Kernel too big? be.exe returned non-zero status -1073741571

Accelerated Computing CUDA CUDA Programming and Performance
1

Is there a way to find out if the size of my kernel is getting too big?

I have a quite complex kernel, lots of branches etc

1>nvopencc ERROR: C:\CUDA\bin64/…/open64/lib//be.exe returned non-zero status -1073741571

The program compiles and works fine in emulation mode

I am using cuda 2.3 on 64 bit.

2

It also shows error 0x2 ?

1>nvopencc ERROR: C:\CUDA\bin64/…/open64/lib//be.exe returned non-zero status -1073741571
1># --error 0x2 –

3

are you sure that problem is due to large size of your kernel?

you can comment some part of your kernel and check if error occurs.

in fact, you can use loop unrolling to flatten the for-loop and such unrolling

would lead to huge binary code, so your problem may not be code size.