1; RUN: opt -S -mtriple=amdgcn-- -amdgpu-lower-module-lds < %s | FileCheck %s
2; RUN: opt -S -mtriple=amdgcn-- -passes=amdgpu-lower-module-lds < %s | FileCheck %s
3
4; CHECK: %llvm.amdgcn.kernel.k2.lds.t = type { i32 }
5
6@lds.1 = internal unnamed_addr addrspace(3) global [2 x i8] undef, align 1
7
8; Use constant from different kernels
9;.
10; CHECK: @lds.1 = internal unnamed_addr addrspace(3) global [2 x i8] undef, align 1
11; CHECK: @llvm.amdgcn.kernel.k2.lds = internal addrspace(3) global %llvm.amdgcn.kernel.k2.lds.t undef, align 4
12;.
13define amdgpu_kernel void @k0(i64 %x) {
14; CHECK-LABEL: @k0(
15; CHECK-NEXT:    %ptr = getelementptr inbounds i8, i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([2 x i8], [2 x i8] addrspace(3)* @lds.1, i32 0, i32 0) to i8*), i64 %x
16; CHECK-NEXT:    store i8 1, i8* %ptr, align 1
17; CHECK-NEXT:    ret void
18;
19  %ptr = getelementptr inbounds i8, i8* addrspacecast ([2 x i8] addrspace(3)* @lds.1 to i8*), i64 %x
20  store i8 1, i8 addrspace(0)* %ptr, align 1
21  ret void
22}
23
24define amdgpu_kernel void @k1(i64 %x) {
25; CHECK-LABEL: @k1(
26; CHECK-NEXT:    %ptr = getelementptr inbounds i8, i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([2 x i8], [2 x i8] addrspace(3)* @lds.1, i32 0, i32 0) to i8*), i64 %x
27; CHECK-NEXT:    store i8 1, i8* %ptr, align 1
28; CHECK-NEXT:    ret void
29;
30  %ptr = getelementptr inbounds i8, i8* addrspacecast ([2 x i8] addrspace(3)* @lds.1 to i8*), i64 %x
31  store i8 1, i8 addrspace(0)* %ptr, align 1
32  ret void
33}
34
35@lds.2 = internal unnamed_addr addrspace(3) global i32 undef, align 4
36
37; Use constant twice from the same kernel
38define amdgpu_kernel void @k2(i64 %x) {
39; CHECK-LABEL: @k2(
40; CHECK-NEXT:    %ptr1 = bitcast i32 addrspace(3)* getelementptr inbounds (%llvm.amdgcn.kernel.k2.lds.t, %llvm.amdgcn.kernel.k2.lds.t addrspace(3)* @llvm.amdgcn.kernel.k2.lds, i32 0, i32 0) to i8 addrspace(3)*
41; CHECK-NEXT:    store i8 1, i8 addrspace(3)* %ptr1, align 4
42; CHECK-NEXT:    %ptr2 = bitcast i32 addrspace(3)* getelementptr inbounds (%llvm.amdgcn.kernel.k2.lds.t, %llvm.amdgcn.kernel.k2.lds.t addrspace(3)* @llvm.amdgcn.kernel.k2.lds, i32 0, i32 0) to i8 addrspace(3)*
43; CHECK-NEXT:    store i8 2, i8 addrspace(3)* %ptr2, align 4
44; CHECK-NEXT:    ret void
45;
46  %ptr1 = bitcast i32 addrspace(3)* @lds.2 to i8 addrspace(3)*
47  store i8 1, i8 addrspace(3)* %ptr1, align 4
48  %ptr2 = bitcast i32 addrspace(3)* @lds.2 to i8 addrspace(3)*
49  store i8 2, i8 addrspace(3)* %ptr2, align 4
50  ret void
51}
52
53@lds.3 = internal unnamed_addr addrspace(3) global [32 x i8] undef, align 1
54
55; Use constant twice from the same kernel but a different other constant.
56define amdgpu_kernel void @k3(i64 %x) {
57; CHECK-LABEL: @k3(
58; CHECK-NEXT:    %ptr1 = addrspacecast i64 addrspace(3)* bitcast (i8 addrspace(3)* getelementptr inbounds (%llvm.amdgcn.kernel.k3.lds.t, %llvm.amdgcn.kernel.k3.lds.t addrspace(3)* @llvm.amdgcn.kernel.k3.lds, i32 0, i32 0, i32 16) to i64 addrspace(3)*) to i64*
59; CHECK-NEXT:    store i64 1, i64* %ptr1, align 1
60; CHECK-NEXT:    %ptr2 = addrspacecast i64 addrspace(3)* bitcast (i8 addrspace(3)* getelementptr inbounds (%llvm.amdgcn.kernel.k3.lds.t, %llvm.amdgcn.kernel.k3.lds.t addrspace(3)* @llvm.amdgcn.kernel.k3.lds, i32 0, i32 0, i32 24) to i64 addrspace(3)*) to i64*
61; CHECK-NEXT:    store i64 2, i64* %ptr2, align 1
62;
63  %ptr1 = addrspacecast i64 addrspace(3)* bitcast (i8 addrspace(3)* getelementptr inbounds ([32 x i8], [32 x i8] addrspace(3)* @lds.3, i32 0, i32 16) to i64 addrspace(3)*) to i64*
64  store i64 1, i64* %ptr1, align 1
65  %ptr2 = addrspacecast i64 addrspace(3)* bitcast (i8 addrspace(3)* getelementptr inbounds ([32 x i8], [32 x i8] addrspace(3)* @lds.3, i32 0, i32 24) to i64 addrspace(3)*) to i64*
66  store i64 2, i64* %ptr2, align 1
67  ret void
68}
69