General

Profile

Erik Lindahl

  • Email: erik@kth.se
  • Registered on:
  • Last connection: 09/08/2017

Issues

Projects

Activity

09/14/2017

09:51 AM GROMACS Bug #2243 (Resolved): Reference SIMD exp doesn't flush to zero for very large negative inputs
Applied in changeset commit:36311d95c2d8b0ebd889d96b681795c4c34634dd.

09/12/2017

02:26 PM GROMACS Revision 36311d95: Improve accuracy of SIMD exp for small args
Introduce a separate check for small arguments and
set the result to zero below this cutoff. This change
also moves t...

09/08/2017

01:03 AM GROMACS Bug #2243 (Fix uploaded): Reference SIMD exp doesn't flush to zero for very large negative inputs
This change got more complex than I first thought, but now it's at least fixed properly. I disliked our previous hack...
11:49 AM GROMACS Bug #2243 (Accepted): Reference SIMD exp doesn't flush to zero for very large negative inputs

09/05/2017

06:29 PM GROMACS Bug #2164 (Resolved): SIMD sqrt in double-precision build does not work correctly
Applied in changeset commit:e34ead15b327872ba9362de9416178cce53158b3.

09/04/2017

11:41 PM GROMACS Revision e34ead15: More SIMD math argument checking, added unsafe options
This change adds more argument checking and safeguards
for sqrt, exp2, and exp-related SIMD math functions, and
prope...

08/28/2017

10:30 PM GROMACS Feature #2163 (Resolved): Use better input/output value choices for SIMD tests
Applied in changeset commit:b9057925955f066c1c88719f6bbff29cfc5fe666.
10:17 PM GROMACS Revision b9057925: Improved SIMD test data to use all bits
The SIMD test data now uses all available bits in the
current precision floating-point numbers, to catch errors
where...

08/02/2017

09:50 PM GROMACS Feature #2224: Proposed feature: conditional stop
Hi,
I'm somewhat hesitant :-)
The main reason is that we have a history of repeatedly making the simple art of ...

07/17/2017

06:57 PM GROMACS Task #2216 (New): GROMACS SIMD acceleration: generation 3
The architecture-agnostic SIMD interface works quite well, but there are a number of challenges we should start discu...

Also available in: Atom