Segmentation fault in testing_zheevd_gpu.cpp

[nunocardoso]

Hi,

Why the dimension, N, in testing_zheevd_gpu.cpp is not working for every N?

I have run the code for 8192 and it gives segamentation fault, however if I put 8196 it runs withou problems...
The same happens for example with 1600, it gives segmentation fault and runs for example 1602.

Can you help me?
I need to run the code for a matrix 9600x9600 double complex, and for this case the testing code also gives segmentation fault.

I'm using MAGMA 1.2 with CUDA 4.1.

Thanks,
Nuno