Cooperative Groups for flexible thread handling

CUDA 9.0 introduces a new CUDA programming feature named Cooperative Groups. This introduces a new CUDA programming design pattern for CUDA collective operations by specifying group-wise operations. Using this, programmers can write CUDA code that controls CUDA threads explicitly.

To begin with, let's see what Cooperative Groups is and its programming advantages.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
13.58.150.59