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
|
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=kaveri -run-pass=amdgpu-regbankselect -mattr=+unaligned-access-mode %s -verify-machineinstrs -o - | FileCheck -check-prefixes=GFX7 %s
# RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 -run-pass=amdgpu-regbankselect -mattr=+unaligned-access-mode %s -verify-machineinstrs -o - | FileCheck -check-prefixes=GFX1010 %s
---
name: test_uniform_load_without_noclobber
legalized: true
tracksRegLiveness: true
body: |
bb.1.entry:
liveins: $sgpr0_sgpr1, $sgpr2_sgpr3
; GFX7-LABEL: name: test_uniform_load_without_noclobber
; GFX7: liveins: $sgpr0_sgpr1, $sgpr2_sgpr3
; GFX7-NEXT: {{ $}}
; GFX7-NEXT: %in_addr:sgpr(p1) = COPY $sgpr0_sgpr1
; GFX7-NEXT: %out_addr:sgpr(p1) = COPY $sgpr2_sgpr3
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD %in_addr(p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX7-NEXT: [[C:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 16
; GFX7-NEXT: [[PTR_ADD:%[0-9]+]]:vgpr(p1) = G_PTR_ADD %in_addr, [[C]](s64)
; GFX7-NEXT: [[LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<4 x s32>) from unknown-address + 16, align 4, addrspace 1)
; GFX7-NEXT: [[C1:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 32
; GFX7-NEXT: [[PTR_ADD1:%[0-9]+]]:vgpr(p1) = G_PTR_ADD %in_addr, [[C1]](s64)
; GFX7-NEXT: [[LOAD2:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD1]](p1) :: (load (<4 x s32>) from unknown-address + 32, align 4, addrspace 1)
; GFX7-NEXT: [[C2:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 48
; GFX7-NEXT: [[PTR_ADD2:%[0-9]+]]:vgpr(p1) = G_PTR_ADD %in_addr, [[C2]](s64)
; GFX7-NEXT: [[LOAD3:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD2]](p1) :: (load (<4 x s32>) from unknown-address + 48, align 4, addrspace 1)
; GFX7-NEXT: %load:vgpr(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
; GFX7-NEXT: %load0_3:vgpr(<4 x s32>), %load4_7:vgpr(<4 x s32>), %load8_11:vgpr(<4 x s32>), %load12_15:vgpr(<4 x s32>) = G_UNMERGE_VALUES %load(<16 x s32>)
; GFX7-NEXT: G_STORE %load0_3(<4 x s32>), %out_addr(p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX7-NEXT: %cst16:sgpr(s64) = G_CONSTANT i64 16
; GFX7-NEXT: %out_addr_plus_16:sgpr(p1) = G_PTR_ADD %out_addr, %cst16(s64)
; GFX7-NEXT: G_STORE %load4_7(<4 x s32>), %out_addr_plus_16(p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX7-NEXT: %cst32:sgpr(s64) = G_CONSTANT i64 32
; GFX7-NEXT: %out_addr_plus_32:sgpr(p1) = G_PTR_ADD %out_addr, %cst32(s64)
; GFX7-NEXT: G_STORE %load8_11(<4 x s32>), %out_addr_plus_32(p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX7-NEXT: %cst48:sgpr(s64) = G_CONSTANT i64 48
; GFX7-NEXT: %out_addr_plus_48:sgpr(p1) = G_PTR_ADD %out_addr, %cst48(s64)
; GFX7-NEXT: G_STORE %load12_15(<4 x s32>), %out_addr_plus_48(p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX7-NEXT: S_ENDPGM 0
; GFX1010-LABEL: name: test_uniform_load_without_noclobber
; GFX1010: liveins: $sgpr0_sgpr1, $sgpr2_sgpr3
; GFX1010-NEXT: {{ $}}
; GFX1010-NEXT: %in_addr:sgpr(p1) = COPY $sgpr0_sgpr1
; GFX1010-NEXT: %out_addr:sgpr(p1) = COPY $sgpr2_sgpr3
; GFX1010-NEXT: [[COPY:%[0-9]+]]:vgpr(p1) = COPY %in_addr(p1)
; GFX1010-NEXT: [[LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD %in_addr(p1) :: (load (<4 x s32>), align 4, addrspace 1)
; GFX1010-NEXT: [[C:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 16
; GFX1010-NEXT: [[PTR_ADD:%[0-9]+]]:vgpr(p1) = G_PTR_ADD %in_addr, [[C]](s64)
; GFX1010-NEXT: [[LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<4 x s32>) from unknown-address + 16, align 4, addrspace 1)
; GFX1010-NEXT: [[C1:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 32
; GFX1010-NEXT: [[PTR_ADD1:%[0-9]+]]:vgpr(p1) = G_PTR_ADD %in_addr, [[C1]](s64)
; GFX1010-NEXT: [[LOAD2:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD1]](p1) :: (load (<4 x s32>) from unknown-address + 32, align 4, addrspace 1)
; GFX1010-NEXT: [[C2:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 48
; GFX1010-NEXT: [[PTR_ADD2:%[0-9]+]]:vgpr(p1) = G_PTR_ADD %in_addr, [[C2]](s64)
; GFX1010-NEXT: [[LOAD3:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD2]](p1) :: (load (<4 x s32>) from unknown-address + 48, align 4, addrspace 1)
; GFX1010-NEXT: %load:vgpr(<16 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>), [[LOAD2]](<4 x s32>), [[LOAD3]](<4 x s32>)
; GFX1010-NEXT: %load0_3:vgpr(<4 x s32>), %load4_7:vgpr(<4 x s32>), %load8_11:vgpr(<4 x s32>), %load12_15:vgpr(<4 x s32>) = G_UNMERGE_VALUES %load(<16 x s32>)
; GFX1010-NEXT: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY %out_addr(p1)
; GFX1010-NEXT: G_STORE %load0_3(<4 x s32>), [[COPY1]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX1010-NEXT: %cst16:sgpr(s64) = G_CONSTANT i64 16
; GFX1010-NEXT: %out_addr_plus_16:sgpr(p1) = G_PTR_ADD %out_addr, %cst16(s64)
; GFX1010-NEXT: [[COPY2:%[0-9]+]]:vgpr(p1) = COPY %out_addr_plus_16(p1)
; GFX1010-NEXT: G_STORE %load4_7(<4 x s32>), [[COPY2]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX1010-NEXT: %cst32:sgpr(s64) = G_CONSTANT i64 32
; GFX1010-NEXT: %out_addr_plus_32:sgpr(p1) = G_PTR_ADD %out_addr, %cst32(s64)
; GFX1010-NEXT: [[COPY3:%[0-9]+]]:vgpr(p1) = COPY %out_addr_plus_32(p1)
; GFX1010-NEXT: G_STORE %load8_11(<4 x s32>), [[COPY3]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX1010-NEXT: %cst48:sgpr(s64) = G_CONSTANT i64 48
; GFX1010-NEXT: %out_addr_plus_48:sgpr(p1) = G_PTR_ADD %out_addr, %cst48(s64)
; GFX1010-NEXT: [[COPY4:%[0-9]+]]:vgpr(p1) = COPY %out_addr_plus_48(p1)
; GFX1010-NEXT: G_STORE %load12_15(<4 x s32>), [[COPY4]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
; GFX1010-NEXT: S_ENDPGM 0
%in_addr:_(p1) = COPY $sgpr0_sgpr1
%out_addr:_(p1) = COPY $sgpr2_sgpr3
%load:_(<16 x s32>) = G_LOAD %in_addr(p1) :: (load (<16 x s32>), align 4, addrspace 1)
%load0_3:_(<4 x s32>), %load4_7:_(<4 x s32>), %load8_11:_(<4 x s32>), %load12_15:_(<4 x s32>) = G_UNMERGE_VALUES %load(<16 x s32>)
G_STORE %load0_3(<4 x s32>), %out_addr(p1) :: (store (<4 x s32>), align 4, addrspace 1)
%cst16:_(s64) = G_CONSTANT i64 16
%out_addr_plus_16:_(p1) = G_PTR_ADD %out_addr, %cst16(s64)
G_STORE %load4_7(<4 x s32>), %out_addr_plus_16(p1) :: (store (<4 x s32>), align 4, addrspace 1)
%cst32:_(s64) = G_CONSTANT i64 32
%out_addr_plus_32:_(p1) = G_PTR_ADD %out_addr, %cst32(s64)
G_STORE %load8_11(<4 x s32>), %out_addr_plus_32(p1) :: (store (<4 x s32>), align 4, addrspace 1)
%cst48:_(s64) = G_CONSTANT i64 48
%out_addr_plus_48:_(p1) = G_PTR_ADD %out_addr, %cst48(s64)
G_STORE %load12_15(<4 x s32>), %out_addr_plus_48(p1) :: (store (<4 x s32>), align 4, addrspace 1)
S_ENDPGM 0
...
---
name: test_s_load_constant_v8i32_align1
legalized: true
tracksRegLiveness: true
body: |
bb.0.entry:
liveins: $sgpr0_sgpr1, $sgpr2_sgpr3
; GFX7-LABEL: name: test_s_load_constant_v8i32_align1
; GFX7: liveins: $sgpr0_sgpr1, $sgpr2_sgpr3
; GFX7-NEXT: {{ $}}
; GFX7-NEXT: %ptr:sgpr(p4) = COPY $sgpr0_sgpr1
; GFX7-NEXT: %out:sgpr(p1) = COPY $sgpr2_sgpr3
; GFX7-NEXT: [[LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD %ptr(p4) :: (load (<4 x s32>), align 1, addrspace 4)
; GFX7-NEXT: [[C:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 16
; GFX7-NEXT: [[PTR_ADD:%[0-9]+]]:vgpr(p4) = G_PTR_ADD %ptr, [[C]](s64)
; GFX7-NEXT: [[LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<4 x s32>) from unknown-address + 16, align 1, addrspace 4)
; GFX7-NEXT: %load:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; GFX7-NEXT: %load0_3:vgpr(<4 x s32>), %load4_7:vgpr(<4 x s32>) = G_UNMERGE_VALUES %load(<8 x s32>)
; GFX7-NEXT: G_STORE %load0_3(<4 x s32>), %out(p1) :: (store (<4 x s32>), align 32, addrspace 1)
; GFX7-NEXT: %cst_16:sgpr(s64) = G_CONSTANT i64 16
; GFX7-NEXT: %out_plus_16:sgpr(p1) = G_PTR_ADD %out, %cst_16(s64)
; GFX7-NEXT: G_STORE %load4_7(<4 x s32>), %out_plus_16(p1) :: (store (<4 x s32>), align 32, addrspace 1)
; GFX7-NEXT: S_ENDPGM 0
; GFX1010-LABEL: name: test_s_load_constant_v8i32_align1
; GFX1010: liveins: $sgpr0_sgpr1, $sgpr2_sgpr3
; GFX1010-NEXT: {{ $}}
; GFX1010-NEXT: %ptr:sgpr(p4) = COPY $sgpr0_sgpr1
; GFX1010-NEXT: %out:sgpr(p1) = COPY $sgpr2_sgpr3
; GFX1010-NEXT: [[COPY:%[0-9]+]]:vgpr(p4) = COPY %ptr(p4)
; GFX1010-NEXT: [[LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD %ptr(p4) :: (load (<4 x s32>), align 1, addrspace 4)
; GFX1010-NEXT: [[C:%[0-9]+]]:vgpr(s64) = G_CONSTANT i64 16
; GFX1010-NEXT: [[PTR_ADD:%[0-9]+]]:vgpr(p4) = G_PTR_ADD %ptr, [[C]](s64)
; GFX1010-NEXT: [[LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<4 x s32>) from unknown-address + 16, align 1, addrspace 4)
; GFX1010-NEXT: %load:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[LOAD]](<4 x s32>), [[LOAD1]](<4 x s32>)
; GFX1010-NEXT: %load0_3:vgpr(<4 x s32>), %load4_7:vgpr(<4 x s32>) = G_UNMERGE_VALUES %load(<8 x s32>)
; GFX1010-NEXT: [[COPY1:%[0-9]+]]:vgpr(p1) = COPY %out(p1)
; GFX1010-NEXT: G_STORE %load0_3(<4 x s32>), [[COPY1]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
; GFX1010-NEXT: %cst_16:sgpr(s64) = G_CONSTANT i64 16
; GFX1010-NEXT: %out_plus_16:sgpr(p1) = G_PTR_ADD %out, %cst_16(s64)
; GFX1010-NEXT: [[COPY2:%[0-9]+]]:vgpr(p1) = COPY %out_plus_16(p1)
; GFX1010-NEXT: G_STORE %load4_7(<4 x s32>), [[COPY2]](p1) :: (store (<4 x s32>), align 32, addrspace 1)
; GFX1010-NEXT: S_ENDPGM 0
%ptr:_(p4) = COPY $sgpr0_sgpr1
%out:_(p1) = COPY $sgpr2_sgpr3
%load:_(<8 x s32>) = G_LOAD %ptr(p4) :: (load (<8 x s32>), align 1, addrspace 4)
%load0_3:_(<4 x s32>), %load4_7:_(<4 x s32>) = G_UNMERGE_VALUES %load(<8 x s32>)
G_STORE %load0_3(<4 x s32>), %out(p1) :: (store (<4 x s32>), align 32, addrspace 1)
%cst_16:_(s64) = G_CONSTANT i64 16
%out_plus_16:_(p1) = G_PTR_ADD %out, %cst_16(s64)
G_STORE %load4_7(<4 x s32>), %out_plus_16(p1) :: (store (<4 x s32>), basealign 32, addrspace 1)
S_ENDPGM 0
...
|
