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
|
; FIXME: Need to add support for mubuf stores to enable this on SI.
; XUN: llc < %s -march=amdgcn -mcpu=tahiti -show-mc-encoding -verify-machineinstrs -global-isel | FileCheck --check-prefix=SI --check-prefix=GCN --check-prefix=SIVI %s
; RUN: llc < %s -march=amdgcn -mcpu=bonaire -show-mc-encoding -verify-machineinstrs -global-isel | FileCheck --check-prefix=CI --check-prefix=GCN --check-prefix=SICIVI %s
; RUN: llc < %s -march=amdgcn -mcpu=tonga -show-mc-encoding -verify-machineinstrs -global-isel | FileCheck --check-prefix=VI --check-prefix=GCN --check-prefix=SIVI --check-prefix=SICIVI %s
; RUN: llc -march=amdgcn -mcpu=gfx900 -show-mc-encoding -verify-machineinstrs -global-isel < %s | FileCheck -check-prefix=GFX9 -check-prefix=GFX9_10 -check-prefix=GCN -check-prefix=VIGFX9_10 -check-prefix=SIVIGFX9_10 %s
; RUN: llc -march=amdgcn -mcpu=gfx1010 -show-mc-encoding -verify-machineinstrs -global-isel < %s | FileCheck -check-prefix=GFX10 -check-prefix=GFX9_10 -check-prefix=GCN -check-prefix=VIGFX9_10 -check-prefix=SIVIGFX9_10 %s
; SMRD load with an immediate offset.
; GCN-LABEL: {{^}}smrd0:
; SICI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x1 ; encoding: [0x01
; VIGFX9_10: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x4
define amdgpu_kernel void @smrd0(i32 addrspace(4)* %ptr) {
entry:
%0 = getelementptr i32, i32 addrspace(4)* %ptr, i64 1
%1 = load i32, i32 addrspace(4)* %0
store i32 %1, i32 addrspace(1)* undef
ret void
}
; SMRD load with the largest possible immediate offset.
; GCN-LABEL: {{^}}smrd1:
; SICI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0xff ; encoding: [0xff,0x{{[0-9]+[137]}}
; VIGFX9_10: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x3fc
define amdgpu_kernel void @smrd1(i32 addrspace(4)* %ptr) {
entry:
%0 = getelementptr i32, i32 addrspace(4)* %ptr, i64 255
%1 = load i32, i32 addrspace(4)* %0
store i32 %1, i32 addrspace(1)* undef
ret void
}
; SMRD load with an offset greater than the largest possible immediate.
; GCN-LABEL: {{^}}smrd2:
; SI: s_movk_i32 s[[OFFSET:[0-9]]], 0x400
; SI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], s[[OFFSET]] ; encoding: [0x0[[OFFSET]]
; CI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x100
; VIGFX9_10: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x400
; GCN: s_endpgm
define amdgpu_kernel void @smrd2(i32 addrspace(4)* %ptr) {
entry:
%0 = getelementptr i32, i32 addrspace(4)* %ptr, i64 256
%1 = load i32, i32 addrspace(4)* %0
store i32 %1, i32 addrspace(1)* undef
ret void
}
; SMRD load with a 64-bit offset
; GCN-LABEL: {{^}}smrd3:
; FIXME: There are too many copies here because we don't fold immediates
; through REG_SEQUENCE
; XSI: s_load_dwordx2 s[{{[0-9]:[0-9]}}], s[{{[0-9]:[0-9]}}], 0xb ; encoding: [0x0b
; TODO: Add VI checks
; XGCN: s_endpgm
define amdgpu_kernel void @smrd3(i32 addrspace(4)* %ptr) {
entry:
%0 = getelementptr i32, i32 addrspace(4)* %ptr, i64 4294967296 ; 2 ^ 32
%1 = load i32, i32 addrspace(4)* %0
store i32 %1, i32 addrspace(1)* undef
ret void
}
; SMRD load with the largest possible immediate offset on VI
; GCN-LABEL: {{^}}smrd4:
; SI: s_mov_b32 [[OFFSET:s[0-9]+]], 0xffffc
; SI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], [[OFFSET]]
; CI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x3ffff
; VI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0xffffc
; GFX9_10: s_mov_b32 [[OFFSET:s[0-9]+]], 0xffffc
; GFX9_10: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], [[OFFSET]]
define amdgpu_kernel void @smrd4(i32 addrspace(4)* %ptr) {
entry:
%0 = getelementptr i32, i32 addrspace(4)* %ptr, i64 262143
%1 = load i32, i32 addrspace(4)* %0
store i32 %1, i32 addrspace(1)* undef
ret void
}
; SMRD load with an offset greater than the largest possible immediate on VI
; GCN-LABEL: {{^}}smrd5:
; SIVIGFX9_10: s_mov_b32 [[OFFSET:s[0-9]+]], 0x100000
; SIVIGFX9_10: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], [[OFFSET]]
; CI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x40000
; GCN: s_endpgm
define amdgpu_kernel void @smrd5(i32 addrspace(4)* %ptr) {
entry:
%0 = getelementptr i32, i32 addrspace(4)* %ptr, i64 262144
%1 = load i32, i32 addrspace(4)* %0
store i32 %1, i32 addrspace(1)* undef
ret void
}
; GFX9_10 can use a signed immediate byte offset
; GCN-LABEL: {{^}}smrd6:
; SICIVI: s_add_u32 s{{[0-9]}}, s{{[0-9]}}, -4
; SICIVI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x0
; GFX9_10: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], -0x4
define amdgpu_kernel void @smrd6(i32 addrspace(1)* %out, i32 addrspace(4)* %ptr) #0 {
entry:
%tmp = getelementptr i32, i32 addrspace(4)* %ptr, i64 -1
%tmp1 = load i32, i32 addrspace(4)* %tmp
store i32 %tmp1, i32 addrspace(1)* %out
ret void
}
; Don't use a negative SGPR offset
; GCN-LABEL: {{^}}smrd7:
; GCN: s_add_u32 s{{[0-9]}}, s{{[0-9]}}, 0xffe00000
; SICIVI: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x0
; GFX9_10: s_load_dword s{{[0-9]}}, s[{{[0-9]:[0-9]}}], 0x0
define amdgpu_kernel void @smrd7(i32 addrspace(1)* %out, i32 addrspace(4)* %ptr) #0 {
entry:
%tmp = getelementptr i32, i32 addrspace(4)* %ptr, i64 -524288
%tmp1 = load i32, i32 addrspace(4)* %tmp
store i32 %tmp1, i32 addrspace(1)* %out
ret void
}
|
