[llvm] [LoopVectorizer] Add support for partial reductions (PR #92418)

[Sam Tebbs via llvm-commits]

================
@@ -2266,6 +2275,39 @@ class VPReductionPHIRecipe : public VPHeaderPHIRecipe,
   bool isInLoop() const { return IsInLoop; }
 };
 
+/// A recipe for forming partial reductions. In the loop, an accumulator and
+/// vector operand are added together and passed to the next iteration as the
+/// next accumulator. After the loop body, the accumulator is reduced to a
+/// scalar value.
+class VPPartialReductionRecipe : public VPRecipeWithIRFlags {
+  unsigned Opcode;
+
+public:
+  template <typename IterT>
+  VPPartialReductionRecipe(unsigned ReductionOpcode,
+                           iterator_range<IterT> Operands)
+      : VPRecipeWithIRFlags(VPDef::VPPartialReductionSC, Operands),
+        Opcode(ReductionOpcode) {
+    assert(isa<VPReductionPHIRecipe>(getOperand(1)->getDefiningRecipe()) &&
+           "Unexpected operand order for partial reduction recipe");
+  }
+  ~VPPartialReductionRecipe() override = default;
+  VPPartialReductionRecipe *clone() override {
+    return new VPPartialReductionRecipe(Opcode, operands());
+  }
+
+  VP_CLASSOF_IMPL(VPDef::VPPartialReductionSC)
+
+  /// Generate the reduction in the loop
+  void execute(VPTransformState &State) override;
+
+#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+  /// Print the recipe.
+  void print(raw_ostream &O, const Twine &Indent,
+             VPSlotTracker &SlotTracker) const override;
+#endif
+};
----------------
SamTebbs33 wrote:

Yep I was working on that on Friday afternoon, pushed now! It took a little while since my rebase somehow added a `VPSingleDefRecipe::computeCost` implementation that caused some build issues.

https://github.com/llvm/llvm-project/pull/92418