More optimizations

• Speedup backprop
• Fix gpu backend for different CUDA versions

juno_samples/rodinia/bfs/benches/bfs_bench.rs

@@ -13,41 +13,22 @@ fn bfs_bench(c: &mut Criterion) {
 
     let mut r = runner!(bfs);
 
-    group.bench_function("bfs bench 4096", |b| {
-        let input = "data/graph4096.txt";
-        let (nodes, source, edges) = parse_graph(input.into()).unwrap();
-        let n = nodes.len() as u64;
-        let m = edges.len() as u64;
-        let nodes = HerculesImmBox::from(&nodes as &[Node]);
-        let edges = HerculesImmBox::from(&edges as &[u32]);
-        b.iter(|| {
-            async_std::task::block_on(async { r.run(n, m, nodes.to(), source, edges.to()).await });
-        })
-    });
-
-    group.bench_function("bfs bench 65536", |b| {
-        let input = "data/graph65536.txt";
-        let (nodes, source, edges) = parse_graph(input.into()).unwrap();
-        let n = nodes.len() as u64;
-        let m = edges.len() as u64;
-        let nodes = HerculesImmBox::from(&nodes as &[Node]);
-        let edges = HerculesImmBox::from(&edges as &[u32]);
-        b.iter(|| {
-            async_std::task::block_on(async { r.run(n, m, nodes.to(), source, edges.to()).await });
-        })
-    });
-
-    group.bench_function("bfs bench 64M", |b| {
-        let input = "/scratch/aaronjc4/rodinia_3.1/data/bfs/graph64M.txt";
+    let mut bench = |name, input: &'_ str| {
         let (nodes, source, edges) = parse_graph(input.into()).expect("PANIC: Couldn't read input file for 64M benchmark. Currently, this benchmark uses a hard-coded path, so it can only be run on the lab machines.");
         let n = nodes.len() as u64;
         let m = edges.len() as u64;
         let nodes = HerculesImmBox::from(&nodes as &[Node]);
         let edges = HerculesImmBox::from(&edges as &[u32]);
-        b.iter(|| {
-            async_std::task::block_on(async { r.run(n, m, nodes.to(), source, edges.to()).await });
-        })
-    });
+        group.bench_function(name, |b| {
+            b.iter(|| {
+                async_std::task::block_on(async { r.run(n, m, nodes.to(), source, edges.to()).await });
+            })
+        });
+    };
+
+    bench("bfs bench 4096", "data/graph4096.txt");
+    bench("bfs bench 65536", "data/graph65536.txt");
+    bench("bfs bench 64M", "/scratch/aaronjc4/rodinia_3.1/data/bfs/graph64M.txt");
 }
 
 criterion_group!(benches, bfs_bench);