More Sway compiler optimizations

[xunilrj]

Description

This PR is a prequel to #7015, trying to optimize the compiler to alleviate the hit of having more items, impl etc...

We have here two optimizations:
1 - We spend a lot of time counting newlines when converting byte offsets to LineCol. Now, we calculate line offsets just once, and use binary search later to find which line a byte offset is at.
2 - QueryEngine::get_programs_cache_entry was cloning the whole TyProgram. That is why did_change_with_caching was always getting worse, as the program was increasing in size. Now all compilation stages are behind Arc, which makes the clone free.

Analysis

Putting a dbg!(...) like the image below, and calling counts (https://github.com/nnethercote/counts).

cargo bench -- traverse 2>&1 | grep "sway-types/src/span.rs:29:9" | counts

I get the following results:

972102 counts
(  1)   156720 (16.1%, 16.1%): [sway-types/src/span.rs:29:9] self.pos = 0
(  2)    15900 ( 1.6%, 17.8%): [sway-types/src/span.rs:29:9] self.pos = 104
(  3)    15840 ( 1.6%, 19.4%): [sway-types/src/span.rs:29:9] self.pos = 107
(  4)     2280 ( 0.2%, 19.6%): [sway-types/src/span.rs:29:9] self.pos = 19281
(  5)     2280 ( 0.2%, 19.9%): [sway-types/src/span.rs:29:9] self.pos = 19285
(  6)     2280 ( 0.2%, 20.1%): [sway-types/src/span.rs:29:9] self.pos = 19287
(  7)     2280 ( 0.2%, 20.3%): [sway-types/src/span.rs:29:9] self.pos = 19292
(  8)     2280 ( 0.2%, 20.6%): [sway-types/src/span.rs:29:9] self.pos = 19323
(  9)     2280 ( 0.2%, 20.8%): [sway-types/src/span.rs:29:9] self.pos = 19327
( 10)     2280 ( 0.2%, 21.0%): [sway-types/src/span.rs:29:9] self.pos = 19329
( 11)     2280 ( 0.2%, 21.3%): [sway-types/src/span.rs:29:9] self.pos = 19334
( 12)      870 ( 0.1%, 21.4%): [sway-types/src/span.rs:29:9] self.pos = 4285
...

This means that line_col is being called 972k times. 16% is for position zero, which should be trivial. The rest will iterate the whole file source code to count number of lines. Making the real complexity of the work here something like O(qty * self.pos). And some values of self.pos are not trivial at all.

Checklist

• I have linked to any relevant issues.
• I have commented my code, particularly in hard-to-understand areas.
• I have updated the documentation where relevant (API docs, the reference, and the Sway book).
  • If my change requires substantial documentation changes, I have requested support from the DevRel team
• I have added tests that prove my fix is effective or that my feature works.
• I have added (or requested a maintainer to add) the necessary Breaking* or New Feature labels where relevant.
• I have done my best to ensure that my PR adheres to the Fuel Labs Code Review Standards.
• I have requested a review from the relevant team or maintainers.

[codspeed-hq]

CodSpeed Performance Report

Merging #7093 will improve performances by ×140

_{Comparing xunilrj/sway-optimizations-2 (8fed389) with master (7cb7809)}

Summary

⚡ 11 improvements
✅ 11 untouched benchmarks

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	did_change_with_caching	520.9 ms	3.8 ms	×140
⚡	traverse	230.9 ms	48.3 ms	×4.8
⚡	code_action	28.3 ms	7.2 ms	×3.9
⚡	completion	21.1 ms	5.3 ms	×3.9
⚡	document_symbol	19.1 ms	3.1 ms	×6.1
⚡	find_all_references	47.4 ms	5.2 ms	×9.1
⚡	highlight	44.7 ms	5.8 ms	×7.7
⚡	inlay_hints	18.7 ms	2.2 ms	×8.6
⚡	rename	47.4 ms	5.2 ms	×9.1
⚡	parent_decl_at_position	18.8 ms	3.1 ms	×6.1
⚡	tokens_at_position	18.8 ms	3.1 ms	×6.1

[JoshuaBatty]

This is really neat. Amazing performance increase. Hats off!

[tritao]

Crazy speedup, amazing 🚀

Just to note that "compile" and "did_change_with_caching" steps on LSP benchmark seem to have gotten around 7-9% slower though, seems unlikely to have been caused by the changes though given the speedups?

[xunilrj]

Crazy speedup, amazing 🚀

Just to note that "compile" and "did_change_with_caching" steps on LSP benchmark seem to have gotten around 7-9% slower though, seems unlikely to have been caused by the changes though given the speedups?

I also managed to optimize "did_change_with_caching".

[Voxelot]

wut 🤯