diff --git a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
--- a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
+++ b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h
@@ -133,7 +133,9 @@
       bool UseMaskForCond = false, bool UseMaskForGaps = false);
   InstructionCost getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
                                         TTI::TargetCostKind CostKind);
-
+  bool areFunctionArgsABICompatible(const Function *Caller,
+                                    const Function *Callee,
+                                    SmallPtrSetImpl<Argument *> &Args) const;
   /// @}
 };
 
diff --git a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
--- a/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
+++ b/llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp
@@ -1220,6 +1220,26 @@
   return BaseT::getIntrinsicInstrCost(ICA, CostKind);
 }
 
+bool PPCTTIImpl::areFunctionArgsABICompatible(
+    const Function *Caller, const Function *Callee,
+    SmallPtrSetImpl<Argument *> &Args) const {
+  if (!BaseT::areFunctionArgsABICompatible(Caller, Callee, Args))
+    return false;
+
+  const DataLayout &CallerDL = Caller->getParent()->getDataLayout();
+  const DataLayout &CalleeDL = Callee->getParent()->getDataLayout();
+
+  return llvm::none_of(Args, [CallerDL, CalleeDL](Argument *A) {
+    auto *EltTy = cast<PointerType>(A->getType())->getElementType();
+    if (EltTy->isSized()) {
+      unsigned CallerTypeSize = CallerDL.getTypeSizeInBits(EltTy);
+      unsigned CalleeTypeSize = CalleeDL.getTypeSizeInBits(EltTy);
+      return (CallerTypeSize > 128 || CalleeTypeSize > 128);
+    }
+    return false;
+  });
+}
+
 bool PPCTTIImpl::canSaveCmp(Loop *L, BranchInst **BI, ScalarEvolution *SE,
                             LoopInfo *LI, DominatorTree *DT,
                             AssumptionCache *AC, TargetLibraryInfo *LibInfo) {
diff --git a/llvm/test/CodeGen/PowerPC/arg_promotion.ll b/llvm/test/CodeGen/PowerPC/arg_promotion.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/arg_promotion.ll
@@ -0,0 +1,40 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S -argpromotion -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s
+; RUN: opt -S -passes=argpromotion -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s
+
+; Test to check that we do not promote arguments when the
+; type size is greater than 128 bits.
+
+define internal fastcc void @print_acc(<512 x i1>* nocapture readonly %a) nounwind {
+; CHECK-LABEL: @print_acc(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = load <512 x i1>, <512 x i1>* [[A:%.*]], align 64
+; CHECK-NEXT:    [[TMP1:%.*]] = tail call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.ppc.mma.disassemble.acc(<512 x i1> [[TMP0]])
+; CHECK-NEXT:    [[TMP2:%.*]] = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } [[TMP1]], 0
+; CHECK-NEXT:    ret void
+;
+entry:
+  %0 = load <512 x i1>, <512 x i1>* %a, align 64
+  %1 = tail call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.ppc.mma.disassemble.acc(<512 x i1> %0)
+  %2 = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } %1, 0
+  ret void
+}
+
+declare { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.ppc.mma.disassemble.acc(<512 x i1>) nounwind
+
+define dso_local void @test(<512 x i1>* nocapture %a, <16 x i8> %ac) {
+; CHECK-LABEL: @test(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = tail call <512 x i1> @llvm.ppc.mma.xvf32ger(<16 x i8> [[AC:%.*]], <16 x i8> [[AC]])
+; CHECK-NEXT:    store <512 x i1> [[TMP0]], <512 x i1>* [[A:%.*]], align 64
+; CHECK-NEXT:    tail call fastcc void @print_acc(<512 x i1>* nonnull [[A]])
+; CHECK-NEXT:    ret void
+;
+entry:
+  %0 = tail call <512 x i1> @llvm.ppc.mma.xvf32ger(<16 x i8> %ac, <16 x i8> %ac)
+  store <512 x i1> %0, <512 x i1>* %a, align 64
+  tail call fastcc void @print_acc(<512 x i1>* nonnull %a)
+  ret void
+}
+
+declare <512 x i1> @llvm.ppc.mma.xvf32ger(<16 x i8>, <16 x i8>) nounwind