Conversation
| Most Sn transport codes are designed around one of these nestings, which is an inflexibility that leads to software engineering compromises when porting to new architectures and programming paradigms. | ||
|
|
||
| Early research has found that the problem dimensions (zones, groups, directions, scattering order) and the scaling (number of threads and MPI tasks), can make a profound difference in the performance of each of these nestings. | ||
| To our knowledge, this is a capability unique to Kripke, and should provide key insight into how data-layout affects Sn solver performance. |
There was a problem hiding this comment.
@rchen20 We need details in the code modification section on what may be changed, and what should stay the same. If I am not mistaken, the data ordering (GDZ) needs to stay the same,. while the execution policy can be changed and optimizations like loop tiling are allowed.
rchen20
changed the title
|
@pearce8 We can merge this after John Loffeld and Tom take quick look. |
|
@rchen20 I was able to tag one of them here for the review - please check in with them and let me know when I should merge, in case they don't end up putting a formal review of the PR in. |
John Loffeld already took a look at this and told me this was good to go. Feel free to merge it whenever you'd like. |
3 participants
No description provided.