- run: rustup component add rustfmt
- run: cargo fmt --all -- --check
- run: |
- for manifest in `find crates -name Cargo.toml`
+ for manifest in `find crates benches/benchsuite benches/capture -name Cargo.toml`
do
echo check fmt for $manifest
cargo fmt --all --manifest-path $manifest -- --check
if: matrix.os == 'macos-latest'
- run: cargo build --manifest-path crates/credential/cargo-credential-wincred/Cargo.toml
if: matrix.os == 'windows-latest'
+ - name: Check benchmarks
+ env:
+ # Share the target dir to try to cache a few build-time deps.
+ CARGO_TARGET_DIR: target
+ run: |
+ # This only tests one benchmark since it can take over 10 minutes to
+ # download all workspaces.
+ cargo test --manifest-path benches/benchsuite/Cargo.toml --all-targets -- cargo
+ cargo check --manifest-path benches/capture/Cargo.toml
- name: Fetch smoke test
run: ci/fetch-smoke-test.sh
--- /dev/null
+# Cargo Benchmarking
+
+This directory contains some benchmarks for cargo itself. This uses
+[Criterion] for running benchmarks. It is recommended to read the Criterion
+book to get familiar with how to use it. A basic usage would be:
+
+```sh
+cd benches/benchsuite
+cargo bench
+```
+
+The tests involve downloading the index and benchmarking against some
+real-world and artificial workspaces located in the [`workspaces`](workspaces)
+directory.
+
+**Beware** that the initial download can take a fairly long amount of time (10
+minutes minimum on an extremely fast network) and require significant disk
+space (around 4.5GB). The benchsuite will cache the index and downloaded
+crates in the `target/tmp/bench` directory, so subsequent runs should be
+faster. You can (and probably should) specify individual benchmarks to run to
+narrow it down to a more reasonable set, for example:
+
+```sh
+cargo bench -- resolve_ws/rust
+```
+
+This will only download what's necessary for the rust-lang/rust workspace
+(which is about 330MB) and run the benchmarks against it (which should take
+about a minute). To get a list of all the benchmarks, run:
+
+```sh
+cargo bench -- --list
+```
+
+## Viewing reports
+
+The benchmarks display some basic information on the command-line while they
+run. A more complete HTML report can be found at
+`target/criterion/report/index.html` which contains links to all the
+benchmarks and summaries. Check out the Criterion book for more information on
+the extensive reporting capabilities.
+
+## Comparing implementations
+
+Knowing the raw numbers can be useful, but what you're probably most
+interested in is checking if your changes help or hurt performance. To do
+that, you need to run the benchmarks multiple times.
+
+First, run the benchmarks from the master branch of cargo without any changes.
+To make it easier to compare, Criterion supports naming the baseline so that
+you can iterate on your code and compare against it multiple times.
+
+```sh
+cargo bench -- --save-baseline master
+```
+
+Now you can switch to your branch with your changes. Re-run the benchmarks
+compared against the baseline:
+
+```sh
+cargo bench -- --baseline master
+```
+
+You can repeat the last command as you make changes to re-compare against the
+master baseline.
+
+Without the baseline arguments, it will compare against the last run, which
+can be helpful for comparing incremental changes.
+
+## Capturing workspaces
+
+The [`workspaces`](workspaces) directory contains several workspaces that
+provide a variety of different workspaces intended to provide good exercises
+for benchmarks. Some of these are shadow copies of real-world workspaces. This
+is done with the tool in the [`capture`](capture) directory. The tool will
+copy `Cargo.lock` and all of the `Cargo.toml` files of the workspace members.
+It also adds an empty `lib.rs` so Cargo won't error, and sanitizes the
+`Cargo.toml` to some degree, removing unwanted elements. Finally, it
+compresses everything into a `tgz`.
+
+To run it, do:
+
+```sh
+cd benches/capture
+cargo run -- /path/to/workspace/foo
+```
+
+The resolver benchmarks also support the `CARGO_BENCH_WORKSPACES` environment
+variable, which you can point to a Cargo workspace if you want to try
+different workspaces. For example:
+
+```sh
+CARGO_BENCH_WORKSPACES=/path/to/some/workspace cargo bench
+```
+
+## TODO
+
+This is just a start for establishing a benchmarking suite for Cargo. There's
+a lot that can be added. Some ideas:
+
+* Fix the benchmarks so that the resolver setup doesn't run every iteration.
+* Benchmark [this section of
+ code](https://github.com/rust-lang/cargo/blob/a821e2cb24d7b6013433f069ab3bad53d160e100/src/cargo/ops/cargo_compile.rs#L470-L549)
+ which builds the unit graph. The performance there isn't great, and it would
+ be good to keep an eye on it. Unfortunately that would mean doing a bit of
+ work to make `generate_targets` publicly visible, and there is a bunch of
+ setup code that may need to be duplicated.
+* Benchmark the fingerprinting code.
+* Benchmark running the `cargo` executable. Running something like `cargo
+ build` or `cargo check` with everything "Fresh" would be a good end-to-end
+ exercise to measure the overall overhead of Cargo.
+* Benchmark pathological resolver scenarios. There might be some cases where
+ the resolver can spend a significant amount of time. It would be good to
+ identify if these exist, and create benchmarks for them. This may require
+ creating an artificial index, similar to the `resolver-tests`. This should
+ also consider scenarios where the resolver ultimately fails.
+* Benchmark without `Cargo.lock`. I'm not sure if this is particularly
+ valuable, since we are mostly concerned with incremental builds which will
+ always have a lock file.
+* Benchmark just
+ [`resolve::resolve`](https://github.com/rust-lang/cargo/blob/a821e2cb24d7b6013433f069ab3bad53d160e100/src/cargo/core/resolver/mod.rs#L122)
+ without anything else. This can help focus on just the resolver.
+
+[Criterion]: https://bheisler.github.io/criterion.rs/book/
--- /dev/null
+[package]
+name = "benchsuite"
+version = "0.1.0"
+edition = "2018"
+license = "MIT OR Apache-2.0"
+homepage = "https://github.com/rust-lang/cargo"
+repository = "https://github.com/rust-lang/cargo"
+documentation = "https://docs.rs/cargo-platform"
+description = "Benchmarking suite for Cargo."
+
+[dependencies]
+cargo = { path = "../.." }
+# Consider removing html_reports in 0.4 and switching to `cargo criterion`.
+criterion = { version = "0.3.5", features = ["html_reports"] }
+flate2 = { version = "1.0.3", default-features = false, features = ["zlib"] }
+tar = { version = "0.4.35", default-features = false }
+url = "2.2.2"
+
+[[bench]]
+name = "resolve"
+harness = false
--- /dev/null
+use cargo::core::compiler::{CompileKind, RustcTargetData};
+use cargo::core::resolver::features::{CliFeatures, FeatureOpts, FeatureResolver, ForceAllTargets};
+use cargo::core::resolver::{HasDevUnits, ResolveBehavior};
+use cargo::core::{PackageIdSpec, Workspace};
+use cargo::ops::WorkspaceResolve;
+use cargo::Config;
+use criterion::{criterion_group, criterion_main, Criterion};
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::process::Command;
+use url::Url;
+
+// This is an arbitrary commit that existed when I started. This helps
+// ensure consistent results. It can be updated if needed, but that can
+// make it harder to compare results with older versions of cargo.
+const CRATES_IO_COMMIT: &str = "85f7bfd61ea4fee08ec68c468762e886b2aebec6";
+
+fn setup() {
+ create_home();
+ create_target_dir();
+ clone_index();
+ unpack_workspaces();
+}
+
+fn root() -> PathBuf {
+ let mut p = PathBuf::from(env!("CARGO_TARGET_TMPDIR"));
+ p.push("bench");
+ p
+}
+
+fn target_dir() -> PathBuf {
+ let mut p = root();
+ p.push("target");
+ p
+}
+
+fn cargo_home() -> PathBuf {
+ let mut p = root();
+ p.push("chome");
+ p
+}
+
+fn index() -> PathBuf {
+ let mut p = root();
+ p.push("index");
+ p
+}
+
+fn workspaces_path() -> PathBuf {
+ let mut p = root();
+ p.push("workspaces");
+ p
+}
+
+fn registry_url() -> Url {
+ Url::from_file_path(index()).unwrap()
+}
+
+fn create_home() {
+ let home = cargo_home();
+ if !home.exists() {
+ fs::create_dir_all(&home).unwrap();
+ }
+ fs::write(
+ home.join("config.toml"),
+ format!(
+ r#"
+ [source.crates-io]
+ replace-with = 'local-snapshot'
+
+ [source.local-snapshot]
+ registry = '{}'
+ "#,
+ registry_url()
+ ),
+ )
+ .unwrap();
+}
+
+fn create_target_dir() {
+ // This is necessary to ensure the .rustc_info.json file is written.
+ // Otherwise it won't be written, and it is very expensive to create.
+ if !target_dir().exists() {
+ std::fs::create_dir_all(target_dir()).unwrap();
+ }
+}
+
+/// This clones crates.io at a specific point in time into tmp/index.
+fn clone_index() {
+ let index = index();
+ let maybe_git = |command: &str| {
+ let status = Command::new("git")
+ .current_dir(&index)
+ .args(command.split_whitespace().collect::<Vec<_>>())
+ .status()
+ .expect("git should be installed");
+ status.success()
+ };
+ let git = |command: &str| {
+ if !maybe_git(command) {
+ panic!("failed to run git command: {}", command);
+ }
+ };
+ if index.exists() {
+ if maybe_git(&format!(
+ "rev-parse -q --verify {}^{{commit}}",
+ CRATES_IO_COMMIT
+ )) {
+ // Already fetched.
+ return;
+ }
+ } else {
+ fs::create_dir_all(&index).unwrap();
+ git("init --bare");
+ git("remote add origin https://github.com/rust-lang/crates.io-index");
+ }
+ git(&format!("fetch origin {}", CRATES_IO_COMMIT));
+ git("branch -f master FETCH_HEAD");
+}
+
+/// This unpacks the compressed workspace skeletons into tmp/workspaces.
+fn unpack_workspaces() {
+ let ws_dir = Path::new(env!("CARGO_MANIFEST_DIR"))
+ .parent()
+ .unwrap()
+ .join("workspaces");
+ let archives = fs::read_dir(ws_dir)
+ .unwrap()
+ .map(|e| e.unwrap().path())
+ .filter(|p| p.extension() == Some(std::ffi::OsStr::new("tgz")));
+ for archive in archives {
+ let name = archive.file_stem().unwrap();
+ let f = fs::File::open(&archive).unwrap();
+ let f = flate2::read::GzDecoder::new(f);
+ let dest = workspaces_path().join(&name);
+ if dest.exists() {
+ fs::remove_dir_all(&dest).unwrap();
+ }
+ let mut archive = tar::Archive::new(f);
+ archive.unpack(workspaces_path()).unwrap();
+ }
+}
+
+struct ResolveInfo<'cfg> {
+ ws: Workspace<'cfg>,
+ requested_kinds: [CompileKind; 1],
+ target_data: RustcTargetData<'cfg>,
+ cli_features: CliFeatures,
+ specs: Vec<PackageIdSpec>,
+ has_dev_units: HasDevUnits,
+ force_all_targets: ForceAllTargets,
+ ws_resolve: WorkspaceResolve<'cfg>,
+}
+
+/// Vec of `(ws_name, ws_root)`.
+fn workspaces() -> Vec<(String, PathBuf)> {
+ // CARGO_BENCH_WORKSPACES can be used to override, otherwise it just uses
+ // the workspaces in the workspaces directory.
+ let mut ps: Vec<_> = match std::env::var_os("CARGO_BENCH_WORKSPACES") {
+ Some(s) => std::env::split_paths(&s).collect(),
+ None => fs::read_dir(workspaces_path())
+ .unwrap()
+ .map(|e| e.unwrap().path())
+ // These currently fail in most cases on Windows due to long
+ // filenames in the git checkouts.
+ .filter(|p| {
+ !(cfg!(windows)
+ && matches!(p.file_name().unwrap().to_str().unwrap(), "servo" | "tikv"))
+ })
+ .collect(),
+ };
+ // Sort so it is consistent.
+ ps.sort();
+ ps.into_iter()
+ .map(|p| (p.file_name().unwrap().to_str().unwrap().to_owned(), p))
+ .collect()
+}
+
+/// Helper for resolving a workspace. This will run the resolver once to
+/// download everything, and returns all the data structures that are used
+/// during resolution.
+fn do_resolve<'cfg>(config: &'cfg Config, ws_root: &Path) -> ResolveInfo<'cfg> {
+ let requested_kinds = [CompileKind::Host];
+ let ws = cargo::core::Workspace::new(&ws_root.join("Cargo.toml"), config).unwrap();
+ let target_data = RustcTargetData::new(&ws, &requested_kinds).unwrap();
+ let cli_features = CliFeatures::from_command_line(&[], false, true).unwrap();
+ let pkgs = cargo::ops::Packages::Default;
+ let specs = pkgs.to_package_id_specs(&ws).unwrap();
+ let has_dev_units = HasDevUnits::Yes;
+ let force_all_targets = ForceAllTargets::No;
+ // Do an initial run to download anything necessary so that it does
+ // not confuse criterion's warmup.
+ let ws_resolve = cargo::ops::resolve_ws_with_opts(
+ &ws,
+ &target_data,
+ &requested_kinds,
+ &cli_features,
+ &specs,
+ has_dev_units,
+ force_all_targets,
+ )
+ .unwrap();
+ ResolveInfo {
+ ws,
+ requested_kinds,
+ target_data,
+ cli_features,
+ specs,
+ has_dev_units,
+ force_all_targets,
+ ws_resolve,
+ }
+}
+
+/// Creates a new Config.
+///
+/// This is separate from `do_resolve` to deal with the ownership and lifetime.
+fn make_config(ws_root: &Path) -> Config {
+ let shell = cargo::core::Shell::new();
+ let mut config = cargo::util::Config::new(shell, ws_root.to_path_buf(), cargo_home());
+ // Configure is needed to set the target_dir which is needed to write
+ // the .rustc_info.json file which is very expensive.
+ config
+ .configure(
+ 0,
+ false,
+ None,
+ false,
+ false,
+ false,
+ &Some(target_dir()),
+ &[],
+ &[],
+ )
+ .unwrap();
+ config
+}
+
+/// Benchmark of the full `resovle_ws_with_opts` which runs the resolver
+/// twice, the feature resolver, and more. This is a major component of a
+/// regular cargo build.
+fn resolve_ws(c: &mut Criterion) {
+ setup();
+ let mut group = c.benchmark_group("resolve_ws");
+ for (ws_name, ws_root) in workspaces() {
+ let config = make_config(&ws_root);
+ // The resolver info is initialized only once in a lazy fashion. This
+ // allows criterion to skip this workspace if the user passes a filter
+ // on the command-line (like `cargo bench -- resolve_ws/tikv`).
+ //
+ // Due to the way criterion works, it tends to only run the inner
+ // iterator once, and we don't want to call `do_resolve` in every
+ // "step", since that would just be some useless work.
+ let mut lazy_info = None;
+ group.bench_function(&ws_name, |b| {
+ let ResolveInfo {
+ ws,
+ requested_kinds,
+ target_data,
+ cli_features,
+ specs,
+ has_dev_units,
+ force_all_targets,
+ ..
+ } = lazy_info.get_or_insert_with(|| do_resolve(&config, &ws_root));
+ b.iter(|| {
+ cargo::ops::resolve_ws_with_opts(
+ ws,
+ target_data,
+ requested_kinds,
+ cli_features,
+ specs,
+ *has_dev_units,
+ *force_all_targets,
+ )
+ .unwrap();
+ })
+ });
+ }
+ group.finish();
+}
+
+/// Benchmark of the feature resolver.
+fn feature_resolver(c: &mut Criterion) {
+ setup();
+ let mut group = c.benchmark_group("feature_resolver");
+ for (ws_name, ws_root) in workspaces() {
+ let config = make_config(&ws_root);
+ let mut lazy_info = None;
+ group.bench_function(&ws_name, |b| {
+ let ResolveInfo {
+ ws,
+ requested_kinds,
+ target_data,
+ cli_features,
+ specs,
+ has_dev_units,
+ ws_resolve,
+ ..
+ } = lazy_info.get_or_insert_with(|| do_resolve(&config, &ws_root));
+ b.iter(|| {
+ let feature_opts = FeatureOpts::new_behavior(ResolveBehavior::V2, *has_dev_units);
+ FeatureResolver::resolve(
+ ws,
+ target_data,
+ &ws_resolve.targeted_resolve,
+ &ws_resolve.pkg_set,
+ cli_features,
+ specs,
+ requested_kinds,
+ feature_opts,
+ )
+ .unwrap();
+ })
+ });
+ }
+ group.finish();
+}
+
+// Criterion complains about the measurement time being too small, but the
+// measurement time doesn't seem important to me, what is more important is
+// the number of iterations which defaults to 100, which seems like a
+// reasonable default. Otherwise, the measurement time would need to be
+// changed per workspace. We wouldn't want to spend 60s on every workspace,
+// that would take too long and isn't necessary for the smaller workspaces.
+criterion_group!(benches, resolve_ws, feature_resolver);
+criterion_main!(benches);
--- /dev/null
+[package]
+name = "capture"
+version = "0.1.0"
+edition = "2018"
+license = "MIT OR Apache-2.0"
+description = "Tool for capturing a real-world workspace for benchmarking."
+
+[dependencies]
+cargo_metadata = "0.14.0"
+flate2 = { version = "1.0.3", default-features = false, features = ["zlib"] }
+tar = { version = "0.4.35", default-features = false }
+toml = "0.5.8"
--- /dev/null
+//! This tool helps to capture the `Cargo.toml` files of a workspace.
+//!
+//! Run it by passing a list of workspaces to capture.
+//! Use the `-f` flag to allow it to overwrite existing captures.
+//! The workspace will be saved in a `.tgz` file in the `../workspaces` directory.
+
+use flate2::{Compression, GzBuilder};
+use std::fs;
+use std::path::{Path, PathBuf};
+use std::process::Command;
+
+fn main() {
+ let force = std::env::args().any(|arg| arg == "-f");
+ let dest = Path::new(env!("CARGO_MANIFEST_DIR"))
+ .parent()
+ .unwrap()
+ .join("workspaces");
+ if !dest.exists() {
+ panic!("expected {} to exist", dest.display());
+ }
+ for arg in std::env::args().skip(1).filter(|arg| !arg.starts_with("-")) {
+ let source_root = fs::canonicalize(arg).unwrap();
+ capture(&source_root, &dest, force);
+ }
+}
+
+fn capture(source_root: &Path, dest: &Path, force: bool) {
+ let name = Path::new(source_root.file_name().unwrap());
+ let mut dest_gz = PathBuf::from(dest);
+ dest_gz.push(name);
+ dest_gz.set_extension("tgz");
+ if dest_gz.exists() {
+ if !force {
+ panic!(
+ "dest {:?} already exists, use -f to force overwriting",
+ dest_gz
+ );
+ }
+ fs::remove_file(&dest_gz).unwrap();
+ }
+ let vcs_info = capture_vcs_info(source_root, force);
+ let dst = fs::File::create(&dest_gz).unwrap();
+ let encoder = GzBuilder::new()
+ .filename(format!("{}.tar", name.to_str().unwrap()))
+ .write(dst, Compression::best());
+ let mut ar = tar::Builder::new(encoder);
+ ar.mode(tar::HeaderMode::Deterministic);
+ if let Some(info) = &vcs_info {
+ add_ar_file(&mut ar, &name.join(".cargo_vcs_info.json"), info);
+ }
+
+ // Gather all local packages.
+ let metadata = cargo_metadata::MetadataCommand::new()
+ .manifest_path(source_root.join("Cargo.toml"))
+ .features(cargo_metadata::CargoOpt::AllFeatures)
+ .exec()
+ .expect("cargo_metadata failed");
+ let mut found_root = false;
+ for package in &metadata.packages {
+ if package.source.is_some() {
+ continue;
+ }
+ let manifest_path = package.manifest_path.as_std_path();
+ copy_manifest(&manifest_path, &mut ar, name, &source_root);
+ found_root |= manifest_path == source_root.join("Cargo.toml");
+ }
+ if !found_root {
+ // A virtual workspace.
+ let contents = fs::read_to_string(source_root.join("Cargo.toml")).unwrap();
+ assert!(!contents.contains("[package]"));
+ add_ar_file(&mut ar, &name.join("Cargo.toml"), &contents);
+ }
+ let lock = fs::read_to_string(source_root.join("Cargo.lock")).unwrap();
+ add_ar_file(&mut ar, &name.join("Cargo.lock"), &lock);
+ let encoder = ar.into_inner().unwrap();
+ encoder.finish().unwrap();
+ eprintln!("created {}", dest_gz.display());
+}
+
+fn copy_manifest<W: std::io::Write>(
+ manifest_path: &Path,
+ ar: &mut tar::Builder<W>,
+ name: &Path,
+ source_root: &Path,
+) {
+ let relative_path = manifest_path
+ .parent()
+ .unwrap()
+ .strip_prefix(source_root)
+ .expect("workspace member should be under workspace root");
+ let relative_path = name.join(relative_path);
+ let contents = fs::read_to_string(&manifest_path).unwrap();
+ let mut manifest: toml::Value = toml::from_str(&contents).unwrap();
+ let remove = |obj: &mut toml::Value, name| {
+ let table = obj.as_table_mut().unwrap();
+ if table.contains_key(name) {
+ table.remove(name);
+ }
+ };
+ remove(&mut manifest, "lib");
+ remove(&mut manifest, "bin");
+ remove(&mut manifest, "example");
+ remove(&mut manifest, "test");
+ remove(&mut manifest, "bench");
+ remove(&mut manifest, "profile");
+ if let Some(package) = manifest.get_mut("package") {
+ remove(package, "default-run");
+ }
+ let contents = toml::to_string(&manifest).unwrap();
+ add_ar_file(ar, &relative_path.join("Cargo.toml"), &contents);
+ add_ar_file(ar, &relative_path.join("src").join("lib.rs"), "");
+}
+
+fn add_ar_file<W: std::io::Write>(ar: &mut tar::Builder<W>, path: &Path, contents: &str) {
+ let mut header = tar::Header::new_gnu();
+ header.set_entry_type(tar::EntryType::file());
+ header.set_mode(0o644);
+ header.set_size(contents.len() as u64);
+ header.set_mtime(123456789);
+ header.set_cksum();
+ ar.append_data(&mut header, path, contents.as_bytes())
+ .unwrap();
+}
+
+fn capture_vcs_info(ws_root: &Path, force: bool) -> Option<String> {
+ let maybe_git = |command: &str| {
+ Command::new("git")
+ .current_dir(ws_root)
+ .args(command.split_whitespace().collect::<Vec<_>>())
+ .output()
+ .expect("git should be installed")
+ };
+ assert!(ws_root.join("Cargo.toml").exists());
+ let relative = maybe_git("ls-files --full-name Cargo.toml");
+ if !relative.status.success() {
+ if !force {
+ panic!("git repository not detected, use -f to force");
+ }
+ return None;
+ }
+ let p = Path::new(std::str::from_utf8(&relative.stdout).unwrap().trim());
+ let relative = p.parent().unwrap();
+ if !force {
+ let has_changes = !maybe_git("diff-index --quiet HEAD .").status.success();
+ if has_changes {
+ panic!("git repo appears to have changes, use -f to force, or clean the repo");
+ }
+ }
+ let commit = maybe_git("rev-parse HEAD");
+ assert!(commit.status.success());
+ let commit = std::str::from_utf8(&commit.stdout).unwrap().trim();
+ let remote = maybe_git("remote get-url origin");
+ assert!(remote.status.success());
+ let remote = std::str::from_utf8(&remote.stdout).unwrap().trim();
+ let info = format!(
+ "{{\n \"git\": {{\n \"sha1\": \"{}\",\n \"remote\": \"{}\"\n }},\
+ \n \"path_in_vcs\": \"{}\"\n}}\n",
+ commit,
+ remote,
+ relative.display()
+ );
+ eprintln!("recording vcs info:\n{}", info);
+ Some(info)
+}
pub use self::registry::{publish, registry_configuration, RegistryConfig};
pub use self::resolve::{
add_overrides, get_resolved_packages, resolve_with_previous, resolve_ws, resolve_ws_with_opts,
+ WorkspaceResolve,
};
pub use self::vendor::{vendor, VendorOptions};
- [Tests](./tests/index.md)
- [Running Tests](./tests/running.md)
- [Writing Tests](./tests/writing.md)
- - [Profiling](./tests/profiling.md)
+ - [Benchmarking and Profiling](./tests/profiling.md)
- [Design Principles](./design.md)
-# Profiling
+# Benchmarking and Profiling
## Internal profiler
CARGO_PROFILE=3 cargo generate-lockfile
```
-## Informal profiling
+## Benchmarking
+
+### Benchsuite
+
+Head over to the [`benches`
+directory](https://github.com/rust-lang/cargo/tree/master/benches) for more
+information about the benchmarking suite.
+
+### Informal benchmarking
The overhead for starting a build should be kept as low as possible
(preferably, well under 0.5 seconds on most projects and systems). Currently,
* Scanning the local project.
* Building the unit dependency graph.
-We currently don't have any automated systems or tools for measuring or
-tracking the startup time. We informally measure these on changes that are
-likely to affect the performance. Usually this is done by measuring the time
-for `cargo build` to finish in a large project where the build is fresh (no
-actual compilation is performed). [Hyperfine] is a command-line tool that can
-be used to roughly measure the difference between different commands and
-settings.
+One way to test this is to use [hyperfine]. This is a tool that can be used to
+measure the difference between different commands and settings. Usually this
+is done by measuring the time it takes for `cargo build` to finish in a large
+project where the build is fresh (no actual compilation is performed). Just
+run `cargo build` once before using hyperfine.
-[Hyperfine]: https://github.com/sharkdp/hyperfine
+[hyperfine]: https://github.com/sharkdp/hyperfine