[llvm] [SROA] Propagate no-signed-zeros(nsz) fast-math flag on the phi node using function attribute (PR #83381)

[Sushant Gokhale via llvm-commits]

================
@@ -1113,6 +1114,17 @@ void PromoteMem2Reg::RenamePass(BasicBlock *BB, BasicBlock *Pred,
         for (unsigned i = 0; i != NumEdges; ++i)
           APN->addIncoming(IncomingVals[AllocaNo], Pred);
 
+        // For the  sequence `return X > 0.0 ? X : -X`, it is expected that this
+        // results in fabs intrinsic. However, without no-signed-zeros(nsz) flag
+        // on the phi node generated at this stage, fabs folding does not
+        // happen. So, we try to infer nsz flag from the function attributes to
+        // enable this fabs folding.
+        bool FnHasNSZAttr = APN->getFunction()
+                                ->getFnAttribute("no-signed-zeros-fp-math")
+                                .getValueAsBool();
----------------
sushgokh wrote:

you mean check this way, right? 
`if(... && <check nsz flag> && ...)`

I did the above way previously

https://github.com/llvm/llvm-project/pull/83381