[clang] [llvm] [inlineasm] Add special support for "rm" inline asm constraints (PR #181973)

[Bill Wendling via cfe-commits]

================
@@ -207,28 +636,83 @@ static bool processCallBrInst(Function &F, CallBrInst *CBR, DominatorTree *DT) {
   return Changed;
 }
 
-static SmallVector<CallBrInst *, 2> findCallBrs(Function &F) {
-  SmallVector<CallBrInst *, 2> CBRs;
-  for (BasicBlock &BB : F)
-    if (auto *CBR = dyn_cast<CallBrInst>(BB.getTerminator()))
-      if (!CBR->getType()->isVoidTy() && !CBR->use_empty())
-        CBRs.push_back(CBR);
-  return CBRs;
-}
-
-static bool runImpl(Function &F, ArrayRef<CallBrInst *> CBRs,
-                    DominatorTree *DT) {
+static bool runImpl(Function &F, ArrayRef<CallBase *> IAs, DominatorTree *DT,
+                    const TargetMachine *TM) {
   bool Changed = false;
+  bool isOptLevelNone = TM->getOptLevel() == CodeGenOptLevel::None;
----------------
bwendling wrote:

> Perhaps modifying the front end then to only produce "m" when "rm" appear in sources but `TM->getOptLevel() == CodeGenOptLevel::None` is perhaps also a way to avoid transforming the instructions. Then the middle end could fold redundant loads and kill dead stores.

If we go this route, we could simply have two paths in the front-end: one for `-O0` (the original behavior), and one for `-On` (the desired behavior). That would obviate the need for any of this code.

**Pros:** This should "Just Work(tm)" (to borrow a phrase from Todd Howard)! And it's easier to extend to other constraint combos, like `"irm"`, and others that may come up in the future.

**Cons:** This would "break" the use of `-O2 -mllvm -regalloc=fast`, but then this PR doesn't support that either. We also lose the ability to test this in the middle- and back-ends, so a lot of surgery would need to be done on the ASM test files so that we get good coverage.

https://github.com/llvm/llvm-project/pull/181973