[llvm] [RISCV] Add combine for shadd family of instructions. (PR #130829)

[Stefan Pintilie via llvm-commits]

================
@@ -14547,9 +14634,12 @@ static SDValue performADDCombine(SDNode *N,
     return V;
   if (SDValue V = transformAddImmMulImm(N, DAG, Subtarget))
     return V;
-  if (!DCI.isBeforeLegalize() && !DCI.isCalledByLegalizer())
+  if (!DCI.isBeforeLegalize() && !DCI.isCalledByLegalizer()) {
     if (SDValue V = transformAddShlImm(N, DAG, Subtarget))
       return V;
+    if (SDValue V = combineShlAddIAdd(N, DAG, Subtarget))
----------------
stefanp-synopsys wrote:

That's a good point.
I would say that running `transformAddShlImm()` first might provide a little advantage. For example, given that c0 and c1 cooperate:
```
(add (addi (add (shl x, c0), (shl y, c1)), c2), z)
-- > Run transformAddShlImm() 
(add (addi (shl (sh*add x, y) c0), c2, z)
--> Run transformAddShlImm()
(addi (sh*add (sh*add x, y), z), c2)
```
Running `transformAddShlImm()` first would not really help because it produces `sh*add` and `addi` which are not helpful for the previous transformation.

I do believe that this is the order in which they run. 

https://github.com/llvm/llvm-project/pull/130829