OpenACC and Fortran

I am beginner with openACC, when I add openACC directives to my Fortran program, the execution time increases !! Is it normal?

Can someone explain to me or give me a detailed tutorial to understand loop gang, loop vector, reduction, copyin, copyout …

I need your collaboration
THX

Hi inti_sar,

It’s not uncommon to see an initial slowdown due to extra data movement between the device and host, if the program has insufficient parallelization or a small work load.

Can someone explain to me or give me a detailed tutorial to understand loop gang, loop vector, reduction, copyin, copyout …

That a bit too long of a question to try and answer via the user forum, but there’s many resources out there for you. I’d recommend starting at the OpenACC resource page: https://www.openacc.org/resources

There are several courses and articles you can review, as well as book recommendations.

If you still have specific questions, I’m happy to help!

-Mat

Forgiving the self-promotion, there are a lot of nice tips for implementing OpenACC here:

https://devblogs.nvidia.com/solar-storm-modeling-gpu-openacc/