Add Ampere (sm_80/sm_86) extension - #5
Open
alesha-pro wants to merge 2 commits into
Open
Conversation
Vocabulary entries, 4 source docs, 3 hardware pages and a Hopper to Ampere backport migration page. All additive, upstream files touched only to register the new tags and aliases. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Also make the cp-async snippet self-contained and fix the mma snippet so both compile standalone under nvcc -arch=sm_86. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds an Ampere layer so the same knowledge base is usable on A100 (sm_80) and GA10x cards (sm_86: RTX 3090/3080, RTX A6000, A40).
Everything is additive. Upstream files are touched only to register the new vocabulary, so future merges stay trivial.
New pages
wiki/migration/hopper-to-ampere-backport.md, the entry point: instruction replacement table, capacity re-planning, scheduling paradigm, ncu checklistwiki/hardware/:cp-async,mma-sync-ampere,ampere-memory-modelsources/docs/: Ampere tuning guide, GA102 whitepaper, PTX ISA for Ampere, CUTLASS SM80 supportVocabulary and tooling
data/tags.yaml: architecturessm80,sm86; hardware featurescp-async,mma-sync,l2-persistencedata/aliases.yaml: alias groups for the same termsscripts/query.pynow also scoresarchitectures,from_archandto_arch, so--architecture sm86and the migration page are actually reachableVerification
scripts/validate.pypasses on the full tree (2795 files). Snippets compile undernvcc -arch=sm_86on a 4x RTX 3090 machine. Facts come from the NVIDIA Ampere tuning guide (CC 8.0/8.6 table), the GA102 whitepaper v2.1, the PTX ISA and CUTLASS.The Blackwell-first rule is untouched, since it constrains Hopper-only pages, and every new page carries the frontmatter its type requires.
Happy to adjust the naming or split this up if it does not fit the direction you want for the repo.