Optimize write_metadata.

[nnethercote]

write_metadata currently generates metadata unnecessarily in some
cases, and also compresses it unnecessarily in some cases. This commit
fixes that. It speeds up three of the rustc-benchmarks by 1--4%.

r? @eddyb, who deserves much of the credit because he (a) identified the problem from the profile data I provided in #37086, and (b) explained to me how to fix it. Thank you, @eddyb!

[nnethercote]

Measurements for a stage1 compiler doing debug builds:

futures-rs-test  4.147s vs  4.175s --> 0.993x faster (variance: 1.012x, 1.009x)
helloworld       0.223s vs  0.224s --> 0.992x faster (variance: 1.011x, 1.006x)
html5ever-2016-  5.182s vs  5.129s --> 1.010x faster (variance: 1.006x, 1.008x)
hyper.0.5.0      4.772s vs  4.754s --> 1.004x faster (variance: 1.012x, 1.007x)
inflate-0.1.0    4.524s vs  4.512s --> 1.003x faster (variance: 1.003x, 1.020x)
issue-32062-equ  0.323s vs  0.322s --> 1.003x faster (variance: 1.009x, 1.023x)
issue-32278-big  1.653s vs  1.661s --> 0.995x faster (variance: 1.004x, 1.009x)
jld-day15-parse  1.460s vs  1.462s --> 0.999x faster (variance: 1.016x, 1.009x)
piston-image-0. 11.318s vs 11.308s --> 1.001x faster (variance: 1.007x, 1.063x)
regex.0.1.30     2.414s vs  2.312s --> 1.044x faster (variance: 1.013x, 1.019x)
rust-encoding-0  1.990s vs  1.979s --> 1.005x faster (variance: 1.008x, 1.007x)
syntex-0.42.2   29.514s vs 29.446s --> 1.002x faster (variance: 1.013x, 1.008x)
syntex-0.42.2-i 15.318s vs 14.959s --> 1.024x faster (variance: 1.059x, 1.006x)

So it's a big win for regex, a moderate win for syntax-incr, a small win for html5ever, and makes little difference to the others. (I did multiple timing runs and these outcomes were consistent.)

[eddyb]

This could be written on one line, with the closure passed to it directly, and with no explicit type on the closure argument.
.max().unwrap() would come on the same line after the }) where the } is the closing bracket of the closure.
Also, you might want to do .unwrap_or(MetadataKind::None) instead of .unwrap() (cc @alexcrichton).

[nnethercote]

I fixed the indentation.

@alexcrichton: the specific question is this: in write_metadata, can cx.sess().crate_types be empty?

[eddyb]

@alexcrichton r=me if the unwrap isn't wrong here.

[alexcrichton]

@bors: r=eddyb

On Tuesday, October 18, 2016, Eduard-Mihai Burtescu <
[email protected]> wrote:

@eddyb commented on this pull request.

In src/librustc_trans/base.rs
#37267 (review):

•    Uncompressed,
  

•    Compressed
  

• }

• let kind = cx.sess().crate_types.borrow().iter().map(|ty| {

•    match *ty {
  

•        config::CrateTypeExecutable |
  

•        config::CrateTypeStaticlib |
  

•        config::CrateTypeCdylib => MetadataKind::None,
  

•        config::CrateTypeRlib => MetadataKind::Uncompressed,
  

•        config::CrateTypeDylib |
  

•        config::CrateTypeProcMacro => MetadataKind::Compressed,
  

•    }
  

• }).max().unwrap();

@alexcrichton https://github.com/alexcrichton r=me if the unwrap isn't
wrong here.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#37267 (review),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAD95KDALZ8ApEOShtYia3TaL_UNoQbaks5q1ZbjgaJpZM4KagGJ
.

[bors]

📌 Commit 016f69f has been approved by eddyb