I’m using OpenACC directives with “pgcpp 14.10-0 64-bit target on x86-64 Windows -tp haswell The Portland Group - PGI Compilers and Tools” version. When I compile hello world examples with pgcpp, compiler outputs the information such as “acclerator kernel generated”, “loop is vectorized”, etc. Now, I am struggling compiling a large code set which is written in c++ and I’m trying to run on gpu using OpenACC directives. I parallelized many for loops but in the compiler output I don’t see any message that any of those kernels are generated. I just see that very few “Accelerator region ignored” messages. When I run the exactly same code in my friednd’s pc who uses “pgcpp 14.9-0 64-bit target on x86-64 Windows -tp sandybridge The Portland Group - PGI Compilers and Tools”, I get the messages such as “acclerator kernel generated”, “loop is vectorized”. Does any faced a similar strange problem? The only difference between two of us just compiler version. Could that somehow cause that?
P.S. Needless to say, I use -Minfo flag to see the acceleretor kernel related messages:)