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
|
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=amdgpu-always-inline -amdgpu-enable-lower-module-lds=false %s | FileCheck --check-prefix=ALL %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -passes=amdgpu-always-inline -amdgpu-enable-lower-module-lds=false %s | FileCheck --check-prefix=ALL %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-stress-function-calls -passes=amdgpu-always-inline -amdgpu-enable-lower-module-lds=false %s | FileCheck --check-prefix=ALL %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-stress-function-calls -passes=amdgpu-always-inline -amdgpu-enable-lower-module-lds=false %s | FileCheck --check-prefix=ALL %s
target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-p7:160:256:256:32-p8:128:128-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"
@lds0 = addrspace(3) global i32 undef, align 4
@lds1 = addrspace(3) global [512 x i32] undef, align 4
@nested.lds.address = addrspace(1) global ptr addrspace(3) @lds0, align 4
@gds0 = addrspace(2) global i32 undef, align 4
@alias.lds0 = alias i32, ptr addrspace(3) @lds0
@lds.cycle = addrspace(3) global i32 ptrtoint (ptr addrspace(3) @lds.cycle to i32), align 4
; ALL-LABEL: define i32 @load_lds_simple() #0 {
define i32 @load_lds_simple() {
%load = load i32, ptr addrspace(3) @lds0, align 4
ret i32 %load
}
; ALL-LABEL: define i32 @load_gds_simple() #0 {
define i32 @load_gds_simple() {
%load = load i32, ptr addrspace(2) @gds0, align 4
ret i32 %load
}
; ALL-LABEL: define i32 @load_lds_const_gep() #0 {
define i32 @load_lds_const_gep() {
%load = load i32, ptr addrspace(3) getelementptr inbounds ([512 x i32], ptr addrspace(3) @lds1, i64 0, i64 4), align 4
ret i32 %load
}
; ALL-LABEL: define i32 @load_lds_var_gep(i32 %idx) #0 {
define i32 @load_lds_var_gep(i32 %idx) {
%gep = getelementptr inbounds [512 x i32], ptr addrspace(3) @lds1, i32 0, i32 %idx
%load = load i32, ptr addrspace(3) %gep, align 4
ret i32 %load
}
; ALL-LABEL: define ptr addrspace(3) @load_nested_address(i32 %idx) #0 {
define ptr addrspace(3) @load_nested_address(i32 %idx) {
%load = load ptr addrspace(3), ptr addrspace(1) @nested.lds.address, align 4
ret ptr addrspace(3) %load
}
; ALL-LABEL: define i32 @load_lds_alias() #0 {
define i32 @load_lds_alias() {
%load = load i32, ptr addrspace(3) @alias.lds0, align 4
ret i32 %load
}
; ALL-LABEL: define i32 @load_lds_cycle() #0 {
define i32 @load_lds_cycle() {
%load = load i32, ptr addrspace(3) @lds.cycle, align 4
ret i32 %load
}
; ALL-LABEL: define i1 @icmp_lds_address() #0 {
define i1 @icmp_lds_address() {
%cmp = icmp eq ptr addrspace(3) @lds0, null
ret i1 %cmp
}
; ALL-LABEL: define i32 @transitive_call() #0 {
define i32 @transitive_call() {
%call = call i32 @load_lds_simple()
ret i32 %call
}
; ALL-LABEL: define i32 @recursive_call_lds(i32 %arg0) #0 {
define i32 @recursive_call_lds(i32 %arg0) {
%load = load i32, ptr addrspace(3) @lds0, align 4
%add = add i32 %arg0, %load
%call = call i32 @recursive_call_lds(i32 %add)
ret i32 %call
}
; Test we don't break the IR and have both alwaysinline and noinline
; FIXME: We should really not override noinline.
; ALL-LABEL: define i32 @load_lds_simple_noinline() #0 {
define i32 @load_lds_simple_noinline() noinline {
%load = load i32, ptr addrspace(3) @lds0, align 4
ret i32 %load
}
; ALL-LABEL: define i32 @recursive_call_lds_noinline(i32 %arg0) #0 {
define i32 @recursive_call_lds_noinline(i32 %arg0) noinline {
%load = load i32, ptr addrspace(3) @lds0, align 4
%add = add i32 %arg0, %load
%call = call i32 @recursive_call_lds(i32 %add)
ret i32 %call
}
; ALL: attributes #0 = { alwaysinline }
|
