[llvm] [AArch64] Give a higher cost for more expensive SVE FCMP instructions (PR #153816)

[Sander de Smalen via llvm-commits]

================
@@ -4462,6 +4490,12 @@ InstructionCost AArch64TTIImpl::getCmpSelInstrCost(
              (VecPred == FCmpInst::FCMP_ONE || VecPred == FCmpInst::FCMP_UEQ))
       Factor = 3; // fcmxx+fcmyy+or
 
+    if (isa<ScalableVectorType>(ValTy) &&
----------------
sdesmalen-arm wrote:

nit: it would be nice if would also do this for fixed-length types when we know that the type will be lowered using SVE? (using `useSVEForFixedLengthVectors`), although that may require more changes to this function (e.g. for the case above)

https://github.com/llvm/llvm-project/pull/153816