Set up cava benchmark

• Set up cava benchmark w/ criterion.
• GPU schedule for cava.
• Convert a bunch of benchmarks from using hercules refs to use hercules boxes.
• Define AlignedAlloc global allocator in hercules_rt that aligns everything to 32 bytes.
• Bump nom version.

juno_samples/cava/benches/cava_bench.rs

+#![feature(concat_idents)]
+
 use criterion::{criterion_group, criterion_main, Criterion};
 
-use juno_cava::{cava_harness, CavaInputs};
+use hercules_rt::{runner, HerculesImmBoxTo};
+
+use juno_cava::*;
+
+juno_build::juno!("cava");
+
+fn cava_bench(c: &mut Criterion) {
+    let mut group = c.benchmark_group("cava bench");
+    group.sample_size(10);
+
+    let args = CavaInputs {
+        input: "examples/raw_tulip-small.bin".to_string(),
+        output: None,
+        verify: false,
+        output_verify: None,
+        cam_model: "cam_models/NikonD7000".to_string(),
+        crop_rows: Some(144),
+        crop_cols: Some(192),
+    };
+    let (raw_image, cam_model) = make_raw_image_and_cam_model(&args);
+    let (rows, cols, num_ctrl_pts, image, tstw, ctrl_pts, weights, coefs, tonemap) =
+        prepare_hercules_inputs(&raw_image, &cam_model);
+    let mut r = runner!(cava);
+
+    group.bench_function("cava bench small", |b| {
+        b.iter(|| {
+            async_std::task::block_on(r.run(
+                rows as u64,
+                cols as u64,
+                num_ctrl_pts as u64,
+                image.to(),
+                tstw.to(),
+                ctrl_pts.to(),
+                weights.to(),
+                coefs.to(),
+                tonemap.to(),
+            ));
+        })
+    });
+
+    let args = CavaInputs {
+        input: "examples/raw_tulips.bin".to_string(),
+        output: None,
+        verify: true,
+        output_verify: None,
+        cam_model: "cam_models/NikonD7000".to_string(),
+        crop_rows: None,
+        crop_cols: None,
+    };
+    let (raw_image, cam_model) = make_raw_image_and_cam_model(&args);
+    let (rows, cols, num_ctrl_pts, image, tstw, ctrl_pts, weights, coefs, tonemap) =
+        prepare_hercules_inputs(&raw_image, &cam_model);
+    let mut r = runner!(cava);
 
-fn cava_bench_small(c: &mut Criterion) {
-    c.bench_function("cava_bench_small", |b| {
+    group.bench_function("cava bench full", |b| {
         b.iter(|| {
-            cava_harness(CavaInputs {
-                input: "examples/raw_tulip-small.bin".to_string(),
-                output: None,
-                verify: false,
-                output_verify: None,
-                cam_model: "cam_models/NikonD7000".to_string(),
-                crop_rows: Some(144),
-                crop_cols: Some(192),
-            })
+            async_std::task::block_on(r.run(
+                rows as u64,
+                cols as u64,
+                num_ctrl_pts as u64,
+                image.to(),
+                tstw.to(),
+                ctrl_pts.to(),
+                weights.to(),
+                coefs.to(),
+                tonemap.to(),
+            ));
         })
     });
 }
 
-criterion_group!(benches, cava_bench_small);
+criterion_group!(benches, cava_bench);
 criterion_main!(benches);