Bench for edge detection

Cargo.lock

@@ -1338,6 +1338,7 @@ version = "0.1.0"
 dependencies = [
  "async-std",
  "clap",
+ "criterion",
  "hercules_rt",
  "juno_build",
  "opencv",

juno_samples/edge_detection/Cargo.toml

@@ -27,3 +27,10 @@ async-std = "*"
 clap = { version = "*", features = ["derive"] }
 with_builtin_macros = "0.1.0"
 opencv = { version = "*", features = ["clang-runtime"], optional = true }
+
+[dev-dependencies]
+criterion = { version = "0.5", features = ["html_reports"] }
+
+[[bench]]
+name = "edge_detection_bench"
+harness = false

juno_samples/edge_detection/benches/edge_detection_bench.rs

+#![feature(concat_idents)]
+use std::slice::from_raw_parts;
+
+use criterion::{criterion_group, criterion_main, Criterion};
+
+use opencv::core::{Mat, Size, CV_32F, CV_8U};
+use opencv::imgproc::{cvt_color_def, ColorConversionCodes};
+use opencv::prelude::{MatTraitConst, VideoCaptureTrait, VideoCaptureTraitConst};
+use opencv::videoio::{VideoCapture, VideoCaptureProperties, VideoWriter, VideoWriterTrait};
+
+use hercules_rt::{runner, HerculesImmBox, HerculesImmBoxTo};
+
+use juno_edge_detection::*;
+
+juno_build::juno!("edge_detection");
+
+fn edge_detection_bench(c: &mut Criterion) {
+    let mut group = c.benchmark_group("edge detection bench");
+    group.sample_size(10);
+
+    let input = "examples/formula1_scaled.mp4";
+
+    let gs: usize = 7;
+    let gaussian_filter: Vec<f32> = vec![
+        0.000036, 0.000363, 0.001446, 0.002291, 0.001446, 0.000363, 0.000036, 0.000363, 0.003676,
+        0.014662, 0.023226, 0.014662, 0.003676, 0.000363, 0.001446, 0.014662, 0.058488, 0.092651,
+        0.058488, 0.014662, 0.001446, 0.002291, 0.023226, 0.092651, 0.146768, 0.092651, 0.023226,
+        0.002291, 0.001446, 0.014662, 0.058488, 0.092651, 0.058488, 0.014662, 0.001446, 0.000363,
+        0.003676, 0.014662, 0.023226, 0.014662, 0.003676, 0.000363, 0.000036, 0.000363, 0.001446,
+        0.002291, 0.001446, 0.000363, 0.000036,
+    ];
+    let gaussian_filter_h = HerculesImmBox::from(gaussian_filter.as_slice());
+
+    let sz: usize = 3;
+    let structure: Vec<f32> = vec![1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0];
+    let structure_h = HerculesImmBox::from(structure.as_slice());
+
+    let sb: usize = 3;
+    let sx: Vec<f32> = vec![-1.0, 0.0, 1.0, -2.0, 0.0, 2.0, -1.0, 0.0, 1.0];
+    let sx_h = HerculesImmBox::from(sx.as_slice());
+
+    let sy: Vec<f32> = vec![-1.0, -2.0, -1.0, 0.0, 0.0, 0.0, 1.0, 2.0, 1.0];
+    let sy_h = HerculesImmBox::from(sy.as_slice());
+
+    let theta: f32 = 0.1;
+
+    let mut video = VideoCapture::from_file_def(&input).expect("Error loading video");
+    assert!(video.is_opened().unwrap());
+
+    let fps = video
+        .get(VideoCaptureProperties::CAP_PROP_FPS.into())
+        .expect("Error getting fps");
+
+    let _num_frames = video
+        .get(VideoCaptureProperties::CAP_PROP_FRAME_COUNT.into())
+        .expect("Error getting number of frames") as usize;
+    let width = video
+        .get(VideoCaptureProperties::CAP_PROP_FRAME_WIDTH.into())
+        .expect("Error getting width") as usize;
+    let height = video
+        .get(VideoCaptureProperties::CAP_PROP_FRAME_HEIGHT.into())
+        .expect("Error getting height") as usize;
+    let num_frames = 5;
+
+    let mut r = runner!(edge_detection);
+
+    group.bench_function("edge detection bench", |b| {
+        b.iter(|| {
+            for i in 0..num_frames {
+                let frame = load_frame(&mut video);
+                let ptr = frame.ptr_def().unwrap() as *const f32;
+
+                assert!(frame.rows() as usize == height);
+                assert!(frame.cols() as usize == width);
+
+                let input = unsafe { from_raw_parts(ptr, height * width) };
+
+                let input_h = HerculesImmBox::from(input);
+
+                let result = async_std::task::block_on(async {
+                    r.run(
+                        height as u64,
+                        width as u64,
+                        gs as u64,
+                        sz as u64,
+                        sb as u64,
+                        input_h.to(),
+                        gaussian_filter_h.to(),
+                        structure_h.to(),
+                        sx_h.to(),
+                        sy_h.to(),
+                        theta,
+                    )
+                    .await
+                });
+            }
+        })
+    });
+}
+
+criterion_group!(benches, edge_detection_bench);
+criterion_main!(benches);

juno_samples/edge_detection/src/lib.rs

@@ -34,7 +34,7 @@ pub struct EdgeDetectionInputs {
     pub frames: Option<usize>,
 }
 
-fn load_frame(video: &mut VideoCapture) -> Mat {
+pub fn load_frame(video: &mut VideoCapture) -> Mat {
     let mut frame = Mat::default();
 
     let Ok(true) = video.read(&mut frame) else {
@@ -67,7 +67,7 @@ fn load_frame(video: &mut VideoCapture) -> Mat {
     result
 }
 
-fn frame_from_slice(frame: &[f32], height: usize, width: usize) -> Mat {
+pub fn frame_from_slice(frame: &[f32], height: usize, width: usize) -> Mat {
     let result = Mat::from_slice(frame)
         .expect("Failed to create matrix from result")
         .reshape(1, height as i32)