This is the mail archive of the
mailing list for the GCC project.
Re: [patch, fortran, RFC] First steps towards inlining matmul
- From: dominiq at lps dot ens dot fr (Dominique Dhumieres)
- To: tkoenig at netcologne dot de
- Cc: gcc-patches at gcc dot gnu dot org, fortran at gcc dot gnu dot org, jvdelisle at charter dot com
- Date: Sun, 05 Apr 2015 15:55:19 +0200
- Subject: Re: [patch, fortran, RFC] First steps towards inlining matmul
- Authentication-results: sourceware.org; auth=none
> > So, what do you think about this?
> I am curious about what performance gain results from this?
> I can see saving a library call to our runtime libraries.
> Do you have some timing results?
IMO the inlining of MATMUL should be restricted to small matrices (less than 4x4, 9x9
or 16x16 depending of your field!-). I have a variant of the polyhedron test induct.f90
in which I have replaced three dot products by the equivalent matmul and which executes
ten times slower than the original test.