How many threads can reside in a CUDA core at the same time?

Accelerated Computing CUDA CUDA Programming and Performance
1

I am getting some conflicting answers to this question.

In the programming guide (V8.0), it is said that “…all threads of a block are expected to reside on the same processor core and must share the limited memory resources of that core”, so it seems that a core can have more than one thread residing on it. However, in this following post

[url]http://yosefk.com/blog/simd-simt-smt-parallelism-in-nvidia-gpus.html[/url]

it is said that “…an NVIDIA GPU contains several largely independent processors called ‘Streaming Multiprocessors’ (SMs), each SM hosts several ‘cores’, and each ‘core’ runs a thread”.

I wonder which of the two is correct? Or maybe they are both correct, but there is something wrong with my understanding?

Thank you!

2

They are both correct. The first statement is using processor core to refer to SM, so the two statements are consistent. All threads of a block reside on the same SM, and they share the resources of that SM.

3

That clears the air. Thanks a lot!