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

[Sam Tebbs via llvm-commits]

================
@@ -5028,10 +5041,19 @@ calculateRegisterUsage(VPlan &Plan, ArrayRef<ElementCount> VFs,
           // even in the scalar case.
           RegUsage[ClassID] += 1;
         } else {
+          // The output from scaled phis and scaled reductions actually has
+          // fewer lanes than the VF.
+          unsigned ScaleFactor = getVFScaleFactor(R);
+          ElementCount VF = VFs[J].divideCoefficientBy(ScaleFactor);
+          LLVM_DEBUG(if (VF != VFs[J]) {
+            dbgs() << "LV(REG): Scaled down VF from " << VFs[J] << " to " << VF
----------------
SamTebbs33 wrote:

Unfortunately the VF that has partial reductions is pruned before the register usage is calculated. I can add one as part of https://github.com/llvm/llvm-project/pull/132190 once this is merged, though.

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