Preserve runtime allocations during constant propagation (#21897) - #21897
Preserve runtime allocations during constant propagation (#21897)#21897mcremon-meta wants to merge 1 commit into
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21897
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit ab0a089 with merge base a1eee71 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
|
@mcremon-meta has exported this pull request. If you are a Meta employee, you can view the originating Diff in D116349009. |
This PR needs a
|
Summary: Treat executorch.exir.memory.alloc as intrinsically non-constant so constant propagation never executes and lifts runtime-planned workspaces into serialized model data. This applies even when callers replace the default skip targets. Add mirrored framework coverage for default and custom skip sets, plus a Cortex-M fused depthwise-convolution regression that verifies scratch remains a uint8 runtime allocation rather than a lifted constant. Reviewed By: angelayi Differential Revision: D116349009
meta-codesync
Bot
changed the title
72d1fd5 to
ab0a089
Compare
2 participants
Summary:
Treat executorch.exir.memory.alloc as intrinsically non-constant so constant propagation never executes and lifts runtime-planned workspaces into serialized model data. This applies even when callers replace the default skip targets.
Add mirrored framework coverage for default and custom skip sets, plus a Cortex-M fused depthwise-convolution regression that verifies scratch remains a uint8 runtime allocation rather than a lifted constant.
Reviewed By: angelayi
Differential Revision: D116349009