Synchronizing stream between calls to NPP component labeling functions

I’m running the following to find connected components within an image and find their sizes:

int newNumber;
nppiCompressMarkerLabelsUF_32u_C1IR_Ctx(..., &newNumber,...);

From what I understand, in order for the newNumber parameter to the third function to be ready, I’d have to insert a call to cudaStreamSynchronize(stream) to synchronize the calls before passing that parameter to the next function (otherwise the value passed might not have been changed yet- I’m assuming a call to cudaMemcpyAsync() is used).
However, if I know for certain that there is a maximum number of connected components that can be present in the image, can I pass that maximum to the labels info function as long as I provided a large enough host memory for the infos?
Or does the CompressMarkerLabels function synchronize the stream since it’s returning a host value?
Also, out of curiosity, why isn’t there a nppiCompressedMarkerLabelsUFInfo_32u_C1R function (without the NppStreamContext parameter)?


Did you manage to figure it out?