NVIDIA's OpenCL 1.1 and clCreateSubBuffer

Accelerated Computing CUDA CUDA Programming and Performance
1

dear all,

I’m currently working with NVIDIA’s preview drivers for OpenCL 1.1 on a Tesla T10 and I’m trying to use “clCreateSubBuffer”

however, each time I run my code, I get a segmentation fault on that command

is there a solution to this problem?

should I just avoid using clCreateSubBuffer when using NVIDIA’s implementation of OpenCL?

the code looks something like this:

cl_mem buffer = clCreateBuffer(context, CL_MEM_READ_WRITE, size, NULL, &status);

cl_buffer_region region;

region.origin = 0;

region.size = size_subBuffer;

cl_mem subBuffer = clCreateSubBuffer(buffer, CL_MEM_READ_WRITE, CL_BUFFER_CREATE_TYPE_REGION, &region, &status);

where size_subBuffer < size

2

Just tried it on a GTX285 and a Tesla C2050 with the same result. It works (well, it doesn’t seg fault) with the

CPU_DEVICE_TYPE using the AMD implementation so it is presumably some error with the NVIDIA implementation.

–

jason

3

+1 Ubuntu 15.10 driver 352.