SCM

Forums for pkdgrav2: a parallel n-body gravity code

View Forums

My Monitored Forums
Forum Description Threads Posts Last Post Moderation Level
 open-discussion General Discussion 1 1 2010-09-27 08:53 No Moderation
 help Get Public Help 1 1 2010-09-27 08:53 No Moderation
 developers Project Developer Discussion 1 1 2010-09-27 08:53 No Moderation
 release-file-2.2.9-1-compilation-fix-for-non-simd-machines Small fix for non-SIMD builds. If your code compiles currently, you DO NOT need to download this update. On Intel and AMD architectures you should not have had a problem building the code so far. 0 0 1970-01-01 00:00 No Moderation
 new-release-of-pkdgrav2--non-fortran-version- This latest release uses single precision multipole code which achieves the same errors down to at least theta=0.3. Note however, that at the same theta the code is much more accurate than previously. To get similar force errors as in older pkdgrav2m versions you can add about 0.2 to the theta you set for this version. Eg. theta=0.5 now has errors comparable to theta=0.3 in original pkdgrav2. The single precision currently only saves memory in the tree, but will speed up the code in the *next* release. This release also avoids the use of a Fortran compiler, which makes porting much easier. - Joachim 0 0 1970-01-01 00:00 No Moderation
 version-2.2.11--an-updated-simd-release-which-is-2x-faster- This release version uses SIMD (SSE or Altivec) instructions now for multipoles and also to evaluate the opening criteria. The resulting code is now 2x as fast at the same accuracy. Previously we only used SIMD to evaluate the particle-particle interactions and could not do the same for multipole interactions. New formulae allowing so called "scaled multipoles" to be used (release 2.2.10) paved the way for this full SIMD version. This completes a certain development theme and there will not be a next release till late december. You can see what things we are working on from the link "Next Release" under the Task Manager. - Joachim 0 0 1970-01-01 00:00 No Moderation
 bug-fix-release-2.2.11-1.-sorry--released-2.2.11-in-a-rush- There was a new problem introduced in version 2.2.11 to do with opening of top tree cells causing parallel simulations to crash. Furthermore there was an incorrect assert() which was removed. All changes were to the file walk2.c. 0 0 1970-01-01 00:00 No Moderation