-
Notifications
You must be signed in to change notification settings - Fork 310
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
[ENH] Streams support in cugraph #957
Comments
afender
changed the title
[ENH] Stream support in cugraph
[ENH] Streams support in cugraph
Jun 19, 2020
BradReesWork
added
feature request
New feature or request
and removed
? - Needs Triage
Need team to review and classify
labels
Jun 24, 2020
blocked by #815 |
Merged
rapids-bot bot
pushed a commit
that referenced
this issue
Feb 4, 2021
### Description Let the egonet graph of a node x be the subgraph that includes the neighborhood of x and all edges between them. Here is a basic description (1-hop, single seed) : - Add center node x to the graph. - Go through all the neighbors y of this center node x, add edge (x, y) to the graph. - For each neighbor y of center node x, go through all the neighbors z of center node x, if there is an edge between y and z in the original graph, add edge (y, z) to our new graph. ### Proposed solution Rather than doing custom one/two hops features, we propose a generic k-hops solution leveraging BFS with cutoff to identify neighbors within a given radius. In addition to the single source version (matching what's available in Nx), we propose to handle multiple sources (seeds) at once which allows better performances. This PR also enables a path in the experimental stack for returning multiple graphs (edge list format) from CUDA prims to python without using the legacy classes. As future work, we want to enable concurrency for the cutoff BFS for each seed. This is dependent of #957 Close #475 Authors: - Alex Fender (@afender) - @Iroy30 Approvers: - @Iroy30 - Brad Rees (@BradReesWork) URL: #1365
rapids-bot bot
pushed a commit
that referenced
this issue
Mar 5, 2021
- Stream synchronization behavior: switched to per-thread default stream instead of the legacy default stream - Update raft tag - EgoNet upgrade to use `uvector` instead of `device_vector` - EgoNet upgrade to execute on a different stream for each seed - Perf analysis timers/app for EgoNet Concurrency is limited by the number of available blocks on the device. Thrust-based codes may request a lot of blocks without a way to control this. In practice, smaller graphs leverage concurrency better than larger ones where tasks may end up waiting for available resources. We may wait on #1407 before reconciling and merging this Close #957 Authors: - Alex Fender (@afender) Approvers: - Seunghwa Kang (@seunghwak) - Andrei Schaffer (@aschaffer) URL: #1435
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
RAFT handle #815 comes with streams in it.
We should consider supporting these streams in cuGraph.
The text was updated successfully, but these errors were encountered: