[PoC][Clang][CodeGen] Do not use getelementptr for scalable struct.

[HsiangKai]

Summary

We do not support getelementptr for scalable struct. Avoid using getelementptr when dealing with scalable struct arguments.

Use multiple insertvalue and store the whole scalable struct at the end.

[frasercrmck]

Are there any tests for this?

clang/lib/CodeGen/CGCall.cpp

1298–1299

This comment might need to be sunk as it doesn't apply to structs containing scalable vectors?

[HsiangKai]

In D99482#2661334, @frasercrmck wrote:

Are there any tests for this?

The only scalable struct types are Zvlsseg types in Clang. We have use cases in D99593. I have no separate test cases in this commit.

[HsiangKai]

Address comments.

clang/lib/CodeGen/CGCall.cpp

static llvm::Value *CreateCoercedLoad(Address Src, llvm::Type *Ty,
   return CGF.Builder.CreateLoad(Tmp);
 }
 
 // Function to store a first-class aggregate into memory.  We prefer to
 // store the elements rather than the aggregate to be more friendly to
 // fast-isel.
 // FIXME: Do we need to recurse here?
 void CodeGenFunction::EmitAggregateStore(llvm::Value *Val, Address Dest,
                                          bool DestIsVolatile) {
-  // Prefer scalar stores to first-class aggregate stores.
   if (llvm::StructType *STy = dyn_cast<llvm::StructType>(Val->getType())) {
+    if (STy->containsScalableVectorType())
+      Builder.CreateStore(Val, Dest, DestIsVolatile);
+    else {
+      // Prefer scalar stores to first-class aggregate stores.
       for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
         Address EltPtr = Builder.CreateStructGEP(Dest, i);
         llvm::Value *Elt = Builder.CreateExtractValue(Val, i);
         Builder.CreateStore(Elt, EltPtr, DestIsVolatile);
       }
+    }
   } else {
     Builder.CreateStore(Val, Dest, DestIsVolatile);
   }
 }
 
 /// CreateCoercedStore - Create a store to \arg DstPtr from \arg Src,
 /// where the source and destination may have different types.  The
 /// destination is known to be aligned to \arg DstAlign bytes.
case ABIArgInfo::Direct: {
         if (SrcSize <= DstSize) {
           AddrToStoreInto = Builder.CreateElementBitCast(Ptr, STy);
         } else {
           AddrToStoreInto =
             CreateTempAlloca(STy, Alloca.getAlignment(), "coerce");
         }
 
         assert(STy->getNumElements() == NumIRArgs);
+        if (STy->containsScalableVectorType()) {
+          llvm::Value *Val = llvm::UndefValue::get(STy);
+          for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
+            auto AI = Fn->getArg(FirstIRArg + i);
+            AI->setName(Arg->getName() + ".coerce" + Twine(i));
+            Val = Builder.CreateInsertValue(Val, AI, i);
+          }
+          Builder.CreateStore(Val, AddrToStoreInto);
+        } else {
           for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
             auto AI = Fn->getArg(FirstIRArg + i);
             AI->setName(Arg->getName() + ".coerce" + Twine(i));
             Address EltPtr = Builder.CreateStructGEP(AddrToStoreInto, i);
             Builder.CreateStore(AI, EltPtr);
           }
+        }
 
         if (SrcSize > DstSize) {
           Builder.CreateMemCpy(Ptr, AddrToStoreInto, DstSize);
         }
 
       } else {
         // Simple case, just do a coerced store of the argument into the alloca.
         assert(NumIRArgs == 1);
case ABIArgInfo::Direct: {
           Builder.CreateMemCpy(TempAlloca, Src, SrcSize);
           Src = TempAlloca;
         } else {
           Src = Builder.CreateBitCast(Src,
                                       STy->getPointerTo(Src.getAddressSpace()));
         }
 
         assert(NumIRArgs == STy->getNumElements());
+        if (STy->containsScalableVectorType()) {
+          llvm::Value *Val = Builder.CreateLoad(Src);
+          for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
+            llvm::Value *LI = Builder.CreateExtractValue(Val, i);
+            IRCallArgs[FirstIRArg + i] = LI;
+          }
+        } else {
           for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
             Address EltPtr = Builder.CreateStructGEP(Src, i);
             llvm::Value *LI = Builder.CreateLoad(EltPtr);
             IRCallArgs[FirstIRArg + i] = LI;
           }
+        }
       } else {
         // In the simple case, just pass the coerced loaded value.
         assert(NumIRArgs == 1);
         llvm::Value *Load =
             CreateCoercedLoad(Src, ArgInfo.getCoerceToType(), *this);
 
         if (CallInfo.isCmseNSCall()) {
           // For certain parameter types, clear padding bits, as they may reveal