[llvm] [NewPM][CodeGen] Add `MachineFunctionAnalysis` (PR #88610)

[Arthur Eubanks via llvm-commits]

================
@@ -0,0 +1,46 @@
+//===- MachineFunctionAnalysis.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
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains the definitions of the MachineFunctionAnalysis members.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/CodeGen/MachineFunctionAnalysis.h"
+#include "llvm/CodeGen/MachineFunction.h"
+#include "llvm/CodeGen/MachineModuleInfo.h"
+#include "llvm/Target/TargetMachine.h"
+
+using namespace llvm;
+
+AnalysisKey MachineFunctionAnalysis::Key;
+
+bool MachineFunctionAnalysis::Result::invalidate(
+    Function &, const PreservedAnalyses &PA,
+    FunctionAnalysisManager::Invalidator &) {
+  // Unless it is invalidated explicitly, it should remain preserved.
+  auto PAC = PA.getChecker<MachineFunctionAnalysis>();
+  return !PAC.preservedWhenStateless();
+}
+
+MachineFunctionAnalysis::Result
+MachineFunctionAnalysis::run(Function &F, FunctionAnalysisManager &FAM) {
+  /// Next unique number available for a MachineFunction.
+  static unsigned NextFnNum = 0;
----------------
aeubanks wrote:

this seems problematic. I believe people use multiple LLVMContexts in a process and are able to parallelize optimization/codegen that way.

perhaps we should put this in LLVMContext? unsure

https://github.com/llvm/llvm-project/pull/88610