{"id":51317774,"url":"https://github.com/abhiksark/gluon-by-example","last_synced_at":"2026-07-01T09:04:25.122Z","repository":{"id":363887100,"uuid":"1265223134","full_name":"abhiksark/gluon-by-example","owner":"abhiksark","description":"Learn Triton's Gluon by example — the same GPU kernels written in Triton and Gluon, benchmarked","archived":false,"fork":false,"pushed_at":"2026-06-10T20:00:49.000Z","size":129,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-10T20:22:05.044Z","etag":null,"topics":["cuda","deep-learning","gluon","gpu","gpu-kernels","triton","tutorial"],"latest_commit_sha":null,"homepage":null,"language":"Python","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/abhiksark.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,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-10T15:20:02.000Z","updated_at":"2026-06-10T20:01:52.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/abhiksark/gluon-by-example","commit_stats":null,"previous_names":["abhiksark/gluon-by-example"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/abhiksark/gluon-by-example","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhiksark%2Fgluon-by-example","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhiksark%2Fgluon-by-example/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhiksark%2Fgluon-by-example/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhiksark%2Fgluon-by-example/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/abhiksark","download_url":"https://codeload.github.com/abhiksark/gluon-by-example/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/abhiksark%2Fgluon-by-example/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34999810,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-01T02:00:05.325Z","response_time":130,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"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":["cuda","deep-learning","gluon","gpu","gpu-kernels","triton","tutorial"],"created_at":"2026-07-01T09:04:24.352Z","updated_at":"2026-07-01T09:04:25.114Z","avatar_url":"https://github.com/abhiksark.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Gluon by Example\n\n**Learn [Triton](https://github.com/triton-lang/triton)'s Gluon, the new\nlow-level GPU kernel language, by writing the same kernels in Triton and\nGluon side by side, and benchmarking both.**\n\n\u003e Not MXNet Gluon. Not GluonHQ/JavaFX. This is about\n\u003e `triton.experimental.gluon`, the explicit-layout GPU language in the Triton\n\u003e compiler stack.\n\n![vector add: torch vs Triton vs Gluon](benchmarks/charts/vector_add-nvidia-rtx-a6000.png)\n\n## Quickstart\n\n```bash\ngit clone https://github.com/abhiksark/gluon-by-example\ncd gluon-by-example\npip install -e \".[dev]\"\npytest tests/ -v                          # correctness on your GPU\npython chapters/01-vector-add/bench.py    # benchmark on your GPU\n```\n\nRequires: NVIDIA GPU (Ampere or newer), CUDA PyTorch, Triton ≥ 3.7.\n\n## Chapters\n\n| # | Kernel | Backends | Status |\n|---|--------|----------|--------|\n| [1](chapters/01-vector-add/) | vector add | Triton + Gluon | ✅ |\n| [2](chapters/02-softmax/) | fused softmax | Triton | ✅ |\n| [3](chapters/03-softmax-gluon/) | softmax | Gluon | ✅ |\n| [4](chapters/04-matmul/) | matmul | Triton | ✅ |\n| [5](chapters/05-matmul-gluon/) | matmul | Gluon (mma_v2) | ✅ |\n| 6 | flash attention | Triton | planned |\n| 7 | flash attention | Gluon (TMA + warp specialization) | planned |\n\n## Which Gluon features run on which GPU?\n\nVerified against the official Gluon tutorial gates (Triton main, 2026-06):\n\n| Gluon feature | Requires | RTX 30/40-series, A6000 | RTX 5090 | H100 | B200 |\n|---|---|---|---|---|---|\n| Core: layouts, `cp.async`, `mma_v2` | CC ≥ 8.0 | ✅ | ✅ | ✅ | ✅ |\n| TMA, warp specialization | CC major ≥ 9 | ❌ | ✅ | ✅ | ✅ |\n| wgmma | CC major == 9 (Hopper only) | ❌ | ❌ | ✅ | ❌ |\n| tcgen05 / tensor memory | CC major == 10 (sm_100) | ❌ | ❌ | ❌ | ✅ |\n\n(Yes, wgmma is Hopper-*only* and tcgen05 is datacenter-Blackwell-only.\nConsumer Blackwell has neither: it has TMA. This table took actual source\nreading to assemble; that's why it's here.)\n\n## Gluon resources\n\n- [Official Gluon tutorials](https://github.com/triton-lang/triton/tree/main/python/tutorials/gluon)\n- [Gluon docs](https://triton-lang.org/main/gluon/index.html)\n- [Triton repo](https://github.com/triton-lang/triton)\n\n## Layout\n\n```\nchapters/    one directory per chapter: explainer README + bench script\nsrc/         installable package: gluon_by_example.{triton_impl,gluon_impl}\ntests/       pytest, parametrized over backends, vs PyTorch references\nbenchmarks/  committed CSV results + charts, tagged by GPU\ntools/       shared chart generator\n```\n\n*Gluon is experimental and its API moves fast. Each chapter records the\nTriton version it was written against. Currently: 3.7.0.*\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fabhiksark%2Fgluon-by-example","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fabhiksark%2Fgluon-by-example","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fabhiksark%2Fgluon-by-example/lists"}