[Mlir-commits] [llvm] [mlir] Split the llvm::ThreadPool into an abstract base class and an impleme… (PR #82094)

[Mehdi Amini]

joker-eph wrote:

Ah right I see what you mean now, indeed it would be mostly doable with a factory function (with the caveat that you mention about making it a heap allocation and all the updated).

There is also a minor efficiency aspect where if we care about the performance of LLVM_ENABLE_THREADS=OFF the current implementation has a shortcut in the base class (less encapsulation for the task structure in `asyncImpl()`).

https://github.com/llvm/llvm-project/pull/82094