Unweaving warp specialization on modern tensor core GPUs
rohany.github.io
rohany.github.io
And here is my understanding where it differs:
1. multi-stage pipeline requires careful hand-tuning, even at PTX level to make sure your async wait is weaved properly to maximize overlap.
2. since these register files now is huge, multi-stage pipeline is difficult to write at intrinsics level to make efficient use of these huge register files.
3. Warp specialization delegated most of these scheduling dynamically, hence it is better adapted to hardware (and have more information to make scheduling decisions at runtime). Although this is a bit moot because we write different code for different hardware anyway.
Anything more I am missing?
The fact that we tend to need different warp specialization strategies for different hardware is a consequence of the capabilities of that hardware (i.e. different asynchronous instruction types), and contributes to the complexity of targeting that new hardware.
I guess this post assumes the need to use all the gpu resources from within a single block.
Yes, that is correct. However, most MMA-style kernels that utilize the Tensor Core usually need enough resources per block that only 1 block fits on each SM.