[llvm-dev] [RFC][LangRef][AliasAnalysis] Clarify noalias
affects only modified objects (original) (raw)
Finkel, Hal J. via llvm-dev llvm-dev at lists.llvm.org
Thu Feb 20 17🔞49 PST 2020
- Previous message: [llvm-dev] [RFC][LangRef][AliasAnalysis] Clarify `noalias` affects only modified objects
- Next message: [llvm-dev] Check if an SCC is a Loop
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
For AA-based memory-dependency analysis, this makes sense to me. read/read aliasing doesn't affect anything.
My only questions is: is there any concern that this might confuse instrumentation passes, or similar, which might do something even for read/read cases? Does anyone know of any other use cases that might be affected by this?
-Hal
Hal Finkel Lead, Compiler Technology and Programming Languages Leadership Computing Facility Argonne National Laboratory
From: Johannes Doerfert
Sent: Thursday, February 20, 2020 5:46 PM
To: LLVM-Dev
Cc: Finkel, Hal J.
Subject: [RFC][LangRef][AliasAnalysis] Clarify noalias
affects only modified objects
I would like to clarify that noalias
on arguments matches restrict
in C99. In particular, objects need to be modified for the noalias
to
be "effective". We basically apply this reasoning already in various
parts but I want to make it explicit now.
The wording change and the reasons are explained in https://reviews.llvm.org/D74935
Please let me know if there are concerns.
Thanks, Johannes -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200221/a6130d91/attachment.html>
- Previous message: [llvm-dev] [RFC][LangRef][AliasAnalysis] Clarify `noalias` affects only modified objects
- Next message: [llvm-dev] Check if an SCC is a Loop
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]