[llvm] [AMDGPU] Create an AMDGPUIfConverter pass (PR #106415)

[Jason Eckhardt via llvm-commits]

Juan Manuel Martinez =?utf-8?q?Caamaño?= <juamarti at amd.com>,
Juan Manuel Martinez =?utf-8?q?Caamaño?= <juamarti at amd.com>,
Juan Manuel Martinez =?utf-8?q?Caamaño?= <juamarti at amd.com>,
Juan Manuel Martinez =?utf-8?q?Caamaño?= <juamarti at amd.com>,
Juan Manuel Martinez =?utf-8?q?Caamaño?= <juamarti at amd.com>
Message-ID:
In-Reply-To: <llvm.org/llvm/llvm-project/pull/106415 at github.com>


nvjle wrote:

I realize this is a draft patch, but this refactoring of the early if-conversion code is a great idea!

As this matures, does it makes sense to separate the target-independent portion into an isolated patch? Or at least provide a more generic title so other downstream users don't overlook the target-independent changes.

FWIW, I'm using it downstream in an NVIDIA backend-- specifically, the `EarlyIfPredicator`--  and I know of at least one other downstream backend using it. I'd be happy to review the target-independent code and try it locally on our tree. I believe the `EarlyIfPredicator` is not currently used in-tree , but the `EarlyIfConverter` is.

Thanks again for spearheading this work.


https://github.com/llvm/llvm-project/pull/106415