Combine all changes

[TimothyMakkison]

All my local/pending changes, I have a couple more ideas but this should give a general feel for the final performance

Method	Mean	Error	StdDev	Gen0	Gen1	Allocated
Default_CodeFormatter_Tests	64.72 ms	0.548 ms	0.512 ms	1666.6667	1000.0000	18.33 MB
Default_CodeFormatter_Complex	137.83 ms	2.730 ms	4.708 ms	3000.0000	1000.0000	30.78 MB

[Hona]

This looks incredible @TimothyMakkison

Did you have a benchmark before and after any of your changes so we can estimate the total change?

[TimothyMakkison]

This looks incredible @TimothyMakkison

Thanks 😄 optimisations are my favourite changes, very satisfying. OOC how did you find this PR?

Did you have a benchmark before and after any of your changes so we can estimate the total change?

See #1552 for benchmark performance before I made any changes. The before section of #1550 is the current performance.

Note that my changes only apply to the formatting logic, csharpier uses a caching system and only formats files that changed. This will be most beneficial when working on several medium to large files.

IIRC the bulk of the time running dotnet format . is mostly IO

Original #1552

Method	Mean	Error	StdDev	Median	Gen0	Gen1	Gen2	Allocated
Default_CodeFormatter_Tests	233.3 ms	4.63 ms	8.23 ms	229.7 ms	11000.0000	4000.0000	1000.0000	101.41 MB
Default_CodeFormatter_Complex	433.7 ms	8.53 ms	7.56 ms	433.4 ms	20000.0000	5000.0000	1000.0000	182.44 MB

[Hona]

I've been meaning to learn about performance and improving that in .NET. I was curious if you had any resources/courses that you'd recommend?

I have been 'watching' this repo on GitHub so I just see the list of PRs in my GH notifications, and I've been seeing all your awesome PRs in awe.

Thanks for that info and the old link - looks like pretty awesome speedups. I can't wait :)

[TimothyMakkison]

I don't know what your day job is, but most day to day CRUD apps are bottlenecked by: IO, HTTP calls, bad mulithreading and database calls/SQL.

These optimisation are more applicable for libraries or high performance scenarios.

I'm not an expert but I'd recommend:

• Being familiar with Span<T>/ReadOnlySpan<T>, it can be used instead of copying string or T[].
  • IMO ReadOnlySpan<T> can be semantically better than arrays.
• Learning how to throw together a benchmark in BenchmarkDotNet
• Be able to use a profiler
• Looking at the internals of async, Linq, Func/closures.
  • It's pretty cool looking at how code is lowered and turned into IL code. Try using Sharplab or your IDEs built IL viewer
• Reading Stephen Toub's performance improvements in .NET
  • These are great because they show real world examples of how to identify, profile and apply performance techniques such as stackalloc, ValueListBuilder etc
• Andrew Locke has some good blog posts
• CySharp have some great repos. Ive learned a lot from going through them. Neuecc is allergic to memory usage 😅
• Nick Chapsas on youtube
• The book Pro .NET Memory Management: For Better Code, Performance, and Scalability goes into a lot of depth on the CLR (I haven't read it, I've heard its pretty thorough)
  • IIRC they are pretty chill about piracy, so you can see if it's your cup of tea before buying a copy.
• SIMD is cool but most problems don't have a direct use case for it😞