This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch, fortran, RFC] First steps towards inlining matmul
- From: Thomas Koenig <tkoenig at netcologne dot de>
- To: Dominique Dhumieres <dominiq at lps dot ens dot fr>
- 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 16:33:07 +0200
- Subject: Re: [patch, fortran, RFC] First steps towards inlining matmul
- Authentication-results: sourceware.org; auth=none
- References: <20150405135519 dot AF21A105 at mailhost dot lps dot ens dot fr>
Hi Dominique,
> IMO the inlining of MATMUL should be restricted to small matrices (less than 4x4, 9x9
> or 16x16 depending of your field!-)
The problem with the library function we have is that it is quite
general; it can deal with all the complexity of assumed-shape array
arguments. Inlining allows the compiler to take advantage of
contiguous memory, compile-time knowledge of array sizes, etc.
to allow for vectorization or even complete unrolling.
Of course, if you use c=matmul(a,b) where all three are assumed-shape
arrays, your advantage over the library function will be minimal at
best.
It will be interesting to see at what size calling BLAS directly
will be better.
Thomas