llvm dev - Jul 2015 - [LLVMdev] AliasSetTracker and UnknownInst's (callsites mostly) problem

Hi all,

There is a problem about how AliasSetTracker merging AliasSet's when meet
UnknownInst.

When adding new pointer it looks for existing AliasSet's aliased with new
pointer. And merging them together. It is ok for pointers: if %A mayalias
%B and %B mayalias %C then %A mayalias %C.

But the same logic when adding callsite is wrong (
findAliasSetForUnknownInst).

Callsite may be known to readonly access one pointer and change the content
of other pointer.
Why to merge together two noalias pointers if they have some mod/ref
relationships with common callsite?

If modrefinfo(%A, %call1) == REF and modrefinfo(%B, %call1) == MOD
it doesn't mean %A may/must alias %B.
AliasSet's for %A and %B must stay independent with corrected access
lattice: aliasset(%A).Access |= RefAccess, aliasset(%B).Access |= ModAccess.
And I think new UnknownInst must be added to both AliasSet's.

Or maybe UnknownInst's must be collected in central part - AliasSetTracker,
not AliasSet's.

Am I right?

-- 
Best regards,
Andrew Zhogin
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150728/4e0ece51/attachment.html>

This was discussed a while back in the thread titled "alias set collapse 
and LICM".

Philip

On 07/28/2015 06:13 AM, Andrew Zhogin wrote:
> Hi all,
>
> There is a problem about how AliasSetTracker merging AliasSet's when 
> meet UnknownInst.
>
> When adding new pointer it looks for existing AliasSet's aliased with 
> new pointer. And merging them together. It is ok for pointers: if %A 
> mayalias %B and %B mayalias %C then %A mayalias %C.
>
> But the same logic when adding callsite is wrong 
> (findAliasSetForUnknownInst).
>
> Callsite may be known to readonly access one pointer and change the 
> content of other pointer.
> Why to merge together two noalias pointers if they have some mod/ref 
> relationships with common callsite?
>
> If modrefinfo(%A, %call1) == REF and modrefinfo(%B, %call1) == MOD
> it doesn't mean %A may/must alias %B.
> AliasSet's for %A and %B must stay independent with corrected access 
> lattice: aliasset(%A).Access |= RefAccess, aliasset(%B).Access |= 
> ModAccess.
> And I think new UnknownInst must be added to both AliasSet's.
>
> Or maybe UnknownInst's must be collected in central part - 
> AliasSetTracker, not AliasSet's.
>
> Am I right?
>
> -- 
> Best regards,
> Andrew Zhogin
>
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20150728/b4571eb8/attachment.html>