CL_INVALID_KERNEL_ARGS with OpenCL 1.1 Nvidia device


I have an issue with one of my program.
It’s an image processing algorithm. It works with these configurations:

(1)- Laptop : CPU Intel Core 2 Duo P8600 with OpenCL 1.1 Intel (Build 31360.31441) and OpenCL 1.2 AMD-APP (923.1) platforms
(2)- Desktop : CPU Intel Core 2 Duo E8400 with Intel OpenCL 1.1 Intel (Build 31360.31441) platform.

But not with:
(3)- Desktop : GPU Nvidia GTX260 I OpenCL 1.1 CUDA 4.2.1 platform.

Configurations number 2 and 3 are on the same computer.

So, the error CL_INVALID_KERNEL_ARGS is returned by clEnqueueNDRangeKernel.
I checked several times the arguments of the kernels passed in clEnqueueNDRangeKernel, and they are all correct and initialized.

This issue is already known ?

I want to profile my program and I want to use the Nvidia Visual Profiler. So if you know an other good profiling opencl software. I tried gDebugger but it’s a bit messy.