Repository navigation

Move bmm calls to stable cshim - #220

Draft

asglover wants to merge 18 commits into

mainfrom

move-bmm-calls-to-stable-cshim

Commits on Sep 13, 2026

Add stable C shim BMM prototype and migration design
asglover
committed
69463cb
Route Torch grouped GEMM through the stable BMM C shim
asglover
committed
19635e8
Expand grouped GEMM integration coverage
asglover
committed
bab7e4d
Package stable libraries in wheels and resolve editable library paths
asglover
committed
387ecf7
Keep Torch adapter helpers local to each extension
asglover
committed
d2eac4e
Verify compiled grouped GEMM accepts changing counts
asglover
committed
ee1829c
Initialize Inductor codecache for fresh-process AOTI validation
asglover
committed
e4df36c
Document production CUDA validation of the stable BMM shim
asglover
committed
4e3dcc8
Document AMD hardware validation of the stable BMM shim
asglover
committed
7399be6
human changelog
asglover
committed
99d06ea
minimize diff, remove files that should not be tracked
asglover
committed
4314273
specify cuda c shim
asglover
committed
cc02c70

Commits on Sep 14, 2026

remove target version to minimize PR, this will go on a separate PR
asglover
committed
62eae5c
simply PR my moving build issues to a separate pr
asglover
committed
15f766a
simplify pr and move unrelated fixes to separate prs
asglover
committed
e644be8
simplify pr. We're not exposing gemm as an external interface, just test it through symmetric tensor product. The test that we don't rely on blas is over testing.
asglover
committed
157de65
typo
asglover
committed
9ab016f
fmt
asglover
committed
64af319