Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/tikv/minstant
Performant time measuring in Rust
https://github.com/tikv/minstant
high-performance high-performance-computing timing tsc
Last synced: 3 days ago
JSON representation
Performant time measuring in Rust
- Host: GitHub
- URL: https://github.com/tikv/minstant
- Owner: tikv
- License: apache-2.0
- Created: 2020-07-08T13:35:35.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2024-03-18T22:04:36.000Z (8 months ago)
- Last Synced: 2024-11-09T00:34:56.382Z (11 days ago)
- Topics: high-performance, high-performance-computing, timing, tsc
- Language: Rust
- Homepage:
- Size: 226 KB
- Stars: 170
- Watchers: 14
- Forks: 21
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# minstant
[](https://github.com/tikv/minstant/actions)
[](https://travis-ci.org/tikv/minstant)
[](https://docs.rs/minstant/)
[](https://crates.io/crates/minstant)
[](https://github.com/tikv/minstant/blob/master/LICENSE)
A drop-in replacement for [`std::time::Instant`](https://doc.rust-lang.org/std/time/struct.Instant.html) that measures time with high performance and high accuracy powered by [TSC](https://en.wikipedia.org/wiki/Time_Stamp_Counter).
## Usage
```toml
[dependencies]
minstant = "0.1"
```
```rust
let start = minstant::Instant::now();
// Code snipppet to measure
let duration: std::time::Duration = start.elapsed();
```
## Motivation
This library is used by a high performance tracing library [`minitrace-rust`](https://github.com/tikv/minitrace-rust). The main purpose is to use [TSC](https://en.wikipedia.org/wiki/Time_Stamp_Counter) on x86 processors to measure time at high speed without losing much accuracy.
## Platform Support
Currently, only the Linux on `x86` or `x86_64` is backed by [TSC](https://en.wikipedia.org/wiki/Time_Stamp_Counter). On other platforms, `minstant` falls back to `std::time`. If TSC is unstable, it will also fall back to `std::time`.
If speed is privileged over accuracy when fallback occurs, you can use `fallback-coarse` feature to use coarse time:
```toml
[dependencies]
minstant = { version = "0.1", features = ["fallback-coarse"] }
```
## Benchmark
Benchmark platform is `Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz` on CentOS 7.
```sh
> cargo criterion
Instant::now()/minstant time: [10.449 ns 10.514 ns 10.619 ns]
Instant::now()/quanta time: [31.467 ns 31.628 ns 31.822 ns]
Instant::now()/std time: [26.831 ns 26.924 ns 27.016 ns]
minstant::Anchor::new() time: [46.987 ns 47.243 ns 47.498 ns]
minstant::Instant::as_unix_nanos() time: [15.287 ns 15.318 ns 15.350 ns]
```
