Page 1 of 1

CuBLAS not found, even though Nvidia examples work fine

PostPosted: Thu Oct 30, 2014 11:34 am
by govg
OS : Centos 6
GPU : Quadro K2000
Compute : 3.0
CUDA : 6.0
CULA : Dense, Academic

Hi, I recently installed CUDA, and then CULA on my university machine. I compiled and ran all the Nvidia examples provided, including the CUBLAS ones, (matrixMulCublas etc.) and they ran well, showing accurate results. When I try compiling the examples from CULA, I get the following error:
"ld error : -lcublas not found"

I have set the LD_LIBRARY_PATH to include both /usr/local/cuda/lib64 as well as CULA libraries, and I can not understand what the problem is.

Also, I switched ${CC} to nvcc in the makefile for basic usage, (build64) and it built and ran without displaying anything. But trying this on others didn't work, as most would show some error based on gcc specific tags.