[REVIEW] 2D cython/python infrastructure- BFS & SSSP #1177
Conversation
|
Please update the changelog in order to start CI tests. View the gpuCI docs here. |
Iroy30
changed the title
…st_mnmg_pagerank_bfs_sssp
…st_mnmg_pagerank_bfs_sssp
…n alnog the minor direction
|
rerun tests |
Codecov Report
@@ Coverage Diff @@
## branch-0.16 #1177 +/- ##
===============================================
- Coverage 57.04% 56.79% -0.25%
===============================================
Files 61 62 +1
Lines 2514 2546 +32
===============================================
+ Hits 1434 1446 +12
- Misses 1080 1100 +20
Continue to review full report at Codecov.
|
| dg = cugraph.DiGraph() | ||
| dg.from_dask_cudf_edgelist(ddf, "src", "dst") | ||
|
|
||
| expected_dist = cugraph.sssp(g, 0) |
There was a problem hiding this comment.
Second problem looks similar to the issue that was fixed with pagerank values... note that the index restarts... perhaps the predecessor and distance aren't getting gathered from the correct GPU?
cpp/src/utilities/cython.cu
Outdated
| graph_container.graph_ptr_union.GraphCSCViewDoublePtr->get_vertex_identifiers( | ||
| reinterpret_cast<int32_t*>(identifiers)); | ||
| } else if (graph_container.graph_type == graphTypeEnum::graph_t) { | ||
| if (graph_container.edgeType == numberTypeEnum::int32Type) { |
There was a problem hiding this comment.
This would be cleaner (for pagerank, BFS and SSSP) to use the technique used by louvain... define a functor for each algorithm and then call detail::call_function<true, void>(handle, graph_container, functor);
There was a problem hiding this comment.
Although we could defer this work to 0.17
| graph : cugraph.DiGraph | ||
| cuGraph graph descriptor, should contain the connectivity information | ||
| as dask cudf edge list dataframe(edge weights are not used for this | ||
| algorithm). Undirected Graph not currently supported. |
There was a problem hiding this comment.
"Undirected Graph not currently supported."
Could we add that undirected graphs should be passed as Digraph with edges in both directions (apply to other MG APIs as well).
|
|
||
| Returns | ||
| ------- | ||
| df : cudf.DataFrame |
There was a problem hiding this comment.
Should this be a dask data frame or are we assuming that 3 global V sets will fit in one GPU?
No description provided.