1; RUN: opt < %s -loop-unroll -codegenprepare -S -mtriple=x86_64 | FileCheck %s 2 3; This test is a worst-case scenario for bitreversal/byteswap detection. 4; After loop unrolling (the unrolled loop is unreadably large so it has been kept 5; rolled here), we have a binary tree of OR operands (as bitreversal detection 6; looks straight through shifts): 7; 8; OR 9; | \ 10; | LSHR 11; | / 12; OR 13; | \ 14; | LSHR 15; | / 16; OR 17; 18; This results in exponential runtime. The loop here is 32 iterations which will 19; totally hang if we don't deal with this case cleverly. 20 21@b = common global i32 0, align 4 22 23; CHECK: define i32 @fn1 24define i32 @fn1() #0 { 25entry: 26 %b.promoted = load i32, i32* @b, align 4, !tbaa !2 27 br label %for.body 28 29for.body: ; preds = %for.body, %entry 30 %or4 = phi i32 [ %b.promoted, %entry ], [ %or, %for.body ] 31 %i.03 = phi i32 [ 0, %entry ], [ %inc, %for.body ] 32 %shr = lshr i32 %or4, 1 33 %or = or i32 %shr, %or4 34 %inc = add nuw nsw i32 %i.03, 1 35 %exitcond = icmp eq i32 %inc, 32 36 br i1 %exitcond, label %for.end, label %for.body 37 38for.end: ; preds = %for.body 39 store i32 %or, i32* @b, align 4, !tbaa !2 40 ret i32 undef 41} 42 43attributes #0 = { norecurse nounwind ssp uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="core2" "target-features"="+cx16,+fxsr,+mmx,+sse,+sse2,+sse3,+ssse3" "unsafe-fp-math"="false" "use-soft-float"="false" } 44 45!llvm.module.flags = !{!0} 46!llvm.ident = !{!1} 47 48!0 = !{i32 1, !"PIC Level", i32 2} 49!1 = !{!"clang version 3.8.0"} 50!2 = !{!3, !3, i64 0} 51!3 = !{!"int", !4, i64 0} 52!4 = !{!"omnipotent char", !5, i64 0} 53!5 = !{!"Simple C/C++ TBAA"} 54