CUDA x86 on MacOS and/or Linux

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

Is CUDA x86 still supported?
is it supported in the community version, or only in the Accelerator edition?
Is NVIDIA nvcc needed along with the PGI compiler to make it work?
is it possible to get a trial version of the Accelerator edition if needed for CUDA x86?

$ pgc++ -Mcudax86 cudax86.cpp
"/opt/pgi/osx86-64/2016/cuda/7.0/include/host_config.h", line 110: catastrophic
          error: #error directive: -- clang and clang++ are the only supported
          host compilers on Mac OS X!
  #error -- clang and clang++ are the only supported host compilers on Mac OS X!
2

Hi Sutter,

Is CUDA x86 still supported?

Yes, but only on Linux.

is it supported in the community version, or only in the Accelerator edition?

The Community Edition contains all the same compiler features as the Professional Edition. The main difference between the two versions is the available support options.

Is NVIDIA nvcc needed along with the PGI compiler to make it work?

No. PGI ships all components needed to compile code targeting NVIDIA GPUs.

Is it possible to get a trial version of the Accelerator edition if needed for CUDA x86?

This is not necessary. The Community Edition for Linux includes support for CUDA x86.

Hope this helps,
Mat