diff --git a/lib/Target/ARM64/ARM64FastISel.cpp b/lib/Target/ARM64/ARM64FastISel.cpp
index 0d4001ebf67..f8b54d033f7 100644
--- a/lib/Target/ARM64/ARM64FastISel.cpp
+++ b/lib/Target/ARM64/ARM64FastISel.cpp
@@ -1250,10 +1250,16 @@ bool ARM64FastISel::ProcessCallArgs(SmallVectorImpl<Value *> &Args,
       assert(VA.isMemLoc() && "Assuming store on stack.");
 
       // Need to store on the stack.
+      unsigned ArgSize = VA.getLocVT().getSizeInBits() / 8;
+
+      unsigned BEAlign = 0;
+      if (ArgSize < 8 && !Subtarget->isLittleEndian())
+        BEAlign = 8 - ArgSize;
+
       Address Addr;
       Addr.setKind(Address::RegBase);
       Addr.setReg(ARM64::SP);
-      Addr.setOffset(VA.getLocMemOffset());
+      Addr.setOffset(VA.getLocMemOffset() + BEAlign);
 
       if (!EmitStore(ArgVT, Arg, Addr))
         return false;
diff --git a/test/CodeGen/ARM64/fast-isel-call.ll b/test/CodeGen/ARM64/fast-isel-call.ll
index 637ce2898f9..8d756ae5461 100644
--- a/test/CodeGen/ARM64/fast-isel-call.ll
+++ b/test/CodeGen/ARM64/fast-isel-call.ll
@@ -1,4 +1,5 @@
 ; RUN: llc < %s -O0 -fast-isel-abort -mtriple=arm64-apple-darwin | FileCheck %s
+; RUN: llc < %s -O0 -fast-isel-abort -mtriple=arm64_be-linux-gnu | FileCheck %s --check-prefix=CHECK-BE
 
 define void @call0() nounwind {
 entry:
@@ -89,3 +90,11 @@ entry:
 }
 
 declare i32 @func2(i64 zeroext, i32 signext, i16 zeroext, i8 signext, i1 zeroext, i1 zeroext)
+
+declare void @callee_b0f(i8 %bp10, i8 %bp11, i8 %bp12, i8 %bp13, i8 %bp14, i8 %bp15, i8 %bp17, i8 %bp18, i8 %bp19)
+define void @caller_b1f() {
+entry:
+  ; CHECK-BE: strb w{{.*}}, [sp, #7]
+  call void @callee_b0f(i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 42)
+  ret void
+}