[llvm] [AMDGPU][ISEL] Use target specific hook to block narrowing insert-extract sub-vector (PR #201056)

[via llvm-commits]

================
@@ -27193,8 +27192,6 @@ static SDValue narrowInsertExtractVectorBinOp(SDNode *N, EVT SubVT,
     return SDValue();
 
   EVT VecVT = BinOp.getValueType();
-  if (!TLI.isNarrowingProfitable(N, VecVT, SubVT))
----------------
Shoreshen wrote:

Hi @krzysz00 , the default of this is returning false. This will block lots of target's splitting and create regression on them

https://github.com/llvm/llvm-project/pull/201056