[llvm] [CodeGen] Add scalarization fallback for multi-result scalable intrinsics (PR #218634)

[Mattéo Rizza Murgier via llvm-commits]

================
@@ -808,6 +809,37 @@ bool PreISelIntrinsicLowering::lowerIntrinsics(Module &M) const {
         return lowerUnaryVectorIntrinsicAsLoop(M, CI);
       });
       break;
+    case Intrinsic::modf:
+    case Intrinsic::sincos:
+    case Intrinsic::sincospi:
+      Changed |= forEachCall(F, [&](CallInst *CI) {
+        Type *Ty = CI->getArgOperand(0)->getType();
+        if (!TM || !isa<ScalableVectorType>(Ty))
+          return false;
+        const TargetLowering *TL = TM->getSubtargetImpl(F)->getTargetLowering();
+        unsigned Op = TL->IntrinsicIDToISD(F.getIntrinsicID());
----------------
matteo-rm wrote:

`Op` is still needed for the `isOperationExpand` check so it has to be computed either way. Were you suggesting we use  `getLibcallForIntrinsic` here?  It currently doesn't map vector types so it returns `UNKNOWN_LIBCALL` for these intrinsics; but I can expand it if that's what you had in mind.

https://github.com/llvm/llvm-project/pull/218634