1 ; RUN: llc -mtriple=amdgcn--amdhsa -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s 2 3 declare i32 @llvm.amdgcn.workitem.id.x() #0 4 ; GCN-LABEL: {{^}}convergent_inlineasm: 5 ; GCN: BB#0: 6 ; GCN: v_cmp_ne_i32_e64 7 ; GCN: BB#1: 8 define void @convergent_inlineasm(i64 addrspace(1)* nocapture %arg) { 9 bb: 10 %tmp = call i32 @llvm.amdgcn.workitem.id.x() 11 %tmp1 = tail call i64 asm "v_cmp_ne_i32_e64 $0, 0, $1", "=s,v"(i32 1) #1 12 %tmp2 = icmp eq i32 %tmp, 8 13 br i1 %tmp2, label %bb3, label %bb5 14 15 bb3: ; preds = %bb 16 %tmp4 = getelementptr i64, i64 addrspace(1)* %arg, i32 %tmp 17 store i64 %tmp1, i64 addrspace(1)* %arg, align 8 18 br label %bb5 19 20 bb5: ; preds = %bb3, %bb 21 ret void 22 } 23 24 ; GCN-LABEL: {{^}}nonconvergent_inlineasm: 25 ; GCN: BB#1: 26 ; GCN: v_cmp_ne_i32_e64 27 ; GCN: BB1_2: 28 define void @nonconvergent_inlineasm(i64 addrspace(1)* nocapture %arg) { 29 bb: 30 %tmp = call i32 @llvm.amdgcn.workitem.id.x() 31 %tmp1 = tail call i64 asm "v_cmp_ne_i32_e64 $0, 0, $1", "=s,v"(i32 1) 32 %tmp2 = icmp eq i32 %tmp, 8 33 br i1 %tmp2, label %bb3, label %bb5 34 35 bb3: ; preds = %bb 36 %tmp4 = getelementptr i64, i64 addrspace(1)* %arg, i32 %tmp 37 store i64 %tmp1, i64 addrspace(1)* %arg, align 8 38 br label %bb5 39 40 bb5: ; preds = %bb3, %bb 41 ret void 42 } 43 44 attributes #0 = { nounwind readnone } 45 attributes #1 = { convergent nounwind readnone } 46
