[llvm] [LV, VP]VP intrinsics support for the Loop Vectorizer (PR #76172)

[Florian Hahn via llvm-commits]

================
@@ -1156,6 +1156,45 @@ static VPActiveLaneMaskPHIRecipe *addVPLaneMaskPhiAndUpdateExitBranch(
   return LaneMaskPhi;
 }
 
+/// Replaces (ICMP_ULE, WideCanonicalIV, backedge-taken-count) pattern using
+/// the given idiom \p Idiom.
+static void replaceHeaderPredicateWithIdiom(
+    VPlan &Plan, VPValue &Idiom,
+    function_ref<bool(VPUser &, unsigned)> Cond = {}) {
+  auto *FoundWidenCanonicalIVUser =
+      find_if(Plan.getCanonicalIV()->users(),
+              [](VPUser *U) { return isa<VPWidenCanonicalIVRecipe>(U); });
+  if (FoundWidenCanonicalIVUser == Plan.getCanonicalIV()->users().end())
+    return;
+  auto *WideCanonicalIV =
+      cast<VPWidenCanonicalIVRecipe>(*FoundWidenCanonicalIVUser);
+  // Walk users of WideCanonicalIV and replace all compares of the form
+  // (ICMP_ULE, WideCanonicalIV, backedge-taken-count) with
+  // the given idiom VPValue.
+  VPValue *BTC = Plan.getOrCreateBackedgeTakenCount();
+  for (VPUser *U : SmallVector<VPUser *>(WideCanonicalIV->users())) {
+    auto *CompareToReplace = dyn_cast<VPInstruction>(U);
+    if (!CompareToReplace ||
+        CompareToReplace->getOpcode() != Instruction::ICmp ||
+        CompareToReplace->getPredicate() != CmpInst::ICMP_ULE ||
+        CompareToReplace->getOperand(1) != BTC)
+      continue;
+
+    assert(CompareToReplace->getOperand(0) == WideCanonicalIV &&
+           "WidenCanonicalIV must be the first operand of the compare");
+    if (Cond) {
+      CompareToReplace->replaceUsesWithIf(&Idiom, Cond);
+      if (!CompareToReplace->getNumUsers())
+        CompareToReplace->eraseFromParent();
+    } else {
+      CompareToReplace->replaceAllUsesWith(&Idiom);
+      CompareToReplace->eraseFromParent();
+    }
+  }
+  if (!WideCanonicalIV->getNumUsers())
+    WideCanonicalIV->eraseFromParent();
----------------
fhahn wrote:

> > When introducing the EVL, the transform should do something like Plan->setUF(1)
> 
> It does. It adds UF=1 to Ufs, but VPlan print does not print the list of Ufs, it just checks the size and if UFs is not empty, prints "UF>=1". "UF=1" is printed only if UFs is empty.

I might be missing something but the code here doesn't seem to do `setUF(1)`. If `UFs` are set, it should only print those (https://github.com/llvm/llvm-project/blob/main/llvm/lib/Transforms/Vectorize/VPlan.cpp#L972)

https://github.com/llvm/llvm-project/pull/76172