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
|
; RUN: llc < %s -mattr=-avx -fast-isel -mcpu=core2 -O0 -regalloc=fast -asm-verbose=0 -fast-isel-abort | FileCheck %s
; RUN: llc < %s -mattr=+avx -fast-isel -mcpu=core2 -O0 -regalloc=fast -asm-verbose=0 -fast-isel-abort | FileCheck %s --check-prefix=AVX
; RUN: llc < %s -fast-isel -O0 -regalloc=fast -asm-verbose=0 -fast-isel-abort -mtriple=x86_64-none-nacl | FileCheck %s --check-prefix=NACL64
; RUN: llc < %s -fast-isel -O0 -regalloc=fast -asm-verbose=0 -fast-isel-abort -mtriple=x86_64-none-nacl -relocation-model=pic | FileCheck %s --check-prefix=NACL64_PIC
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
target triple = "x86_64-apple-darwin10.0.0"
; Make sure that fast-isel folds the immediate into the binop even though it
; is non-canonical.
define i32 @test1(i32 %i) nounwind ssp {
%and = and i32 8, %i
ret i32 %and
}
; CHECK: test1:
; CHECK: andl $8,
; rdar://9289512 - The load should fold into the compare.
define void @test2(i64 %x) nounwind ssp {
entry:
%x.addr = alloca i64, align 8
store i64 %x, i64* %x.addr, align 8
%tmp = load i64* %x.addr, align 8
%cmp = icmp sgt i64 %tmp, 42
br i1 %cmp, label %if.then, label %if.end
if.then: ; preds = %entry
br label %if.end
if.end: ; preds = %if.then, %entry
ret void
; CHECK: test2:
; CHECK: movq %rdi, -8(%rsp)
; CHECK: cmpq $42, -8(%rsp)
}
@G = external global i32
define i64 @test3() nounwind {
%A = ptrtoint i32* @G to i64
ret i64 %A
; CHECK: test3:
; CHECK: movq _G@GOTPCREL(%rip), %rax
; CHECK-NEXT: ret
}
; NACL64 version uses i32 for 32-bit pointers.
define i32 @test3_nacl64() nounwind {
%A = ptrtoint i32* @G to i32
ret i32 %A
; NACL64_PIC: test3_nacl64:
; NACL64_PIC: movl G@GOTPCREL(%rip), %eax
; NACL64_PIC-NEXT: popq %r11
; NACL64_PIC-NEXT: nacljmp %r11, %r15
}
; rdar://9289558
@rtx_length = external global [153 x i8]
define i32 @test4(i64 %idxprom9) nounwind {
%arrayidx10 = getelementptr inbounds [153 x i8]* @rtx_length, i32 0, i64 %idxprom9
%tmp11 = load i8* %arrayidx10, align 1
%conv = zext i8 %tmp11 to i32
ret i32 %conv
; CHECK: test4:
; CHECK: movq _rtx_length@GOTPCREL(%rip), %rax
; CHECK-NEXT: movzbl (%rax,%rdi), %eax
; CHECK-NEXT: ret
}
; PR3242 - Out of range shifts should not be folded by fastisel.
define void @test5(i32 %x, i32* %p) nounwind {
%y = ashr i32 %x, 50000
store i32 %y, i32* %p
ret void
; CHECK: test5:
; CHECK: movl $50000, %ecx
; CHECK: sarl %cl, %edi
; CHECK: ret
}
; rdar://9289501 - fast isel should fold trivial multiplies to shifts.
define i64 @test6(i64 %x) nounwind ssp {
entry:
%mul = mul nsw i64 %x, 8
ret i64 %mul
; CHECK: test6:
; CHECK: shlq $3, %rdi
}
define i32 @test7(i32 %x) nounwind ssp {
entry:
%mul = mul nsw i32 %x, 8
ret i32 %mul
; CHECK: test7:
; CHECK: shll $3, %edi
}
; rdar://9289507 - folding of immediates into 64-bit operations.
define i64 @test8(i64 %x) nounwind ssp {
entry:
%add = add nsw i64 %x, 7
ret i64 %add
; CHECK: test8:
; CHECK: addq $7, %rdi
}
define i64 @test9(i64 %x) nounwind ssp {
entry:
%add = mul nsw i64 %x, 7
ret i64 %add
; CHECK: test9:
; CHECK: imulq $7, %rdi, %rax
}
; rdar://9297011 - Don't reject udiv by a power of 2.
define i32 @test10(i32 %X) nounwind {
%Y = udiv i32 %X, 8
ret i32 %Y
; CHECK: test10:
; CHECK: shrl $3,
}
define i32 @test11(i32 %X) nounwind {
%Y = sdiv exact i32 %X, 8
ret i32 %Y
; CHECK: test11:
; CHECK: sarl $3,
}
; rdar://9297006 - Trunc to bool.
define void @test12(i8 %tmp) nounwind ssp noredzone {
entry:
%tobool = trunc i8 %tmp to i1
br i1 %tobool, label %if.then, label %if.end
if.then: ; preds = %entry
call void @test12(i8 0) noredzone
br label %if.end
if.end: ; preds = %if.then, %entry
ret void
; CHECK: test12:
; CHECK: testb $1,
; CHECK-NEXT: je L
; CHECK-NEXT: movl $0, %edi
; CHECK-NEXT: callq
}
declare void @test13f(i1 %X)
define void @test13() nounwind {
call void @test13f(i1 0)
ret void
; CHECK: test13:
; CHECK: movl $0, %edi
; CHECK-NEXT: callq
}
; rdar://9297003 - fast isel bails out on all functions taking bools
define void @test14(i8 %tmp) nounwind ssp noredzone {
entry:
%tobool = trunc i8 %tmp to i1
call void @test13f(i1 zeroext %tobool) noredzone
ret void
; CHECK: test14:
; CHECK: andb $1,
; CHECK: callq
}
declare void @llvm.memcpy.p0i8.p0i8.i64(i8*, i8*, i64, i32, i1)
; rdar://9289488 - fast-isel shouldn't bail out on llvm.memcpy
define void @test15(i8* %a, i8* %b) nounwind {
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* %b, i64 4, i32 4, i1 false)
ret void
; CHECK: test15:
; CHECK-NEXT: movl (%rsi), %eax
; CHECK-NEXT: movl %eax, (%rdi)
; CHECK-NEXT: ret
}
; Handling for varargs calls
declare void @test16callee(...) nounwind
define void @test16() nounwind {
; CHECK: test16:
; CHECK: movl $1, %edi
; CHECK: movb $0, %al
; CHECK: callq _test16callee
call void (...)* @test16callee(i32 1)
br label %block2
block2:
; CHECK: movabsq $1
; CHECK: cvtsi2sdq {{.*}} %xmm0
; CHECK: movb $1, %al
; CHECK: callq _test16callee
; AVX: movabsq $1
; AVX: vmovsd LCP{{.*}}_{{.*}}(%rip), %xmm0
; AVX: movb $1, %al
; AVX: callq _test16callee
call void (...)* @test16callee(double 1.000000e+00)
ret void
}
declare void @foo() unnamed_addr ssp align 2
; Verify that we don't fold the load into the compare here. That would move it
; w.r.t. the call.
define i32 @test17(i32 *%P) ssp nounwind {
entry:
%tmp = load i32* %P
%cmp = icmp ne i32 %tmp, 5
call void @foo()
br i1 %cmp, label %if.then, label %if.else
if.then: ; preds = %entry
ret i32 1
if.else: ; preds = %entry
ret i32 2
; CHECK: test17:
; CHECK: movl (%rdi), %eax
; CHECK: callq _foo
; CHECK: cmpl $5, %eax
; CHECK-NEXT: je
}
; Check that 0.0 is materialized using xorps
define void @test18(float* %p1) {
store float 0.0, float* %p1
ret void
; CHECK: test18:
; CHECK: xorps
}
; Without any type hints, doubles use the smaller xorps instead of xorpd.
define void @test19(double* %p1) {
store double 0.0, double* %p1
ret void
; CHECK: test19:
; CHECK: xorps
}
; Check that we fast-isel sret
%struct.a = type { i64, i64, i64 }
define void @test20() nounwind ssp {
entry:
%tmp = alloca %struct.a, align 8
call void @test20sret(%struct.a* sret %tmp)
ret void
; CHECK: test20:
; CHECK: leaq (%rsp), %rdi
; CHECK: callq _test20sret
}
declare void @test20sret(%struct.a* sret)
; Check that -0.0 is not materialized using xor
define void @test21(double* %p1) {
store double -0.0, double* %p1
ret void
; CHECK: test21:
; CHECK-NOT: xor
; CHECK: movsd LCPI
}
; Check that immediate arguments to a function
; do not cause massive spilling and are used
; as immediates just before the call.
define void @test22() nounwind {
entry:
call void @foo22(i32 0)
call void @foo22(i32 1)
call void @foo22(i32 2)
call void @foo22(i32 3)
ret void
; CHECK: test22:
; CHECK: movl $0, %edi
; CHECK: callq _foo22
; CHECK: movl $1, %edi
; CHECK: callq _foo22
; CHECK: movl $2, %edi
; CHECK: callq _foo22
; CHECK: movl $3, %edi
; CHECK: callq _foo22
}
declare void @foo22(i32)
; PR13563
define void @test23(i8* noalias sret %result) {
%a = alloca i8
%b = call i8* @foo23()
ret void
; CHECK: test23:
; CHECK: call
; CHECK: movq %rdi, %rax
; CHECK: ret
; NACL64: test23:
; NACL64: call
; NACL64: movl %edi, %eax
; NACL64: popq %r11
; NACL64: nacljmp %r11, %r15
}
declare i8* @foo23()
|
