On Thu, May 11, 2023, Siwakorn Sukharom via AMBER wrote:
>I want to build Amber22 on cray system, and I would like to use cray-libsci
>for blas and lapack as suggested in reference manual
>
>
>"If you are building Amber for a high performance computing environment, it
>is highly recommended to make use of an optimized BLAS implementation"
An admittedly non-answer:
That sentence is both old and non-Cray-specific. I'm guessing(?) that
actual Amber MD runs would never see the 50-100% speedup one might get on
"large matrix operations". More niche codes, like RISM or PB, use a larger
fraction of linear algebra.
That said, I hope the Cray gurus on the list can chime in with their
experience. It might be helpful to state the exact computer and compilers
that you are using.
....regards...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu May 11 2023 - 09:30:03 PDT
