-
Notifications
You must be signed in to change notification settings - Fork 55
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[MLIR][OpenMP] Introduce host_eval clause to omp.target #178
[MLIR][OpenMP] Introduce host_eval clause to omp.target #178
Conversation
44b6230
to
719e50c
Compare
719e50c
to
bc6485b
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you for this PR, @skatrak. I have one nit in the code and 2-3 suggestions in the docs part of the PR. Of course, the suggestions for the docs are entirely subjective, so please take with a grain of salt. Otherwise, LGTM.
bc6485b
to
ad391a1
Compare
Thanks for the review @bhandarkar-pranav, your comments should be addressed now. |
ad391a1
to
595f1ae
Compare
This patch defines a map-like clause named `host_eval` used to capture host values for use inside of target regions on restricted cases: - As `num_teams` or `thread_limit` of a nested `omp.target` operation. - As `num_threads` of a nested `omp.parallel` operation or as bounds or steps of a nested `omp.loop_nest`, if it is a target SPMD kernel. This replaces the following `omp.target` arguments: `trip_count`, `num_threads`, `num_teams_lower`, `num_teams_upper` and `teams_thread_limit`.
595f1ae
to
f33594a
Compare
This patch defines a map-like clause named
host_eval
used to capture host values for use inside of target regions on restricted cases:num_teams
orthread_limit
of a nestedomp.target
operation.num_threads
of a nestedomp.parallel
operation or as bounds or steps of a nestedomp.loop_nest
, if it is a target SPMD kernel.This replaces the following
omp.target
arguments:trip_count
,num_threads
,num_teams_lower
,num_teams_upper
andteams_thread_limit
.Since it's an entry block argument-generating clause, custom region parsers/printers and the
BlockArgOpenMPOpInterface
are updated.