rustc_codegen_ssa: use wall time for codegen_to_LLVM_IR time-passes entry#81369
Conversation
…ntry Use elapsed wall time spent on codegen_to_LLVM_IR for all CGUs as a whole, rather than the sum for each CGU (the distinction matters for parallel builds, where some CGUs are processed in parallel).
|
r? @lcnr (rust-highfive has picked a reviewer for you, use r? to override) |
|
this PR looks good. I don't have an opinion on whether tracking the shared wall time is better than adding the wall time of each CGU 🤔 Why do you prefer looking at the shared wall time here? |
I think it's what most people looking at We could also maintain consistent meaning of the time in an entry by tracking |
|
@bors r+ rollup |
|
📌 Commit 59195a2 has been approved by |
…l-time, r=lcnr rustc_codegen_ssa: use wall time for codegen_to_LLVM_IR time-passes entry Use elapsed wall time spent on codegen_to_LLVM_IR for all CGUs as a whole, rather than the sum for each CGU (the distinction matters for parallel builds, where some CGUs are processed in parallel).
Rollup of 14 pull requests Successful merges: - rust-lang#80812 (Update RELEASES.md for 1.50.0) - rust-lang#80876 (Add `unwrap_unchecked()` methods for `Option` and `Result`) - rust-lang#80900 (Fix ICE with `ReadPointerAsBytes` validation error) - rust-lang#81191 (BTreeMap: test all borrowing interfaces and test more chaotic order behavior) - rust-lang#81195 (Account for generics when suggesting bound) - rust-lang#81299 (Fix some bugs reported by eslint) - rust-lang#81325 (typeck: Don't suggest converting LHS exprs) - rust-lang#81353 (Fix spelling in documentation for error E0207) - rust-lang#81369 (rustc_codegen_ssa: use wall time for codegen_to_LLVM_IR time-passes entry) - rust-lang#81389 (rustdoc: Document CommonMark extensions.) - rust-lang#81399 (Update books) - rust-lang#81401 (tidy: Some code cleanup.) - rust-lang#81407 (Refine "remove semicolon" suggestion in trait selection) - rust-lang#81412 (Fix assertion in `MaybeUninit::array_assume_init()` for zero-length arrays) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup
Use elapsed wall time spent on codegen_to_LLVM_IR for all CGUs as a
whole, rather than the sum for each CGU (the distinction matters for
parallel builds, where some CGUs are processed in parallel).