[cig-commits] [commit] devel: switched to option -qstrict for the IBM xlf compiler (b43c63c)

[cig_noreply at geodynamics.org]

Repository : https://github.com/geodynamics/specfem3d

On branch  : devel
Link       : https://github.com/geodynamics/specfem3d/compare/ee55a07d2b0334368c3d84832e2c66cb81659902...b43c63c4cea034e3e6f647cebc591fb4c2ab3894

>---------------------------------------------------------------

commit b43c63c4cea034e3e6f647cebc591fb4c2ab3894
Author: Dimitri Komatitsch <komatits at users.noreply.github.com>
Date:   Thu Nov 20 16:08:45 2014 +0100

    switched to option -qstrict for the IBM xlf compiler


>---------------------------------------------------------------

b43c63c4cea034e3e6f647cebc591fb4c2ab3894
 flags.guess | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/flags.guess b/flags.guess
index 5efc76d..bcd180d 100644
--- a/flags.guess
+++ b/flags.guess
@@ -166,8 +166,11 @@ case $my_FC in
         # one should also use additional flags:
         # CFLAGS = -Wl,-relax
         #
-# deleted -qxflag=dvz because it requires handler function __xl_dzx and thus linking will fail 
-        DEF_FFLAGS="-O4 -qnostrict -qassert=contig -qhot -q64 -qtune=auto -qarch=auto -qcache=auto -qfree=f90 -qsuffix=f=f90 -qhalt=w -qlanglvl=2003std -g -qsuppress=1518-234 -qsuppress=1518-317 -qsuppress=1518-318 -qsuppress=1500-036 -Q -Q+rank,swap_all -Wl,-relax"
+# deleted -qxflag=dvz because it requires handler function __xl_dzx and thus linking will fail.
+# The -qstrict option prevents some minor differences of results between xlf and
+# other compilers. There is a small decrease in performance, but generally
+# it is small or negligible compared to other issues like slow I/O.
+        DEF_FFLAGS="-O4 -qstrict -qassert=contig -qhot -q64 -qtune=auto -qarch=auto -qcache=auto -qfree=f90 -qsuffix=f=f90 -qhalt=w -qlanglvl=2003std -g -qsuppress=1518-234 -qsuppress=1518-317 -qsuppress=1518-318 -qsuppress=1500-036 -Q -Q+rank,swap_all -Wl,-relax"
         # Options -qreport -qsource -qlist create a *.lst file containing detailed information about vectorization.
         # On IBM BlueGene at IDRIS (France) use:
         # -qtune=auto -qarch=450d -qsave     instead of -qtune=auto -qarch=auto