5

Overall goal is to use NumbaPro to run some functions on the GPU (on OSX 10.8.3).

Before starting, I just wanted to get everything set up. According to this page I installed CUDA, registered as a CUDA developer, downloaded the Compiler SDK and set up the NUMBAPRO_NVVM=/path/to/libnvvm.dylib environment variable.

However, running this basic test function:

from numbapro import autojit

@autojit(target='gpu')
def my_function(x):
    if x == 0.0:
        return 1.0
    else:
        return x*x*x

print my_function(4.4)
exit()

Brings up this error:

File ".../anaconda/lib/python2.7/site-packages/numba/decorators.py", line 207, in compile_function
    compiled_function = dec(f)
File "...lib/python2.7/site-packages/numbapro/cudapipeline/decorators.py", line 35, in _jit_decorator
File "...lib/python2.7/site-packages/numbapro/cudapipeline/decorators.py", line 128, in __init__
File "...lib/python2.7/site-packages/numbapro/cudapipeline/environment.py", line 31, in generate_ptx
File "...lib/python2.7/site-packages/numbapro/cudapipeline/environment.py", line 186, in _link_llvm_math_intrinsics
KeyError: 1

I've tried @vectorize'ing instead of autojit, same error. @autojit by itself with no target works fine.

Any ideas?

Saullo G. P. Castro
  • 56,802
  • 26
  • 179
  • 234
thearrow3456
  • 101
  • 1
  • 5
  • 1
    You may also want to ask the NumbaPro people. I note on the site you linked it states: "GPU support is rapidly improving but still an area where you may encounter difficulties. Please let us know if you have any trouble with our GPU support." As a product that expects you to [pay for it's use after 30 days](http://docs.continuum.io/numbapro/eula_numbapro.html#trial-period), I would assume they would offer support of some sort. – Robert Crovella Mar 25 '13 at 17:05
  • This really isn't a CUDA question. I am tempted to remove the CUDA tag altogether.... – talonmies Mar 25 '13 at 17:10
  • @RobertCrovella, I'm aware NumbaPro is a paid product, I'm using the [academic license.](https://store.continuum.io/cshop/academicanaconda) I'll try emailing them, but I was hoping for a quick-and-easy fix. – thearrow3456 Mar 25 '13 at 19:50
  • @talonmies, You're right. Done. – thearrow3456 Mar 25 '13 at 19:51

1 Answers1

5

For posterity's sake, I asked Continuum Support. They responded:

It seems that you are running a CUDA GPU with compute capability 1.x. NVVM only supports CC2.0 and above. We definitely should have a better error reporting and make it clear in the NumbaPro documentation for the supported compute capability.

thearrow3456
  • 101
  • 1
  • 5