[cfe-dev] [OT?] real-world interest of the polly optimiser

[C Bergström via cfe-dev]

On Mon, May 22, 2017 at 10:08 PM, Michael Kruse via cfe-dev <
cfe-dev at lists.llvm.org> wrote:

> Hi René,
>
> 2017-05-19 10:38 GMT+02:00 René J.V. Bertin via cfe-dev
> <cfe-dev at lists.llvm.org>:
> > Hi,
> >
> > Apologies if this isn't the best place.
>
> Polly has its own mailing list here:
> https://groups.google.com/forum/#!forum/polly-dev
> polly-dev at googlegroups.com
>
>
> > I've been looking for some information (understandable by the average
> user) about the real-world benefits of the polly optimiser, but have found
> only either very broad and vague claims or specialist research papers.
>
> As a researcher, I can tell about the research we are doing. We
> currently have a paper under review about optimizing gemm where we get
> 85\% of vendor-provided BLAS implementation, which is 20x the speed of
> the program compiled by clang without Polly.
>

Sorry, but please don't
1) Provide numbers when comparing against a weak baseline

Please do
2) If you do have a valid performance comparison or claim - please do
provide enough information so that a complete picture is presented.

You statement just came across as something like either llvm's loop
optimizer sucks so bad that polly is required and or somehow it's hitting a
corner case which is a sweetspot for polly.
------------
Also I'd kindly ask that if you do have such specific performance examples
of clang doing a rather poor job, please file a bug report and include as
much detail as you have time. It's unlikely that polly is doing anything
that a traditional loop optimizer can't do and or at least attempted.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20170522/43ddfd79/attachment.html>