More optimizations

• Speedup backprop
• Fix gpu backend for different CUDA versions

juno_samples/rodinia/backprop/benches/backprop_bench.rs

@@ -68,7 +68,7 @@ fn backprop_bench(c: &mut Criterion) {
     };
 
     bench("backprop bench small", 65536);
-    bench("backprop bench large", 33554432);
+    bench("backprop bench large", 1048576);
 }
 
 criterion_group!(benches, backprop_bench);