Round 1 of rodinia schedule optimization

• Write better CPU schedules for all rodinia benchmarks.
• Add const-inline pass (inter-procedural constant propagation).
• Add module name to generated function names to avoid linker name collisions.

hercules_cg/src/gpu.rs

@@ -724,8 +724,13 @@ namespace cg = cooperative_groups;
         write!(w, "\tcudaError_t err;\n")?;
         write!(
             w,
-            "\t{}_gpu<<<{}, {}, {}>>>({});\n",
-            self.function.name, num_blocks, num_threads, dynamic_shared_offset, pass_args
+            "\t{}_{}_gpu<<<{}, {}, {}>>>({});\n",
+            self.module_name,
+            self.function.name,
+            num_blocks,
+            num_threads,
+            dynamic_shared_offset,
+            pass_args
         )?;
         write!(w, "\terr = cudaGetLastError();\n")?;
         write!(