More optimizations

• Speedup backprop
• Fix gpu backend for different CUDA versions

juno_samples/rodinia/bfs/src/gpu.sch

@@ -38,6 +38,12 @@ fixpoint {
 }
 simpl!(collect);
 
+fork-tile[32, 0, false, true](traverse, collect);
+let (outer, inner) = fork-reshape[[1], [0]](traverse);
+let traverse_body = outline(inner);
+let (outer, inner) = fork-reshape[[1], [0]](collect);
+let collect_body = outline(inner);
+
 fork-tile[1024, 0, false, true](traverse, collect);
 fork-split(traverse, collect);