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
|
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -run-pass=none -o - %s | FileCheck %s
--- |
; ModuleID = 'test/CodeGen/AMDGPU/memcpy-scoped-aa.ll'
source_filename = "test/CodeGen/AMDGPU/memcpy-scoped-aa.ll"
target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5-G1-ni:7"
target triple = "amdgcn-amd-amdhsa"
define i32 @test_memcpy(ptr addrspace(1) nocapture %p, ptr addrspace(1) nocapture readonly %q) #0 {
%p0 = bitcast ptr addrspace(1) %p to ptr addrspace(1)
%add.ptr = getelementptr inbounds i32, ptr addrspace(1) %p, i64 4
%p1 = bitcast ptr addrspace(1) %add.ptr to ptr addrspace(1)
tail call void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) noundef nonnull align 4 dereferenceable(16) %p0, ptr addrspace(1) noundef nonnull align 4 dereferenceable(16) %p1, i64 16, i1 false), !alias.scope !0, !noalias !3
%1 = bitcast ptr addrspace(1) %q to ptr addrspace(1)
%2 = load <2 x i32>, ptr addrspace(1) %1, align 4, !alias.scope !3, !noalias !0
%v01 = extractelement <2 x i32> %2, i32 0
%v12 = extractelement <2 x i32> %2, i32 1
%add = add i32 %v01, %v12
ret i32 %add
}
define i32 @test_memcpy_inline(ptr addrspace(1) nocapture %p, ptr addrspace(1) nocapture readonly %q) #0 {
%p0 = bitcast ptr addrspace(1) %p to ptr addrspace(1)
%add.ptr = getelementptr inbounds i32, ptr addrspace(1) %p, i64 4
%p1 = bitcast ptr addrspace(1) %add.ptr to ptr addrspace(1)
tail call void @llvm.memcpy.inline.p1.p1.i64(ptr addrspace(1) noundef nonnull align 4 dereferenceable(16) %p0, ptr addrspace(1) noundef nonnull align 4 dereferenceable(16) %p1, i64 16, i1 false), !alias.scope !0, !noalias !3
%1 = bitcast ptr addrspace(1) %q to ptr addrspace(1)
%2 = load <2 x i32>, ptr addrspace(1) %1, align 4, !alias.scope !3, !noalias !0
%v01 = extractelement <2 x i32> %2, i32 0
%v12 = extractelement <2 x i32> %2, i32 1
%add = add i32 %v01, %v12
ret i32 %add
}
; Function Attrs: argmemonly nofree nounwind willreturn
declare void @llvm.memcpy.p1.p1.i64(ptr addrspace(1) noalias nocapture writeonly, ptr addrspace(1) noalias nocapture readonly, i64, i1 immarg) #1
; Function Attrs: argmemonly nofree nounwind willreturn
declare void @llvm.memcpy.inline.p1.p1.i64(ptr addrspace(1) noalias nocapture writeonly, ptr addrspace(1) noalias nocapture readonly, i64 immarg, i1 immarg) #1
; Function Attrs: convergent nounwind willreturn
declare { i1, i32 } @llvm.amdgcn.if.i32(i1) #2
; Function Attrs: convergent nounwind willreturn
declare { i1, i32 } @llvm.amdgcn.else.i32.i32(i32) #2
; Function Attrs: convergent nounwind readnone willreturn
declare i32 @llvm.amdgcn.if.break.i32(i1, i32) #3
; Function Attrs: convergent nounwind willreturn
declare i1 @llvm.amdgcn.loop.i32(i32) #2
; Function Attrs: convergent nounwind willreturn
declare void @llvm.amdgcn.end.cf.i32(i32) #2
attributes #0 = { "target-cpu"="gfx1010" }
attributes #1 = { argmemonly nofree nounwind willreturn "target-cpu"="gfx1010" }
attributes #2 = { convergent nounwind willreturn }
attributes #3 = { convergent nounwind readnone willreturn }
!0 = !{!1}
!1 = distinct !{!1, !2, !"bax: %p"}
!2 = distinct !{!2, !"bax"}
!3 = !{!4}
!4 = distinct !{!4, !2, !"bax: %q"}
...
---
name: test_memcpy
machineMetadataNodes:
- '!9 = distinct !{!9, !7, !"Dst"}'
- '!6 = distinct !{!6, !7, !"Src"}'
- '!11 = !{!4, !6}'
- '!5 = !{!1, !6}'
- '!8 = !{!4, !9}'
- '!10 = !{!1, !9}'
- '!7 = distinct !{!7, !"MemcpyLoweringDomain"}'
body: |
bb.0 (%ir-block.0):
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; CHECK-LABEL: name: test_memcpy
; CHECK: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr3
; CHECK: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1
; CHECK: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]]
; CHECK: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY5]], 16, 0, implicit $exec :: (load (s128) from %ir.p1, align 4, !alias.scope !5, !noalias !8, addrspace 1)
; CHECK: [[COPY6:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]]
; CHECK: GLOBAL_STORE_DWORDX4 [[COPY6]], killed [[GLOBAL_LOAD_DWORDX4_]], 0, 0, implicit $exec :: (store (s128) into %ir.p0, align 4, !alias.scope !10, !noalias !11, addrspace 1)
; CHECK: [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]]
; CHECK: [[GLOBAL_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = GLOBAL_LOAD_DWORDX2 killed [[COPY7]], 0, 0, implicit $exec :: (load (s64) from %ir.1, align 4, !alias.scope !3, !noalias !0, addrspace 1)
; CHECK: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[GLOBAL_LOAD_DWORDX2_]].sub0
; CHECK: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[GLOBAL_LOAD_DWORDX2_]].sub1
; CHECK: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 killed [[COPY8]], killed [[COPY9]], 0, implicit $exec
; CHECK: $vgpr0 = COPY [[V_ADD_U32_e64_]]
; CHECK: SI_RETURN implicit $vgpr0
%3:vgpr_32 = COPY $vgpr3
%2:vgpr_32 = COPY $vgpr2
%1:vgpr_32 = COPY $vgpr1
%0:vgpr_32 = COPY $vgpr0
%17:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
%18:vreg_64 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1
%9:vreg_64 = COPY %18
%8:vreg_128 = GLOBAL_LOAD_DWORDX4 %9, 16, 0, implicit $exec :: (load (s128) from %ir.p1, align 4, !alias.scope !5, !noalias !8, addrspace 1)
%10:vreg_64 = COPY %18
GLOBAL_STORE_DWORDX4 %10, killed %8, 0, 0, implicit $exec :: (store (s128) into %ir.p0, align 4, !alias.scope !10, !noalias !11, addrspace 1)
%12:vreg_64 = COPY %17
%11:vreg_64 = GLOBAL_LOAD_DWORDX2 killed %12, 0, 0, implicit $exec :: (load (s64) from %ir.1, align 4, !alias.scope !3, !noalias !0, addrspace 1)
%13:vgpr_32 = COPY %11.sub0
%14:vgpr_32 = COPY %11.sub1
%15:vgpr_32 = V_ADD_U32_e64 killed %13, killed %14, 0, implicit $exec
$vgpr0 = COPY %15
SI_RETURN implicit $vgpr0
...
---
name: test_memcpy_inline
machineMetadataNodes:
- '!6 = distinct !{!6, !7, !"Src"}'
- '!7 = distinct !{!7, !"MemcpyLoweringDomain"}'
- '!9 = distinct !{!9, !7, !"Dst"}'
- '!11 = !{!4, !6}'
- '!5 = !{!1, !6}'
- '!8 = !{!4, !9}'
- '!10 = !{!1, !9}'
body: |
bb.0 (%ir-block.0):
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
; CHECK-LABEL: name: test_memcpy_inline
; CHECK: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr3
; CHECK: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; CHECK: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; CHECK: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; CHECK: [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY2]], %subreg.sub0, [[COPY1]], %subreg.sub1
; CHECK: [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY4]], %subreg.sub0, [[COPY3]], %subreg.sub1
; CHECK: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]]
; CHECK: [[GLOBAL_LOAD_DWORDX4_:%[0-9]+]]:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY5]], 16, 0, implicit $exec :: (load (s128) from %ir.p1, align 4, !alias.scope !5, !noalias !8, addrspace 1)
; CHECK: [[COPY6:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]]
; CHECK: GLOBAL_STORE_DWORDX4 [[COPY6]], killed [[GLOBAL_LOAD_DWORDX4_]], 0, 0, implicit $exec :: (store (s128) into %ir.p0, align 4, !alias.scope !10, !noalias !11, addrspace 1)
; CHECK: [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]]
; CHECK: [[GLOBAL_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = GLOBAL_LOAD_DWORDX2 killed [[COPY7]], 0, 0, implicit $exec :: (load (s64) from %ir.1, align 4, !alias.scope !3, !noalias !0, addrspace 1)
; CHECK: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[GLOBAL_LOAD_DWORDX2_]].sub0
; CHECK: [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[GLOBAL_LOAD_DWORDX2_]].sub1
; CHECK: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 killed [[COPY8]], killed [[COPY9]], 0, implicit $exec
; CHECK: $vgpr0 = COPY [[V_ADD_U32_e64_]]
; CHECK: SI_RETURN implicit $vgpr0
%3:vgpr_32 = COPY $vgpr3
%2:vgpr_32 = COPY $vgpr2
%1:vgpr_32 = COPY $vgpr1
%0:vgpr_32 = COPY $vgpr0
%17:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %3, %subreg.sub1
%18:vreg_64 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1
%9:vreg_64 = COPY %18
%8:vreg_128 = GLOBAL_LOAD_DWORDX4 %9, 16, 0, implicit $exec :: (load (s128) from %ir.p1, align 4, !alias.scope !5, !noalias !8, addrspace 1)
%10:vreg_64 = COPY %18
GLOBAL_STORE_DWORDX4 %10, killed %8, 0, 0, implicit $exec :: (store (s128) into %ir.p0, align 4, !alias.scope !10, !noalias !11, addrspace 1)
%12:vreg_64 = COPY %17
%11:vreg_64 = GLOBAL_LOAD_DWORDX2 killed %12, 0, 0, implicit $exec :: (load (s64) from %ir.1, align 4, !alias.scope !3, !noalias !0, addrspace 1)
%13:vgpr_32 = COPY %11.sub0
%14:vgpr_32 = COPY %11.sub1
%15:vgpr_32 = V_ADD_U32_e64 killed %13, killed %14, 0, implicit $exec
$vgpr0 = COPY %15
SI_RETURN implicit $vgpr0
...
|
