Hi Jack, > On Mar 6, 2015, at 8:01 AM, Jack Perdue <[email protected]> wrote: > > > FYI, there are newer goolf and foss in the > 'devel' tree. > > gompi/goolf-1.7.20 with GCC 4.8.4 and OpenMPI 1.8.4 > gompi/foss-2015a with GCC 4.9.2 and OpenMPI 1.8.4 >
Ah, yes. I should have looked at the devel tree since v2.0 is about to come out. > As for predicting what future numbers will be used > by EB... good luck. 8^) They have rough heuristic > that used to be used. Very rough. > Dang. Thanks. > Personally, here at TAMU where I am creating toolchains > for the IBM XL compilers, I've just gone to using a time stamp. > e.g. > > xlompi-215.1.1.447 - XL compiler on top of GCC 4.4.7 with OpenMPI 1.8.4 > xlmpich-2015.1.1.484 - XL compiler on top of GCC 4.8.4 with MPICH 3.1.3 > xlmpich2-2015.1.1.492 - XL compiler on top of GCC 4.9.2 with MPICH2 1.5 > xlmvapich-2015.1.1.484 - XL compiler on top of GCC 4.8.4 with MVAPICH 1.9 > > I figure when it comes time to bump the XL or MPI version > number I'll just use the date at that time (e.g a new toolchain today > would have 2015.3.6[.gccver] as a version). Interesting way of numbering. I like the idea of a serial number based on the date, however I would also like to reflect the version numbers. Our current Tcl module names are insane, for example: netcdf/netcdf4-4.3_hdf5-1.8.11_hdf4-4.2.9_szip-2.1_zlib-1.2.78_jpeglib-8d_intel-13.0.0 refers to NetCDF 4.3 with HDF5 1.8.11, HDF4 4.2.9, .... Very YUCK! (Yes, this was put in place before my time). It confuses me, needless to say the users! Thanks Timothy
