On Mon, Mar 16, 2015 at 12:34 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
> Okay - then I am fully open to other suggestions for fixing this.
>
> The rounding issues are there - they aren't going away so we need a test
> system that can handle this.
>
The rounding issues are easy to solve -- just increase the tolerance on the
tests (by as little as possible to make sure they pass on a wide range of
hardware). Again, if it was *just* the rounding issues, I'd have gone
ahead and fixed this a long time ago. I didn't want to touch the CUDA
tests until all of the failures were obvious hardware-related round-off
issues.
All the best,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Mar 16 2015 - 10:00:03 PDT
