Error compiling cuda code on OSX

Accelerated Computing HPC Compilers Legacy PGI Compilers
1

Hi,

I have some code written for cuda that i want to compile with pgi compiler, but i experience issue with it.
Each time i want to compile some .cu files, i get the following error :

"/opt/pgi/osx86-64/2012/cuda/4.0/include/math_functions.h", line 397: error: 
          more than one instance of overloaded function "__isinf" has "C"
          linkage
  extern __host__ __device__ int                    __isinf(double) __THROW;
                                                    ^

"/opt/pgi/osx86-64/2012/cuda/4.0/include/math_functions.h", line 402: error: 
          more than one instance of overloaded function "__isnan" has "C"
          linkage
  extern __host__ __device__ int                    __isnan(double) __THROW;

I get this error even when i try to compile the nvidia examples which are included with pig workstation. (And my code compile just fine with nvcc).

PGI version is :

20:14:45 - zacharie******@MacBook ~/tmp/CUDA-x86-SDK $ pgcpp -V

pgcpp 12.2-0 64-bit target on Apple OS/X -tp penryn 
Copyright 1989-2000, The Portland Group, Inc.  All Rights Reserved.
Copyright 2000-2012, STMicroelectronics, Inc.  All Rights Reserved.

Running on OSX Lion 10.7.3.

Does anybody have a suggestion on that ?

Thank you,

Zacharie

2

I’ve just solve the problem, probably not in the cleanest way, but it seems working.

I had to comment the line in the .h and it did the job.

Zacharie.

3

Zacharie-

Thanks for pointing out this issue. I’ve fixed it for release 12.3.