Make std::thread::available_concurrency support process-limited number of CPUs - #89310
Merged
bors merged 2 commits intoNov 7, 2021
Merged
Conversation
Contributor
|
r? @m-ou-se (rust-highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
joshtriplett
changed the title
std::thread::available_concurrency support process-limited number of CPUs
7 tasks
This comment has been minimized.
This comment has been minimized.
joshtriplett
force-pushed
the
available-concurrency-affinity
branch
from
September 27, 2021 20:33
d6affdc to
33e4614
Compare
yoshuawuyts
approved these changes
Sep 28, 2021
yoshuawuyts
left a comment
yoshuawuyts
left a comment
Member
There was a problem hiding this comment.
Compared this against num_cpus/linux.rs, and this looks good!
yoshuawuyts
reviewed
Sep 28, 2021
m-ou-se
reviewed
Oct 3, 2021
joshtriplett
force-pushed
the
available-concurrency-affinity
branch
from
October 4, 2021 03:45
33e4614 to
433439e
Compare
Member
|
I believe this PR should be ready to be merged? |
Member
|
@bors r+ |
Collaborator
|
📌 Commit 433439e has been approved by |
bors
added
S-waiting-on-bors
matthiaskrgr
added a commit
to matthiaskrgr/rust
that referenced
this pull request
Oct 15, 2021
…affinity, r=m-ou-se Make `std::thread::available_concurrency` support process-limited number of CPUs Use `libc::sched_getaffinity` and count the number of CPUs in the returned mask. This handles cases where the process doesn't have access to all CPUs, such as when limited via `taskset` or similar. This also covers cgroup cpusets.
matthiaskrgr
added a commit
to matthiaskrgr/rust
that referenced
this pull request
Oct 15, 2021
…affinity, r=m-ou-se Make `std::thread::available_concurrency` support process-limited number of CPUs Use `libc::sched_getaffinity` and count the number of CPUs in the returned mask. This handles cases where the process doesn't have access to all CPUs, such as when limited via `taskset` or similar. This also covers cgroup cpusets.
Member
bors
added
S-waiting-on-author
Member
Author
|
Fixed the missing macros on Android in rust-lang/libc#2470 . Once that goes in, I'll do a libc release and update. |
joshtriplett
force-pushed
the
available-concurrency-affinity
branch
from
October 24, 2021 16:03
433439e to
645e46e
Compare
Member
Author
|
@bors r=m-ou-se |
…r of CPUs Use libc::sched_getaffinity and count the number of CPUs in the returned mask. This handles cases where the process doesn't have access to all CPUs, such as when limited via taskset or similar.
joshtriplett
force-pushed
the
available-concurrency-affinity
branch
from
October 30, 2021 23:38
645e46e to
7c9611d
Compare
Member
Author
|
Updated to libc 0.2.106 which includes the fix for the issue that broke CI on macOS. @bors r=m-ou-se |
Collaborator
|
📌 Commit 7c9611d has been approved by |
bors
added
S-waiting-on-bors
Collaborator
|
⌛ Testing commit 7c9611d with merge bd73825db1627e06ead2c3e17513f3aaa9b2c46d... |
Collaborator
|
The job Click to see the possible cause of the failure (guessed by this bot) |
Collaborator
|
💔 Test failed - checks-actions |
bors
added
S-waiting-on-review
Member
Author
Collaborator
|
💡 This pull request was already approved, no need to approve it again.
|
Collaborator
|
📌 Commit 7c9611d has been approved by |
bors
added
S-waiting-on-bors
Member
Author
|
@bors retry |
Collaborator
Collaborator
|
☀️ Test successful - checks-actions |
Collaborator
|
Finished benchmarking commit (fecfc0e): comparison url. Summary: This benchmark run did not return any relevant changes. If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf. @rustbot label: -perf-regression |
brawer
added a commit
to alltheplaces/osm-diffs
that referenced
this pull request
Aug 24, 2026
Every internal thread pool (prune.rs/assemble.rs's own num_workers, Rayon's default global pool for conflate.match) sizes itself from std::thread::available_parallelism() -- which historically ignored Docker/podman's --cpus (a CFS bandwidth quota, not a cpuset) and just reported the host's full core count. Rust's std did fix this for cgroup v2 (rust-lang/rust#89310), and this toolchain (1.98) is well past that fix, but it's exactly the kind of library behavior worth confirming empirically rather than trusting -- this is that confirmation mechanism, for whenever a --cpus sweep (see docs/PRODUCTION.md) actually happens: log the value once, structured, and check it matches N under a real --cpus=N. While touching this call site: also moves the pipeline version out of the message text and into a structured field, matching this module's own documented convention (logging.rs's own doc comment: "Call sites that have actual structured data to log ... should attach it via the log crate's key-value API instead of interpolating it into the message text") -- the version was doing exactly that before. Uses the bare CARGO_PKG_VERSION value (no "v" prefix), consistent with how version already appears elsewhere (e.g. pipeline::provenance's tool_component()), not the "vX.Y.Z" tag format. Extends logging::init()'s existing one-time startup log record rather than adding a second one -- already the natural, only-runs-once home for this kind of environment fact. Verified: cargo test (253 passed, extended the existing startup-record test to check both new structured fields instead of parsing the old message string)/clippy/fmt clean. Manually ran against the zugerland.osm.pbf fixture and confirmed the real JSON shape: {"fields":{"available_parallelism":10,"version":"0.8.2"},..., "message":"pipeline starting up",...} Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
brawer
added a commit
to alltheplaces/osm-diffs
that referenced
this pull request
Aug 24, 2026
Every internal thread pool (prune.rs/assemble.rs's own num_workers, Rayon's default global pool for conflate.match) sizes itself from std::thread::available_parallelism() -- which historically ignored Docker/podman's --cpus (a CFS bandwidth quota, not a cpuset) and just reported the host's full core count. Rust's std did fix this for cgroup v2 (rust-lang/rust#89310), and this toolchain (1.98) is well past that fix, but it's exactly the kind of library behavior worth confirming empirically rather than trusting -- this is that confirmation mechanism, for whenever a --cpus sweep (see docs/PRODUCTION.md) actually happens: log the value once, structured, and check it matches N under a real --cpus=N. While touching this call site: also moves the pipeline version out of the message text and into a structured field, matching this module's own documented convention (logging.rs's own doc comment: "Call sites that have actual structured data to log ... should attach it via the log crate's key-value API instead of interpolating it into the message text") -- the version was doing exactly that before. Keeps the "vX.Y.Z" form (concat!("v", env!("CARGO_PKG_VERSION")), a zero-cost compile-time concatenation of two string literals, not a runtime format!) to match both the previously-logged string and the release tag format (see RELEASING.md) -- not the bare semver pipeline::provenance's tool_component() uses, which is a different field for a different purpose. Extends logging::init()'s existing one-time startup log record rather than adding a second one -- already the natural, only-runs-once home for this kind of environment fact. Verified: cargo test (253 passed, extended the existing startup-record test to check both new structured fields instead of parsing the old message string)/clippy/fmt clean. Manually ran against the zugerland.osm.pbf fixture and confirmed the real JSON shape: {"fields":{"available_parallelism":10,"version":"v0.8.2"},..., "message":"pipeline starting up",...} Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
9 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Use
libc::sched_getaffinityand count the number of CPUs in the returned mask. This handles cases where the process doesn't have access to all CPUs, such as when limited viatasksetor similar.This also covers cgroup cpusets.