{"id":16597856,"url":"https://github.com/szilard/gbm-meltdown","last_synced_at":"2026-04-29T23:33:13.617Z","repository":{"id":74911253,"uuid":"116605805","full_name":"szilard/GBM-meltdown","owner":"szilard","description":"The Effect of the Linux Kernel Page-Table Isolation (KPTI) Patch (Meltdown Vulnerability) on GBMs","archived":false,"fork":false,"pushed_at":"2018-01-10T13:05:01.000Z","size":18,"stargazers_count":3,"open_issues_count":0,"forks_count":1,"subscribers_count":3,"default_branch":"master","last_synced_at":"2025-03-06T19:44:04.810Z","etag":null,"topics":["gbm","h2o","kpti","lightgbm","linux-kernel","machine-learning","meltdown","xgboost"],"latest_commit_sha":null,"homepage":"","language":"R","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/szilard.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"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}},"created_at":"2018-01-07T23:00:36.000Z","updated_at":"2018-06-04T07:44:40.000Z","dependencies_parsed_at":"2023-07-07T05:31:01.831Z","dependency_job_id":null,"html_url":"https://github.com/szilard/GBM-meltdown","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/szilard/GBM-meltdown","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szilard%2FGBM-meltdown","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szilard%2FGBM-meltdown/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szilard%2FGBM-meltdown/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szilard%2FGBM-meltdown/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/szilard","download_url":"https://codeload.github.com/szilard/GBM-meltdown/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szilard%2FGBM-meltdown/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32448400,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-29T22:27:22.272Z","status":"ssl_error","status_checked_at":"2026-04-29T22:10:49.234Z","response_time":110,"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":["gbm","h2o","kpti","lightgbm","linux-kernel","machine-learning","meltdown","xgboost"],"created_at":"2024-10-12T00:06:51.324Z","updated_at":"2026-04-29T23:33:13.600Z","avatar_url":"https://github.com/szilard.png","language":"R","funding_links":[],"categories":[],"sub_categories":[],"readme":"\n## GBM Meltdown\n\n### The Effect of the Linux Kernel Page-Table Isolation (KPTI) Patch (Meltdown Vulnerability) on GBMs\n\n\nMeltdown is a serious vulnerability in Intel CPUs made public January 3, 2018. The KPTI patch that\nmitigates it is believed to cause a performance hit of 5-30% in various software. \n\nGBM training is expected not to be significantly affected as it makes few system calls, which I'm\nchecking below in a setup similar to this [simple benchmark](https://github.com/szilard/GBM-perf/). \n\nThe data is 1M records of the airline dataset, the GBMs have 100 trees, depth 10 and learning rate 0.1.\nIt is run on Ubuntu 16.04 on EC2 r4.8xlarge (Intel Xeon CPU E5-2686 v4, 32 cores, 250GB RAM), \nkernel versions 4.14.10 (before patch) and 4.14.12 (after patch). \nThe software versions for h2o, xgboost and lightgbm are the same as in the benchmark linked above. \nThe code run is [here](run/) (same as in the previous benchmark).\n\n\nGBM implementation    |   Time before (sec)  | Time after (sec)\n----------------------|----------------------|------------------\nh2o                   |   21.9 - 24.5        |  24.0 - 25.3\nxgboost               |   23.3 - 23.7        |  23.3 - 23.6\nlightgbm              |   5.5 - 5.9          |  5.4-5.7\n\nDetailed results [here](results-GBM.txt). So, **GBM training is not really affected by the KPTI patch.**\n\n\n#### Other Results\n\nglmnet (linear model, not a GBM) is also unaffected (time before 5.1 - 5.7 sec, after 5.1 - 5.7 sec).\n\nNo impact on Simon Urbanek's [R benchmark](https://r.research.att.com/benchmarks/) \n(containing various matrix operations, FFT, sorting etc.) either. Time before 46.6 sec, time after 46.1 sec\n(or 6.4 sec vs 6.3 sec with optimized BLAS) (on m5.xlarge), \nsee the breakdown e.g. random matrix, sorting, matrix cross-product, linear regression, \neigenvalue, determinant, Cholesky etc. [here](results-R-Simon.txt).\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fszilard%2Fgbm-meltdown","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fszilard%2Fgbm-meltdown","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fszilard%2Fgbm-meltdown/lists"}