volatile in Cuda Fortran

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

Programming Guide Version 2.0 (p.21) say:
“Only after the execution of a __syncthreads() are writes to
shared variables guaranteed to be visible by other threads. Unless the variable is declared as volatile, the compiler is free to optimize the reads and writes to shared memory as long as the previous statement is met.”
How realize “volatile” in cuda fortran for shered-memory?

2

How realize “volatile” in cuda fortran for shered-memory?

We don’t support volatile for device variables. However, I just submitted a feature request (TPR#17460) and sent it to our engineers.

Thanks,
Mat

3

Hello,

Engineering reports that TPR 17460 was fixed in our 11.8 release.

Thanks again for the report.

regards,
dave