About Altera OpenCL Compilation

Thank you so much for the detailed answer!

I have one more question regarding the local memory: if I have a kernel where all of its workgroups needs to read a common block of data (stored in global memory) which was updated by a previous launched kernel, is it possible to load this block of data only in the first workgroup to some kind of local memory space, and then share it across all others workgroups, so that the kernel don't have to load it multiple times? I know that the local memory can not be shared across workgroups, but I am just wondering if there is a mechanism in opencl that allow this kind of memory optimization. Basically I am wondering if there is a kind of memory that can be shared across workgroups of a single kernel that is faster than global memory. Thanks!