[Mlir-commits] [mlir] [mlir][gpu] Use `known_block_size` to set `maxntid` for NVVM target (PR #77301)

[Krzysztof Drewniak]

krzysz00 wrote:

@grypp Having been poking around all the relevant code ... shouldn't `known_block_size` correspond to `reqdntid` and not `maxntid`? That annotation is an array of the statically-known block sizes that will be used at launch.

(By analogy, the ROCDL version uses `!reqd_work_group_size`, not any of the weaker properties)

https://github.com/llvm/llvm-project/pull/77301