 as sh4 context pointer
		"mov x28, %0				\n\t"
		// Use x27 as cycle_counter
		"mov w27, %2				\n\t"	// SH4_TIMESLICE

	"run_loop:						\n\t"
		"ldr w0, [x28, %3]			\n\t"	// CpuRunning
		"cmp w0, #0					\n\t"
		"b.eq end_run_loop			\n\t"

	"slice_loop:					\n\t"
		"ldr w0, [x28, %1]			\n\t"	// pc
		"bl bm_GetCode				\n\t"
		"blr x0						\n\t"
		"cmp w27, #0				\n\t"
		"b.gt slice_loop			\n\t"

		"add w27, w27, %2			\n\t"	// SH4_TIMESLICE
		"bl UpdateSystem_INTC		\n\t"
		"b run_loop					\n\t"

	"end_run_loop:					\n\t"
		"ldp s14, s15, [sp, #128]	\n\t"
		"ldp s12, s13, [sp, #112]	\n\t"
		"ldp s10, s11, [sp, #96]	\n\t"
		"ldp s8, s9, [sp, #80]		\n\t"
		"ldp x27, x28, [sp, #64]	\n\t"
		"ldp x25, x26, [sp, #48]	\n\t"
		"ldp x23, x24, [sp, #32]	\n\t"
		"ldp x21, x22, [sp, #16]	\n\t"
		"ldp x19, x20, [sp], #144	\n\t"
		:
		: "r"(reinterpret_cast<uintptr_t>(&ctx->cntx)),
		  "i"(offsetof(Sh4Context, pc)),
		  "i"(SH4_TIMESLICE),
		  "i"(offsetof(Sh4Context, CpuRunning))
		: "memory"
	);
}

void ngen_init_arm64()
{
}

RuntimeBlockInfo* ngen_AllocateBlock()
{
	return new DynaRBI();
}

static void ngen_blockcheckfail(u32 pc) {
	printf("arm64 JIT: SMC invalidation at %08X\n", pc);
	rdv_BlockCheckFail(pc);
}

class Arm64Assembler : public MacroAssembler
{
	typedef void (MacroAssembler::*Arm64Op_RRO)(const Register&, const Register&, const Operand&);
	typedef void (MacroAssembler::*Arm64Op_RROF)(const Register&, const Register&, const Operand&, enum FlagsUpdate);

public:
	Arm64Assembler() : MacroAssembler((u8 *)emit_GetCCPtr(), 64 * 1024), regalloc(this)
	{
		call_regs.push_back(&w0);
		call_regs.push_back(&w1);
		call_regs.push_back(&w2);
		call_regs.push_back(&w3);
		call_regs.push_back(&w4);
		call_regs.push_back(&w5);
		call_regs.push_back(&w6);
		call_regs.push_back(&w7);

		call_regs64.push_back(&x0);
		call_regs64.push_back(&x1);
		call_regs64.push_back(&x2);
		call_regs64.push_back(&x3);
		call_regs64.push_back(&x4);
		call_regs64.push_back(&x5);
		call_regs64.push_back(&x6);
		call_regs64.push_back(&x7);

		call_fregs.push_back(&s0);
		call_fregs.push_back(&s1);
		call_fregs.push_back(&s2);
		call_fregs.push_back(&s3);
		call_fregs.push_back(&s4);
		call_fregs.push_back(&s5);
		call_fregs.push_back(&s6);
		call_fregs.push_back(&s7);
	}

	void ngen_BinaryOp_RRO(shil_opcode* op, Arm64Op_RRO arm_op, Arm64Op_RROF arm_op2)
	{
		Operand op3 = Operand(0);
		if (op->rs2.is_imm())
		{
			op3 = Operand(op->rs2._imm);
		}
		else if (op->rs2.is_r32i())
		{
			op3 = Operand(regalloc.MapRegister(op->rs2));
		}
		if (arm_op != NULL)
			((*this).*arm_op)(regalloc.MapRegister(op->rd), regalloc.MapRegister(op->rs1), op3);
		else
			((*this).*arm_op2)(regalloc.MapRegister(op->rd), regalloc.MapRegister(op->rs1), op3, LeaveFlags);
	}

	template <typename R, typename... P>
	void ngen_CallRuntime(R (*function)(P...))
	{
		if (!frame_reg_saved)
		{
			Str(x30, MemOperand(sp, -16, PreIndex));
			frame_reg_saved = true;
		}
		Literal<uintptr_t> *function_address = function_literals[(void*)function];
		if (function_address == NULL)
		{
			function_address = new Literal<uintptr_t>(reinterpret_cast<uintptr_t>(function), GetLiteralPool(), RawLiteral::kDeletedOnPoolDestruction);
			function_literals[(void*)function] = function_address;
		}
		Ldr(x9, function_address);
		Blr(x9);
	}

	const Register& GenMemAddr(const shil_opcode& op, const Register* raddr = NULL)
	{
		const Register* ret_reg = raddr == NULL ? &w0 : raddr;

		if (op.rs3.is_imm())
		{
			Add(*ret_reg, regalloc.MapRegister(op.rs1), op.rs3._imm);
		}
		else if (op.rs3.is_r32i())
		{
			Add(*ret_reg, regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs3));
		}
		else if (!op.rs3.is_null())
		{
			die("invalid rs3");
		}
		else
		{
			if (raddr == NULL)
				ret_reg = &regalloc.MapRegister(op.rs1);
			else
				Mov(*ret_reg, regalloc.MapRegister(op.rs1));
		}

		return *ret_reg;
	}

	void ngen_Compile(RuntimeBlockInfo* block, bool force_checks, bool reset, bool staging, bool optimise)
	{
		//printf("REC-ARM64 compiling %08x\n", block->addr);
		if (force_checks)
			CheckBlock(block);

		// run register allocator
		regalloc.DoAlloc(block);

		// scheduler
		Sub(w27, w27, block->guest_cycles);

		for (size_t i = 0; i < block->oplist.size(); i++)
		{
			shil_opcode& op  = block->oplist[i];
			regalloc.OpBegin(&op, i);

			switch (op.op)
			{
			case shop_ifb:	// Interpreter fallback
				if (op.rs1._imm)	// if NeedPC()
				{
					Mov(w10, op.rs2._imm);
					Str(w10, sh4_context_mem_operand(&next_pc));
				}
				Mov(*call_regs[0], op.rs3._imm);

				ngen_CallRuntime(OpDesc[op.rs3._imm]->oph);
				break;

			case shop_jcond:
			case shop_jdyn:
				if (op.rs2.is_imm())
				{
					Add(w10, regalloc.MapRegister(op.rs1), op.rs2._imm);
					Mov(regalloc.MapRegister(op.rd), w10);
				}
				else
					Mov(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1));
				break;

			case shop_mov32:
				verify(op.rd.is_reg());
				verify(op.rs1.is_reg() || op.rs1.is_imm());

				if (regalloc.IsAllocf(op.rd))
				{
					if (op.rs1.is_imm())
						Fmov(regalloc.MapVRegister(op.rd), (float&)op.rs1._imm);
					else if (regalloc.IsAllocf(op.rs1))
						Fmov(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1));
					else
						Fmov(regalloc.MapVRegister(op.rd), regalloc.MapRegister(op.rs1));
				}
				else
				{
					if (op.rs1.is_imm())
						Mov(regalloc.MapRegister(op.rd), op.rs1._imm);
					else if (regalloc.IsAllocg(op.rs1))
						Mov(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1));
					else
						Fmov(regalloc.MapRegister(op.rd), regalloc.MapVRegister(op.rs1));
				}
				break;

			case shop_mov64:
				verify(op.rd.is_reg());
				verify(op.rs1.is_reg() || op.rs1.is_imm());

				shil_param_to_host_reg(op.rs1, x15);
				host_reg_to_shil_param(op.rd, x15);
				break;

			case shop_readm:
			{
				u32 size = op.flags & 0x7f;
				bool is_float = op.rs2.is_r32f() || op.rd.is_r32f();

				if (op.rs1.is_imm())
				{
					bool isram = false;
					void* ptr = _vmem_read_const(op.rs1._imm, isram, size);

					if (isram)
					{
						Ldr(x1, reinterpret_cast<uintptr_t>(ptr));
						switch (size)
						{
						case 2:
							Ldrsh(regalloc.MapRegister(op.rd), MemOperand(x1, xzr, SXTW));
							break;

						case 4:
							if (is_float)
								Ldr(regalloc.MapVRegister(op.rd), MemOperand(x1));
							else
								Ldr(regalloc.MapRegister(op.rd), MemOperand(x1));
							break;

						default:
							die("Invalid size");
							break;
						}
					}
					else
					{
						// Not RAM
						Mov(w0, op.rs1._imm);

						switch(size)
						{
						case 1:
							ngen_CallRuntime((void (*)())ptr);
							Sxtb(w0, w0);
							break;

						case 2:
							ngen_CallRuntime((void (*)())ptr);
							Sxth(w0, w0);
							break;

						case 4:
							ngen_CallRuntime((void (*)())ptr);
							break;

						case 8:
							die("SZ_64F not supported");
							break;
						}

						if (regalloc.IsAllocg(op.rd))
							Mov(regalloc.MapRegister(op.rd), w0);
						else
							Fmov(regalloc.MapVRegister(op.rd), w0);
					}
				}
				else
				{
#if 0	// Direct memory access. Need to handle SIGSEGV and rewrite block as needed (?)
					const Register& raddr = GenMemAddr(&op);

					if (_nvmem_enabled())
					{
						Add(w1, raddr, sizeof(Sh4Context));
						Bfc(w1, 29, 3);		// addr &= ~0xE0000000

						switch(size)
						{
						case 1:
							Ldrsb(regalloc.MapRegister(op.rd), MemOperand(x28, x1, SXTW));
							break;

						case 2:
							Ldrsh(regalloc.MapRegister(op.rd), MemOperand(x28, x1, SXTW));
							break;

						case 4:
							if (!is_float)
								Ldr(regalloc.MapRegister(op.rd), MemOperand(x28, x1));
							else
								Ldr(regalloc.MapVRegister(op.rd), MemOperand(x28, x1));
							break;

						case 8:
							// TODO use regalloc
							Ldr(x0, MemOperand(x28, x1));
							Str(x0, sh4_context_mem_operand(op.rd.reg_ptr()));
							break;
						}
					}
					else
					{
						// TODO
						die("Not implemented")
					}
#endif

					GenMemAddr(op, call_regs[0]);

					switch (size)
					{
					case 1:
						ngen_CallRuntime(ReadMem8);
						Sxtb(w0, w0);
						break;

					case 2:
						ngen_CallRuntime(ReadMem16);
						Sxth(w0, w0);
						break;

					case 4:
						ngen_CallRuntime(ReadMem32);
						break;

					case 8:
						ngen_CallRuntime(ReadMem64);
						break;

					default:
						die("1..8 bytes");
						break;
					}

					if (size != 8)
						host_reg_to_shil_param(op.rd, w0);
					else
						host_reg_to_shil_param(op.rd, x0);
				}
			}
			break;

			case shop_writem:
			{
				GenMemAddr(op, call_regs[0]);

				u32 size = op.flags & 0x7f;
				if (size != 8)
					shil_param_to_host_reg(op.rs2, *call_regs[1]);
				else
					shil_param_to_host_reg(op.rs2, *call_regs64[1]);

				switch (size)
				{
				case 1:
					ngen_CallRuntime(WriteMem8);
					break;

				case 2:
					ngen_CallRuntime(WriteMem16);
					break;

				case 4:
					ngen_CallRuntime(WriteMem32);
					break;

				case 8:
					ngen_CallRuntime(WriteMem64);
					break;

				default:
					die("1..8 bytes");
					break;
				}
			}
			break;

			case shop_sync_sr:
				ngen_CallRuntime(UpdateSR);
				break;
			case shop_sync_fpscr:
				ngen_CallRuntime(UpdateFPSCR);
				break;

			case shop_swaplb:
				// TODO Optimize
				Mov(w9, Operand(regalloc.MapRegister(op.rs1), LSR, 16));
				Rev16(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1));
				Bfc(regalloc.MapRegister(op.rd), 16, 16);
				Orr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), Operand(w9, LSL, 16));
				break;

			case shop_neg:
				Neg(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1));
				break;
			case shop_not:
				Mvn(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1));
				break;

			case shop_and:
				ngen_BinaryOp_RRO(&op, &MacroAssembler::And, NULL);
				break;
			case shop_or:
				ngen_BinaryOp_RRO(&op, &MacroAssembler::Orr, NULL);
				break;
			case shop_xor:
				ngen_BinaryOp_RRO(&op, &MacroAssembler::Eor, NULL);
				break;
			case shop_add:
				ngen_BinaryOp_RRO(&op, NULL, &MacroAssembler::Add);
				break;
			case shop_sub:
				ngen_BinaryOp_RRO(&op, NULL, &MacroAssembler::Sub);
				break;
			case shop_shl:
				if (op.rs2.is_imm())
					Lsl(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), op.rs2._imm);
				else if (op.rs2.is_reg())
					Lsl(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
				break;
			case shop_shr:
				if (op.rs2.is_imm())
					Lsr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), op.rs2._imm);
				else if (op.rs2.is_reg())
					Lsr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
				break;
			case shop_sar:
				if (op.rs2.is_imm())
					Asr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), op.rs2._imm);
				else if (op.rs2.is_reg())
					Asr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
				break;
			case shop_ror:
				if (op.rs2.is_imm())
					Ror(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), op.rs2._imm);
				else if (op.rs2.is_reg())
					Ror(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
				break;

			case shop_adc:
				Cmp(regalloc.MapRegister(op.rs3), 1);	// C = rs3
				Adcs(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2)); // (C,rd)=rs1+rs2+rs3(C)
				Cset(regalloc.MapRegister(op.rd2), cs);	// rd2 = C
				break;
			case shop_sbc:
				Cmp(wzr, regalloc.MapRegister(op.rs3));	// C = ~rs3
				Sbcs(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2)); // (C,rd) = rs1 - rs2 - ~rs3(C)
				Cset(regalloc.MapRegister(op.rd2), cc);	// rd2 = ~C
				break;

			case shop_rocr:
				Ubfm(w0, regalloc.MapRegister(op.rs1), 0, 0);										// w0 = rs1[0]
				Mov(regalloc.MapRegister(op.rd), Operand(regalloc.MapRegister(op.rs1), LSR, 1));	// rd = rs1 >> 1
				Orr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), Operand(regalloc.MapRegister(op.rs2), LSL, 31));	// rd |= C << 31
				Mov(regalloc.MapRegister(op.rd2), w0);												// rd2 = w0 (new C)
				break;
			case shop_rocl:
				Tst(regalloc.MapRegister(op.rs1), 0x80000000);	// Z = ~rs1[31]
				Orr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs2), Operand(regalloc.MapRegister(op.rs1), LSL, 1)); // rd = rs1 << 1 | rs2(C)
				Cset(regalloc.MapRegister(op.rd2), ne);			// rd2 = ~Z(C)
				break;

			case shop_shld:
			case shop_shad:
				// TODO optimize
				Cmp(regalloc.MapRegister(op.rs2), 0);
				Csel(w1, regalloc.MapRegister(op.rs2), wzr, ge);	// if shift >= 0 then w1 = shift else w1 = 0
				Mov(w0, wzr);	// wzr not supported by csneg
				Csneg(w2, w0, regalloc.MapRegister(op.rs2), ge);	// if shift < 0 then w2 = -shift else w2 = 0
				Cmp(w2, 32);
				Csel(w2, 31, w2, eq);								// if shift == -32 then w2 = 31
				Lsl(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), w1);		// Left shift by w1
				if (op.op == shop_shld)													// Right shift by w2
					// Logical shift
					Lsr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), w2);
				else
					// Arithmetic shift
					Asr(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), w2);
				break;

			case shop_test:
			case shop_seteq:
			case shop_setge:
			case shop_setgt:
			case shop_setae:
			case shop_setab:
				{
					if (op.op == shop_test)
					{
						if (op.rs2.is_imm())
							Tst(regalloc.MapRegister(op.rs1), op.rs2._imm);
						else
							Tst(regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
					}
					else
					{
						if (op.rs2.is_imm())
							Cmp(regalloc.MapRegister(op.rs1), op.rs2._imm);
						else
							Cmp(regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
					}

					static const Condition shop_conditions[] = { eq, eq, ge, gt, hs, hi };

					Cset(regalloc.MapRegister(op.rd), shop_conditions[op.op - shop_test]);
				}
				break;
			case shop_setpeq:
				Eor(w1, regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));

				Mov(regalloc.MapRegister(op.rd), wzr);
				Mov(w2, wzr);	// wzr not supported by csinc (?!)
				Tst(w1, 0xFF000000);
				Csinc(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), w2, ne);
				Tst(w1, 0x00FF0000);
				Csinc(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), w2, ne);
				Tst(w1, 0x0000FF00);
				Csinc(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), w2, ne);
				Tst(w1, 0x000000FF);
				Csinc(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rd), w2, ne);
				break;

			case shop_mul_u16:
				Uxth(w10, regalloc.MapRegister(op.rs1));
				Uxth(w11, regalloc.MapRegister(op.rs2));
				Mul(regalloc.MapRegister(op.rd), w10, w11);
				break;
			case shop_mul_s16:
				Sxth(w10, regalloc.MapRegister(op.rs1));
				Sxth(w11, regalloc.MapRegister(op.rs2));
				Mul(regalloc.MapRegister(op.rd), w10, w11);
				break;
			case shop_mul_i32:
				Mul(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
				break;
			case shop_mul_u64:
			case shop_mul_s64:
				{
					const Register& rd_xreg = Register::GetXRegFromCode(regalloc.MapRegister(op.rd).GetCode());
					if (op.op == shop_mul_u64)
						Umull(rd_xreg, regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
					else
						Smull(rd_xreg, regalloc.MapRegister(op.rs1), regalloc.MapRegister(op.rs2));
					const Register& rd2_xreg = Register::GetXRegFromCode(regalloc.MapRegister(op.rd2).GetCode());
					Lsr(rd2_xreg, rd_xreg, 32);
				}
				break;

			case shop_pref:
				Mov(w0, regalloc.MapRegister(op.rs1));
				if (op.flags != 0x1337)
				{
					Lsr(w1, regalloc.MapRegister(op.rs1), 26);
					Cmp(w1, 0x38);
				}

				if (CCN_MMUCR.AT)
				{
					Ldr(x9, reinterpret_cast<uintptr_t>(&do_sqw_mmu));
				}
				else
				{
					Sub(x9, x28, offsetof(Sh4RCB, cntx) - offsetof(Sh4RCB, do_sqw_nommu));
					Ldr(x9, MemOperand(x9));
					Sub(x1, x28, offsetof(Sh4RCB, cntx) - offsetof(Sh4RCB, sq_buffer));
				}
				if (!frame_reg_saved)
				{
					Str(x30, MemOperand(sp, -16, PreIndex));
					frame_reg_saved = true;
				}
				if (op.flags == 0x1337)
					Blr(x9);
				else
				{
					Label no_branch;
					B(&no_branch, ne);
					Blr(x9);
					Bind(&no_branch);
				}
				break;

			case shop_ext_s8:
				Sxtb(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1));
				break;
			case shop_ext_s16:
				Sxth(regalloc.MapRegister(op.rd), regalloc.MapRegister(op.rs1));
				break;

			//
			// FPU
			//

			case shop_fadd:
				Fadd(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1), regalloc.MapVRegister(op.rs2));
				break;
			case shop_fsub:
				Fsub(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1), regalloc.MapVRegister(op.rs2));
				break;
			case shop_fmul:
				Fmul(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1), regalloc.MapVRegister(op.rs2));
				break;
			case shop_fdiv:
				Fdiv(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1), regalloc.MapVRegister(op.rs2));
				break;

			case shop_fabs:
				Fabs(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1));
				break;
			case shop_fneg:
				Fneg(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1));
				break;
			case shop_fsqrt:
				Fsqrt(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs1));
				break;

			case shop_fmac:
				Fmadd(regalloc.MapVRegister(op.rd), regalloc.MapVRegister(op.rs3), regalloc.MapVRegister(op.rs2), regalloc.MapVRegister(op.rs1));
				break;

			case shop_fsrra:
				Fsqrt(s0, regalloc.MapVRegister(op.rs1));
				Fmov(s1, 1.f);
				Fdiv(regalloc.MapVRegister(op.rd), s1, s0);
				break;

			case shop_fsetgt:
			case shop_fseteq:
				Fcmp(regalloc.MapVRegister(op.rs1), regalloc.MapVRegister(op.rs2));
				Cset(regalloc.MapRegister(op.rd), op.op == shop_fsetgt ? gt : eq);
				break;

			case shop_fsca:
				Mov(x1, reinterpret_cast<uintptr_t>(&sin_table));
				Add(x1, x1, Operand(regalloc.MapRegister(op.rs1), UXTH, 3));
				// TODO use regalloc
				//Ldr(regalloc.MapVRegister(op.rd, 0), MemOperand(x1, 4, PostIndex));
				//Ldr(regalloc.MapVRegister(op.rd, 1), MemOperand(x1));
				Ldr(x2, MemOperand(x1));
				Str(x2, sh4_context_mem_operand(op.rd.reg_ptr()));
				break;

			case shop_fipr:
				Add(x9, x28, sh4_context_mem_operand(op.rs1.reg_ptr()).GetOffset());
				Ld1(v0.V4S(), MemOperand(x9));
				if (op.rs1._reg != op.rs2._reg)
				{
					Add(x9, x28, sh4_context_mem_operand(op.rs2.reg_ptr()).GetOffset());
					Ld1(v1.V4S(), MemOperand(x9));
					Fmul(v0.V4S(), v0.V4S(), v1.V4S());
				}
				else
					Fmul(v0.V4S(), v0.V4S(), v0.V4S());
				Faddp(v1.V4S(), v0.V4S(), v0.V4S());
				Faddp(regalloc.MapVRegister(op.rd), v1.V2S());
				break;

			case shop_ftrv:
				Add(x9, x28, sh4_context_mem_operand(op.rs1.reg_ptr()).GetOffset());
				Ld1(v0.V4S(), MemOperand(x9));
				Add(x9, x28, sh4_context_mem_operand(op.rs2.reg_ptr()).GetOffset());
				Ld1(v1.V4S(), MemOperand(x9, 16, PostIndex));
				Ld1(v2.V4S(), MemOperand(x9, 16, PostIndex));
				Ld1(v3.V4S(), MemOperand(x9, 16, PostIndex));
				Ld1(v4.V4S(), MemOperand(x9, 16, PostIndex));
				Fmul(v5.V4S(), v1.V4S(), s0, 0);
				Fmla(v5.V4S(), v2.V4S(), s0, 1);
				Fmla(v5.V4S(), v3.V4S(), s0, 2);
				Fmla(v5.V4S(), v4.V4S(), s0, 3);
				Add(x9, x28, sh4_context_mem_operand(op.rd.reg_ptr()).GetOffset());
				St1(v5.V4S(), MemOperand(x9));
				break;

			case shop_frswap:
				Add(x9, x28, sh4_context_mem_operand(op.rs1.reg_ptr()).GetOffset());
				Add(x10, x28, sh4_context_mem_operand(op.rd.reg_ptr()).GetOffset());
				Ld4(v0.V2D(), v1.V2D(), v2.V2D(), v3.V2D(), MemOperand(x9));
				Ld4(v4.V2D(), v5.V2D(), v6.V2D(), v7.V2D(), MemOperand(x10));
				St4(v4.V2D(), v5.V2D(), v6.V2D(), v7.V2D(), MemOperand(x9));
				St4(v0.V2D(), v1.V2D(), v2.V2D(), v3.V2D(), MemOperand(x10));
				break;

			case shop_cvt_f2i_t:
				Fcvtzs(regalloc.MapRegister(op.rd), regalloc.MapVRegister(op.rs1));
				break;
			case shop_cvt_i2f_n:
			case shop_cvt_i2f_z:
				Scvtf(regalloc.MapVRegister(op.rd), regalloc.MapRegister(op.rs1));
				break;

			default:
				shil_chf[op.op](&op);
				break;
			}
			regalloc.OpEnd(&op);
		}

		switch (block->BlockType)
		{

		case BET_StaticJump:
		case BET_StaticCall:
			// next_pc = block->BranchBlock;
			Ldr(w10, block->BranchBlock);
			Str(w10, sh4_context_mem_operand(&next_pc));
			break;

		case BET_Cond_0:
		case BET_Cond_1:
			{
				// next_pc = next_pc_value;
				// if (*jdyn == 0)
				//   next_pc = branch_pc_value;

				Mov(w10, block->NextBlock);

				if (block->has_jcond)
					Ldr(w11, sh4_context_mem_operand(&Sh4cntx.jdyn));
				else
					Ldr(w11, sh4_context_mem_operand(&sr.T));

				Cmp(w11, block->BlockType & 1);
				Label branch_not_taken;

				B(ne, &branch_not_taken);
				Mov(w10, block->BranchBlock);
				Bind(&branch_not_taken);

				Str(w10, sh4_context_mem_operand(&next_pc));
			}
			break;

		case BET_DynamicJump:
		case BET_DynamicCall:
		case BET_DynamicRet:
			// next_pc = *jdyn;
			Ldr(w10, sh4_context_mem_operand(&Sh4cntx.jdyn));
			Str(w10, sh4_context_mem_operand(&next_pc));
			break;

		case BET_DynamicIntr:
		case BET_StaticIntr:
			if (block->BlockType == BET_DynamicIntr)
			{
				// next_pc = *jdyn;
				Ldr(w10, sh4_context_mem_operand(&Sh4cntx.jdyn));
			}
			else
			{
				// next_pc = next_pc_value;
				Mov(w10, block->NextBlock);
			}
			Str(w10, sh4_context_mem_operand(&next_pc));

			ngen_CallRuntime(UpdateINTC);
			break;

		default:
			die("Invalid block end type");
		}

		if (frame_reg_saved)
			Ldr(x30, MemOperand(sp, 16, PostIndex));
		Ret();

		Label code_end;
		Bind(&code_end);

		FinalizeCode();

		block->code = GetBuffer()->GetStartAddress<DynarecCodeEntryPtr>();
		block->host_code_size = GetBuffer()->GetSizeInBytes();
		block->host_opcodes = GetLabelAddress<u32*>(&code_end) - GetBuffer()->GetStartAddress<u32*>();

		emit_Skip(block->host_code_size);
		CacheFlush((void*)block->code, GetBuffer()->GetEndAddress<void*>());
#if 0
		Instruction* instr_start = GetBuffer()->GetStartAddress<Instruction*>();
		Instruction* instr_end = GetLabelAddress<Instruction*>(&code_end);
		Decoder decoder;
		Disassembler disasm;
		decoder.AppendVisitor(&disasm);
		Instruction* instr;
		for (instr = instr_start; instr < instr_end; instr += kInstructionSize) {
			decoder.Decode(instr);
			printf("VIXL\t %p:\t%s\n",
			           reinterpret_cast<void*>(instr),
			           disasm.GetOutput());
		}
#endif
	}

	void ngen_CC_Start(shil_opcode* op)
	{
		CC_pars.clear();
	}

	void ngen_CC_Param(shil_opcode& op, shil_param& prm, CanonicalParamType tp)
	{
		switch (tp)
		{

		case CPT_u32:
		case CPT_ptr:
		case CPT_f32:
		{
			CC_PS t = { tp, &prm };
			CC_pars.push_back(t);
		}
		break;

		case CPT_u64rvL:
		case CPT_u32rv:
			host_reg_to_shil_param(prm, w0);
			break;

		case CPT_u64rvH:
			Lsr(x10, x0, 32);
			host_reg_to_shil_param(prm, w10);
			break;

		case CPT_f32rv:
			host_reg_to_shil_param(prm, s0);
			break;
		}
	}

	void ngen_CC_Call(shil_opcode*op, void* function)
	{
		int regused = 0;
		int fregused = 0;

		// Args are pushed in reverse order by shil_canonical
		for (int i = CC_pars.size(); i-- > 0;)
		{
			verify(fregused < call_fregs.size() && regused < call_regs.size());
			shil_param& prm = *CC_pars[i].prm;
			switch (CC_pars[i].type)
			{
			// push the params

			case CPT_u32:
				shil_param_to_host_reg(prm, *call_regs[regused++]);

				break;

			case CPT_f32:
				if (prm.is_reg()) {
					Fmov(*call_fregs[fregused], regalloc.MapVRegister(prm));
				}
				else {
					verify(prm.is_null());
				}
				fregused++;
				break;

			case CPT_ptr:
				verify(prm.is_reg());
				// push the ptr itself
				Mov(*call_regs64[regused++], reinterpret_cast<uintptr_t>(prm.reg_ptr()));

				break;
			case CPT_u32rv:
			case CPT_u64rvL:
			case CPT_u64rvH:
			case CPT_f32rv:
				// return values are handled in ngen_CC_param()
				break;
			}
		}
		ngen_CallRuntime((void (*)())function);
	}

	MemOperand sh4_context_mem_operand(void *p)
	{
		u32 offset = (u8*)p - (u8*)&p_sh4rcb->cntx;
		verify((offset & 3) == 0 && offset <= 16380);	// FIXME 64-bit regs need multiple of 8 up to 32760
		return MemOperand(x28, offset);
	}

private:
	void CheckBlock(RuntimeBlockInfo* block)
	{
		s32 sz = block->sh4_code_size;

		Label blockcheck_fail;
		Label blockcheck_success;

		u8* ptr = GetMemPtr(block->addr, sz);
		if (ptr == NULL)
			// FIXME Can a block cross a RAM / non-RAM boundary??
			return;

		Ldr(x9, reinterpret_cast<uintptr_t>(ptr));

		while (sz > 0)
		{
			if (sz >= 8)
			{
				Ldr(x10, MemOperand(x9, 8, PostIndex));
				Ldr(x11, *(u64*)ptr);
				Cmp(x10, x11);
				sz -= 8;
				ptr += 8;
			}
			else if (sz >= 4)
			{
				Ldr(w10, MemOperand(x9, 4, PostIndex));
				Ldr(w11, *(u32*)ptr);
				Cmp(w10, w11);
				sz -= 4;
				ptr += 4;
			}
			else
			{
				Ldrh(w10, MemOperand(x9, 2, PostIndex));
				Mov(w11, *(u16*)ptr);
				Cmp(w10, w11);
				sz -= 2;
				ptr += 2;
			}
			B(ne, &blockcheck_fail);
		}
		B(&blockcheck_success);

		Bind(&blockcheck_fail);
		Ldr(w0, block->addr);
		TailCallRuntime(ngen_blockcheckfail);

		Bind(&blockcheck_success);
	}

	void shil_param_to_host_reg(const shil_param& param, const Register& reg)
	{
		if (param.is_imm())
		{
			Mov(reg, param._imm);
		}
		else if (param.is_reg())
		{
			if (param.is_r64f())
			{
				// TODO use regalloc
				Ldr(reg, sh4_context_mem_operand(param.reg_ptr()));
			}
			else if (param.is_r32f())
				Fmov(reg, regalloc.MapVRegister(param));
			else
				Mov(reg, regalloc.MapRegister(param));
		}
		else
		{
			verify(param.is_null());
		}
	}

	void host_reg_to_shil_param(const shil_param& param, const CPURegister& reg)
	{
		if (reg.Is64Bits())
		{
			// TODO use regalloc
			Str((const Register&)reg, sh4_context_mem_operand(param.reg_ptr()));
		}
		else if (regalloc.IsAllocg(param))
		{
			if (reg.IsRegister())
				Mov(regalloc.MapRegister(param), (const Register&)reg);
			else
				Fmov(regalloc.MapRegister(param), (const VRegister&)reg);
		}
		else
		{
			if (reg.IsVRegister())
				Fmov(regalloc.MapVRegister(param), (const VRegister&)reg);
			else
				Fmov(regalloc.MapVRegister(param), (const Register&)reg);
		}
	}

	struct CC_PS
	{
		CanonicalParamType type;
		shil_param* prm;
	};
	vector<CC_PS> CC_pars;
	std::vector<const WRegister*> call_regs;
	std::vector<const XRegister*> call_regs64;
	std::vector<const VRegister*> call_fregs;
	Arm64RegAlloc regalloc;
	bool frame_reg_saved = false;
	std::map<void*, Literal<uintptr_t>*> function_literals;
};

static Arm64Assembler* compiler;

void ngen_Compile_arm64(RuntimeBlockInfo* block, bool force_checks, bool reset, bool staging, bool optimise)
{
	verify(emit_FreeSpace() >= 16 * 1024);

	compiler = new Arm64Assembler();

	compiler->ngen_Compile(block, force_checks, reset, staging, optimise);

	delete compiler;
	compiler = NULL;
}

void ngen_CC_Start_arm64(shil_opcode* op)
{
	compiler->ngen_CC_Start(op);
}

void ngen_CC_Param_arm64(shil_opcode* op, shil_param* par, CanonicalParamType tp)
{
	compiler->ngen_CC_Param(*op, *par, tp);
}

void ngen_CC_Call_arm64(shil_opcode*op, void* function)
{
	compiler->ngen_CC_Call(op, function);
}

void ngen_CC_Finish_arm64(shil_opcode* op)
{

}

void Arm64RegAlloc::Preload(u32 reg, eReg nreg)
{
	assembler->Ldr(Register(nreg, 32), assembler->sh4_context_mem_operand(GetRegPtr(reg)));
}
void Arm64RegAlloc::Writeback(u32 reg, eReg nreg)
{
	assembler->Str(Register(nreg, 32), assembler->sh4_context_mem_operand(GetRegPtr(reg)));
}
void Arm64RegAlloc::Preload_FPU(u32 reg, eFReg nreg)
{
	assembler->Ldr(VRegister(nreg, 32), assembler->sh4_context_mem_operand(GetRegPtr(reg)));
}
void Arm64RegAlloc::Writeback_FPU(u32 reg, eFReg nreg)
{
	assembler->Str(VRegister(nreg, 32), assembler->sh4_context_mem_operand(GetRegPtr(reg)));
}


extern "C" void do_sqw_nommu_area_3(u32 dst, u8* sqb)
{
	__asm__ volatile
	(
		"movz x11, #0x0C00, lsl #16 \n\t"
		"add x11, x1, x11			\n\t"	// get ram ptr from x1, part 1
		"and x12, x0, #0x20			\n\t"	// SQ# selection, isolate
		"ubfx x0, x0, #5, #20		\n\t"	// get ram offset
		"add x1, x12, x1			\n\t"	// SQ# selection, add to SQ ptr
		"add x11, x11, #512			\n\t"	// get ram ptr from x1, part 2
		"add x11, x11, x0, lsl #5	\n\t"	// ram + offset
		"ld2 { v0.2D, v1.2D }, [x1]	\n\t"
		"st2 { v0.2D, v1.2D }, [x11] \n\t"
		"ret						\n"

		: : : "memory"
	);
}
#endif	// FEAT_SHREC == DYNAREC_JIT