> On 11 Oct 2018, at 9:42, Åke Sandgren <[email protected]> wrote: > > > > On 10/11/2018 09:32 AM, Jakob Schiøtz wrote: >> >> >>> On 11 Oct 2018, at 9:17, Miguel Dias Costa <[email protected]> >>> wrote: >>> >>> I suppose the question then is whether there is an issue with libvdwxc >>> always using FFTW and the rest of GPAW using the FFTW wrappers to Intel >>> MKL, but I don't see why that would be a problem. >> >> Unfortunately, this is where the problem is. It causes core dumps. >> According to the libvdwxc developer, this is because MKL and FFTW use the >> same internal function names for incompatible functions (I assume some >> internal functions called by the official API functions). >> >>> Or did I misunderstand? > > Do you have any example of how libvdwxc fails with the mkl FFTW wrapper > that EB uses? Preferably standalone to make it easier to test?
I cannot really test that, because libvdwxc configures FFTW in its configure.sh script, and I do not know how to control it. But according to the developer, it uses a function in the MPI FFTW, which is not provided in the MKL wrappers. We have an open issue discussing it on the gpaw gitlab page. However, the correct way of solving it *may* be to change GPAW to load FFTW the way it loads all other libraries, instead of doing some kind of bizarre magic to load it from Python. That might solve the issue. We will look into it. Best regards Jakob > > -- > Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden > Internet: [email protected] Phone: +46 90 7866134 Fax: +46 90-580 14 > Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se -- Jakob Schiøtz, professor, Ph.D. Department of Physics Technical University of Denmark DK-2800 Kongens Lyngby, Denmark http://www.fysik.dtu.dk/~schiotz/
