Any method for time overlap?

Accelerated Computing CUDA CUDA Programming and Performance
1

In the programming guide[Steams].it says:

Each stream copies its portion of input array hostPtr to array inputDevPtr in

device memory, processes inputDevPtr on the device by calling myKernel(), and

copies the result outputDevPtr back to the same portion of hostPtr. Processing

hostPtr using two streams allows for the memory copies of one stream to overlap

with the kernel execution of the other stream.

Its for two task in two streams.

Are there any method to make copy and kernel launch in only one stream be overlapped.

What I mean is only one stream. and copy only portion (or currently requied) data for kernel launch which can make the tow job be overlaped?

2

a stream represents a sequential series of events, so it wouldn’t make sense to overlap events. If you want things to overlap, you need to use multiple streams.

3

Thank you,~

But if I do copy and kenel launch in two different streams.

the data used in kernel won’t wait for it to be transferd, right?

So we don’t have a mechanism to make the portional copy and prtional kernel launch which will wait for the required data copied?