rust-stream

rust-stream

This is an implementation of BabelStream in Rust.

Currently, we support three CPU threading API as devices:

• Plain - basic single-threaded for version, see plain_stream.rs
• Rayon - Parallel with high level API, see rayon_stream.rs
• Crossbeam - Parallel with partitions per thread, see crossbeam_stream.rs
• Arc - Parallel with Vec per thread (static partitions) wrapped in Mutex contained in Arcs, see crossbeam_stream.rs
• Unsafe - Parallel with unsafe pointer per thread (static partitions) to Vec, see crossbeam_stream.rs

In addition, this implementation also supports the following extra flags:

---

--init    Initialise each benchmark array at allocation time on the main thread
--malloc  Use libc malloc instead of the Rust's allocator for benchmark array allocation
--pin     Pin threads to distinct cores, this has NO effect in Rayon devices

Max thread count is controlled by the environment variable BABELSTREAM_NUM_THREADS which is compatible for all devices (avoid setting RAYON_NUM_THREADS, the implementation will issue a warning if this happened).

There is an ongoing investigation on potential performance issues under NUMA situations. As part of the experiment, this implementation made use of the provisional Allocator traits which requires rust unstable. We hope a NUMA aware allocator will be available once the allocator API reaches rust stable.

Build & Run

Prerequisites:

• Rust toolchain

Once the toolchain is installed, enable the nightly channel:

> rustup install nightly
> rustup default nightly # optional, this sets `+nightly` automatically for cargo calls later

With cargo on path, compile and run the benchmark with:

> cd rust-stream/
> cargo +nightly build --release # or simply `cargo build --release` if nightly channel is the default 
> ./target/release/rust-stream --help
rust-stream 3.4.0

USAGE:
    rust-stream [FLAGS] [OPTIONS]

FLAGS:
        --csv             Output as csv table
        --float           Use floats (rather than doubles)
    -h, --help            Prints help information
        --init            Initialise each benchmark array at allocation time on the main thread
        --list            List available devices
        --malloc          Use libc malloc instead of the Rust's allocator for benchmark array allocation
        --mibibytes       Use MiB=2^20 for bandwidth calculation (default MB=10^6)
        --nstream-only    Only run nstream
        --pin             Pin threads to distinct cores, this has NO effect in Rayon devices
        --triad-only      Only run triad
    -V, --version         Prints version information

OPTIONS:
    -s, --arraysize <arraysize>    Use <arraysize> elements in the array [default: 33554432]
        --device <device>          Select device at <device> [default: 0]
    -n, --numtimes <numtimes>      Run the test <numtimes> times (NUM >= 2) [default: 100]