search

andersbll / cudarray

CUDA-based NumPy
MIT License
234 stars 61 forks source link

Install not working on OS X #7

Closed elyase closed 9 years ago

elyase commented 9 years ago

I clone the project, then:

~/D/cudarray master ❯ make
g++ -O3 -fPIC -Wall -Wfatal-errors -I./include -I/usr/local/cuda/include -c -o src/nnet/conv_bc01_matmul.o src/nnet/conv_bc01_matmul.cpp
In file included from src/nnet/conv_bc01_matmul.cpp:1:
./include/cudarray/common.hpp:8:10: fatal error: 'cuda_runtime_api.h' file not
      found
#include <cuda_runtime_api.h>
         ^
1 error generated.
make: *** [src/nnet/conv_bc01_matmul.o] Error 1
andersbll commented 9 years ago

It can't find the CUDA header files. Make sure the CUDA path is set to where the CUDA SDK is installed. If you don't want CUDA acceleration you should install without CUDA.

elyase commented 9 years ago

Oh thanks that worked! I somehow assumed that you also had to do make install without CUDA.

andersbll commented 9 years ago

Ok, I'll add to the 'Without CUDA' installation description that you don't have to build libcudarray. :)

elyase commented 9 years ago

I think it is already clear enough but it might save blind people like me some time,:-) Thanks!

elyase commented 9 years ago

Some what off-topic but do you get high CPU usage in deeppy? I get only about 138% in my quad-core.

andersbll commented 9 years ago

Make sure that Numpy's matrix multiplication operations are parallelized. If you are running the convolution examples you will get poor multi-core performance because the convolution code is not yet parallelized.