1*9880d681SAndroid Build Coastguard Worker; RUN: llc -O2 -march=hexagon < %s | FileCheck %s 2*9880d681SAndroid Build Coastguard Worker; CHECK: cmp.eq(r{{[0-9]+}}, #0) 3*9880d681SAndroid Build Coastguard Worker; Check that the result of the builtin is not stored directly, i.e. that 4*9880d681SAndroid Build Coastguard Worker; there is an instruction that converts it to {0,1} from {0,-1}. Right now 5*9880d681SAndroid Build Coastguard Worker; the instruction is "r4 = !cmp.eq(r0, #0)". 6*9880d681SAndroid Build Coastguard Worker 7*9880d681SAndroid Build Coastguard Worker@var = common global i32 0, align 4 8*9880d681SAndroid Build Coastguard Workerdeclare i32 @llvm.hexagon.C2.cmpgtup(i64,i64) nounwind 9*9880d681SAndroid Build Coastguard Worker 10*9880d681SAndroid Build Coastguard Workerdefine void @foo(i64 %a98, i64 %a100) nounwind { 11*9880d681SAndroid Build Coastguard Workerentry: 12*9880d681SAndroid Build Coastguard Worker %a101 = tail call i32 @llvm.hexagon.C2.cmpgtup(i64 %a98, i64 %a100) 13*9880d681SAndroid Build Coastguard Worker %tobool250 = icmp eq i32 %a101, 0 14*9880d681SAndroid Build Coastguard Worker %a102 = zext i1 %tobool250 to i8 15*9880d681SAndroid Build Coastguard Worker %detected.0 = xor i8 %a102, 1 16*9880d681SAndroid Build Coastguard Worker %conv253 = zext i8 %detected.0 to i32 17*9880d681SAndroid Build Coastguard Worker store i32 %conv253, i32* @var, align 4 18*9880d681SAndroid Build Coastguard Worker ret void 19*9880d681SAndroid Build Coastguard Worker} 20