[clang] [llvm] [SPIRV] GPU intrinsics (PR #131190)

[Dmitry Sidorov via llvm-commits]

================
@@ -0,0 +1,501 @@
+//===- LowerGPUIntrinsic.cpp ----------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Lower the llvm.gpu intrinsics to target specific code sequences.
+// Can be called from clang if building for a specific GPU or from the backend
+// as part of a SPIRV lowering pipeline. Initial pass can lower to amdgcn or
----------------
MrSidims wrote:

First of all I agree, that having a common interface between various frontends is a good idea. Yet as from what I see now - the intrinsics added in the PR are not common, but tuned only for AMDGPU and NVPTX. And just like @michalpaszkowski I don't understand, what place SPIR-V has for those intrinsics. 

>The spirv backend should stash these intrinsics in the SPV file like any other ones. It might need a patch to do so or we might get default handling for llvm prefixed intrinsics. That's the point really - we pass information about the SIMT computation through the SPV onward to whatever is dealing with that later.
>I don't know what distinction you're drawing between abstract function calls and intrinsics, terminology is not consistent across compilers here.

Please correct me if I understood this part incorrectly. Is your proposal to store `llvm.gpu` intrinsic call in SPIR-V module as is if there is no appropriate SPIR-V instruction for it?

https://github.com/llvm/llvm-project/pull/131190