Iterator performance regression 1.81 -> 1.82

[Sunshine40]

Code

I tried this code:

use std::time::Instant;

fn main() {
    let start = Instant::now();

    (0..=1000)
    .flat_map(|a| {
        (0..=1000).flat_map(move |b| {
            (0..=1000)
                .filter(move |&c| a * a + b * b == c * c && a + b + c == 1000)
                .map(move |c| (a, b, c))
        })
    })
    .for_each(|(a, b, c)| {
        println!("a: {} b: {} c: {}", a, b, c);
    });

    let duration = start.elapsed();
    println!("{} seconds", duration.as_secs_f64());
}

I expected to see this happen: code behaves the same across compiler versions

Instead, this happened: code runs 80% slower when compiled with Rust 1.82+

Version it worked on

It most recently worked on: Rust 1.81

Version with regression

rustc --version --verbose:

rustc 1.82.0 (f6e511eec 2024-10-15)
binary: rustc
commit-hash: f6e511eec7342f59a25f7c0534f1dbea00d01b14
commit-date: 2024-10-15
host: x86_64-pc-windows-msvc
release: 1.82.0
LLVM version: 19.1.1

More details in this post

https://internals.rust-lang.org/t/iterator-performance-regression-1-81-1-82/22473

[Sunshine40]

I wonder if this is somehow related to #132636?

[steffahn]

searched toolchains nightly-2024-07-20 through nightly-2024-08-30


********************************************************************************
Regression in nightly-2024-08-01
********************************************************************************

fetching https://static.rust-lang.org/dist/2024-07-31/channel-rust-nightly-git-commit-hash.txt
nightly manifest 2024-07-31: 40 B / 40 B [====================================================================] 100.00 % 165.01 KB/s converted 2024-07-31 to f8060d282d42770fadd73905e3eefb85660d3278
fetching https://static.rust-lang.org/dist/2024-08-01/channel-rust-nightly-git-commit-hash.txt
nightly manifest 2024-08-01: 40 B / 40 B [====================================================================] 100.00 % 249.29 KB/s converted 2024-08-01 to 28a58f2fa7f0c46b8fab8237c02471a915924fe5
looking for regression commit between 2024-07-31 and 2024-08-01
fetching (via remote github) commits from max(f8060d282d42770fadd73905e3eefb85660d3278, 2024-07-29) to 28a58f2fa7f0c46b8fab8237c02471a915924fe5
ending github query because we found starting sha: f8060d282d42770fadd73905e3eefb85660d3278
get_commits_between returning commits, len: 7
  commit[0] 2024-07-30: Auto merge of #128083 - Mark-Simulacrum:bump-bootstrap, r=albertlarsan68
  commit[1] 2024-07-30: Auto merge of #128413 - matthiaskrgr:rollup-nrfcvdq, r=matthiaskrgr
  commit[2] 2024-07-31: Auto merge of #128075 - Oneirical:try-your-damnetest, r=jieyouxu
  commit[3] 2024-07-31: Auto merge of #128420 - weihanglo:update-cargo, r=weihanglo
  commit[4] 2024-07-31: Auto merge of #127513 - nikic:llvm-19, r=cuviper
  commit[5] 2024-07-31: Auto merge of #128435 - matthiaskrgr:rollup-l76vu3i, r=matthiaskrgr
  commit[6] 2024-07-31: Auto merge of #126991 - cjgillot:gvn-prof, r=oli-obk
ERROR: no CI builds available between f8060d282d42770fadd73905e3eefb85660d3278 and 28a58f2fa7f0c46b8fab8237c02471a915924fe5 within last 167 days

so this might be

• Update to LLVM 19 #127513

@rustbot label -E-needs-bisection

[nikic]

This sounds like llvm/llvm-project#81689 to me.

[apiraino]

WG-prioritization assigning priority (Zulip discussion).

@rustbot label -I-prioritize +P-medium