vine: count the available resources for a task on a worker

[JinZhou5042]

Proposed Changes

Calculate the usable resources for a task on a worker and check it earlier. The avoids resource allocations if the worker doesn't have usable resources for a task at all.

This speeds up my workflows slightly, for about 1.5%

Merge Checklist

The following items must be completed before PRs can be merged.
Check these off to verify you have completed all steps.

• make test Run local tests prior to pushing.
• make format Format source code to comply with lint policies. Note that some lint errors can only be resolved manually (e.g., Python)
• make lint Run lint on source code prior to pushing.
• Manual Update: Update the manual to reflect user-visible changes.
• Type Labels: Select a github label for the type: bugfix, enhancement, etc.
• Product Labels: Select a github label for the product: TaskVine, Makeflow, etc.
• PR RTM: Mark your PR as ready to merge.

[btovar]

This change is fragile to memory leaks. There are some returns that do not free the struct rmsummary created. Unless there is a strong reason, I'd advise against it.

[JinZhou5042]

The reason I do this is that even though most allocations work fine in #4035 and tasks run pretty densely, the manager still wastes a lot of time checking tasks when there aren’t any usable cores.

The most common issue in check_worker_against_task is that the worker doesn’t have a free core. But it still goes through the whole resource allocation process, and this keeps happening for a bunch of tasks each time, based on the depth set by q->attempt_schedule_depth.

Task scheduling seems to be an expensive part of the main loop, we could save a lot of time by making this more efficient, and then the manager has more time to do other stuff like output retrieving and temp file replication.

[btovar]

But this change is not reducing the number of computations? It is just moving code around and creates mem leaks?

What about adding a flag to the worker like has_free_resources. On reaping a task, has_free_resources is set to 1. On scheduling a task to the worker is set to 0 if any resource in use matches the total * overcommitment (unless it starts at 0, like gpus).

[JinZhou5042]

This is a smart way!

[JinZhou5042]

An alternative in #4039