Inline StrSearcher::haystack()

[alexbool]

No description provided.

[alexbool]

r? @BurntSushi

[BurntSushi]

@bors r+

[bors]

📌 Commit ca52d80 has been approved by BurntSushi

[BurntSushi]

@bors r- (see: #42716 (comment))

[alexcrichton]

Ah I feel the same as #42716, let's be sure we have numbers to justify this but otherwise let's hold off on #[inline]

[alexbool]

@alexcrichton I discovered this by accident when inspecting assembly of one of my programs in release mode. I was surprised to find lots of calls to that function here and there when it's just a getter to the field. I thought this method is an easy catch for the optimizer to inline (provided the attribute is there), basically 2 instructions to make a call to the function and 6 instructions in the function itself become 2 mov instructions. I have neither benchmarks nor numbers. Other impls of that trait have #[inline] on this method, so I filed this PR for the sake of consistency too.

[alexcrichton]

Oh don't worry, I understand how #[inline] works and how it's necessary for many small functions, and how it often can help performance quite a bit.

All I personally want is an anecdote or benchmarking showing the necessity of such tags. We have not always been so principled in the past, but it's very easy to go overboard with #[inline].

Sounds like you hit this when benchmarking, though, so sounds good to me, thanks!

@bors: r+

[bors]

📌 Commit ca52d80 has been approved by alexcrichton

[Mark-Simulacrum]

@bors rollup