[WIP][skip-ci] SCC part 1 - trimming vertices with 0 in or out degrees #3325

wu6u3 · 2023-03-09T20:10:26Z

No description provided.

rapids-bot · 2023-03-09T20:10:30Z

Pull requests from external contributors require approval from a rapidsai organization member with write or admin permissions before CI can begin.

seunghwak · 2023-03-09T20:38:18Z

cpp/src/cores/trim.cu

+ remaining_vertices.begin(),
+ [in_degrees] __device__(
+ auto v) { return out_degrees[v] > edge_t{0}; })),
+ handle.get_stream());


You need to trim 0 in|out degree vertices from the graph.

You can refer to https://github.com/rapidsai/cugraph/blob/branch-23.04/cpp/src/community/triangle_count_impl.cuh#L261 (see line 234-298).

You can copy-and-paste-and-modify this code. The return value of this function should be a graph object and a renumber_map. See the link above and let me know if you have questions (code related to renumbering can be confusing if you are new to cuGraph, so don't hesitate to ask me questions).

Also note that extract_if_e will be renamed to extract_transform_e with some changes in function input parameters. See #3323.

seunghwak · 2023-03-23T19:52:46Z

cpp/src/cores/trim.cu

+ graph_view);
+ auto in_zero_core_first =
+ thrust::make_transform_iterator(core_numbers.begin(), is_zero_or_greater_t<edge_t>{});
+ rmm::device_uvector<uint8_t> in_zero_core_flags(core_numbers.size(), handle.get_stream());


Why are you finding 0-core? Vertex degrees can't be negative, so 0-core is the entire graph.

Here, we are trying to extract a subgraph where every vertex has non-zero in&out degree (as zero indegree or out degree vertices are singleton components).

You can do something like the following.

auto in_degrees = graph_view.compute_in_degrees(handle); auto out_degrees = graph_view.compute_out_degrees(handle); rmm::device_uvector<uint8_t> include_flags(in_degrees.size(), handle.get_stream()); auto degree_pair_first = thrust::make_zip_iterator(thrust::make_tuple(in_degrees.begin(), out_degrees.begin())); thrust::transform(handle.get_thrust_policy(), degree_pair_first, degree_pair_last, []__device__(auto pair) { return (thrust::get<0>(pair) >= 1 && thrust::get<1>(pair) >= 1) ? uint8_t{1} : uint8_t{0}; }); edge_src_property_t<decltype(graph_view), uint8_t> edge_src_include_flags(handle, graph_view); edge_dst_property_t<decltype(graph_view), uint8_t> edge_dst_include_flags(handle, graph_view); update_edge_src_property(...); update_edge_dst_property(...);

seunghwak · 2023-05-18T21:20:06Z

cpp/src/components/trim.cu

+#include <cugraph/graph_view.hpp>
+#include <cugraph/partition_manager.hpp>
+#include <cugraph/utilities/error.hpp>
+#include <cugraph/vertex_partition_device_view.cuh>


Need a space between cugraph includes and thrust includes.

seunghwak · 2023-05-18T21:20:25Z

cpp/src/components/trim.cu

+#include <thrust/copy.h>
+#include <thrust/iterator/discard_iterator.h>
+
+


One empty line is enough between two include groups.

seunghwak · 2023-05-18T21:20:46Z

cpp/src/components/trim.cu

@@ -0,0 +1,186 @@
+#pragma once


#pragma once unnecessary for trim.cu.

seunghwak · 2023-05-18T21:21:48Z

cpp/src/components/trim.cu

@@ -0,0 +1,186 @@
+#pragma once


And add a license statement at the beginning.

/* * Copyright (c) 2023, NVIDIA CORPORATION. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governin_from_mtxg permissions and * limitations under the License. */

seunghwak · 2023-05-18T21:22:35Z

cpp/src/components/trim.cu

+ std::optional<edge_property_view_t<edge_t, weight_t const*>> edge_weight_view,
+ bool do_expensive_check)
+{
+


no need for an empty line here.

seunghwak · 2023-05-18T21:29:01Z

cpp/src/components/trim.cu

+ raft::update_device(d_subgraph_offsets_out.data(),
+ h_subgraph_offsets_out.data(),
+ h_subgraph_offsets_out.size(),
+ handle.get_stream());


You need to extract a subgraph here and create a new graph from the extracted edges.

seunghwak · 2023-05-18T21:33:38Z

cpp/src/components/trim.cu

+ auto subgraph_out_view = subgraph_out.view(); 
+ auto [src, dst, wgt, offsets] = extract_induced_subgraphs(
+ handle,
+ subgraph_out_view,
+ subgraph_edge_weights_out,
+ raft::device_span<size_t const>{d_subgraph_offsets_out.data(), d_subgraph_offsets_out.size()},
+ raft::device_span<vertex_t const>{out_one_core_flags.data(), out_one_core_flags.size()},
+ do_expensive_check);


We're first extracting a one core (based on the out-degree) from the original graph.

Then, you create a new graph (from the edges that belong to the one core).

Then, extract a one core (based on the in-degree) from the new graph.

This code's return value should be based on the original vertex's renumbering, so you unrenumber at this end.

Please ping me if this is confusing.

seunghwak · 2023-05-18T21:33:58Z

cpp/src/components/trim.cu

+ minor_ptrs,
+ major_ptrs.size(),
+ (*renumber_map_out).data(),
+ (*renumber_map_out).size());


This unrenumber should happen after extracting 1 core.

seunghwak · 2023-05-18T21:35:00Z

cpp/src/components/trim.cu

+ std::tie(subgraph_in, subgraph_edge_weights_in, std::ignore, std::ignore, renumber_map_in) = 
+ cugraph::create_graph_from_edgelist<vertex_t,
+ edge_t,
+ weight_t,
+ edge_t,
+ int32_t,
+ false,
+ multi_gpu>(handle, in_one_core_flags, edge_src_in_one_cores, edge_dst_in_one_cores, edge_weight_view);
+ auto subgraph_in_view = subgraph_in.view();
+ auto [src_, dst_, wgt_, offsets_] = extract_induced_subgraphs(
+ handle,
+ subgraph_in_view,
+ subgraph_edge_weights_in,
+ raft::device_span<size_t const>{d_subgraph_offsets_in.data(), d_subgraph_offsets_in.size()},
+ raft::device_span<vertex_t const>{subgraph_in.data(), subgraph_in.size()},
+ do_expensive_check);


This is unnecessary.

seunghwak · 2023-05-18T21:36:01Z

cpp/src/components/trim.cu

+ core_number(
+ handle, graph_view, core_numbers.data(), k_core_degree_type_t::IN, size_t{1}, size_t{1});
+
+ auto in_one_core_first =
+ thrust::make_transform_iterator(core_numbers.begin(), is_one_or_greater_t<edge_t>{});
+ rmm::device_uvector<uint8_t> in_one_core_flags(core_numbers.size(), handle.get_stream());
+ thrust::copy(handle.get_thrust_policy(),
+ in_one_core_first,
+ in_one_core_first + core_numbers.size(),
+ in_one_core_flags.begin());
+
+ edge_src_property_t<decltype(graph_view), uint8_t> edge_src_in_one_cores(handle,
+ graph_view);
+ edge_dst_property_t<decltype(graph_view), uint8_t> edge_dst_in_one_cores(handle,
+ graph_view);
+
+ update_edge_src_property(
+ handle, graph_view, in_one_core_flags.begin(), edge_src_in_one_cores);
+ update_edge_dst_property(
+ handle, graph_view, in_one_core_flags.begin(), edge_dst_in_one_cores);
+ rmm::device_uvector<size_t> d_subgraph_offsets_in(2, handle.get_stream());
+ std::vector<size_t> h_subgraph_offsets_in{{0, in_one_core_flags.size()}};
+ raft::update_device(d_subgraph_offsets_in.data(),
+ h_subgraph_offsets_in.data(),
+ h_subgraph_offsets_in.size(),
+ handle.get_stream());


Why did you do all these if you're not calling extract_subgraph to extract 1 (indegree) core?

…to tests

copy-pr-bot · 2023-09-25T14:24:53Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

BradReesWork · 2023-09-25T14:31:11Z

/okay to test

seunghwak · 2024-04-04T17:55:02Z

This PR won't be merged.

wu6u3tw and others added 6 commits February 9, 2023 02:24

trim

4886374

add trim

af38716

test

57e88fa

tests

99ccd1c

change

1fedd9c

test

5c4f7bc

wu6u3 requested review from a team as code owners March 9, 2023 20:10

seunghwak mentioned this pull request Mar 9, 2023

Strongly connected components Part 1 (trimming zero in|out degree vertices) #3261

Closed

seunghwak changed the title ~~Adding SCC tests~~ [WIP][skip-ci] SCC part 1 - trimming vertices with 0 in or out degrees Mar 9, 2023

seunghwak reviewed Mar 9, 2023

View reviewed changes

wu6u3 and others added 8 commits March 23, 2023 14:24

update

91c2d05

trim

eae2691

add trim

b41133f

change

32fd94e

test

69b445e

update

7022bda

update tests

3adbbf4

update tests

25de0c5

seunghwak reviewed Mar 23, 2023

View reviewed changes

BradReesWork assigned wu6u3 Mar 28, 2023

BradReesWork added improvement Improvement / enhancement to an existing function non-breaking Non-breaking change labels Mar 28, 2023

BradReesWork changed the base branch from branch-23.04 to branch-23.06 March 30, 2023 12:43

BradReesWork added this to the 23.06 milestone Mar 30, 2023

wu6u3 added 2 commits April 20, 2023 14:40

add core 1 and extract graph

b402d98

merge

293953b

wu6u3 requested a review from a team as a code owner April 20, 2023 18:44

merge

9e735d4

add renumber

2cc86b8

seunghwak reviewed May 18, 2023

View reviewed changes

ChuckHastings modified the milestones: 23.06, 23.08 May 23, 2023

BradReesWork changed the base branch from branch-23.06 to branch-23.08 May 30, 2023 12:00

wu6u3 added 3 commits June 1, 2023 17:51

modify order of operation

42e4cc3

modify order of operation

b14d669

Merge branch 'branch-23.08' of https://github.com/rapidsai/cugraph in…

a8dcaef

…to tests

BradReesWork modified the milestones: 23.08, 23.10 Jul 24, 2023

Merge branch 'branch-23.08' into tests

a2fb1dc

BradReesWork changed the base branch from branch-23.08 to branch-23.10 September 25, 2023 14:25

Merge branch 'branch-23.10' into tests

c926f77

ChuckHastings modified the milestones: 23.10, 23.12 Sep 25, 2023

BradReesWork changed the base branch from branch-23.10 to branch-23.12 September 26, 2023 13:54

naimnv added the DO NOT MERGE Hold off on merging; see PR for details label Oct 20, 2023

BradReesWork marked this pull request as draft November 6, 2023 21:09

ChuckHastings modified the milestones: 23.12, 24.02 Nov 14, 2023

BradReesWork changed the base branch from branch-23.12 to branch-24.04 January 23, 2024 19:04

BradReesWork modified the milestones: 24.02, 24.04 Jan 23, 2024

BradReesWork modified the milestones: 24.04, 24.06 Mar 12, 2024

BradReesWork changed the base branch from branch-24.04 to branch-24.06 March 19, 2024 17:42

seunghwak closed this Apr 4, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP][skip-ci] SCC part 1 - trimming vertices with 0 in or out degrees #3325

[WIP][skip-ci] SCC part 1 - trimming vertices with 0 in or out degrees #3325

wu6u3 commented Mar 9, 2023

rapids-bot bot commented Mar 9, 2023

seunghwak Mar 9, 2023

seunghwak Mar 23, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

seunghwak May 18, 2023

copy-pr-bot bot commented Sep 25, 2023

BradReesWork commented Sep 25, 2023

seunghwak commented Apr 4, 2024

		#include <thrust/copy.h>
		#include <thrust/iterator/discard_iterator.h>

[WIP][skip-ci] SCC part 1 - trimming vertices with 0 in or out degrees #3325

[WIP][skip-ci] SCC part 1 - trimming vertices with 0 in or out degrees #3325

Conversation

wu6u3 commented Mar 9, 2023

rapids-bot bot commented Mar 9, 2023

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

copy-pr-bot bot commented Sep 25, 2023

BradReesWork commented Sep 25, 2023

seunghwak commented Apr 4, 2024