To determine parameters of the matrix multiplication, we check RAW dependencies that can be expressed using only reduction dependencies. Consequently, we should check the reduction dependencies, if this is the case.
Details
- Reviewers
Meinersbur grosser jdoerfert - Commits
- rGb196055c0cb8: Check reduction dependencies in case of the matrix multiplication optimization
rPLO294836: Check reduction dependencies in case of the matrix multiplication optimization
rL294836: Check reduction dependencies in case of the matrix multiplication optimization
Diff Detail
- Repository
- rL LLVM
Event Timeline
Otherwise, this looks good to me.
Some of the lower changes seem to be mostly stylistic. In case they are, I suggest to commit them ahead of time and then add the actual functionality change separately. In case it introduces a bug, the surface is smaller.
lib/Transform/ScheduleOptimizer.cpp | ||
---|---|---|
772 ↗ | (On Diff #87968) | Would it make sense to just compute the union of both? |
lib/Transform/ScheduleOptimizer.cpp | ||
---|---|---|
782 ↗ | (On Diff #87968) | This change seems unrelated |
2017-02-10 13:33 GMT+05:00 Tobias Grosser via Phabricator <reviews@reviews.llvm.org>:
grosser added a comment.
Otherwise, this looks good to me.
Some of the lower changes seem to be mostly stylistic. In case they are, I suggest to commit them ahead of time and then add the actual functionality change separately. In case it introduces a bug, the surface is smaller.
2017-02-10 13:43 GMT+05:00 Sven Verdoolaege <skimo-polly@kotnet.org>:
You're computing the domain of a schedule just to get its space?!??
Get the domain of the space instead.
2017-02-10 22:26 GMT+05:00 Michael Kruse via Phabricator <reviews@reviews.llvm.org>:
This change seems unrelated
Thanks for the comments! Stylistic changes can be found in r294831.