Merge remote-tracking branch 'origin/master'

Cargo.toml

@@ -11,6 +11,11 @@ path = "benches/fill_f32.rs"
 name = "fill_f32_slice"
 path = "benches/fill_f32_slice.rs"
 
+[[bin]]
+name = "ones_f32"
+path = "benches/ones_f32.rs"
+
+
 [dependencies]
 cpu-time = "1.0.0"
 bitflags = "2.6.0"

benches/ones_f32.py

@@ -0,0 +1,25 @@
+import numpy as np
+import torch
+
+from perfprofiler import *
+
+class TensorOnes(TimingSuite):
+    def __init__(self, n):
+        self.n = n
+
+    @measure_performance("NumPy")
+    def run(self):
+        np.ones(self.n, dtype="float32")
+
+    @measure_performance("PyTorch CPU")
+    def run(self):
+        torch.ones(self.n, dtype=torch.float32)
+
+    @measure_rust_performance("Chela CPU", target="ones_f32")
+    def run(self, executable):
+        return self.run_rust(executable, self.n)
+
+if __name__ == "__main__":
+    sizes = [2 ** n for n in range(9, 25)]
+    results = TensorOnes.profile_each(sizes, n=10)
+    plot_results(sizes, results, "tensor.ones() CPU time vs length")

benches/ones_f32.rs

@@ -0,0 +1,16 @@
+use chela::*;
+use std::env;
+use cpu_time::ProcessTime;
+
+fn profile(size: usize) -> u128 {
+    let start = ProcessTime::now();
+    let tensor: Tensor<f32> = Tensor::ones(size);
+    start.elapsed().as_nanos()
+}
+
+fn main() {
+    let args: Vec<String> = env::args().collect();
+    let size = args[1].parse::<usize>().unwrap();
+
+   println!("{}", profile(size));
+}

benches/perfprofiler/profile.py

@@ -59,5 +59,4 @@ def wrapper(self, *args, **kwargs):
 
     return decorator
 
-
 __all__ = ["measure_performance", "measure_rust_performance"]

benches/perfprofiler/util.py

@@ -15,7 +15,6 @@ def compile_rust(target: str) -> str:
     out = out.split(b"\n")[-3]
     return json.loads(out)["executable"]
 
-
 def get_method_class(method) -> str:
     return method.__qualname__.split(".")[0]