Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Spgemm hash promotion #171

Closed
ndellingwood opened this issue Mar 7, 2018 · 0 comments
Closed

Spgemm hash promotion #171

ndellingwood opened this issue Mar 7, 2018 · 0 comments

Comments

@ndellingwood
Copy link
Contributor

As worked on by @mndevec and submitted in PR's #134 and #139.
Issue added for documentation and CHANGELOG tracking.

Quoting from #134:

Updating SpGEMM methods. These will be in the next KokkosKernels promotion to Trilinos.

1- New parameter selection on GPUs. This adjust the shared memory size, block size, vectorsize.
2- New meta-algorithm called SPGEMM_KK. This is the default algorithm from now on. ---> SPGEMM_KK_MEMORY = KKMEM is outdated.
It chooses the best algorithm to run based on the input structure.
For example, it may choose dense accumulators or sparse on CPUs.
On GPUs, it adjusts the old KKMEM parameters as above, or it may choose to run a block-based algorithm rather than a warp-based one (KKMEM).
3- Minor updates: changing the test driver calls. It now requires "--" before parameter name that was missing previously.

4- The new parameters name is SPGEMM_KK. In the xml file, any of the below will run this default algorithm. Please use the ones below after the new kokkoskernels promotion:

  • "SPGEMM_DEFAULT", "SPGEMM_KK", "SPGEMM_KK_MEMSPEED"
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants