If i create a kernel like this :

[codebox]__kernel void Test(int iN, float fValue, __global float *pValue)


int iID = get_global_id(0);

int iSize = get_global_size(0);

for(unsigned int iPath = iID; iPath < iN; iPath += iSize)


        float af[100] = {0};

	pValue[iPath] = fValue;



i Have a crash in clBuildProgram :

Assertion Failed

Program: …

File : C:\nv\sw\undev\gpu_drv\opencl_a\co…\PTXBackend.cpp

Expression : 0 && "UNknown LLVM intrinsic!é

Whereas i remember in CUDA i have not the same problem

if i reduce the size of the tab for example

float af[4] = {0} i have not the problem

Thx in advance for your help


anyone to solve this problem ?

maybe you should post a bug report on the nvdeveloper website where you downloaded the opencl sdk?..

Good idea