This repository was archived by the owner on Apr 28, 2023. It is now read-only.
insertCopiesUnder: drop iteration over tensor dimensions of size 1 up front #355
Conversation
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
skimo-openhub
force-pushed
the
pr/copies_under
branch
from
4f534b1 to
3195113
Compare
added 2 commits
April 26, 2018 16:41
The promoted declaration information associated to a given identifier of a promoted tensor reference group is accessed in two locations, but only one of them has an explicit check to see if this information is available. Extract out a promotedDecl that performs the operation (with check) consistently. The main motivation, however, is to reduce the clutter inside mapCopiesToThreads.
… front When mapping the code for copying arrays to/from shared memory to threads, the band members of the copying schedule that iterate over a tensor dimension of size 1 do not get mapped to a thread identifier. This was handled by code in mapCopiesToThreads that is relatively complicated and that needs to take into account situations that may never arise in practice. It is much simpler to make sure that those band members do not even appear in the copying schedule, by removing them while the copying schedule is being constructed in insertCopiesUnder. Note that insertCopiesUnder now relies on promotedDecls_, so the call to insertCopiesUnder needs to be moved after the construction of promotedDecls_.
skimo-openhub
force-pushed
the
pr/copies_under
branch
from
3195113 to
d984f4e
Compare
ftynse
approved these changes
Apr 26, 2018
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
When mapping the code for copying arrays to/from shared memory to threads,
the band members of the copying schedule that iterate over
a tensor dimension of size 1 do not get mapped to a thread identifier.
This was handled by code in mapCopiesToThreads that is relatively
complicated and that needs to take into account situations
that may never arise in practice.
It is much simpler to make sure that those band members
do not even appear in the copying schedule, by removing
them while the copying schedule is being constructed in
insertCopiesUnder.