[llvm-branch-commits] [llvm] [LV] Reduce register usage for scaled reductions (PR #133090)

[Sam Tebbs via llvm-branch-commits]

================
@@ -5026,10 +5026,23 @@ calculateRegisterUsage(VPlan &Plan, ArrayRef<ElementCount> VFs,
             // even in the scalar case.
             RegUsage[ClassID] += 1;
           } else {
+            // The output from scaled phis and scaled reductions actually have
+            // fewer lanes than the VF.
+            auto VF = VFs[J];
+            if (auto *ReductionR = dyn_cast<VPReductionPHIRecipe>(R))
----------------
SamTebbs33 wrote:

Yeah that's a nice idea. We could add a `VPScaledRecipe` class. I agree with doing it afterwards.

https://github.com/llvm/llvm-project/pull/133090