I am using OpenCL to run programs on NVIDIA TESLA S2050. The call to clBuildProgram to compile the OpenCL kernel seems to take slightly more than a minute (60+ seconds), irrespective of what kernel it is. Even the SDK examples take the same time (> 1 minute) to build the program.
Has anyone else encountered this issue? The CUDA driver and toolkit versions are both 3.1. The TESLA is part of a cluster environment, where the Linux OS is CentOS release 5.5. But, I am using only one node-one GPU.
Thanks!
I am using OpenCL to run programs on NVIDIA TESLA S2050. The call to clBuildProgram to compile the OpenCL kernel seems to take slightly more than a minute (60+ seconds), irrespective of what kernel it is. Even the SDK examples take the same time (> 1 minute) to build the program.
Has anyone else encountered this issue? The CUDA driver and toolkit versions are both 3.1. The TESLA is part of a cluster environment, where the Linux OS is CentOS release 5.5. But, I am using only one node-one GPU.
Thanks!
To add, my driver version is 256.40, and I am NOT running in the debug mode.
To add, my driver version is 256.40, and I am NOT running in the debug mode.