https://bugs.llvm.org/show_bug.cgi?id=44419
Bug ID: 44419
Summary: LLVM 10 regression in comparison of gep inbounds
difference
Product: libraries
Version: trunk
Hardware: All
OS: All
Status: NEW
Severity: enhancement
Priority: P
Component: Scalar Optimizations
Assignee: unassignedb...@nondot.org
Reporter: nikita....@gmail.com
CC: llvm-bugs@lists.llvm.org
define i1 @test([0 x i32]* nonnull %arg, i64 %arg1, i64 %arg2) {
%tmp = getelementptr inbounds [0 x i32], [0 x i32]* %arg, i64 0, i64 0
%tmp3 = getelementptr inbounds [0 x i32], [0 x i32]* %arg, i64 0, i64 %arg1
%tmp6 = ptrtoint i32* %tmp3 to i64
%tmp7 = ptrtoint i32* %tmp to i64
%tmp8 = sub nuw i64 %tmp6, %tmp7
%tmp9 = lshr exact i64 %tmp8, 2
%tmp24 = icmp ugt i64 %tmp9, %arg2
ret i1 %tmp24
}
opt -instcombine on LLVM 9:
define i1 @test([0 x i32]* nonnull %arg, i64 %arg1, i64 %arg2) {
%tmp24 = icmp ugt i64 %arg1, %arg2
ret i1 %tmp24
}
opt -instcombine on LLVM 10:
define i1 @test([0 x i32]* nonnull %arg, i64 %arg1, i64 %arg2) {
%tmp3.idx = and i64 %arg1, 4611686018427387903
%tmp24 = icmp ugt i64 %tmp3.idx, %arg2
ret i1 %tmp24
}
Am I missing some subtlety here that prevents inbounds from implying that the
masking is unnecessary?
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs