[llvm] [AArch64][SVE] Use ADD/ADR instead of MUL/MLA for x*N (PR #198566)

[Tomas Matheson via llvm-commits]

================
@@ -24214,6 +24412,10 @@ static SDValue performIntrinsicCombine(SDNode *N,
                        N->getOperand(1));
   case Intrinsic::aarch64_sve_ext:
     return LowerSVEIntrinsicEXT(N, DAG);
+  case Intrinsic::aarch64_sve_mul:
+    return performAllActiveSVEMulIntrinsicCombine(N, DAG);
+  case Intrinsic::aarch64_sve_mla:
+    return performAllActiveSVEMlaIntrinsicCombine(N, DAG);
----------------
tommat01 wrote:

Removed from this patch

https://github.com/llvm/llvm-project/pull/198566