[clang] [llvm] [AMDGPU][clang][CodeGen][opt] Add late-resolved feature identifying predicates (PR #134016)

[Alex Voicu via cfe-commits]

================
@@ -0,0 +1,157 @@
+//===- AMDGPUExpandFeaturePredicates.cpp - Feature Predicate Expander Pass ===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+// This file implements a pass that deals with expanding AMDGCN generic feature
+// predicates into target specific quantities / sequences. In this context, a
+// generic feature predicate is an implementation detail global variable that
+// is inserted by the FE as a consequence of using either the __builtin_cpu_is
+// or the __builtin_amdgcn_is_invocable special builtins on an abstract target
+// (AMDGCNSPIRV). These placeholder globals are used to guide target specific
+// lowering, once the concrete target is known, by way of constant folding their
+// value all the way into a terminator (i.e. a controlled block) or into a no
+// live use scenario. The pass makes a best effort attempt to look through
+// calls, i.e. a constant evaluatable passthrough of a predicate value will
----------------
AlexVlx wrote:

The comment is vestigial from a prior iteration of this, thank you for catching that, I have to correct it.

https://github.com/llvm/llvm-project/pull/134016