{"id":25614382,"url":"https://github.com/stckvrflw/pem-spgemm","last_synced_at":"2026-04-17T13:31:40.478Z","repository":{"id":278781872,"uuid":"932537910","full_name":"stckvrflw/pem-spgemm","owner":"stckvrflw","description":"pemSpGEMM - An Improved SpGEMM Algorithm","archived":false,"fork":false,"pushed_at":"2025-07-17T15:48:55.000Z","size":1633,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-07-17T18:04:10.044Z","etag":null,"topics":["cpp","cuda"],"latest_commit_sha":null,"homepage":"","language":"C","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/stckvrflw.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-02-14T04:27:23.000Z","updated_at":"2025-07-17T15:48:58.000Z","dependencies_parsed_at":"2025-02-21T17:33:57.455Z","dependency_job_id":"e34bee9f-ebaa-40d0-9e02-6d901635e54a","html_url":"https://github.com/stckvrflw/pem-spgemm","commit_stats":null,"previous_names":["stckvrflw/pem-spgemm"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/stckvrflw/pem-spgemm","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stckvrflw%2Fpem-spgemm","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stckvrflw%2Fpem-spgemm/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stckvrflw%2Fpem-spgemm/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stckvrflw%2Fpem-spgemm/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/stckvrflw","download_url":"https://codeload.github.com/stckvrflw/pem-spgemm/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stckvrflw%2Fpem-spgemm/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31931243,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-17T12:37:54.787Z","status":"ssl_error","status_checked_at":"2026-04-17T12:37:25.095Z","response_time":62,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["cpp","cuda"],"created_at":"2025-02-22T02:19:04.661Z","updated_at":"2026-04-17T13:31:40.473Z","avatar_url":"https://github.com/stckvrflw.png","language":"C","funding_links":[],"categories":[],"sub_categories":[],"readme":"# pem-spgemm\n#### **BETA**\nFinal Assignment Project - SpGEMM algorithm in CUDA  \nBy Petrus E. Manurung  \n2025\n\nAn Improved Sparse General Matrix-Matrix Multiplication (SpGEMM) algorithm.  \nImproving upon TileSpGEMM by eliminating atomics and better cache utilization on step 2 and step 3.  \nAnother improvement includes native GPU implementation of conversion from .mtx to Tiled CSR intermediate format. \n\nLibraries used:\n* [thrust][thrust]\n* [rmm][rapidsrmm]\n* [fast_matrix_market][fmm]\n* [nsparse][nsparse]\n\nOther resources:\n* Sparse matrices from [suiteSparse][suitesparse]\n\nReference:\n1. [TileSpGEMM -- **Niu et al.**](https://doi.org/10.1145/3503221.3508431)\n\n\nEnvironment:\n* CPU       : 11th Gen Intel(R) Core(TM) i7-11800H\n* GPU       : NVIDIA Corporation GA104M [GeForce RTX 3080 Mobile / Max-Q 8GB/16GB]\n* OS        : Gentoo Linux\n* Kernel    : 6.13.8-zen1\n* CUDA      : 12.8\n* driver    : 570.144\n* gcc       : 14.2.1 20241221\n\nHow to compile:  \n1. clone this repository\n2. get rapidsrmm v24.12.00 from [rapidsrmm] and extract to pem-spgemm (cloned repo)\n3. get fastmatrixmarket v1.7.6 from [fmm] and extract to pem-spgemm (cloned repo)\n4. run \"make\"\n\nHow to use:\n* A^2   : ./pemspgemm \"path-to-.mtx-file\" [0/1] \n* A*At  : ./pemspgemm \"path-to-.mtx-file\" [0/1] 1  \n*** 0 to skip saving result (in COO) to file, 1 to save to /tmp  \n*** since /tmp is in RAM, make sure there is enough space.  \n(e.g. result from A^2 of webbase-1M can cost more than 1.5GiB)  \n*** no quote on path to mtx-file  \n\nTo reproduce: GPU with sm_86  \nif using different GPU, change the \"code\" part in NVCC_FLAGS in the Makefile.  \nKeep \"compute_61\" unchanged.\n\nBenchmark result is saved in 'pemspgemm_benchmark_result.csv' file  \nheader for the csv:  \nmatrix,flop,C_nnz,compression_ratio,A_conversion_kernel_time,B_conversion_kernel_time,total_conversion_overhead_time,step1_time,step2_time,step3_time,pem_spgemm_time,pem_spgemm_kernel_time,pem_spgemm_malloc_time,Gflops\n\n[ansorge]: https://github.com/RichardAns/CUDA-Programs\n[thrust]: https://developer.nvidia.com/thrust\n[rapidsrmm]: https://github.com/rapidsai/rmm\n[cusparse]: https://developer.nvidia.com/cusparse\n[fmm]: https://github.com/alugowski/fast_matrix_market\n[suitesparse]: https://sparse.tamu.edu\n[nsparse]: https://github.com/EBD-CREST/nsparse","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fstckvrflw%2Fpem-spgemm","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fstckvrflw%2Fpem-spgemm","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fstckvrflw%2Fpem-spgemm/lists"}