https://gcc.gnu.org/bugzilla/show_bug.cgi?id=51119
--- Comment #40 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> ---
(In reply to Joost VandeVondele from comment #37)
> (In reply to Joost VandeVondele from comment #36)
> > #pragma GCC optimize ( "-Ofast -fvariable-expansion-in-unroller
> > -funroll-loops" )
>
Using: (I found it necessary to split into separate lines)
#pragma GCC optimize ( "-Ofast" )
#pragma GCC optimize ( "-funroll-loops" )
#pragma GCC optimize ( "-fvariable-expansion-in-unroller" )
$ gfc -static -Ofast -finline-matmul-limit=0 compare.f90
[jerry@quasar pr51119]$ ./a.out
=========================================================
================ MEASURED GIGAFLOPS =
=========================================================
Matmul Matmul
fixed Matmul variable
Size Loops explicit refMatmul assumed explicit
=========================================================
2 2000 0.055 0.048 0.042 0.055
4 2000 0.366 0.236 0.299 0.368
8 2000 0.628 0.673 1.610 1.833
16 2000 2.876 2.765 2.821 2.930
32 2000 4.681 3.382 4.812 4.763
64 2000 6.742 2.817 6.760 6.764
128 2000 8.532 3.194 7.852 8.539
256 477 9.420 3.319 9.053 9.420
512 59 8.435 2.358 8.319 8.390
1024 7 8.493 1.368 8.379 8.444
2048 1 8.499 1.666 8.385 8.448
