I am trying to allocate page locked memory. I have tryed both of these function calls:
cudaHostAlloc((void **)&state_to_GPU_d,16sizeof(int),cudaHostAllocDefault)
and
cudaMallocHost ((void **)&state_to_GPU_d, 16sizeof(int))
The variable state_to_GPU_d is simply an int pointer. I have also tried running the SDK program simpleStreams which uses cudaMallocHost on an int pointer as well. For each execution I get the error “feature is not yet implemented”.
I’m using a GeForce GTX 260, compatibility 1.3, and my CUDA driver is 185.18.08. Can someone tell me what I can do to get my code to work?
How could 185.18.08 be installed and not be running? Whenever I try to reinstall the lastest driver the system says I already have 185.18.08. But if there is an older driver, how could it run in place of the latest driver?
NVRM version: NVIDIA UNIX x86 Kernel Module 180.44 Mon Mar 23 14:59:10 PST 2009
GCC version: gcc version 4.3.3 (Ubuntu 4.3.3-5ubuntu4)
Does this mean that 185.18.08 is not installed or that it’s just not the active driver? I will install 185.18.08 again but how do I ensure that it will be the active driver?