configure script hardcodes location of cuda that makes it fail on OSX

[delip]

Cuda installation on OSX is at $CUDA_TOOLKIT_PATH/lib (not lib64), and on OSX the shared libraries are end in .dylib (not .so).

  if [ -e "$CUDA_TOOLKIT_PATH/lib64/libcudart.so.7.0" ]; then
    break
  fi
  echo "Invalid path to CUDA 7.0 toolkit. ${CUDA_TOOLKIT_PATH}/lib64/libcudart.so.7.0 cannot be found"

[vrv]

Hey @delip, we currently don't have support for GPUs on Mac OS X.

Out of curiosity, what GPU are you trying to get working with TensorFlow?

[dhess]

I'm not OP, but in my case, it would be the D700 and D500 as found on Mac Pro (Late 2013) models (which require OpenCL, of course).

[zheng-xq]

We don't support OpenCL right now. You need a NV GPU with at least compute 3.0 capability.

[dhess]

I understand. I was just replying to @vrv's question about GPUs people are using on OS X.

[delip]

Hey @vrv, I'm on a Mac Pro (late 2013) like Drew. Also, a lot of my friends
develop on their macbook pros.

On Wed, Nov 11, 2015 at 1:03 AM, Drew Hess notifications@github.com wrote:

I understand. I was just replying to @vrv https://github.com/vrv's
question about GPUs people are using on OS X.

—
Reply to this email directly or view it on GitHub
#111 (comment)
.

[vrv]

@delip: so do you also have AMD GPUs?

[delip]

That's correct. I also tried this on the latest iMac 5K which also contains
a AMD GPU (AMD Radeon R9 M290X)

On Wed, Nov 11, 2015 at 5:24 PM, Vijay Vasudevan notifications@github.com
wrote:

@delip https://github.com/delip: so do you also have AMD GPUs?

—
Reply to this email directly or view it on GitHub
#111 (comment)
.

[vrv]

Gotcha, so that would require OpenCL support, which we're tracking here: #22. I'll de-dupe with that feature request.