About Stack and Heap

I have read that each thread get 256 kb of stack. Is there other stacks or not ? And their size ?

I have the same question for the heap because I found nothing about this in the guide (but it’s possible that I have bad eyes :P).

Where have you read about the stack?

There’s no dynamic memory allocation within the threads. The register file (8KB) gets partitioned among all the active threads (as many as are allowed by occupancy).


That’s in emulation mode, when code runs on CPU.

When code runs on GPU there’s no such thing as stack, at least as of CUDA 1.0.