[REVIEW] FEA subgraph modifications

.github/ISSUE_TEMPLATE/question.md

@@ -0,0 +1,10 @@
+---
+name: Question
+about: I have a question for Gunrock team.
+title: ''
+labels: question
+assignees: ''
+
+---
+
+

.gitignore

@@ -48,3 +48,6 @@ gunrock/util/gitsha1.c
 
 # Ignore build directory
 build
+
+# Ignore vscode files
+*.vscode*

.gitmodules

@@ -6,10 +6,6 @@
 	path = externals/moderngpu
 	url = https://github.com/NVlabs/moderngpu.git
 	branch = master
-[submodule "docs"]
-	path = docs
-	url = https://github.com/gunrock/docs.git
-	branch = master
 [submodule "externals/rapidjson"]
 	path = externals/rapidjson
 	url = https://github.com/Tencent/rapidjson.git

CMakeLists.txt

@@ -359,6 +359,12 @@ endif (CUDA_VERBOSE_PTXAS)
 set(CUDA_NVCC_FLAGS --expt-extended-lambda -rdc=true -lcudadevrt)
 set(CUDA_LIBRARIES ${CUDA_LIBRARIES} ${CUDA_curand_LIBRARY})
 
+# begin /* Suppress all warnings from nvcc */
+if (NOT ENABLE_WARNINGS)
+  set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -w")
+endif (NOT ENABLE_WARNINGS)
+# end /* Suppress all warnings from nvcc */
+
 if(GUNROCK_BUILD_LIB)
   if(GUNROCK_BUILD_SHARED_LIBS)
     set(LIB_TYPE SHARED)
@@ -393,3 +399,12 @@ if(GUNROCK_BUILD_TESTS)
     add_subdirectory(unittests)
   endif(GUNROCK_GOOGLE_TESTS)
 endif(GUNROCK_BUILD_TESTS)
+
+###################################################################################################
+# - install targets -------------------------------------------------------------------------------
+
+install(TARGETS gunrock LIBRARY
+    DESTINATION lib)
+
+install(FILES ${CMAKE_SOURCE_DIR}/gunrock/gunrock.h
+    DESTINATION include/gunrock)

README.md

@@ -1,6 +1,6 @@
 <!-- start: gunrock's header -->
 <p align="center">
-  <a href="https://github.com/gunrock/gunrock/"><img src="https://github.com/gunrock/docs/raw/master/source/images/GunrockLogo150px.png"></a>
+  <a href="https://github.com/gunrock/gunrock/"><img src="https://github.com/gunrock/docs/raw/master/docs/_media/logo.png"></a>
   <br>
   <a href="https://github.com/gunrock/gunrock/releases/tag/v1.0"><img src="https://img.shields.io/badge/gunrock-v1.0-blue.svg?logo=data%3Aimage%2Fpng%3Bbase64%2CiVBORw0KGgoAAAANSUhEUgAAAB4AAAAeCAYAAAA7MK6iAAAIjklEQVRIDQXBaZCeBWEA4Oc9vnN3v93sbrKbhGxCIMVASERQSCt0BCo4lOGonVirMOPR45cVbS3t1B8OYy%2B1WKdFtKPiqB162LHjdAZqKQUBwyEBYgKJuZYku9kju%2Fvtd79XnycoAHD0NznzEDe1OT1gOmDhLLMhN5RxLcEtrOxn%2BX6VbTOu6Z52oVh2auOHqf2CE%2BucmWXuFfbuo3w5UcpShWLAxhUuv5UQAAAAACDMSdus3EP4Hu%2BeHLG0dMgLSdfJZNID0SKQIisQAAAAADEAAAAEyDHIyAqCNp0%2FNDR8wTOdFZXpHT7U6rizvezLC4cllZJ%2FCiLdSkha0E%2BICiLkAQEAQgAAUJAVJDlRyoaGLc1Rn2w1fbd81FxnXXWy7%2FbGwOPdno%2BEQ462m%2F5%2BYUknWnNie8OXLt%2FttjgSV2OKgjAjKSgCIAYAAWnCANWQNGCo5g%2FKQx7pJ6Rt6mXCgG7f7k7hyaCqESY25jnlOmluZ8Rntl%2FiM3nuZNj0%2Bf7A9VM0VkJ%2FJnBeQFAA4Pg9nH%2FQ1u0XPRbGrhaaTbquKwrKAWEoa0f%2BNkztDsruynPz1dx00TZb2%2BKuYs1lacfvGfH%2BQU46ICqIQ0oRSSoNYu%2Fa9j5vhG7CTbiKiyM0Bn4UF24p2uJ01bVZxFBIJ%2Fd1I8ZfO%2BPB5pC7J%2Bv%2BJu6abrcdmeq4cXSTQ%2FWL%2Fn1r1W3ry3Z3VrxQKhPHfnluWHS8sLnZF8e5b%2FdfJfYQUmxgvO1TxUXXhjO257Nms9Sv1yJPNxOfPtTy8K4S6Tqq1M77XGPJ596scckMtYAwB852vTkWumsstNCueWb6sPxMxXxn3AuTVfvfmiU%2BuR1QYfi08aJKa83Z%2BYDLLnXGLL2dTl21m60xbw%2BQEKHICQLCgCIj7WKKWz%2FAwlmLy684FJd9bDz2aLVuS3fC%2FtWLfjgdEQ9eIgjIBiQVP6oM%2B3y07ImJqn8oWh5Jh6wPX%2FTj6otkNSZWqG0kCwkCBEA%2BoKhgM1DZRK%2Fmt6YWvHzhcgfDlGzem6Vf8fFNbeLxKgVqo6xlbu8OiOpuHSncmq9zYcK925%2BXTR9BmWqVtf20c6IUCyzt5cIK6Qq%2FmgB5mVLTydqamc6E%2BzYOW1%2Bd94PJpsyNxHOXUQyo%2FFK9XvapKCAdODh6qTsao5pnliQ5TKJMXCPqU%2F0pgybFOzgyx%2FgYgwZvfI3tBwivo5SSFVpB4B8bMesjVAsyxPuewQYWtrh5ZdWmIGQs8%2B1G03JWoSiT97GGq6n9gupegoL27xDexHCFIKdUo17iyIs0R9hWQUEJQUY0SjTJH%2F0x8ZEjxGPEw%2FaVSuQhWd9PLjxHsIPyDGmd7pWEK1S%2BSVAAQUSW0cuohGQpIsp10mXOj7IwyVDKaoe1YdZCFl4nTr%2FIq8%2Fx3tT1EmLazQ36pbvZNM5YQr6H1hDD9yEDIE0pVQlS79jctTdpONYLHUoQVsgykoJWSIpKh61TlGrEuxYpL%2FvndMydKT%2BtRvZa9vZwxY6RXc4AgPMAwCCmcYnfbxzzdR2iCvVr%2FUcauTdtE2cAgIC4xc5tBIvPenfvohfTuocX1316JLerXHasVve90et9NKhQpBRVqk8RncAIUlbGbV%2BsO520%2FKDZ8LuXFv660%2FMnYehwnDpbhGoAAEnmsuGqL8TLSy6pDZMH%2FictkTQdLzcYrLoiWicaJgvI%2BoR7cRVCDJG95eb8BHHNQzPj1MY92Pm5D2Xs6XftKVDkxGXiGnlOnhBuNRNv3u6%2Fl05L8rrHxmvuyBN3lyKimkeee5l0jCBkKuKVxzn8OtWYVoc7PuzIe66h1XfP8IKji3M%2BGEdmssgbjc1eDwrDUUWQ9l1szbsuDO3JE6KyY0F7nlOH7BqP%2FCzpG89LJGt%2BVrvS%2FnSa7gVmOpSmeOoJ5kYYnmRhlf1T7Oj504Vlf1muokzQ83R%2F0vt23kgZ8NYrDE74zsiI%2B4fqvPC2a%2BJkQD7leKVqYv0tz8a8t1xzw9wF%2B6oTXhuKqK7R3cy7tpFMsulaSigWOPiMv5q82imLDtT6nhjEHm0lhAAJo6sqzZLfGBTUQ6f2rjsU9k5TW2DhJFnXV7MYIcMtXxmbZ%2FIseY0iYKzKf%2F2QH38DCCL6GaUZjy%2BPubcbeXRsiuYiZ%2F8PyzhNOXNHHtkSRAxantyWEWYvMXqU8SNMJf6tnHuyh1rFzdlRj6ydYnlg2%2BJ5cVyisoFzT9E7zmKTuExcod6j%2BBbrSzQmSAI8y9x%2Fsrrss6UhxilOXfCV76XEWz4BKDBE6YS%2FWDjm%2FWmOUZ%2FMpn1sva8crFnJx93SSryaphRtDq8zUyJ6gasOsDTPW%2Fewa5uP9BbdMF%2FySPudPiuyP1%2FjdOILG97p2BUNgqIAABaPc%2FJpH53a4btFSjHwcrXk%2B1nVl%2FQVS3Wb0oNWZm7kVJXNJyl3WDuIOvWbfKeouL8Y0AspBeQ5%2FVnPhzv82pW3A9GDd5HNk80TNknOUCx4PR6zp6i4shPaH%2F2rJzqRp4JLfWLHiLlNkYOrMe2IxhKlGo07ya9xIFzz0ELLA6UJt49Edvfa9tTKzm3d4capfXphDYjX3wQgz6gOMb2Tsy0%2FD4Z8cGrcZP8Os%2B0VU8MDzi14OBzy8XLX6Y19s0lDqyiUiznjeeFAlLPa8P1Ok32hL3fKDlTLvlaettJbwxxCguIIACBAyHJLY%2Fltp6KS8dpGryUr9vVbxDXygiKnFJKFFMj6BCiV6VQ8kHf93cbcvyzGfnvzuOsnh72oC0BQFAAAQHaexZdsTIY80e%2B5pjrsm0Hk0W7b3WHhtohLstyGIFbNGASslAKv5plyNnBzuc4gZSLyxeeP%2B%2FNzLYZKAARFAQAA9M%2FReYOhCqsVyrvonaNzjlpB7wphccpEc9HQxlBvap%2BlXl%2BaLpBMu6856%2Bag7LGdS%2F73G7P8ZJGpCgD%2FD%2Ffbv1cjCgq6AAAAAElFTkSuQmCC"></a>
   <a href="http://daisy.ece.ucdavis.edu:8080/job/gunrock/job/master/"><img src="http://daisy.ece.ucdavis.edu:8080/buildStatus/icon?job=gunrock/master" alt="Build Status"></a>
@@ -15,16 +15,16 @@
 <h1 id="gunrock-gpu-graph-analytics" align="center">Gunrock: GPU Graph Analytics</h1>
 <!-- end: gunrock's header -->
 
-**Gunrock** is a CUDA library for graph-processing designed specifically for the GPU. It uses a **high-level**, **bulk-synchronous**, **data-centric abstraction** focused on operations on vertex or edge frontiers. Gunrock achieves a balance between performance and expressiveness by coupling high-performance GPU computing primitives and optimization strategies, particularly in the area of fine-grained load balancing, with a high-level programming model that allows programmers to quickly develop new graph primitives that scale from one to many GPUs on a node with small code size and minimal GPU programming knowledge. For more details, see [Gunrock's Overview](http://gunrock.github.io/docs/#overview).
+**Gunrock** is a CUDA library for graph-processing designed specifically for the GPU. It uses a **high-level**, **bulk-synchronous**, **data-centric abstraction** focused on operations on vertex or edge frontiers. Gunrock achieves a balance between performance and expressiveness by coupling high-performance GPU computing primitives and optimization strategies, particularly in the area of fine-grained load balancing, with a high-level programming model that allows programmers to quickly develop new graph primitives that scale from one to many GPUs on a node with small code size and minimal GPU programming knowledge. For more details, see [Gunrock's Overview](https://gunrock.github.io/docs/#/gunrock/overview).
 
 
 | **Service**                    | **System**         | **Environment**                              | **Status**                                                                                                                                                                                       |
 |--------------------------------|--------------------|----------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-| [Jenkins](https://jenkins.io/) | Ubuntu 18.04.2 LTS | CUDA 10.1, NVIDIA Driver 418.39, GCC/G++ 7.3 | <a href="http://daisy.ece.ucdavis.edu:8080/blue/organizations/jenkins/gunrock/activity"><img src="http://daisy.ece.ucdavis.edu:8080/buildStatus/icon?job=gunrock/master" alt="Build Status"></a> |
+| [Jenkins](https://jenkins.io/) | Ubuntu 18.04 LTS | CUDA 10.2, NVIDIA Driver 440.33, GCC/G++ 7.4.0 | <a href="http://daisy.ece.ucdavis.edu:8080/blue/organizations/jenkins/gunrock/activity"><img src="http://daisy.ece.ucdavis.edu:8080/buildStatus/icon?job=gunrock/master" alt="Build Status"></a> |
 
 ## Quick Start Guide
 
-Before building Gunrock make sure you have **CUDA 7.5 or higher** (recommended CUDA 9 or higher) installed on your Linux system. We also support building Gunrock on docker images using the provided docker files under `docker` subdirectory. For complete build guide, see [Building Gunrock](https://gunrock.github.io/docs/#building-gunrock).
+Before building Gunrock make sure you have **CUDA Toolkit 9 or higher** (recommended CUDA 10.2) installed on your Linux system. We also support building Gunrock on docker images using the provided docker files under `docker` subdirectory. For complete build guide, see [Building Gunrock](https://gunrock.github.io/docs/#/gunrock/building_gunrock).
 
 ```shell
 git clone --recursive https://github.com/gunrock/gunrock/
@@ -36,12 +36,12 @@ make test
 
 ## Getting Started with Gunrock
 
-- To learn more about Gunrock and its programming model, see [Gunrock's Overview](http://gunrock.github.io/docs/#overview).
-- For information on building Gunrock, see [Building Gunrock](http://gunrock.github.io/docs/#building-gunrock).
-- Tutorial: [How to write a graph primitive within Gunrock?](https://gunrock.github.io/docs/developers.html)
-- Comprehensive [list](https://gunrock.github.io/docs/#gunrock-39-s-application-cases) of test applications and primitives for most the functionality of Gunrock.
-- [API Reference documentation](http://gunrock.github.io/gunrock) (generated using doxygen).
-- Find our [publications](http://gunrock.github.io/docs/#publications), [presentations](http://gunrock.github.io/docs/#presentations), and [results and analysis](http://gunrock.github.io/docs/#results-and-analysis).
+- To learn more about Gunrock and its programming model, see [Gunrock's Overview](https://gunrock.github.io/docs/#/gunrock/overview).
+- For information on building Gunrock, see [Building Gunrock](https://gunrock.github.io/docs/#/gunrock/building_gunrock).
+- Tutorial: [How to write a graph primitive within Gunrock?](https://gunrock.github.io/docs/#/devs/)
+- Comprehensive [list](https://gunrock.github.io/docs/#/gunrock/gunrock_applications) of test applications and primitives for most the functionality of Gunrock.
+- [API Reference documentation](https://gunrock.github.io/gunrock) (generated using doxygen).
+- Find our [publications](https://gunrock.github.io/docs/#/gunrock/publications_and_presentations), [presentations](https://gunrock.github.io/docs/#/gunrock/publications_and_presentations?id=presentations), and [performance analysis](https://gunrock.github.io/docs/#/analysis/results).
 
 ## Copyright and License
 

examples/BaseMakefile.mk

@@ -61,7 +61,7 @@ BOOST_INC =
 BOOST_LINK =
 ifeq ($(use_boost), 1)
     BOOST_INC = -I"/usr/local/include"
-    BOOST_LINK = -Xcompiler -DBOOST_FOUND -L"/usr/local/lib" -Xlinker -lboost_system -Xlinker -lboost_chrono -Xlinker -lboost_timer -Xlinker -lboost_filesystem
+    BOOST_LINK = -Xcompiler -DBOOST_FOUND -L"/usr/local/lib" -Xlinker -lboost_system -Xlinker -lboost_chrono -Xlinker -lboost_timer -Xlinker -lboost_filesystem -I"../../externals/rapidjson/include"
 else
     BOOST_INC = -I"../../externals/rapidjson/include"
 endif

examples/shared_libs/CMakeLists.txt

@@ -2,12 +2,15 @@
 #  Gunrock: Sub-Project C Shared Libraries
 # ------------------------------------------------------------------------
 add_executable(shared_lib_pr shared_lib_pr.c)
+set_target_properties(shared_lib_pr PROPERTIES LINKER_LANGUAGE CXX)
 target_link_libraries(shared_lib_pr gunrock)
 
 add_executable(shared_lib_sm shared_lib_sm.c)
+set_target_properties(shared_lib_sm PROPERTIES LINKER_LANGUAGE CXX)
 target_link_libraries(shared_lib_sm gunrock)
 
 add_executable(shared_lib_hits shared_lib_hits.c)
+set_target_properties(shared_lib_hits PROPERTIES LINKER_LANGUAGE CXX)
 target_link_libraries(shared_lib_hits gunrock)
 
 add_test(NAME SHARED_LIB_TEST_PAGERANK COMMAND shared_lib_pr)

examples/shared_libs/shared_lib_sm.c

@@ -7,27 +7,27 @@
 #include <gunrock/gunrock.h>
 
 int main(int argc, char *argv[]) {
-
   int num_data_nodes = 5, num_data_edges = 10;
-  int data_row_offsets[6]  = {0, 2, 6, 7, 9, 10};
+  int data_row_offsets[6] = {0, 2, 6, 7, 9, 10};
   int data_col_indices[10] = {1, 3, 0, 2, 3, 4, 1, 0, 1, 1};
 
   int num_query_nodes = 3, num_query_edges = 6;
-  int query_row_offsets[4]  = {0, 2, 4, 6};
+  int query_row_offsets[4] = {0, 2, 4, 6};
   int query_col_indices[6] = {1, 2, 0, 2, 0, 1};
+  unsigned int device = 0x01;  // CPU
 
-  int *sm_counts = (int *)malloc(sizeof(int) * num_data_nodes);
+  unsigned long *sm_counts = (unsigned long *)malloc(sizeof(unsigned long));
+  unsigned long *list_sm;
 
-  double elapsed =  sm(num_data_nodes, num_data_edges, data_row_offsets, 
-		       data_col_indices, num_query_nodes, num_query_edges,
-                       query_row_offsets, query_col_indices, 1, sm_counts); 
+  double elapsed =
+      sm(num_data_nodes, num_data_edges, data_row_offsets, data_col_indices,
+         num_query_nodes, num_query_edges, query_row_offsets, query_col_indices,
+         1, sm_counts, list_sm, device);
 
-  int node;
-  for (node = 0; node < num_data_nodes; ++node) {
-      printf("Node_ID [%d] : Number matched subgraphs: [%d]\n", node, sm_counts[node]);
-  }
+  printf("Number matched subgraphs: [%d]\n", sm_counts[0]);
 
   if (sm_counts) free(sm_counts);
+  if (list_sm) free(list_sm);
 
   return 0;
 }

examples/sm/test_sm.cu

@@ -67,7 +67,7 @@ struct main_struct {
         !quiet);
 
     // counts of matched subgraphs
-    VertexT *ref_subgraph_match = new VertexT[data_graph.nodes];
+    unsigned long *ref_subgraph_match = new unsigned long[data_graph.nodes];
     if (!quick) {
       util::PrintMsg("__________________________", !quiet);
 

gunrock/CMakeLists.txt

@@ -34,7 +34,7 @@ set(CUFILES
   app/pr/pr_app.cu
   app/hits/hits_app.cu
   app/sm/sm_app.cu
-  #app/sssp/sssp_app.cu
+  app/sssp/sssp_app.cu
   #app/louvain/louvain_app.cu
   #app/color/color_app.cu
   util/test_utils.cu

gunrock/app/enactor_types.cuh

@@ -15,7 +15,7 @@
 #pragma once
 
 #include <time.h>
-// #include <moderngpu/context.hxx>
+#include <moderngpu/context.hxx>
 #include <gunrock/util/multithreading.cuh>
 #include <gunrock/util/kernel_runtime_stats.cuh>
 #include <gunrock/util/parameters.h>
@@ -235,7 +235,6 @@ class EnactorSlice {
   typedef oprtr::OprtrParameters<GraphT, FrontierT, LabelT> OprtrParametersT;
 
   cudaStream_t stream, stream2;
-  // mgpu::standard_context_t context;
   EnactorStats<SizeT> enactor_stats;
   FrontierT frontier;
   OprtrParametersT oprtr_parameters;
@@ -273,10 +272,9 @@ class EnactorSlice {
 
     GUARD_CU(enactor_stats.Init(target));
     GUARD_CU(frontier.Init(num_queues, types, frontier_name, target));
-    GUARD_CU(oprtr_parameters.Init());
+    GUARD_CU(oprtr_parameters.Init(stream));
 
-    oprtr_parameters.stream = stream;
-    // oprtr_parameters.context = context;
+    // Create moderngpu context
     oprtr_parameters.frontier = &frontier;
     oprtr_parameters.cuda_props = cuda_properties;
     oprtr_parameters.advance_mode = advance_mode;

gunrock/app/pr/pr_test.cuh

@@ -397,9 +397,14 @@ typename GraphT::SizeT Validate_Results(
   if (ref_node_ids != NULL && ref_ranks != NULL) {
     double ref_total_rank = 0;
     double max_diff = 0;
-    VertexT max_diff_pos = nodes;
+    // TODO: Temporary workaround, max_diff_pos if set to nodes
+    // ends up causing a SegFault when ref_node_id[max_diff_pos] = some random
+    // memory access, which could be greater than num_nodes. And later in the
+    // PrintMsg, we perform `unorder_ranks[ref_node_ids[max_diff_pos]]`, which
+    // is where the SegFault will happen randomly.
+    VertexT max_diff_pos = nodes-1;
     double max_rdiff = 0;
-    VertexT max_rdiff_pos = nodes;
+    VertexT max_rdiff_pos = nodes-1;
     for (VertexT v_ = 0; v_ < nodes; v_++) {
       VertexT v = ref_node_ids[v_];
       if (util::lessThanZero(v) || v >= nodes) {

gunrock/app/problem_base.cuh

@@ -176,9 +176,15 @@ struct ProblemBase {
   cudaError_t Release(util::Location target = util::LOCATION_ALL) {
     cudaError_t retval = cudaSuccess;
     // Cleanup graph slices on the heap
-    if (sub_graphs.GetPointer(util::HOST) != NULL && num_gpus != 1) {
+    if (sub_graphs.GetPointer(util::HOST) != NULL) {
+      // We should have the same number of sub_graphs as GPUs
+      assert(num_gpus == sub_graphs.GetSize());
+
       for (int i = 0; i < num_gpus; ++i) {
-        if (target & util::DEVICE) GUARD_CU(util::SetDevice(gpu_idx[i]));
+        if (num_gpus != 1 && (target & util::DEVICE))
+        {
+          GUARD_CU(util::SetDevice(gpu_idx[i]));
+        }
         GUARD_CU(sub_graphs[i].Release(target));
       }
       GUARD_CU(sub_graphs.Release(target));

gunrock/app/sm/sm_app.cu

@@ -48,23 +48,27 @@ cudaError_t UseParameters(util::Parameters &parameters) {
  * @param[in]  query_col_indices CSR-formatted graph input query column indices
  * @param[in]  num_runs          Number of runs to perform SM
  * @param[out] subgraphs         Return number of subgraphs
+ * @param[out] list_subgraphs    Return list of subgraphs
+ * @param[in]  allocated_on      Input and output target device, by default CPU
  * \return     double            Return accumulated elapsed times for all runs
  */
 double sm(
-    const int            num_nodes,
-    const int            num_edges,
-    const int           *row_offsets,
-    const int           *col_indices,
-    const int            num_query_nodes,
-    const int            num_query_edges,
-    const int           *query_row_offsets,
-    const int           *query_col_indices,
-    const int            num_runs,
-          int           *subgraphs)
+    const int                 num_nodes,
+    const int                 num_edges,
+    const int                *row_offsets,
+    const int                *col_indices,
+    const int                 num_query_nodes,
+    const int                 num_query_edges,
+    const int                *query_row_offsets,
+    const int                *query_col_indices,
+    const int                 num_runs,
+    unsigned long            *subgraphs,
+    unsigned long            *list_subgraphs,
+    gunrock::util::Location   allocated_on = gunrock::util::HOST)
 {
     return sm_template(num_nodes, num_edges, row_offsets, col_indices,
         num_query_nodes, num_query_edges, query_row_offsets,
-        query_col_indices, num_runs, subgraphs);
+        query_col_indices, num_runs, subgraphs, list_subgraphs, allocated_on);
 }
 
 // Leave this at the end of the file