1 ; RUN: opt -basicaa -load-combine -instcombine -S < %s | FileCheck %s 2 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 3 target triple = "x86_64-unknown-linux-gnu" 4 5 declare void @llvm.assume(i1) nounwind 6 7 ; 'load' before the 'call' gets optimized: 8 define i64 @test1(i32* nocapture readonly %a, i1 %b) { 9 ; CHECK-LABEL: @test1 10 11 ; CHECK-DAG: load i64, i64* %1, align 4 12 ; CHECK-DAG: tail call void @llvm.assume(i1 %b) 13 ; CHECK: ret i64 14 15 %load1 = load i32, i32* %a, align 4 16 %conv = zext i32 %load1 to i64 17 %arrayidx1 = getelementptr inbounds i32, i32* %a, i64 1 18 %load2 = load i32, i32* %arrayidx1, align 4 19 tail call void @llvm.assume(i1 %b) 20 %conv2 = zext i32 %load2 to i64 21 %shl = shl nuw i64 %conv2, 32 22 %add = or i64 %shl, %conv 23 ret i64 %add 24 } 25 26 ; 'call' before the 'load' doesn't get optimized: 27 define i64 @test2(i32* nocapture readonly %a, i1 %b) { 28 ; CHECK-LABEL: @test2 29 30 ; CHECK-DAG: load i64, i64* %1, align 4 31 ; CHECK-DAG: tail call void @llvm.assume(i1 %b) 32 ; CHECK: ret i64 33 34 %load1 = load i32, i32* %a, align 4 35 %conv = zext i32 %load1 to i64 36 %arrayidx1 = getelementptr inbounds i32, i32* %a, i64 1 37 tail call void @llvm.assume(i1 %b) 38 %load2 = load i32, i32* %arrayidx1, align 4 39 %conv2 = zext i32 %load2 to i64 40 %shl = shl nuw i64 %conv2, 32 41 %add = or i64 %shl, %conv 42 ret i64 %add 43 } 44 45
