Hi Ben, On Sep 14, 2015, at 7:48 PM, Kenneth Hoste <[email protected]> wrote: > As Jack already mentioned, iompi is a good starting point (but doesn't > include the BLAS/LAPACK/FFTW part to make it a 'full' toolchain). > […] > As long as you limit yourself to slapping together existing toolchain > components (my suggestion would be to go for Intel compilers + OpenMPI + > OpenBLAS + FFTW), you should be in for a fairly smooth ride (since the EB > framework will take care of the hard part), provided that you don't run into > any surprises when building OpenBLAS with Intel compilers (we've done it > before, should be fine).
Effectively, you pronounce the need for the “ioolf” toolchain (alike goolf) and be sure that you are not alone in this ride. > That being said: there's actually a free version of Intel MKL available too, > see https://software.intel.com/sites/campaigns/nest/. > Only for MKL (+ compilers) though, not Intel MPI, and I'm sure there are some > restrictions in place too w.r.t. allowed usage. That alternative, pending a compatible license, would allow you to reuse the “iomkl” toolchain which already exists. >From there, it should be just versions substitutions and a massive recursive >--try-toolchain to build a new module universe. >> Do you think such a toolchain would be useful to the community? > > I think it certainly would, I'm sure there are others in the same boat as you > are. Yes, they will be useful: ioolf & iomkl toolchains may be super handy for cloud-compatible builds. cheers, Fotis -- echo "sysadmin know better bash than english" | sed s/min/mins/ \ | sed 's/better bash/bash better/' # signal detected in a CERN forum
