[cig-commits] r21688 - seismo/3D/SPECFEM3D/trunk

dkomati1 at geodynamics.org dkomati1 at geodynamics.org
Sun Mar 31 09:44:46 PDT 2013


Author: dkomati1
Date: 2013-03-31 09:44:46 -0700 (Sun, 31 Mar 2013)
New Revision: 21688

Modified:
   seismo/3D/SPECFEM3D/trunk/flags.guess
Log:
better flags for IBM xlf


Modified: seismo/3D/SPECFEM3D/trunk/flags.guess
===================================================================
--- seismo/3D/SPECFEM3D/trunk/flags.guess	2013-03-31 16:44:19 UTC (rev 21687)
+++ seismo/3D/SPECFEM3D/trunk/flags.guess	2013-03-31 16:44:46 UTC (rev 21688)
@@ -122,6 +122,10 @@
         # arrays in the stack and the code crashes if the stack size is too
         # small (which is sometimes the case, but less often these days on large machines)
         #
+        # you will probably need to add " module load bgq-xl " or similar to your .bash_profile to load the compilers
+        #
+        # It could also help to put this in your .bash_profile: export XLFRTEOPTS=aggressive_array_io=yes:buffering=enable
+        #
         # on IBM with xlf one should also set
         #
         # CC = xlc_r
@@ -141,15 +145,12 @@
         #
         if test x"$FLAGS_CHECK" = x; then
 # deleted -qxflag=dvz because it requires handler function __xl_dzx and thus linking will fail 
-            FLAGS_CHECK="-O4 -qnostrict -qsimd=auto -qassert=contig -qhot -q64 -qtune=auto -qarch=auto -qcache=auto -qfree=f90 -qsuffix=f=f90 -qhalt=w -qlanglvl=2003std -qsuppress=1518-317 -Q -Q+rank,swap_all -Wl,-relax -WF,-DFORCE_VECTORIZATION"
-        # Options -qreport -qattr -qsource -qlist create a *.lst file containing detailed information about vectorization.
-        # on MareNostrum at the Barcelona SuperComputing Center (Spain) one can use
-        # -qtune=ppc970 -qarch=ppc64v instead of -qtune=auto -qarch=auto
-        # on "Babel" IBM BlueGene at IDRIS (France) use:
+            FLAGS_CHECK="-O4 -qnostrict -qsimd=auto -qassert=contig -qhot -q64 -qtune=auto -qarch=auto -qcache=auto -qfree=f90 -qsuffix=f=f90 -qhalt=w -qlanglvl=2003std -g -qsuppress=1518-317 -qsuppress=1500-036 -Q -Q+rank,swap_all -Wl,-relax -WF,-DFORCE_VECTORIZATION"
+        # 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
         #
-        # to debug with IBM xlf, one can add this: -g -O0 -C -qddim -qfullpath
-        # to debug, one can add these three flags: -qflttrap=overflow:zerodivide:invalid:enable -qfloat=nans -qinitauto=7FBFFFFF
+        # to debug with IBM xlf, one can add this: -g -O0 -C -qddim -qfullpath -qflttrap=overflow:zerodivide:invalid:enable -qfloat=nans -qinitauto=7FBFFFFF
         #
         fi
         ;;



More information about the CIG-COMMITS mailing list