[llvm] [AArch64][Codegen]Transform saturating smull to sqdmulh (PR #143671)

[David Green via llvm-commits]

================
@@ -9365,6 +9368,13 @@ def : Pat<(v4i32 (mulhu V128:$Rn, V128:$Rm)),
                              (EXTRACT_SUBREG V128:$Rm, dsub)),
            (UMULLv4i32_v2i64 V128:$Rn, V128:$Rm))>;
 
+
+def : Pat<(v8i16 (AArch64sqdmulh (v8i16 V128:$Rn), (v8i16 V128:$Rm))),
+          (SQDMULHv8i16 V128:$Rn, V128:$Rm)>;
----------------
davemgreen wrote:

Can you add patterns and tests for 64bit vectors too (v4i16, v2i32).

https://github.com/llvm/llvm-project/pull/143671