This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: LCD benchmark updated to use omp simd
- From: Tobias Burnus <burnus at net-b dot de>
- To: tprince at computer dot org, "fortran at gcc dot gnu dot org" <fortran at gcc dot gnu dot org>
- Date: Sat, 21 Jun 2014 19:24:19 +0200
- Subject: Re: LCD benchmark updated to use omp simd
- Authentication-results: sourceware.org; auth=none
- References: <53A59725 dot 7080504 at aol dot com>
Tim Prince wrote:
I pushed my latest version of netlib vectors benchmark to
github.com/tprince/lcd tested on gfortran 4.9 (which skips Fortran omp
simd) and fortran-dev, and on intel 14.0 and 15.0 (windows and linux/mic)
There is a paper there too (docx file). You may compare
gfortran/gcc/g++ and, for Intel, also Cilk(tm) Plus, all using same
Fortran driver. Relevant cases use as much combined omp parallel and
simd vectorization as possible (more so in ifort).
I assume that you didn't use Cilk™Plus with GCC as cilk_for is not yet
implemented?*
What do you mean by "more so in ifort" in this regard?
Tobias
* See patch from 25 February, not reviewed during Stage 4 but also not
yet resurrected by Intel after Stage 1 has begun.