cache get_platform

[oliver-sanders]

Spotted that the get_platform_from_name function was using a lot of CPU.

Turns out that it was called 7662 times for this sample workflow:

[task parameters]
    a = 1..5
    b = 1..5
    c = 1..5

[scheduling]
    initial cycle point = 2000
    final cycle point = 20000110T00Z
    runahead limit = P5
    [[queues]]
        [[[default]]]
            limit = 25
    [[graph]]
        P1D = """
            <a> => <b> => <c>
            <b>[-P1D] => <b>
        """

[runtime]
    [[<a>, <b>, <c>]]
        script = true

Embarrassingly this workflow only uses the localhost platform.

Added lru_cache to the function to reduce the impact:

Before:

After:

Caching does have an overhead, but inspecting the call tree it would appear this saved ~3 seconds of a ~35s run, which is ~10% of CPU!

Check List

• I have read CONTRIBUTING.md and added my name as a Code Contributor.
• Contains logically grouped changes (else tidy your branch by rebase).
• Does not contain off-topic changes (use other PRs for other changes).
• Applied any dependency changes to both setup.cfg and conda-environment.yml.
• Tests are included (or explain why tests are not needed).
• CHANGES.md entry included if this is a change that can affect users
• Cylc-Doc pull request opened if required at cylc/cylc-doc/pull/XXXX.
• If this is a bug fix, PRs raised to both master and the relevant maintenance branch.

[oliver-sanders]

Dammit the bad_hosts make a mess of this. Will need to split the function into the cacheable and non-cacheable parts.

[oliver-sanders]

Closing for now pending two changes:

1. A reduction in the number of get_platform calls (done).
2. A refactor of the get_platform logic to separate the cachable and non-cachable parts.