Re: [AMBER-Developers] GPU Test Cases

From: Jason Swails <>
Date: Mon, 16 Mar 2015 12:25:27 -0400

On Mon, Mar 16, 2015 at 12:08 PM, Ross Walker <> wrote:

> The code now does this. One can set IG < -1 in the GPU code and it will
> use the CPU RNG which shoudl address the issue fo tests failures with tests
> that use the random number generator. It's on the very very long todo list
> soemwhere before the next release but if someone want to volunteer to
> update the test cases great.

​I'd have done this long ago if I wasn't afraid of masking a *real* error
by simply clobbering test files on a system where 1/3 of the tests didn't
pass. I recall a bug last year shortly before the Amber 14 release went
live that nobody caught during ~2 weeks of testing since all the tests
already failed, anyway.

As for the other diffs - that needs a rework of dacdif most likely - or we
> build both SPFP and DPFP precision models and use the DPFP model for
> running the tests.

​I don't understand the need for dacdif changes here? DPFP and SPFP follow
different code paths, we shouldn't test just one of them (there have been
bugs in the past targeting only one precision model, if memory serves). We
should test the one we build... isn't that what we already do?

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER-Developers mailing list
Received on Mon Mar 16 2015 - 09:30:04 PDT
Custom Search