[inductor] Make AOT CPU Inductor work in fbcode #106225
Conversation
Summary: This diff has a couple of hacks to make inductor-CPU work for AOT codegen in fbcode: - We need to add the CUDA link flags; AOT-Inductor is specialized for CUDA right now and uses a lot of `at::cuda` stuff. We should do a proper AOT CPU at some point but this unblocks perf measurement. - Add an include path to the cpp_prefix. It's kind of hilarious; we remove the include path for remote execution, but then for AOT we need it back. 🤷 Test Plan: internal test Differential Revision: D47882848 fbshipit-source-id: fb2eb9a70128fb6f0465e0c64ef9c9a5fd2131ac
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/106225
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 83b417a: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D47882848 |
|
LGTM! |
|
@pytorchbot merge -f "all tests passed" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
6 participants
Summary:
This diff has a couple of hacks to make inductor-CPU work for AOT codegen in fbcode:
We need to add the CUDA link flags; AOT-Inductor is specialized for CUDA
right now and uses a lot of
at::cudastuff. We should do a proper AOT CPUat some point but this unblocks perf measurement.
Add an include path to the cpp_prefix. It's kind of hilarious; we remove the
include path for remote execution, but then for AOT we need it back. 🤷
Test Plan: internal test
Differential Revision: D47882848
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @Xia-Weiwen @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @ngimel @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov