Device memory state between kernel calls

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi all,

I am wondering if the device memory between kernel executions remains in the same state? Just in case if we neither flushed anything nor copied a new data and want to reuse the memory in 2nd kernell call. If somebody could point me to the official documentation describing the case that would be also nice.

Thanks,
Andrei

2

They remain as they are … And hence you can re-use a kernel’s result in a subsequent kernel… (this is true if the results are stored in global memory) This is documented in the programming guide.