Matthias,
Don't worry about the xlC 6/7 patches. I got v.8 to work finally except for errors related to memory limits on the head-node, which NERSC told me how to fix.
Thanks,
Jeff
Matthias Troyer wrote:
Hi Jeff,
I just tried again on our IBM system (which was down for a few days) and have no problem with xlC 8. I will nevertheless forward the patches which we needed for xlC 6 and xlC 7, although there we had compiler and not linker problems.
Another possibility to explore is to change the version number 700 to 800 in line 85 of src/alps/lattice/graphproperties.h . There is apparently a problem that when typenames get too long on some IBM systems the linker does not manage to mangle them correctly. We dropped some graph properties for these systems - the change will disable the calculation of superfluid densities and spin stiffnesses, but I guess it should then link.
Matthias
On May 4, 2007, at 2:10 PM, Jeff Hammond wrote:
Matthias,
The default, which I used, is 7.0.0.2, and I just now tried 8.0.0.0 which gave a different error at a later point. Fortunately, Googling gave me the fix (C/F)FLAGS=-qnorm (turn off licensing) and got me through until linking. Linking did not go well (see below) but I suspect the fixes you alluded to will solve the problem.
I'm happy to use either compiler, whichever you say is better. I appreciate your help very much.
Jeff