Hi All,

I am wondring if there is any way to decalre dynamic arrays in CUDA?



I mean decalre arrays withouts specifying the size.

Never specifying the size? Certainly doable with C++, but definitely a performance killer.

I guess you are asking about device side memory allocation, which are possible on compute capability 2.x devices.

Yes. I am wondering if i can allocate array without specifying the size on the device side to store the resuls that computed by the kernel?

The reason I need that:

I am trying to implement a kernel that finds the intersection between two polygons. So, I dont know how many points in the calculated ploygon before excuting the kernel?