Refactor place.cpp Part I

[Bill-hbrhbr]

Description

PR in progress. More to come...

• Created placer context for accessing various data structures.
• Moved timing update related routines to place_timing_update.h/cpp.
• Moved connection delay related routines to place_delay_model.h/cpp.
• Merge t_placer_costs and t_placer_inverse_costs.
• Use t_annealing_state to reduce function argument list's length.

Motivation and Context

Refactor place.cpp to modularize it and reduce its length and complexity.

Types of changes

• Bug fix (change which fixes an issue)
• New feature (change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

[Bill-hbrhbr]

Hi @vaughnbetz @HackerFoo , this is a PR in progress. Please drop by and leave some comments when you are available so that I can fix potential issues as early as possible!

[vaughnbetz]

Looks good Bill, thanks!
The style and commenting looks good. My two main comments (details embedded) are:

• Please run cpu tests regularly and attach them as you do this refactoring. I gave one example above of a particularly tricky run time impact (pointer alias analysis) which is impacted by the scope of the variables used in key loops and if their address has ever been taken. Hard to reason about, so good to check regularly.
• Splitting into multiple files is very good. I think it's good to think about what code should go in each. For example, is place.cpp only for annealing? (I think it should be). place_utils is for utilities that are reasonably general and may be useful for more than one placement algorithm (e.g. anneal and analytic). Putting those guidelines in the comments will help future developers stick with your organization.

[vaughnbetz]

Good to add a comment about what should go in this file. (This is an annealing based, cluster-level placer. This file ...)

[Bill-hbrhbr]

Titan benchmark run001@commit bdaf63c: https://drive.google.com/file/d/1F1U-P6b6LLMa3rCzoPEeeJZjWBgqaiGE/view?usp=sharing

Comment: No runtime or performance degradation.

[vaughnbetz]

Great, thanks Bill. Should this be merged now?

[vaughnbetz]

I suggest addressing the comments I have in the review (which aren't too big, pretty much all commenting I think) and then we merge. After that you can do another PR for more refactoring (if you have the time), but probably best to do this in stages. What do you think?

[Bill-hbrhbr]

Doing this in multiple stages sounds good. I'll just reference this PR in the new PR.

[vaughnbetz]

Sounds good. Let me know when the comments above are addressed and I'll merge this.

[Bill-hbrhbr]

Hi @vaughnbetz, I added more comments for the override delay model class. Regarding comments on what should be included in place.cpp, I think maybe it's more suitable for me to do that once the refactorization is mostly done.

QoR test can be skipped since I only changed comments. This PR can be merged now. The next step for me is to shorten the argument lists for various core placer routiners.

[vaughnbetz]

Great, thanks Bill.