rust-lang / rust-lang/rust

Performance degredation when enabling `wgpu` dep

Open
#146,935 3 comments 0 reactions 1 assignee View on GitHub

@cachebag is already working on this.

Since Sep 23, 2025.

C-bug I-compiletime T-compiler
Dominant language
Rust
Stars
119k
Forks
16.1k
PR metrics pending

Description

https://github.com/GraphiteEditor/Graphite/ is seeing terrible rustc performance when compiling graphene-core (./node-graph/gcore) with the wgpu feature (3m 07s clean) vs without (52.57s clean). The wgpu feature only enables the wgpu dependency and only has a two #[cfg]'ed blocks of ~50 lines to provide an alternative implementation if not enabled. With such a relatively small difference, a compile time increase of over 200% is quite unexpected. Using -Ztime-passes reveals that the passes coherence_checking, type_check_crate and generate_crate_metadata take significantly longer.

Test methodology

With RUSTFLAGS="-Ztime-passes" potentially multiple crates are all printed into a single log. So I'm running a warmup phase to build all dependencies, then selectively delete the build artifacts of graphene-core and build again with -j 1. This seems to separate the time-passes values of graphene-core to be printed after cargo printing Compiling graphene-core v0.1.0. Though, there are a bunch of other time-passes before that I cannot associate to a specific crate.

cargo clean
export CARGO_INCREMENTAL=false RUSTFLAGS="-Ztime-passes"
cargo +nightly b -p graphene-core &> warmup.log
cargo +nightly b -p graphene-core --features wgpu &> warmup_wgpu.log
rm ./target/debug/deps/*graphene_core-* ./target/debug/libgraphene_core.* -v
cargo +nightly b -p graphene-core -j 1 &> measure.log
cargo +nightly b -p graphene-core --features wgpu -j 1 &> measure_wgpu.log
Results

only passes above 1s, large changes marked italic

pass default wgpu
coherence_checking 0.476 44.813
type_check_crate 2.174 93.618
MIR_borrow_checking 3.360 5.863
monomorphization_collector_graph_walk 1.587 3.291
generate_crate_metadata 2.192 48.571
codegen_to_LLVM_IR 2.034 2.131
codegen_crate 16.511 16.062
LLVM_passes 46.112 45.793
finish_ongoing_codegen 29.703 29.860
total 55.406 195.463
time-passes log default 

   Compiling graphene-core v0.1.0 (/home/firestar99/workspace/graphite/node-graph/gcore)
time:   0.001; rss:   38MB ->   40MB (   +2MB)	parse_crate
time:   0.000; rss:   45MB ->   45MB (   +0MB)	crate_injection
time:   0.613; rss:   45MB ->  236MB ( +191MB)	expand_crate
time:   0.613; rss:   45MB ->  236MB ( +191MB)	macro_expand_crate
time:   0.000; rss:  236MB ->  236MB (   +0MB)	maybe_building_test_harness
time:   0.011; rss:  236MB ->  236MB (   +0MB)	AST_validation
time:   0.004; rss:  236MB ->  238MB (   +2MB)	finalize_imports
time:   0.006; rss:  238MB ->  241MB (   +2MB)	compute_effective_visibilities
time:   0.000; rss:  241MB ->  241MB (   +0MB)	lint_reexports
time:   0.004; rss:  241MB ->  241MB (   +1MB)	finalize_macro_resolutions
time:   0.119; rss:  241MB ->  281MB (  +40MB)	late_resolve_crate
time:   0.000; rss:  281MB ->  281MB (   +0MB)	resolve_main
time:   0.011; rss:  281MB ->  282MB (   +0MB)	resolve_check_unused
time:   0.017; rss:  282MB ->  282MB (   +0MB)	resolve_postprocess
time:   0.162; rss:  236MB ->  282MB (  +46MB)	resolve_crate
time:   0.009; rss:  282MB ->  283MB (   +0MB)	write_dep_info
time:   0.008; rss:  283MB ->  283MB (   +0MB)	complete_gated_feature_checking
time:   0.021; rss:  324MB ->  324MB (   +0MB)	drop_ast
time:   0.000; rss:  306MB ->  306MB (   +0MB)	looking_for_entry_point
time:   0.001; rss:  306MB ->  306MB (   +0MB)	looking_for_derive_registrar
time:   0.000; rss:  304MB ->  304MB (   +0MB)	unused_lib_feature_checking
time:   0.036; rss:  306MB ->  305MB (   -1MB)	misc_checking_1
time:   0.476; rss:  305MB ->  395MB (  +90MB)	coherence_checking
time:   0.000; rss:  395MB ->  395MB (   +0MB)	emit_ast_lowering_delayed_lints
time:   2.174; rss:  305MB ->  526MB ( +221MB)	type_check_crate
time:   3.360; rss:  526MB ->  672MB ( +146MB)	MIR_borrow_checking
time:   0.047; rss:  673MB ->  673MB (   +0MB)	module_lints
time:   0.047; rss:  673MB ->  673MB (   +0MB)	lint_checking
time:   0.058; rss:  673MB ->  674MB (   +0MB)	privacy_checking_modules
time:   0.132; rss:  672MB ->  674MB (   +1MB)	misc_checking_3
time:   0.001; rss:  702MB ->  702MB (   +0MB)	monomorphization_collector_root_collections
time:   1.587; rss:  702MB ->  837MB ( +135MB)	monomorphization_collector_graph_walk
time:   0.122; rss:  838MB ->  855MB (  +17MB)	partition_and_assert_distinct_symbols
time:   2.192; rss:  674MB ->  855MB ( +182MB)	generate_crate_metadata
time:   0.000; rss:  855MB ->  855MB (   +0MB)	find_cgu_reuse
time:   2.034; rss:  855MB -> 1035MB ( +180MB)	codegen_to_LLVM_IR
time:  16.511; rss:  855MB -> 1035MB ( +180MB)	codegen_crate
time:   0.000; rss: 1035MB -> 1035MB (   +0MB)	serialize_dep_graph
time:  46.112; rss:  877MB ->  765MB ( -112MB)	LLVM_passes
time:   0.005; rss:  663MB ->  616MB (  -47MB)	join_worker_thread
time:  29.703; rss:  719MB ->  616MB ( -104MB)	finish_ongoing_codegen
time:   0.108; rss:  616MB ->  628MB (  +12MB)	link_rlib
time:   0.019; rss:  628MB ->  628MB (   +0MB)	link_binary_remove_temps
time:   0.129; rss:  616MB ->  628MB (  +12MB)	link_binary
time:   0.129; rss:  616MB ->  616MB (   +1MB)	link_crate
time:   0.129; rss:  616MB ->  616MB (   +1MB)	link
time:  55.406; rss:   28MB ->  139MB ( +111MB)	total
    Finished `dev` profile [optimized + debuginfo] target(s) in 55.75s

time-passes log wgpu 

   Compiling graphene-core v0.1.0 (/home/firestar99/workspace/graphite/node-graph/gcore)
time:   0.001; rss:   38MB ->   40MB (   +2MB)	parse_crate
time:   0.000; rss:   45MB ->   45MB (   +0MB)	crate_injection
time:   0.606; rss:   45MB ->  235MB ( +190MB)	expand_crate
time:   0.607; rss:   45MB ->  235MB ( +190MB)	macro_expand_crate
time:   0.000; rss:  235MB ->  236MB (   +0MB)	maybe_building_test_harness
time:   0.010; rss:  236MB ->  236MB (   +0MB)	AST_validation
time:   0.004; rss:  236MB ->  238MB (   +2MB)	finalize_imports
time:   0.006; rss:  238MB ->  240MB (   +2MB)	compute_effective_visibilities
time:   0.000; rss:  240MB ->  240MB (   +0MB)	lint_reexports
time:   0.004; rss:  240MB ->  241MB (   +1MB)	finalize_macro_resolutions
time:   0.124; rss:  241MB ->  290MB (  +49MB)	late_resolve_crate
time:   0.000; rss:  290MB ->  290MB (   +0MB)	resolve_main
time:   0.011; rss:  290MB ->  291MB (   +0MB)	resolve_check_unused
time:   0.015; rss:  291MB ->  291MB (   +0MB)	resolve_postprocess
time:   0.165; rss:  236MB ->  291MB (  +55MB)	resolve_crate
time:   0.007; rss:  291MB ->  291MB (   +0MB)	write_dep_info
time:   0.008; rss:  292MB ->  292MB (   +0MB)	complete_gated_feature_checking
time:   0.022; rss:  332MB ->  332MB (   +0MB)	drop_ast
time:   0.000; rss:  315MB ->  315MB (   +0MB)	looking_for_entry_point
time:   0.001; rss:  315MB ->  315MB (   +0MB)	looking_for_derive_registrar
time:   0.000; rss:  313MB ->  313MB (   +0MB)	unused_lib_feature_checking
time:   0.032; rss:  315MB ->  314MB (   -1MB)	misc_checking_1
time:  44.813; rss:  314MB ->  424MB ( +110MB)	coherence_checking
time:   0.000; rss:  424MB ->  424MB (   +0MB)	emit_ast_lowering_delayed_lints
time:  93.618; rss:  314MB ->  571MB ( +257MB)	type_check_crate
time:   5.863; rss:  571MB ->  778MB ( +207MB)	MIR_borrow_checking
time:   0.050; rss:  779MB ->  779MB (   +1MB)	module_lints
time:   0.050; rss:  779MB ->  779MB (   +1MB)	lint_checking
time:   0.056; rss:  779MB ->  780MB (   +0MB)	privacy_checking_modules
time:   0.131; rss:  778MB ->  780MB (   +2MB)	misc_checking_3
time:   0.001; rss:  832MB ->  832MB (   +0MB)	monomorphization_collector_root_collections
time:   3.291; rss:  832MB -> 1283MB ( +450MB)	monomorphization_collector_graph_walk
time:   0.133; rss: 1284MB -> 1300MB (  +16MB)	partition_and_assert_distinct_symbols
time:  48.571; rss:  780MB -> 1300MB ( +521MB)	generate_crate_metadata
time:   0.000; rss: 1300MB -> 1300MB (   +0MB)	find_cgu_reuse
time:   2.131; rss: 1300MB -> 1482MB ( +182MB)	codegen_to_LLVM_IR
time:  16.062; rss: 1300MB -> 1482MB ( +182MB)	codegen_crate
time:   0.000; rss: 1482MB -> 1482MB (   +0MB)	serialize_dep_graph
time:  45.793; rss: 1321MB -> 1108MB ( -213MB)	LLVM_passes
time:   0.001; rss:  671MB ->  667MB (   -5MB)	join_worker_thread
time:  29.860; rss: 1094MB ->  667MB ( -427MB)	finish_ongoing_codegen
time:   0.111; rss:  667MB ->  679MB (  +13MB)	link_rlib
time:   0.019; rss:  679MB ->  679MB (   +0MB)	link_binary_remove_temps
time:   0.132; rss:  667MB ->  679MB (  +13MB)	link_binary
time:   0.133; rss:  667MB ->  665MB (   -1MB)	link_crate
time:   0.133; rss:  667MB ->  665MB (   -1MB)	link
time: 195.463; rss:   28MB ->  142MB ( +113MB)	total
    Finished `dev` profile [optimized + debuginfo] target(s) in 3m 15s

raw logs
Meta

rustc --version --verbose:

rustc 1.89.0 (29483883e 2025-08-04)
binary: rustc
commit-hash: 29483883eed69d5fb4db01964cdf2af4d86e9cb2
commit-date: 2025-08-04
host: x86_64-unknown-linux-gnu
release: 1.89.0
LLVM version: 20.1.7

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.