1 ; RUN: llc -march=amdgcn -mcpu=bonaire -print-lsr-output < %s 2>&1 | FileCheck %s
3 ; Test various conditions where OptimizeLoopTermCond doesn't look at a
4 ; memory instruction use and fails to find the address space.
6 target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-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"
8 ; CHECK-LABEL: @local_cmp_user(
10 ; CHECK: %lsr.iv1 = phi i32 [ %lsr.iv.next2, %bb ], [ 2, %entry ]
11 ; CHECK: %lsr.iv = phi i32 [ %lsr.iv.next, %bb ], [ %{{[0-9]+}}, %entry ]
12 ; CHECK: %lsr.iv.next = add i32 %lsr.iv, -1
13 ; CHECK: %lsr.iv.next2 = add i32 %lsr.iv1, -2
17 ; CHECK: inttoptr i32 %lsr.iv.next2 to i8 addrspace(3)*
18 ; CHECK: %c1 = icmp ne i8 addrspace(3)*
19 define amdgpu_kernel void @local_cmp_user(i32 %arg0) nounwind {
24 %i = phi i32 [ 0, %entry ], [ %i.next, %bb ]
26 %c0 = icmp eq i32 %i, %arg0
27 br i1 %c0, label %bb13, label %bb
30 %t = load i8 addrspace(3)*, i8 addrspace(3)* addrspace(3)* undef
31 %p = getelementptr i8, i8 addrspace(3)* %t, i32 %ii
32 %c1 = icmp ne i8 addrspace(3)* %p, null
33 %i.next = add i32 %i, 1
34 br i1 %c1, label %bb11, label %bb13
40 ; CHECK-LABEL: @global_cmp_user(
41 ; CHECK: %lsr.iv1 = phi i64
42 ; CHECK: %lsr.iv = phi i64
43 ; CHECK: %lsr.iv.next = add i64 %lsr.iv, -1
44 ; CHECK: %lsr.iv.next2 = add i64 %lsr.iv1, -2
48 ; CHECK: inttoptr i64 %lsr.iv.next2 to i8 addrspace(1)*
49 ; CHECK: icmp ne i8 addrspace(1)* %t
50 define amdgpu_kernel void @global_cmp_user(i64 %arg0) nounwind {
55 %i = phi i64 [ 0, %entry ], [ %i.next, %bb ]
57 %c0 = icmp eq i64 %i, %arg0
58 br i1 %c0, label %bb13, label %bb
61 %t = load i8 addrspace(1)*, i8 addrspace(1)* addrspace(1)* undef
62 %p = getelementptr i8, i8 addrspace(1)* %t, i64 %ii
63 %c1 = icmp ne i8 addrspace(1)* %p, null
64 %i.next = add i64 %i, 1
65 br i1 %c1, label %bb11, label %bb13
71 ; CHECK-LABEL: @global_gep_user(
72 ; CHECK: %lsr.iv1 = phi i32 [ %lsr.iv.next2, %bb ], [ 0, %entry ]
73 ; CHECK: %lsr.iv = phi i32 [ %lsr.iv.next, %bb ], [ %{{[0-9]+}}, %entry ]
74 ; CHECK: %lsr.iv.next = add i32 %lsr.iv, -1
75 ; CHECK: %lsr.iv.next2 = add i32 %lsr.iv1, 2
79 ; CHECK: %idxprom = sext i32 %lsr.iv1 to i64
80 ; CHECK: getelementptr i8, i8 addrspace(1)* %t, i64 %idxprom
81 define amdgpu_kernel void @global_gep_user(i32 %arg0) nounwind {
86 %i = phi i32 [ 0, %entry ], [ %i.next, %bb ]
88 %c0 = icmp eq i32 %i, %arg0
89 br i1 %c0, label %bb13, label %bb
92 %t = load i8 addrspace(1)*, i8 addrspace(1)* addrspace(1)* undef
93 %p = getelementptr i8, i8 addrspace(1)* %t, i32 %ii
94 %c1 = icmp ne i8 addrspace(1)* %p, null
95 %i.next = add i32 %i, 1
96 br i1 %c1, label %bb11, label %bb13
102 ; CHECK-LABEL: @global_sext_scale_user(
103 ; CHECK: %lsr.iv1 = phi i32 [ %lsr.iv.next2, %bb ], [ 0, %entry ]
104 ; CHECK: %lsr.iv = phi i32 [ %lsr.iv.next, %bb ], [ %{{[0-9]+}}, %entry ]
105 ; CHECK: %lsr.iv.next = add i32 %lsr.iv, -1
106 ; CHECK: %lsr.iv.next2 = add i32 %lsr.iv1, 2
110 ; CHECK: %p = getelementptr i8, i8 addrspace(1)* %t, i64 %ii.ext
111 define amdgpu_kernel void @global_sext_scale_user(i32 %arg0) nounwind {
116 %i = phi i32 [ 0, %entry ], [ %i.next, %bb ]
118 %ii.ext = sext i32 %ii to i64
119 %c0 = icmp eq i32 %i, %arg0
120 br i1 %c0, label %bb13, label %bb
123 %t = load i8 addrspace(1)*, i8 addrspace(1)* addrspace(1)* undef
124 %p = getelementptr i8, i8 addrspace(1)* %t, i64 %ii.ext
125 %c1 = icmp ne i8 addrspace(1)* %p, null
126 %i.next = add i32 %i, 1
127 br i1 %c1, label %bb11, label %bb13