diff --git a/cl/_testrt/concat/out.ll b/cl/_testrt/concat/out.ll
index 764decd6..91d51114 100644
--- a/cl/_testrt/concat/out.ll
+++ b/cl/_testrt/concat/out.ll
@@ -21,27 +21,30 @@ _llgo_0:
   br label %_llgo_1
 
 _llgo_1:                                          ; preds = %_llgo_2, %_llgo_0
-  %5 = phi ptr [ %3, %_llgo_0 ], [ %14, %_llgo_2 ]
-  %6 = phi i64 [ %4, %_llgo_0 ], [ %15, %_llgo_2 ]
-  %7 = phi i64 [ -1, %_llgo_0 ], [ %8, %_llgo_2 ]
-  %mrv = insertvalue %"github.com/goplus/llgo/internal/runtime.String" poison, ptr %5, 0
-  %mrv1 = insertvalue %"github.com/goplus/llgo/internal/runtime.String" %mrv, i64 %6, 1
-  ret %"github.com/goplus/llgo/internal/runtime.String" %mrv1
-  %8 = add i64 %7, 1
-  %9 = icmp slt i64 %8, %1
-  br i1 %9, label %_llgo_2, label %_llgo_3
+  %5 = phi ptr [ %3, %_llgo_0 ], [ %18, %_llgo_2 ]
+  %6 = phi i64 [ %4, %_llgo_0 ], [ %19, %_llgo_2 ]
+  %7 = phi i64 [ -1, %_llgo_0 ], [ %12, %_llgo_2 ]
+  %8 = alloca %"github.com/goplus/llgo/internal/runtime.String", align 8
+  %9 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %8, i32 0, i32 0
+  store ptr %5, ptr %9, align 8
+  %10 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %8, i32 0, i32 1
+  store i64 %6, ptr %10, align 4
+  %11 = load %"github.com/goplus/llgo/internal/runtime.String", ptr %8, align 8
+  %12 = add i64 %7, 1
+  %13 = icmp slt i64 %12, %1
+  br i1 %13, label %_llgo_2, label %_llgo_3
 
 _llgo_2:                                          ; preds = %_llgo_1
-  %10 = call ptr @"github.com/goplus/llgo/internal/runtime.SliceData"(%"github.com/goplus/llgo/internal/runtime.Slice" %0)
-  %11 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %10, i64 %8
-  %12 = load %"github.com/goplus/llgo/internal/runtime.String", ptr %11, align 8
-  %13 = call %"github.com/goplus/llgo/internal/runtime.String" @"github.com/goplus/llgo/internal/runtime.StringCat"(void <badref>, %"github.com/goplus/llgo/internal/runtime.String" %12)
-  %14 = extractvalue %"github.com/goplus/llgo/internal/runtime.String" %13, 0
-  %15 = extractvalue %"github.com/goplus/llgo/internal/runtime.String" %13, 1
+  %14 = call ptr @"github.com/goplus/llgo/internal/runtime.SliceData"(%"github.com/goplus/llgo/internal/runtime.Slice" %0)
+  %15 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %14, i64 %12
+  %16 = load %"github.com/goplus/llgo/internal/runtime.String", ptr %15, align 8
+  %17 = call %"github.com/goplus/llgo/internal/runtime.String" @"github.com/goplus/llgo/internal/runtime.StringCat"(%"github.com/goplus/llgo/internal/runtime.String" %11, %"github.com/goplus/llgo/internal/runtime.String" %16)
+  %18 = extractvalue %"github.com/goplus/llgo/internal/runtime.String" %17, 0
+  %19 = extractvalue %"github.com/goplus/llgo/internal/runtime.String" %17, 1
   br label %_llgo_1
 
 _llgo_3:                                          ; preds = %_llgo_1
-  ret void <badref>
+  ret %"github.com/goplus/llgo/internal/runtime.String" %11
 }
 
 define void @main.init() {
@@ -86,7 +89,7 @@ declare i64 @"github.com/goplus/llgo/internal/runtime.SliceLen"(%"github.com/gop
 
 declare ptr @"github.com/goplus/llgo/internal/runtime.SliceData"(%"github.com/goplus/llgo/internal/runtime.Slice")
 
-declare %"github.com/goplus/llgo/internal/runtime.String" @"github.com/goplus/llgo/internal/runtime.StringCat"(%"github.com/goplus/llgo/internal/runtime.Slice")
+declare %"github.com/goplus/llgo/internal/runtime.String" @"github.com/goplus/llgo/internal/runtime.StringCat"(%"github.com/goplus/llgo/internal/runtime.String", %"github.com/goplus/llgo/internal/runtime.String")
 
 declare %"github.com/goplus/llgo/internal/runtime.String" @"github.com/goplus/llgo/internal/runtime.NewString"(ptr, i64)
 
diff --git a/internal/runtime/llgo_autogen.ll b/internal/runtime/llgo_autogen.ll
index cdb2cbde..ce133348 100644
--- a/internal/runtime/llgo_autogen.ll
+++ b/internal/runtime/llgo_autogen.ll
@@ -265,65 +265,38 @@ _llgo_0:
   ret i64 %3
 }
 
-define %"github.com/goplus/llgo/internal/runtime.String" @"github.com/goplus/llgo/internal/runtime.StringCat"(%"github.com/goplus/llgo/internal/runtime.Slice" %0) {
+define %"github.com/goplus/llgo/internal/runtime.String" @"github.com/goplus/llgo/internal/runtime.StringCat"(%"github.com/goplus/llgo/internal/runtime.String" %0, %"github.com/goplus/llgo/internal/runtime.String" %1) {
 _llgo_0:
-  %1 = alloca %"github.com/goplus/llgo/internal/runtime.String", align 8
   %2 = alloca %"github.com/goplus/llgo/internal/runtime.String", align 8
-  %3 = call i64 @"github.com/goplus/llgo/internal/runtime.SliceLen"(%"github.com/goplus/llgo/internal/runtime.Slice" %0)
-  br label %_llgo_1
-
-_llgo_1:                                          ; preds = %_llgo_2, %_llgo_0
-  %4 = phi i64 [ 0, %_llgo_0 ], [ %13, %_llgo_2 ]
-  %5 = phi i64 [ -1, %_llgo_0 ], [ %6, %_llgo_2 ]
-  %6 = add i64 %5, 1
-  %7 = icmp slt i64 %6, %3
-  br i1 %7, label %_llgo_2, label %_llgo_3
-
-_llgo_2:                                          ; preds = %_llgo_1
-  %8 = call ptr @"github.com/goplus/llgo/internal/runtime.SliceData"(%"github.com/goplus/llgo/internal/runtime.Slice" %0)
-  %9 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %8, i64 %6
-  %10 = load %"github.com/goplus/llgo/internal/runtime.String", ptr %9, align 8
-  store %"github.com/goplus/llgo/internal/runtime.String" %10, ptr %2, align 8
-  %11 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %2, i32 0, i32 1
-  %12 = load i64, ptr %11, align 4
-  %13 = add i64 %4, %12
-  br label %_llgo_1
-
-_llgo_3:                                          ; preds = %_llgo_1
-  %14 = call ptr @"github.com/goplus/llgo/internal/runtime.Alloc"(i64 %4)
-  %15 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %1, i32 0, i32 0
-  store ptr %14, ptr %15, align 8
-  %16 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %1, i32 0, i32 1
-  store i64 %4, ptr %16, align 4
-  %17 = alloca %"github.com/goplus/llgo/internal/runtime.String", align 8
-  %18 = call i64 @"github.com/goplus/llgo/internal/runtime.SliceLen"(%"github.com/goplus/llgo/internal/runtime.Slice" %0)
-  br label %_llgo_4
-
-_llgo_4:                                          ; preds = %_llgo_5, %_llgo_3
-  %19 = phi ptr [ %14, %_llgo_3 ], [ %33, %_llgo_5 ]
-  %20 = phi i64 [ -1, %_llgo_3 ], [ %21, %_llgo_5 ]
-  %21 = add i64 %20, 1
-  %22 = icmp slt i64 %21, %18
-  br i1 %22, label %_llgo_5, label %_llgo_6
-
-_llgo_5:                                          ; preds = %_llgo_4
-  %23 = call ptr @"github.com/goplus/llgo/internal/runtime.SliceData"(%"github.com/goplus/llgo/internal/runtime.Slice" %0)
-  %24 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %23, i64 %21
-  %25 = load %"github.com/goplus/llgo/internal/runtime.String", ptr %24, align 8
-  store %"github.com/goplus/llgo/internal/runtime.String" %25, ptr %17, align 8
-  %26 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %17, i32 0, i32 0
-  %27 = load ptr, ptr %26, align 8
-  %28 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %17, i32 0, i32 1
-  %29 = load i64, ptr %28, align 4
-  %30 = call ptr @memcpy(ptr %19, ptr %27, i64 %29)
-  %31 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %17, i32 0, i32 1
-  %32 = load i64, ptr %31, align 4
-  %33 = getelementptr i8, ptr %19, i64 %32
-  br label %_llgo_4
-
-_llgo_6:                                          ; preds = %_llgo_4
-  %34 = load %"github.com/goplus/llgo/internal/runtime.String", ptr %1, align 8
-  ret %"github.com/goplus/llgo/internal/runtime.String" %34
+  store %"github.com/goplus/llgo/internal/runtime.String" %0, ptr %2, align 8
+  %3 = alloca %"github.com/goplus/llgo/internal/runtime.String", align 8
+  store %"github.com/goplus/llgo/internal/runtime.String" %1, ptr %3, align 8
+  %4 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %2, i32 0, i32 1
+  %5 = load i64, ptr %4, align 4
+  %6 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %3, i32 0, i32 1
+  %7 = load i64, ptr %6, align 4
+  %8 = add i64 %5, %7
+  %9 = call ptr @"github.com/goplus/llgo/internal/runtime.Alloc"(i64 %8)
+  %10 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %2, i32 0, i32 0
+  %11 = load ptr, ptr %10, align 8
+  %12 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %2, i32 0, i32 1
+  %13 = load i64, ptr %12, align 4
+  %14 = call ptr @memcpy(ptr %9, ptr %11, i64 %13)
+  %15 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %2, i32 0, i32 1
+  %16 = load i64, ptr %15, align 4
+  %17 = getelementptr i8, ptr %9, i64 %16
+  %18 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %3, i32 0, i32 0
+  %19 = load ptr, ptr %18, align 8
+  %20 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %3, i32 0, i32 1
+  %21 = load i64, ptr %20, align 4
+  %22 = call ptr @memcpy(ptr %17, ptr %19, i64 %21)
+  %23 = alloca %"github.com/goplus/llgo/internal/runtime.String", align 8
+  %24 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %23, i32 0, i32 0
+  %25 = getelementptr inbounds %"github.com/goplus/llgo/internal/runtime.String", ptr %23, i32 0, i32 1
+  store ptr %9, ptr %24, align 8
+  store i64 %8, ptr %25, align 4
+  %26 = load %"github.com/goplus/llgo/internal/runtime.String", ptr %23, align 8
+  ret %"github.com/goplus/llgo/internal/runtime.String" %26
 }
 
 define ptr @"github.com/goplus/llgo/internal/runtime.StringData"(%"github.com/goplus/llgo/internal/runtime.String" %0) {
diff --git a/internal/runtime/z_string.go b/internal/runtime/z_string.go
index e362da23..e7726c36 100644
--- a/internal/runtime/z_string.go
+++ b/internal/runtime/z_string.go
@@ -55,20 +55,13 @@ func StringData(s String) unsafe.Pointer {
 	return s.data
 }
 
-// StringCat concatenates strings.
-func StringCat(args ...String) (ret String) {
-	n := 0
-	for _, v := range args {
-		n += v.len
-	}
+// StringCat concatenates two strings.
+func StringCat(a, b String) String {
+	n := a.len + b.len
 	dest := Alloc(uintptr(n))
-	ret.data = dest
-	ret.len = n
-	for _, v := range args {
-		c.Memcpy(dest, v.data, uintptr(v.len))
-		dest = c.Advance(dest, v.len)
-	}
-	return
+	c.Memcpy(dest, a.data, uintptr(a.len))
+	c.Memcpy(c.Advance(dest, a.len), b.data, uintptr(b.len))
+	return String{dest, n}
 }
 
 // -----------------------------------------------------------------------------
diff --git a/ssa/expr.go b/ssa/expr.go
index 0c4c52b1..e7486ba9 100644
--- a/ssa/expr.go
+++ b/ssa/expr.go
@@ -55,8 +55,7 @@ func (v Expr) Do(b Builder) Expr {
 		return vt.t.(delayExprTy)()
 	case vkPhisExpr:
 		e := vt.t.(*phisExprTy)
-		// TODO(xsw): to check CreateAggregateRet is correct or not
-		return Expr{b.impl.CreateAggregateRet(e.phis), e.Type}
+		return b.aggregateValue(e.Type, e.phis...)
 	}
 	return v
 }
@@ -460,6 +459,20 @@ func (b Builder) Store(ptr, val Expr) Builder {
 	return b
 }
 
+// aggregateValue yields the value of the aggregate X with the fields
+func (b Builder) aggregateValue(t Type, flds ...llvm.Value) Expr {
+	if debugInstr {
+		log.Printf("AggregateValue %v, %v\n", t, flds)
+	}
+	impl := b.impl
+	tll := t.ll
+	ptr := llvm.CreateAlloca(impl, tll)
+	for i, fld := range flds {
+		impl.CreateStore(fld, llvm.CreateStructGEP(impl, tll, ptr, i))
+	}
+	return Expr{llvm.CreateLoad(b.impl, tll, ptr), t}
+}
+
 // The FieldAddr instruction yields the address of Field of *struct X.
 //
 // The field is identified by its index within the field list of the