Double pointer in CUDA

Accelerated Computing CUDA CUDA Programming and Performance
1

I want to make a double pointer in CUDA.
Actually I want to save a various size table in one array.
for example,

void *tables[0x100]
for(i = 0; i < 10; i ++){
tables[i] = malloc(0x100 * 0x100);
}

I want to make such program using cuda. :(
Do you have any idea about it?