[llvm-dev] Semantics of any out-of-bounds address returned by GEP

Nuno Lopes via llvm-dev llvm-dev at lists.llvm.org
Tue Jan 7 07:19:39 PST 2020

You're right: non-inbounds GEPs may return OOB pointers. OOB pointers cannot
be dereferenced.

OOB pointers can still be used as input to GEP to bring them back to being
within bounds.

OOB pointers can also be used for pointer comparisons.




From: Gautam Chakrabarti via llvm-dev
Sent: 07 January 2020 06:37
To: LLVM Dev <llvm-dev at lists.llvm.org>
Subject: [llvm-dev] Semantics of any out-of-bounds address returned by GEP




I am looking to get some clarification about out-of-bounds address that may
be returned by a GEP.


My understanding is that in the absence of the inbounds keyword, a GEP may
return a pointer that is out-of-bounds of the underlying allocated object.
Is it correct that any such out-of-bounds address must not be dereferenced?


I am a bit confused by the current wordings on this.


https://llvm.org/docs/LangRef.html#getelementptr-instruction says:


"If the inbounds keyword is not present, .... The result value of the
getelementptr may be outside the object pointed to by the base pointer. The
result value may not necessarily be used to access memory though, even if it
happens to point into allocated storage."


This seems to say that the out-of-bounds pointer may or may not be used to
access memory.


arithmetic-and-inttoptr says:


"Also, GEP carries additional pointer aliasing rules. It's invalid to take a
GEP from one object, address into a different separately allocated object,
and dereference it. IR producers (front-ends) must follow this rule, and
consumers (optimizers, specifically alias analysis) benefit from being able
to rely on it."


This quite clearly says the out-of-bounds address shall not be dereferenced
through. This also seems to be the assumption in AliasAnalysis
implementation (as already stated in above documentation).




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200107/7dd7402a/attachment-0001.html>

More information about the llvm-dev mailing list