accelerator does not match the profile error

Hi,

I am running pgcc version 16.7 in linux mint. Few days ago everything was running well.

However, suddenly I am getting this error when I run any openAcc program (see below).
This is happening only for openAcc programs. Cuda Fortran and openMP programs produced using pgi compilers runs fine.

Any ideas?

Thanks.

$ laplace_parallelOpenAcc 1000

Ruuning 1000 iterations
Current file: /home/efblack/Development/OpenACC_Programming_and_Best_PracticeGuide/Chapter4MF/openAcc/laplace.c
function: laplace
line: 9
Current region was compiled for:
NVIDIA Tesla GPU sm30 sm35
Available accelerators:
device[1]: Native X86 (CURRENT DEVICE)
The accelerator does not match the profile for which this program was compiled


$ pgcc --version
pgcc 16.7-0 64-bit target on x86-64 Linux -tp haswell
The Portland Group - PGI Compilers and Tools
Copyright © 2016, NVIDIA CORPORATION. All rights reserved.

Hi efblack2,

That’s the error you get when the runtime can’t find the CUDA driver.

What does the output from the “pgaccelinfo” utility show?

  • Mat

Here is the output of pgaccelinfo and pgaccelinfo -v

(Notice thelibcuda.so not found
and libcoi_host.so not found)

$ pgaccelinfo
No accelerators found.
Try pgaccelinfo -v for more information

$ pgaccelinfo -v
libcuda.so not found

OpenCL Platform: NVIDIA CUDA
OpenCL Vendor: NVIDIA Corporation

Device Number: 0
Device Name: GeForce GT 750M
Available: Yes
Compiler Available: Yes
Device Version: OpenCL 1.2 CUDA
Global Memory Size: 2146762752
Maximum Object Size: 536690688
Global Cache Size: 32768
Max Clock (MHz): 967
Compute Units: 2
Constant Memory Size: 65536
Local Memory Size: 49152
Workgroup Size: 1024
Address Bits: 64
ECC Support: No

Device Number: 1
Device Name: GeForce GT 750M
Available: Yes
Compiler Available: Yes
Device Version: OpenCL 1.2 CUDA
Global Memory Size: 2147287040
Maximum Object Size: 536821760
Global Cache Size: 32768
Max Clock (MHz): 967
Compute Units: 2
Constant Memory Size: 65536
Local Memory Size: 49152
Workgroup Size: 1024
Address Bits: 64
ECC Support: No
libcoi_host.so not found


Cuda Fortran has not problem finding the GPUs

libcuda.so not found

This is the problem, though don’t know why it’s occurring. Can you try reinstalling the CUDA Driver?

http://www.nvidia.com/Download/index.aspx?lang=en-us

The libcuda.so was there, but not in the expected place.
It happen after a fresh install of the OS and cuda toolkit in my computer.

A soft link into the right directory solve the problem.


Thanks a lot.

Edgar