1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; REQUIRES: amdgpu-registered-target 3 4; RUN: opt -passes='require<globals-aa>,loop-simplify,loop-load-elim' -S %s | FileCheck %s 5 6target 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-ni:7:8" 7target triple = "amdgcn-amd-amdhsa" 8 9%struct.foo = type { %struct.pluto, i8, ptr, i32 } 10%struct.pluto = type { i32, i32, i32, ptr, i32, i32, i32 } 11%struct.wombat = type { %struct.barney } 12%struct.barney = type { <2 x float> } 13 14@global = external protected local_unnamed_addr addrspace(4) externally_initialized global %struct.foo, align 8 15@global.1 = internal unnamed_addr addrspace(3) constant [4000 x float] undef, align 16 16 17; Function Attrs: nounwind 18define protected amdgpu_kernel void @widget(i32 %arg, i32 %arg1) #0 { 19; CHECK-LABEL: @widget( 20; CHECK-NEXT: bb: 21; CHECK-NEXT: [[TMP:%.*]] = getelementptr inbounds [4000 x float], ptr addrspace(3) @global.1, i32 0, i32 [[ARG:%.*]] 22; CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr addrspace(4) getelementptr inbounds (%struct.foo, ptr addrspace(4) @global, i64 0, i32 0, i32 3), align 8 23; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_WOMBAT:%.*]], ptr [[TMP2]], i64 undef, i32 0 24; CHECK-NEXT: br label [[BB5:%.*]] 25; CHECK: bb5.loopexit: 26; CHECK-NEXT: br label [[BB5]] 27; CHECK: bb5: 28; CHECK-NEXT: br label [[BB6:%.*]] 29; CHECK: bb6: 30; CHECK-NEXT: [[TMP7:%.*]] = phi i32 [ undef, [[BB5]] ], [ [[TMP19:%.*]], [[BB6]] ] 31; CHECK-NEXT: [[TMP8:%.*]] = mul nsw i32 [[TMP7]], undef 32; CHECK-NEXT: [[TMP9:%.*]] = add i32 [[TMP8]], undef 33; CHECK-NEXT: [[TMP10:%.*]] = sext i32 [[TMP9]] to i64 34; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_WOMBAT]], ptr [[TMP2]], i64 [[TMP10]] 35; CHECK-NEXT: [[TMP13:%.*]] = load i64, ptr [[TMP11]], align 8 36; CHECK-NEXT: [[TMP14:%.*]] = srem i32 1, [[ARG1:%.*]] 37; CHECK-NEXT: [[TMP15:%.*]] = add nuw nsw i32 [[TMP14]], 1 38; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4000 x float], ptr addrspace(3) @global.1, i32 0, i32 [[TMP15]] 39; CHECK-NEXT: [[TMP17:%.*]] = load float, ptr addrspace(3) [[TMP16]], align 4 40; CHECK-NEXT: [[TMP18:%.*]] = load float, ptr addrspace(3) [[TMP]], align 4 41; CHECK-NEXT: store i64 [[TMP13]], ptr [[TMP3]], align 8 42; CHECK-NEXT: [[TMP19]] = add nsw i32 [[TMP7]], 1 43; CHECK-NEXT: [[TMP20:%.*]] = icmp slt i32 [[TMP7]], 3 44; CHECK-NEXT: br i1 [[TMP20]], label [[BB6]], label [[BB5_LOOPEXIT:%.*]] 45; 46bb: 47 %tmp = getelementptr inbounds [4000 x float], ptr addrspace(3) @global.1, i32 0, i32 %arg 48 %tmp2 = load ptr, ptr addrspace(4) getelementptr inbounds (%struct.foo, ptr addrspace(4) @global, i64 0, i32 0, i32 3), align 8 49 %tmp3 = getelementptr inbounds %struct.wombat, ptr %tmp2, i64 undef, i32 0 50 br label %bb5 51 52bb5: ; preds = %bb6, %bb 53 br label %bb6 54 55bb6: ; preds = %bb6, %bb5 56 %tmp7 = phi i32 [ undef, %bb5 ], [ %tmp19, %bb6 ] 57 %tmp8 = mul nsw i32 %tmp7, undef 58 %tmp9 = add i32 %tmp8, undef 59 %tmp10 = sext i32 %tmp9 to i64 60 %tmp11 = getelementptr inbounds %struct.wombat, ptr %tmp2, i64 %tmp10 61 %tmp13 = load i64, ptr %tmp11, align 8 62 %tmp14 = srem i32 1, %arg1 63 %tmp15 = add nuw nsw i32 %tmp14, 1 64 %tmp16 = getelementptr inbounds [4000 x float], ptr addrspace(3) @global.1, i32 0, i32 %tmp15 65 %tmp17 = load float, ptr addrspace(3) %tmp16, align 4 66 %tmp18 = load float, ptr addrspace(3) %tmp, align 4 67 store i64 %tmp13, ptr %tmp3, align 8 68 %tmp19 = add nsw i32 %tmp7, 1 69 %tmp20 = icmp slt i32 %tmp7, 3 70 br i1 %tmp20, label %bb6, label %bb5 71} 72 73attributes #0 = { nounwind } 74