General

Profile

Erik Lindahl

  • Email: erik@kth.se
  • Registered on:
  • Last connection: 05/18/2017

Issues

Projects

Activity

05/23/2017

04:51 PM GROMACS Revision 9944f16b: Updated cool quotes
Removed two recent purely political quotes, based on a gmx-users
discussion a few months ago. We aim for quotes to be...

04/24/2017

09:28 AM GROMACS Feature #2163: Use better input/output value choices for SIMD tests
Agreed. I'll try to implement it in master, but it will likely trigger some false failures as we apply trial-and-erro...

04/22/2017

04:51 PM GROMACS Bug #2164: SIMD sqrt in double-precision build does not work correctly
I've had a look to see what compilers generate internally, and unfortunately there doesn't appear to be any easy fix....
12:23 PM GROMACS Bug #2164: SIMD sqrt in double-precision build does not work correctly
Actually, as far as I can see, all SIMD architectures except x86 have the lookup work right (i.e., it will provide co...
12:03 PM GROMACS Bug #2164: SIMD sqrt in double-precision build does not work correctly
Actually, even that won't help, since the clamping means we don't get enough valid bits in the lookup. I need to thin...
11:53 AM GROMACS Bug #2164: SIMD sqrt in double-precision build does not work correctly
Actually, there is a double version of it, but not until Skylake with AVX512 - in that case we already use it.
How...
10:50 AM GROMACS Bug #2160 (Rejected): improper dihedral
Not quite; it is actually used occasionally for groups that shouldn't be planar, e.g. some - NH2+ cases where some fo...
10:34 AM GROMACS Bug #2136: mdrun -noconfout checkpointing issue
-noconfout sounds like a great candidate for an option to just kill.
We need to learn that the question is not how...
10:03 AM GROMACS Feature #2163: Use better input/output value choices for SIMD tests
Dropping to feature.
There nothing *wrong* with the current tests, but of course it might be possible to test mor...

03/14/2017

09:04 PM GROMACS Revision 6f8f5d2a: Fix gcc compiles of AVX-512 SIMD
A previous change to improve shuffling for AVX-512 was
likely only tested on icc, while gcc requires an explicit
poin...

Also available in: Atom