Closed jeffhammond closed 9 years ago
I know that I can push to upstream but I did not test this code yet. I assume you can validate it by inspection better than I can.
Thanks. There are lots of other little things that can be improved like caching norm2(posa-posb), using omp collapse, forcing static schedule etc.
Indeed. Other OpenMP improvements are certainly possible and I'll work on those as time permits.
Also any improvements to libint2 usage are welcome, as I am not particular happy with it as it is now (esp. re general contraction). I'm planning to put in (Opt)ERD as well one of these days.
On 6/26/15 11:00 AM, Jeff Hammond wrote:
Indeed. Other OpenMP improvements are certainly possible and I'll work on those as time permits.
— Reply to this email directly or view it on GitHub https://github.com/devinamatthews/aquarius/pull/4#issuecomment-115740828.
I'm not planning on working on LIBINT2. My preferred approach there is to find money for Ed to do it :-)
(Opt)ERD is a good plan. I haven't done much there other than addressing licensing issues and build system problems on Mac.
See you guys discussion of integrals. Maybe the libcint https://github.com/sunqm/libcint is something you can consider.
Certainly, I'd like to eventually put in as many alternatives as possible. Thanks for the info.
On 6/26/15 1:27 PM, Qiming Sun wrote:
See you guys discussion of integrals. Maybe the libcint https://github.com/sunqm/libcint is something you can consider.
— Reply to this email directly or view it on GitHub https://github.com/devinamatthews/aquarius/pull/4#issuecomment-115824655.
@sunqm You should create a new issue about libcint integration. It is too easy to lose track of it here on a closed OpenMP ticket :-)
Yes, please do that or I will forget. Closing ticket now.
this matters for large thread counts and when OpenMP runtime overhead is high.