[Llvm-bgq-discuss] clang on BGQ performance

[Hal Finkel]

There are a couple possible issues here, but since you're doing memory streaming with many threads, I suspect the software prefetching might be causing some issues. The prefetching heuristic was not tuned for running multiple hardware threads per core, and I'd not be surprised if it is problematic.

In the next build, I'll provide a way to turn it off, so it will be easy to test.

 -Hal

----- Original Message -----
> From: "John A. Biddiscombe" <biddisco at cscs.ch>
> To: "Jeff Hammond" <jhammond at anl.gov>
> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> Sent: Tuesday, March 25, 2014 10:34:58 AM
> Subject: Re: [Llvm-bgq-discuss] clang on BGQ performance
> 
> Jeff
> 
> > Have you looked at the asm to see what the major difference in
> > codegen
> > between clang and gcc is?
> 
> I'm afraid not. I haven't written significant asm since I was a
> teenager (6502 anyone!).
> I'll start relearning the tools and get back if I discover anything
> interesting ...
> 
> JB
> _______________________________________________
> llvm-bgq-discuss mailing list
> llvm-bgq-discuss at lists.alcf.anl.gov
> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> 

-- 
Hal Finkel
Assistant Computational Scientist
Leadership Computing Facility
Argonne National Laboratory