Add GPU port tracking TUI, @noport sentinel, and double_gyre/benchmark cases. - #83
Open
edoyango wants to merge 8 commits into
Open
Add GPU port tracking TUI, @noport sentinel, and double_gyre/benchmark cases.#83edoyango wants to merge 8 commits into
edoyango wants to merge 8 commits into
Conversation
…@gpu-ported Sentinel regexes now match a trailing "! comment" (but not arbitrary trailing text, reserved for future multi-word markers) instead of silently falling through to being treated as an ordinary comment when anything followed the keyword. noport/toport regions can no longer nest — opening a new one while another is still open (pending loop-attachment or awaiting an explicit !@EnD) is now a structural error, replacing the old per-type stacks with a single open-marker slot. Also removes the unused @gpu-ported-start/-end manual tag: no source file uses it, and it doesn't belong as an independent nesting mechanism now that noport/toport are intentionally non-nesting.
…EnD The shorthand attached a marker to the immediately-following do-loop and auto-closed it at that loop's "end do", mirroring "!$omp loop". But a marker followed by a loop is ambiguous: it can't be told apart from a marker meant to span several statements or loops that happens to start with one. This caused a real false positive in MOM_barotropic.F90, where a "!@start noport" before a loop nest silently closed at that nest's "end do", leaving its actual "!@EnD noport" ~400 lines later orphaned. Every existing marker in the tree already uses an explicit, matching "!@EnD" (7/7 pairs), so the shorthand was unused. Dropping it removes the ambiguity outright and simplifies the state machine.
shallow_span() carved nested-loop bodies out of a marked region, keeping only the "directly named" outer loop's lines and leaving inner loops to their own independent classification. That made sense for the old loop-attach shorthand (a marker named exactly one following loop, so what happened inside it was ambiguous), but noport/toport are now always explicit, user-delimited "!@start"..."!@EnD" regions covering everything in between, on purpose. MOM_barotropic.F90's compact "do j=...; do i=..." style opens two loops per line, so nearly every loop in the file counted as "nested" under the old exclusion. A large noport region spanning many such loops (e.g. lines 6182-6578) ended up with most of its loop bodies silently falling back to their structural "portable" classification (shown red in the HTML report) instead of being excluded, even though they were inside the marked span. Removed the now-dead all_loop_records/parent_start bookkeeping along with shallow_span; a marked span is just the full [start, end] line range.
A full clone of MOM6-gpu/MOM6-examples pulled down every case's input and restart NetCDF files just to run benchmark_ALE. Use a blobless partial clone plus git archive to fetch and check out only that one directory, with comments explaining each git flag since these aren't everyday commands.
GPU Port CoverageOverall: 2704 / 7755 portable executed lines ported (34.9%) No executed, GPU-portable lines in the files touched by this PR (nothing in the diff was exercised by the coverage run, or none of it is GPU-portable). Full per-file / per-routine breakdown: see the "gpu-port-report" job summary and artifact. |
Collaborator
Author
|
the ci seems to be working well! there's a page for both dev/gpu and this pr atm. https://mom6-gpu.github.io/MOM6/ |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
This is a follow up to the gpu porting report PR. It's a few features that may be useful:
I can remove any of these if they're not that useful.