wip

README.md

@@ -15,7 +15,8 @@
 |1,090|[numbapig.py](Python/numbapig.py)|Python, Numba, CUDA, 5120 cores|NVIDIA V100|March, 2020|
 |1,062|[taichipi.py](Python/taichipi.py)|Python, Taichi, 5120 cores|NVIDIA V100|March, 2023|
 |811|prior|Cray XT4|C, MPI, 2048 processes|prior|
-|484|[threadpi.rs](Rust/threadpi.rs)|Rust, 96 threads and cores<br>cargo run --release -- 96|Graviton4|December, 2024|
+|501|[rayonpi.rs](Rust/rayonpi.rs)|Rust Rayon, 96 cores<br>cargo run --release -- 96|Graviton4|December, 2024|
+|484|[threadpi.rs](Rust/threadpi.rs)|Rust threads, 96 cores<br>cargo run --release -- 96|Graviton4|December, 2024|
 |315|[numbapip.py](Python/numbapip.py)|Python, Numba, parallel, fastmath<br>96 cores|Intel 2x Xeon Platinum 8175M|February, 2020|
 |272|[threadpi.c](C/threadpi.c)|C, 96 threads<br>gcc threadpi.c -o threadpi -O3 -ffast-math -pthread|Intel 2x Xeon Platinum 8175M|June, 2019|
 |267|[threadpi.cpp](C++/threadpi.cpp)|C++, 96 threads<br>g++ threadpi.cpp -o threadpi -O3 -ffast-math -pthread|Intel 2x Xeon Platinum 8175M|March, 2020|
@@ -31,6 +32,7 @@
 |16.1|[pi.html](https://pub.pages.cba.mit.edu/pi/JavaScript/pi.html)|JavaScript, 6 workers|Intel i7-8700T|November, 2018|
 |15.7|[clusterpi.js](Node/clusterpi.js)|Node, 6 workers|Intel i7-8700T|December, 2018|
 |9.37|[pi.c](C/pi.c)|C<br>gcc pi.c -o pi -lm -O3 -ffast-math|Intel i7-8700T|November, 2018|
+|5.55|[rayonpi.rs](Rust/rayonpi.rs)|Rust Rayon, 1 core<br>cargo run --release -- 96|Graviton4|December, 2024|
 |4.87|[numbapi.py](Python/numbapi.py)|Python, Numba|Intel i7-8700T|February, 2020|
 |4.63|[pi.c](C/pi.c)|C<br>gcc pi.c -o pi -lm -O3|Intel i7-8700T|December, 2024|
 |3.73|[pi.html](https://pub.pages.cba.mit.edu/pi/JavaScript/pi.html)|JavaScript, 1 worker|Intel i7-8700T|November, 2018|

Rust/rayonpi.rs

+/*
+* rayonpi.rs
+* Neil Gershenfeld 12/24/24
+* Rust Rayon parallel pi calculation benchmark
+* pi = 3.14159265358979323846
+*/
+
+use std::time::SystemTime;
+use std::thread::available_parallelism;
+use rayon::prelude::*;
+
+const NPTS:u64 = 1e9 as u64;
+
+fn main() {
+   let a:f64 = 0.5;
+   let b:f64 = 0.75;
+   let c:f64 = 0.25;
+   //
+   let start = SystemTime::now().duration_since(SystemTime::UNIX_EPOCH).unwrap().as_micros();
+   let pi:f64 = (1..=NPTS).into_iter()
+      .map(|i| a/(((i as f64)-b)*((i as f64)-c)))
+      .sum();
+   let end = SystemTime::now().duration_since(SystemTime::UNIX_EPOCH).unwrap().as_micros();
+   let dt = ((end-start) as f64)/1e6;
+   let mflops = (NPTS as f64)*5.0/((dt as f64)*1e6);
+   println!("NPTS = {NPTS}, cores = 1, pi = {pi}");
+   println!("time = {dt:.3}, estimated MFlops = {mflops:.0}");
+   //
+   let num_cores:u64 = available_parallelism().unwrap().get() as u64;
+   let start = SystemTime::now().duration_since(SystemTime::UNIX_EPOCH).unwrap().as_micros();
+   let npts:u64 = NPTS*num_cores;
+   let pi:f64 = (1..=npts).into_par_iter()
+      .map(|i| a/(((i as f64)-b)*((i as f64)-c)))
+      .sum();
+   let end = SystemTime::now().duration_since(SystemTime::UNIX_EPOCH).unwrap().as_micros();
+   let dt = ((end-start) as f64)/1e6;
+   let mflops = (npts as f64)*5.0/((dt as f64)*1e6);
+   println!("NPTS = {npts}, cores = {num_cores}, pi = {pi}");
+   println!("time = {dt:.3}, estimated MFlops = {mflops:.0}");
+   }