[llvm-bugs] [Bug 40371] New: IndVarSimplify incorrectly replaces a pointer with another
via llvm-bugs
llvm-bugs at lists.llvm.org
Fri Jan 18 02:43:10 PST 2019
https://bugs.llvm.org/show_bug.cgi?id=40371
Bug ID: 40371
Summary: IndVarSimplify incorrectly replaces a pointer with
another
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P
Component: Scalar Optimizations
Assignee: unassignedbugs at nondot.org
Reporter: juneyoung.lee at sf.snu.ac.kr
CC: llvm-bugs at lists.llvm.org
Created attachment 21352
--> https://bugs.llvm.org/attachment.cgi?id=21352&action=edit
test.ll
$ cat test.ll
; while (p != q) p++;
; return p;
define i8* @iterate_until_p_becomes_q(i8* %p, i8* %q) {
label:
br label %while.cond
while.cond:
%p.addr.0 = phi i8* [ %p, %label ], [ %incdec.ptr, %while.body ]
%cmp = icmp ne i8* %p.addr.0, %q
br i1 %cmp, label %while.body, label %while.end
while.body:
%incdec.ptr = getelementptr inbounds i8, i8* %p.addr.0, i32 1
br label %while.cond
while.end:
ret i8* %p.addr.0
}
After `opt test.ll -indvars -S -o -`, the returned value %p.addr.0 is replaced
with %q:
define i8* @iterate_until_p_becomes_q(i8* %p, i8* %q) {
label:
br label %while.cond
while.cond: ; preds = %while.body, %label
%p.addr.0 = phi i8* [ %p, %label ], [ %incdec.ptr, %while.body ]
%cmp = icmp ne i8* %p.addr.0, %q
br i1 %cmp, label %while.body, label %while.end
while.body: ; preds = %while.cond
%incdec.ptr = getelementptr inbounds i8, i8* %p.addr.0, i32 1
br label %while.cond
while.end: ; preds = %while.cond
ret i8* %q
}
This is incorrect because %p and %q may have different provenances. This causes
miscompilation in this case:
char p[4]; // 0x100
char q[4]; // 0x104
char *ptr = iterate_until_p_becomes_q(&p[0], &q[0]); // ptr = &p[4]
ptr--;
*ptr = 'a'; // After the optimization, this raises UB because ptr becomes
q[-1].
Interesting thing is that if %p and %q are defined as non-i8-type (e.g. i16*),
LLVM compiles correctly, by introducing uglygep (https://godbolt.org/z/Rk1dWt
).
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20190118/b8c83891/attachment-0001.html>
More information about the llvm-bugs
mailing list