1 ; RUN: opt -instnamer -O1 -mtriple=amdgcn-- -S -o - %s | FileCheck -check-prefixes=GCN,O1 %s
2 ; RUN: opt -instnamer -O2 -mtriple=amdgcn-- -S -o - %s | FileCheck -check-prefixes=GCN,O2 %s
3 ; RUN: opt -instnamer -O3 -mtriple=amdgcn-- -S -o - %s | FileCheck -check-prefixes=GCN,O3 %s
4 target datalayout = "A5"
10 ; GCN-COUNT-27: = load
12 define protected amdgpu_kernel void @t0(i32 addrspace(1)* %p.coerce) #0 {
14 %p = alloca i32*, align 8, addrspace(5)
15 %p.ascast = addrspacecast i32* addrspace(5)* %p to i32**
16 %p.addr = alloca i32*, align 8, addrspace(5)
17 %p.addr.ascast = addrspacecast i32* addrspace(5)* %p.addr to i32**
18 %t = alloca [27 x i32], align 16, addrspace(5)
19 %t.ascast = addrspacecast [27 x i32] addrspace(5)* %t to [27 x i32]*
20 %sum = alloca i32, align 4, addrspace(5)
21 %sum.ascast = addrspacecast i32 addrspace(5)* %sum to i32*
22 %i = alloca i32, align 4, addrspace(5)
23 %i.ascast = addrspacecast i32 addrspace(5)* %i to i32*
24 %cleanup.dest.slot = alloca i32, align 4, addrspace(5)
25 %0 = addrspacecast i32 addrspace(1)* %p.coerce to i32*
26 store i32* %0, i32** %p.ascast, align 8
27 %p1 = load i32*, i32** %p.ascast, align 8
28 store i32* %p1, i32** %p.addr.ascast, align 8
29 %1 = bitcast [27 x i32] addrspace(5)* %t to i8 addrspace(5)*
30 call void @llvm.lifetime.start.p5i8(i64 48, i8 addrspace(5)* %1)
31 %arraydecay = getelementptr inbounds [27 x i32], [27 x i32]* %t.ascast, i64 0, i64 0
32 %2 = load i32*, i32** %p.addr.ascast, align 8
33 call void @copy(i32* %arraydecay, i32* %2, i32 27)
34 %3 = bitcast i32 addrspace(5)* %sum to i8 addrspace(5)*
35 call void @llvm.lifetime.start.p5i8(i64 4, i8 addrspace(5)* %3)
36 store i32 0, i32* %sum.ascast, align 4
37 %4 = bitcast i32 addrspace(5)* %i to i8 addrspace(5)*
38 call void @llvm.lifetime.start.p5i8(i64 4, i8 addrspace(5)* %4)
39 store i32 0, i32* %i.ascast, align 4
42 for.cond: ; preds = %for.inc, %entry
43 %5 = load i32, i32* %i.ascast, align 4
44 %cmp = icmp slt i32 %5, 27
45 br i1 %cmp, label %for.body, label %for.cond.cleanup
47 for.cond.cleanup: ; preds = %for.cond
48 %6 = bitcast i32 addrspace(5)* %i to i8 addrspace(5)*
49 call void @llvm.lifetime.end.p5i8(i64 4, i8 addrspace(5)* %6)
52 for.body: ; preds = %for.cond
53 %7 = load i32, i32* %i.ascast, align 4
54 %idxprom = sext i32 %7 to i64
55 %arrayidx = getelementptr inbounds [27 x i32], [27 x i32]* %t.ascast, i64 0, i64 %idxprom
56 %8 = load i32, i32* %arrayidx, align 4
57 %9 = load i32, i32* %sum.ascast, align 4
58 %add = add nsw i32 %9, %8
59 store i32 %add, i32* %sum.ascast, align 4
62 for.inc: ; preds = %for.body
63 %10 = load i32, i32* %i.ascast, align 4
64 %inc = add nsw i32 %10, 1
65 store i32 %inc, i32* %i.ascast, align 4
68 for.end: ; preds = %for.cond.cleanup
69 %11 = load i32, i32* %sum.ascast, align 4
70 %12 = load i32*, i32** %p.addr.ascast, align 8
71 store i32 %11, i32* %12, align 4
72 %13 = bitcast i32 addrspace(5)* %sum to i8 addrspace(5)*
73 call void @llvm.lifetime.end.p5i8(i64 4, i8 addrspace(5)* %13)
74 %14 = bitcast [27 x i32] addrspace(5)* %t to i8 addrspace(5)*
75 call void @llvm.lifetime.end.p5i8(i64 48, i8 addrspace(5)* %14)
79 define internal void @copy(i32* %d, i32* %s, i32 %N) {
81 %d8 = bitcast i32* %d to i8*
82 %s8 = bitcast i32* %s to i8*
84 tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %d8, i8* %s8, i32 %N8, i1 false)
88 declare void @llvm.lifetime.start.p5i8(i64 immarg, i8 addrspace(5)* nocapture)
89 declare void @llvm.lifetime.end.p5i8(i64 immarg, i8 addrspace(5)* nocapture)
90 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i1)