Hello,
Currently the noalias attribute can be applied on function parameters and return
value, or via scoped alias on store.
The problem is that there is no easy way to apply the noalias attribute on a
pointer that is loaded indirectly (e.g from the field of an aggregate, or from a
memory location - though for this you can still use scoped alias, but they are a
lot more intrusive)
The idea would be to provide some kind of intrinsic that can transform a pointer
to a noalias pointer, something in the lines of:
define noalias i32* @noalias(i32 *%a) {
ret i32* %a
}
What do you think? Would it fit an intrinsic?
Alexandre
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170914/42c08869/attachment.html>
How is this different from the work that Hal has been doing:
https://reviews.llvm.org/search/query/qa3.P4sG6bA8/#R
https://llvm.org/devmtg/2017-02-04/Restrict-Qualified-Pointers-in-LLVM.pdf
From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of
Alexandre Mutel via llvm-dev
Sent: Thursday, September 14, 2017 4:34 AM
To: llvm-dev at lists.llvm.org
Subject: [llvm-dev] [RFC] noalias intrinsic
Hello,
Currently the noalias attribute can be applied on function parameters and return
value, or via scoped alias on store.
The problem is that there is no easy way to apply the noalias attribute on a
pointer that is loaded indirectly (e.g from the field of an aggregate, or from a
memory location - though for this you can still use scoped alias, but they are a
lot more intrusive)
The idea would be to provide some kind of intrinsic that can transform a pointer
to a noalias pointer, something in the lines of:
define noalias i32* @noalias(i32 *%a) {
ret i32* %a
}
What do you think? Would it fit an intrinsic?
Alexandre
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170915/1eeac375/attachment.html>
Hi, Alexandre, As Ehsan mentions, this is something we've been planning for a while. I'll likely commit the patches to add the intrinsic soon. As soon as the AA implementation patches (and the patch to use the intrinsic in the inliner) finish the review process, we'll be able to handle this effectively. In short, yes. :-) -Hal On 09/14/2017 08:15 PM, Ehsan Amiri via llvm-dev wrote:> > How is this different from the work that Hal has been doing: > > https://reviews.llvm.org/search/query/qa3.P4sG6bA8/#R > > https://llvm.org/devmtg/2017-02-04/Restrict-Qualified-Pointers-in-LLVM.pdf > > *From:*llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] *On Behalf Of > *Alexandre Mutel via llvm-dev > *Sent:* Thursday, September 14, 2017 4:34 AM > *To:* llvm-dev at lists.llvm.org > *Subject:* [llvm-dev] [RFC] noalias intrinsic > > Hello, > > Currently the noalias attribute can be applied on function parameters > and return value, or via scoped alias on store. > > The problem is that there is no easy way to apply the noalias > attribute on a pointer that is loaded indirectly (e.g from the field > of an aggregate, or from a memory location - though for this you can > still use scoped alias, but they are a lot more intrusive) > > The idea would be to provide some kind of intrinsic that can transform > a pointer to a noalias pointer, something in the lines of: > > define noalias i32* @noalias(i32 *%a) { > > ret i32* %a > > } > > What do you think? Would it fit an intrinsic? > > Alexandre > > > > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev-- Hal Finkel Lead, Compiler Technology and Programming Languages Leadership Computing Facility Argonne National Laboratory -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170914/a90c02de/attachment.html>