Re: [AMBER] AmberTools19/AMBER18 - FFTW3 and MKL compiling issues

From: viktor drobot <linux776.gmail.com>
Date: Thu, 15 Aug 2019 17:26:49 +0300

Thank you for the answer. I thought that I could benefit from using mkl
blas/lapack routines for qmmm calculations
Will try Intel/mkl toolchain

чт, 15 авг. 2019 г., 4:13 David Case <david.case.rutgers.edu>:

> On Tue, Aug 13, 2019, viktor drobot wrote:
>
> >Hello, dear AMBER users and developers! I'm trying to compile
> >AmberTools19/AMBER18 on our local cluster. I use gcc 4.8.5 compilers and
> >explicitly set MKL_HOME environment variable for static linking MKL
> >routines.
>
> I don't have a definitive answer, but I think that with the released code,
> MKL is only really useful with Intel compilers. There have been attempts
> by
> developers to use MKL with gnu compilers, but I don't think those have made
> it into the released code base yet.
>
> A question: why do you want to do this? Generally, the benefits of MKL
> (vs. just using the standard FFTW builds) are modest to non-existent.
> That said, there might be special cases where MKL would help enough to be
> worthwhile. So I'm wondering if you have identified a particular need for
> MKL or not.
>
> ....dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 15 2019 - 07:30:02 PDT
Custom Search