not). Code generation uses it.
- Added a policy that uses it.

Tested on pipeline benchmark.

Compiles correcctly for all targets and target combinations,
pending fix in python script for llvm 4.0 -> llvm 3.4

for sgem visc sh, also required ignoring argmemonly and writeonly attributes in llvm 4.0 to llvm 3.4 conversion

- Note for an additional required change in transformFunctionToVoid

fixed a bug in opencl_nvidia sgemm. We were freeing params before writing output file, which was not generating correct output

for allocation nodes.

(benchmark not updated)

Bug fix in nvptx and spir backends related to adding arguments to
node functions.