Fixes for assigning the execution units (NUMA/CPUs, GPUs) to the subtrees. #13
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
added 30 commits
April 13, 2017 11:39
… distributing work to the execution units.
…y one of them is acting on a subtree.
…that proc_bind(spread) will adequately cover the NUMA regions.
… reasonable second-level nested parallelism within a NUMA node.
…en OpenMP is disabled.
added 26 commits
May 11, 2017 18:07
…st memory is not pinned.
…uld not hurt anything except performance.
…re, but should not hurt correctness.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Accesses (reads/writes) to some of the task-shared variables in LDL^T and Cholesky C++ kernels that explicitly control behaviour of tasks and their status have been made atomic.
Partially FIXED assigning the execution units (NUMA/CPUs, GPUs) to the subtrees.
That includes a reorganisation of a mechanism for task distribution across the hardware topology.
NUMA binding is still based on OpenMP's proc_bind(spread), which may not be the most robust approach.
Small code-style fixes.
Introduced more
volatileattributes on CUDA shared memory variables.FIXME: Pivoting in the indefinite case, in CUDA, alongside with uninitialised memory accesses (consult
cuda-memcheck --tool initcheck).