[llvm-dev] Live range priority in Greedy RA

Dangeti Tharun kumar via llvm-dev llvm-dev at lists.llvm.org
Thu Nov 8 02:01:27 PST 2018


Hi,

I had a look at this  'calculateSpillWeightAndHint' function, spill weights
are being calculated considering no of uses, defs, loops etc.
Why not these weights be used as priority, seems they are good enough to
serve as priority, instead doing it by their sizes?

On Tue, Oct 30, 2018 at 5:39 PM jingu kang <jaykang10 at gmail.com> wrote:

> Hi Dangeti,
>
> As far as I know, 'calculateSpillWeightAndHint' considers loop
> induction variable and the spill weight affects evict and spill. If it
> is not enough for you, I guess you can add some heuristic code on the
> function.
>
> Thanks,
> JinGu Kang
> 2018년 10월 30일 (화) 오전 11:51, Dangeti Tharun kumar via llvm-dev
> <llvm-dev at lists.llvm.org>님이 작성:
> >
> > Hi,
> >
> > In the Greedy RA, I see that the enqueue method adds higher priority to
> the live intervals based on their sizes.
> >
> > Isn't it makes sense to give priority to live intervals that start and
> end in a loop? Please let me know if the code is already achieving it in
> some way.
> > Also consider correcting me, if this approach is wrong.
> >
> > --
> > Regards,
> > DTharun
> > _______________________________________________
> > LLVM Developers mailing list
> > llvm-dev at lists.llvm.org
> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>


-- 
Regards,
DTharun
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20181108/d793bd5e/attachment.html>


More information about the llvm-dev mailing list