diff --git a/internal/cabi/_testdata/arch/amd64/demo.ll b/internal/cabi/_testdata/arch/amd64/demo.ll
index 0b36b739..c731716b 100644
--- a/internal/cabi/_testdata/arch/amd64/demo.ll
+++ b/internal/cabi/_testdata/arch/amd64/demo.ll
@@ -3,27 +3,928 @@ source_filename = "../../wrap/demo.c"
 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "amd64-unknown-linux-gnu"
 
-%struct.st1 = type { i32, i32 }
+%struct.point = type { i32, i32 }
+%struct.point1 = type { i32, i32, i32 }
+%struct.point64 = type { i64, i64 }
+%struct.struct32 = type { i32 }
+%struct.point2 = type { i8, i32, i32 }
+%struct.point3 = type { i8, i8, i8 }
+%struct.point4 = type { i8, i8, i8, i32 }
+%struct.point5 = type { i8, i8, i8, i8, i8 }
+%struct.point6 = type { i8, i8, i8, i8, i8, i32 }
+%struct.point7 = type { i8, i8, i8, i8, i8, i32, i8 }
+%struct.data1 = type { i8, i64 }
+%struct.data2 = type { i32, i64 }
+%struct.data3 = type { i64, i8 }
+%struct.fdata1 = type { float }
+%struct.ddata1 = type { double }
+%struct.ddata2 = type { double, double }
+%struct.ddata3 = type { double, double, double }
+%struct.fdata2i = type { float, i32 }
+%struct.fdata2 = type { float, float }
+%struct.fdata3 = type { float, float, float }
+%struct.fdata4 = type { float, float, float, float }
+%struct.fdata5 = type { float, float, float, float, float }
+%struct.fdata2id = type { i8, i8, double }
+%struct.fdata7if = type { [7 x i8], float }
+%struct.fdata4if = type { float, i8, float, float }
+%struct.array = type { [8 x i32] }
+
+@.str = private unnamed_addr constant [20 x i8] c"point64: %lld %lld\0A\00", align 1
+@.str.1 = private unnamed_addr constant [14 x i8] c"struct32: %d\0A\00", align 1
+@.str.2 = private unnamed_addr constant [14 x i8] c"point: %d %d\0A\00", align 1
+@.str.3 = private unnamed_addr constant [18 x i8] c"point1: %d %d %d\0A\00", align 1
+@.str.4 = private unnamed_addr constant [18 x i8] c"point2: %d %d %d\0A\00", align 1
+@.str.5 = private unnamed_addr constant [18 x i8] c"point3: %d %d %d\0A\00", align 1
+@.str.6 = private unnamed_addr constant [21 x i8] c"point4: %d %d %d %d\0A\00", align 1
+@.str.7 = private unnamed_addr constant [24 x i8] c"point5: %d %d %d %d %d\0A\00", align 1
+@.str.8 = private unnamed_addr constant [27 x i8] c"point6: %d %d %d %d %d %d\0A\00", align 1
+@.str.9 = private unnamed_addr constant [30 x i8] c"point7: %d %d %d %d %d %d %d\0A\00", align 1
+@.str.10 = private unnamed_addr constant [16 x i8] c"data1: %d %lld\0A\00", align 1
+@.str.11 = private unnamed_addr constant [16 x i8] c"data2: %d %lld\0A\00", align 1
+@.str.12 = private unnamed_addr constant [16 x i8] c"data3: %lld %d\0A\00", align 1
+@.str.13 = private unnamed_addr constant [9 x i8] c"ff1: %f\0A\00", align 1
+@.str.14 = private unnamed_addr constant [9 x i8] c"dd1: %f\0A\00", align 1
+@.str.15 = private unnamed_addr constant [12 x i8] c"dd2: %f %f\0A\00", align 1
+@.str.16 = private unnamed_addr constant [15 x i8] c"dd3: %f %f %f\0A\00", align 1
+@.str.17 = private unnamed_addr constant [13 x i8] c"ff2i: %f %d\0A\00", align 1
+@.str.18 = private unnamed_addr constant [12 x i8] c"ff2: %f %f\0A\00", align 1
+@.str.19 = private unnamed_addr constant [15 x i8] c"ff3: %f %f %f\0A\00", align 1
+@.str.20 = private unnamed_addr constant [18 x i8] c"ff4: %f %f %f %f\0A\00", align 1
+@.str.21 = private unnamed_addr constant [21 x i8] c"ff5: %f %f %f %f %f\0A\00", align 1
+@.str.22 = private unnamed_addr constant [15 x i8] c"ff6: %d %d %f\0A\00", align 1
+@.str.23 = private unnamed_addr constant [17 x i8] c"ff7if: %d %d %f\0A\00", align 1
+@.str.24 = private unnamed_addr constant [20 x i8] c"ff4if: %f %d %f %f\0A\00", align 1
+@.str.25 = private unnamed_addr constant [16 x i8] c"demo: %d %d %d\0A\00", align 1
+@__const.callback.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.26 = private unnamed_addr constant [27 x i8] c"callback1 array: %d %d %d\0A\00", align 1
+@__const.callback1.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback1.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.27 = private unnamed_addr constant [22 x i8] c"callback1 ret: %d,%d\0A\00", align 1
+@.str.28 = private unnamed_addr constant [28 x i8] c"mycallback array: %d %d %d\0A\00", align 1
+@.str.29 = private unnamed_addr constant [22 x i8] c"mycallback pt: %d %d\0A\00", align 1
+@.str.30 = private unnamed_addr constant [26 x i8] c"mycallback pt1: %d %d %d\0A\00", align 1
 
 ; Function Attrs: noinline nounwind optnone uwtable
-define dso_local i64 @fn1(i64 %0) #0 {
-  %2 = alloca %struct.st1, align 4
-  %3 = alloca %struct.st1, align 4
-  %4 = bitcast %struct.st1* %3 to i64*
-  store i64 %0, i64* %4, align 4
-  %5 = bitcast %struct.st1* %2 to i8*
-  %6 = bitcast %struct.st1* %3 to i8*
-  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %5, i8* align 4 %6, i64 8, i1 false)
-  %7 = bitcast %struct.st1* %2 to i64*
-  %8 = load i64, i64* %7, align 4
-  ret i64 %8
+define dso_local i32 @demo32(i32 noundef %0) #0 {
+  %2 = alloca i32, align 4
+  store i32 %0, i32* %2, align 4
+  %3 = load i32, i32* %2, align 4
+  %4 = add nsw i32 %3, 100
+  ret i32 %4
 }
 
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @demo64(i64 noundef %0) #0 {
+  %2 = alloca i64, align 8
+  store i64 %0, i64* %2, align 8
+  %3 = load i64, i64* %2, align 8
+  %4 = add nsw i64 %3, 100
+  ret i64 %4
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, i64 } @pt64(i64 %0, i64 %1) #0 {
+  %3 = alloca %struct.point64, align 8
+  %4 = alloca %struct.point64, align 8
+  %5 = bitcast %struct.point64* %4 to { i64, i64 }*
+  %6 = getelementptr inbounds { i64, i64 }, { i64, i64 }* %5, i32 0, i32 0
+  store i64 %0, i64* %6, align 8
+  %7 = getelementptr inbounds { i64, i64 }, { i64, i64 }* %5, i32 0, i32 1
+  store i64 %1, i64* %7, align 8
+  %8 = getelementptr inbounds %struct.point64, %struct.point64* %4, i32 0, i32 0
+  %9 = load i64, i64* %8, align 8
+  %10 = getelementptr inbounds %struct.point64, %struct.point64* %4, i32 0, i32 1
+  %11 = load i64, i64* %10, align 8
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str, i64 0, i64 0), i64 noundef %9, i64 noundef %11)
+  %13 = bitcast %struct.point64* %3 to i8*
+  %14 = bitcast %struct.point64* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %13, i8* align 8 %14, i64 16, i1 false)
+  %15 = bitcast %struct.point64* %3 to { i64, i64 }*
+  %16 = load { i64, i64 }, { i64, i64 }* %15, align 8
+  ret { i64, i64 } %16
+}
+
+declare i32 @printf(i8* noundef, ...) #1
+
 ; Function Attrs: argmemonly nofree nounwind willreturn
-declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #1
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #2
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i32 @demo32s(i32 %0) #0 {
+  %2 = alloca %struct.struct32, align 4
+  %3 = alloca %struct.struct32, align 4
+  %4 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  store i32 %0, i32* %4, align 4
+  %5 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 4
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.1, i64 0, i64 0), i32 noundef %6)
+  %8 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %9 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %10 = load i32, i32* %9, align 4
+  %11 = add nsw i32 %10, 100
+  store i32 %11, i32* %8, align 4
+  %12 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %13 = load i32, i32* %12, align 4
+  ret i32 %13
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @pt(i64 %0) #0 {
+  %2 = alloca %struct.point, align 4
+  %3 = alloca %struct.point, align 4
+  %4 = bitcast %struct.point* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 4
+  %7 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 1
+  %8 = load i32, i32* %7, align 4
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.2, i64 0, i64 0), i32 noundef %6, i32 noundef %8)
+  %10 = bitcast %struct.point* %2 to i8*
+  %11 = bitcast %struct.point* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %10, i8* align 4 %11, i64 8, i1 false)
+  %12 = bitcast %struct.point* %2 to i64*
+  %13 = load i64, i64* %12, align 4
+  ret i64 %13
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, i32 } @pt1(i64 %0, i32 %1) #0 {
+  %3 = alloca %struct.point1, align 4
+  %4 = alloca %struct.point1, align 4
+  %5 = alloca { i64, i32 }, align 4
+  %6 = alloca { i64, i32 }, align 8
+  %7 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %5, i32 0, i32 0
+  store i64 %0, i64* %7, align 4
+  %8 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %5, i32 0, i32 1
+  store i32 %1, i32* %8, align 4
+  %9 = bitcast %struct.point1* %4 to i8*
+  %10 = bitcast { i64, i32 }* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 4 %10, i64 12, i1 false)
+  %11 = getelementptr inbounds %struct.point1, %struct.point1* %4, i32 0, i32 0
+  %12 = load i32, i32* %11, align 4
+  %13 = getelementptr inbounds %struct.point1, %struct.point1* %4, i32 0, i32 1
+  %14 = load i32, i32* %13, align 4
+  %15 = getelementptr inbounds %struct.point1, %struct.point1* %4, i32 0, i32 2
+  %16 = load i32, i32* %15, align 4
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.3, i64 0, i64 0), i32 noundef %12, i32 noundef %14, i32 noundef %16)
+  %18 = bitcast %struct.point1* %3 to i8*
+  %19 = bitcast %struct.point1* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 12, i1 false)
+  %20 = bitcast { i64, i32 }* %6 to i8*
+  %21 = bitcast %struct.point1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %20, i8* align 4 %21, i64 12, i1 false)
+  %22 = load { i64, i32 }, { i64, i32 }* %6, align 8
+  ret { i64, i32 } %22
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, i32 } @pt2(i64 %0, i32 %1) #0 {
+  %3 = alloca %struct.point2, align 4
+  %4 = alloca %struct.point2, align 4
+  %5 = alloca { i64, i32 }, align 4
+  %6 = alloca { i64, i32 }, align 8
+  %7 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %5, i32 0, i32 0
+  store i64 %0, i64* %7, align 4
+  %8 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %5, i32 0, i32 1
+  store i32 %1, i32* %8, align 4
+  %9 = bitcast %struct.point2* %4 to i8*
+  %10 = bitcast { i64, i32 }* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 4 %10, i64 12, i1 false)
+  %11 = getelementptr inbounds %struct.point2, %struct.point2* %4, i32 0, i32 0
+  %12 = load i8, i8* %11, align 4
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point2, %struct.point2* %4, i32 0, i32 1
+  %15 = load i32, i32* %14, align 4
+  %16 = getelementptr inbounds %struct.point2, %struct.point2* %4, i32 0, i32 2
+  %17 = load i32, i32* %16, align 4
+  %18 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.4, i64 0, i64 0), i32 noundef %13, i32 noundef %15, i32 noundef %17)
+  %19 = bitcast %struct.point2* %3 to i8*
+  %20 = bitcast %struct.point2* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %19, i8* align 4 %20, i64 12, i1 false)
+  %21 = bitcast { i64, i32 }* %6 to i8*
+  %22 = bitcast %struct.point2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %21, i8* align 4 %22, i64 12, i1 false)
+  %23 = load { i64, i32 }, { i64, i32 }* %6, align 8
+  ret { i64, i32 } %23
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i24 @pt3(i24 %0) #0 {
+  %2 = alloca %struct.point3, align 1
+  %3 = alloca %struct.point3, align 1
+  %4 = alloca i24, align 4
+  %5 = alloca i24, align 4
+  store i24 %0, i24* %4, align 4
+  %6 = bitcast %struct.point3* %3 to i8*
+  %7 = bitcast i24* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %6, i8* align 4 %7, i64 3, i1 false)
+  %8 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 1
+  %16 = sext i8 %15 to i32
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.5, i64 0, i64 0), i32 noundef %10, i32 noundef %13, i32 noundef %16)
+  %18 = bitcast %struct.point3* %2 to i8*
+  %19 = bitcast %struct.point3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %18, i8* align 1 %19, i64 3, i1 false)
+  %20 = bitcast i24* %5 to i8*
+  %21 = bitcast %struct.point3* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %20, i8* align 1 %21, i64 3, i1 false)
+  %22 = load i24, i24* %5, align 4
+  ret i24 %22
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @pt4(i64 %0) #0 {
+  %2 = alloca %struct.point4, align 4
+  %3 = alloca %struct.point4, align 4
+  %4 = bitcast %struct.point4* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 3
+  %15 = load i32, i32* %14, align 4
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.6, i64 0, i64 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %15)
+  %17 = bitcast %struct.point4* %2 to i8*
+  %18 = bitcast %struct.point4* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %17, i8* align 4 %18, i64 8, i1 false)
+  %19 = bitcast %struct.point4* %2 to i64*
+  %20 = load i64, i64* %19, align 4
+  ret i64 %20
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i40 @pt5(i40 %0) #0 {
+  %2 = alloca %struct.point5, align 1
+  %3 = alloca %struct.point5, align 1
+  %4 = alloca i40, align 8
+  %5 = alloca i40, align 8
+  store i40 %0, i40* %4, align 8
+  %6 = bitcast %struct.point5* %3 to i8*
+  %7 = bitcast i40* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %6, i8* align 8 %7, i64 5, i1 false)
+  %8 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 1
+  %16 = sext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 3
+  %18 = load i8, i8* %17, align 1
+  %19 = sext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 4
+  %21 = load i8, i8* %20, align 1
+  %22 = sext i8 %21 to i32
+  %23 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([24 x i8], [24 x i8]* @.str.7, i64 0, i64 0), i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %22)
+  %24 = bitcast %struct.point5* %2 to i8*
+  %25 = bitcast %struct.point5* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %24, i8* align 1 %25, i64 5, i1 false)
+  %26 = bitcast i40* %5 to i8*
+  %27 = bitcast %struct.point5* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %26, i8* align 1 %27, i64 5, i1 false)
+  %28 = load i40, i40* %5, align 8
+  ret i40 %28
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, i32 } @pt6(i64 %0, i32 %1) #0 {
+  %3 = alloca %struct.point6, align 4
+  %4 = alloca %struct.point6, align 4
+  %5 = alloca { i64, i32 }, align 4
+  %6 = alloca { i64, i32 }, align 8
+  %7 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %5, i32 0, i32 0
+  store i64 %0, i64* %7, align 4
+  %8 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %5, i32 0, i32 1
+  store i32 %1, i32* %8, align 4
+  %9 = bitcast %struct.point6* %4 to i8*
+  %10 = bitcast { i64, i32 }* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 4 %10, i64 12, i1 false)
+  %11 = getelementptr inbounds %struct.point6, %struct.point6* %4, i32 0, i32 0
+  %12 = load i8, i8* %11, align 4
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point6, %struct.point6* %4, i32 0, i32 1
+  %15 = load i8, i8* %14, align 1
+  %16 = sext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point6, %struct.point6* %4, i32 0, i32 2
+  %18 = load i8, i8* %17, align 2
+  %19 = sext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point6, %struct.point6* %4, i32 0, i32 3
+  %21 = load i8, i8* %20, align 1
+  %22 = sext i8 %21 to i32
+  %23 = getelementptr inbounds %struct.point6, %struct.point6* %4, i32 0, i32 4
+  %24 = load i8, i8* %23, align 4
+  %25 = sext i8 %24 to i32
+  %26 = getelementptr inbounds %struct.point6, %struct.point6* %4, i32 0, i32 5
+  %27 = load i32, i32* %26, align 4
+  %28 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.8, i64 0, i64 0), i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %22, i32 noundef %25, i32 noundef %27)
+  %29 = bitcast %struct.point6* %3 to i8*
+  %30 = bitcast %struct.point6* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %29, i8* align 4 %30, i64 12, i1 false)
+  %31 = bitcast { i64, i32 }* %6 to i8*
+  %32 = bitcast %struct.point6* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %31, i8* align 4 %32, i64 12, i1 false)
+  %33 = load { i64, i32 }, { i64, i32 }* %6, align 8
+  ret { i64, i32 } %33
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, i64 } @pt7(i64 %0, i64 %1) #0 {
+  %3 = alloca %struct.point7, align 4
+  %4 = alloca %struct.point7, align 4
+  %5 = bitcast %struct.point7* %4 to { i64, i64 }*
+  %6 = getelementptr inbounds { i64, i64 }, { i64, i64 }* %5, i32 0, i32 0
+  store i64 %0, i64* %6, align 4
+  %7 = getelementptr inbounds { i64, i64 }, { i64, i64 }* %5, i32 0, i32 1
+  store i64 %1, i64* %7, align 4
+  %8 = getelementptr inbounds %struct.point7, %struct.point7* %4, i32 0, i32 0
+  %9 = load i8, i8* %8, align 4
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point7, %struct.point7* %4, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point7, %struct.point7* %4, i32 0, i32 2
+  %15 = load i8, i8* %14, align 2
+  %16 = sext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point7, %struct.point7* %4, i32 0, i32 3
+  %18 = load i8, i8* %17, align 1
+  %19 = sext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point7, %struct.point7* %4, i32 0, i32 4
+  %21 = load i8, i8* %20, align 4
+  %22 = sext i8 %21 to i32
+  %23 = getelementptr inbounds %struct.point7, %struct.point7* %4, i32 0, i32 5
+  %24 = load i32, i32* %23, align 4
+  %25 = getelementptr inbounds %struct.point7, %struct.point7* %4, i32 0, i32 6
+  %26 = load i8, i8* %25, align 4
+  %27 = sext i8 %26 to i32
+  %28 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([30 x i8], [30 x i8]* @.str.9, i64 0, i64 0), i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %22, i32 noundef %24, i32 noundef %27)
+  %29 = bitcast %struct.point7* %3 to i8*
+  %30 = bitcast %struct.point7* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %29, i8* align 4 %30, i64 16, i1 false)
+  %31 = bitcast %struct.point7* %3 to { i64, i64 }*
+  %32 = load { i64, i64 }, { i64, i64 }* %31, align 4
+  ret { i64, i64 } %32
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i8, i64 } @fn1(i8 %0, i64 %1) #0 {
+  %3 = alloca %struct.data1, align 8
+  %4 = alloca %struct.data1, align 8
+  %5 = bitcast %struct.data1* %4 to { i8, i64 }*
+  %6 = getelementptr inbounds { i8, i64 }, { i8, i64 }* %5, i32 0, i32 0
+  store i8 %0, i8* %6, align 8
+  %7 = getelementptr inbounds { i8, i64 }, { i8, i64 }* %5, i32 0, i32 1
+  store i64 %1, i64* %7, align 8
+  %8 = getelementptr inbounds %struct.data1, %struct.data1* %4, i32 0, i32 0
+  %9 = load i8, i8* %8, align 8
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.data1, %struct.data1* %4, i32 0, i32 1
+  %12 = load i64, i64* %11, align 8
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.10, i64 0, i64 0), i32 noundef %10, i64 noundef %12)
+  %14 = bitcast %struct.data1* %3 to i8*
+  %15 = bitcast %struct.data1* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %14, i8* align 8 %15, i64 16, i1 false)
+  %16 = bitcast %struct.data1* %3 to { i8, i64 }*
+  %17 = load { i8, i64 }, { i8, i64 }* %16, align 8
+  ret { i8, i64 } %17
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i32, i64 } @fn2(i32 %0, i64 %1) #0 {
+  %3 = alloca %struct.data2, align 8
+  %4 = alloca %struct.data2, align 8
+  %5 = bitcast %struct.data2* %4 to { i32, i64 }*
+  %6 = getelementptr inbounds { i32, i64 }, { i32, i64 }* %5, i32 0, i32 0
+  store i32 %0, i32* %6, align 8
+  %7 = getelementptr inbounds { i32, i64 }, { i32, i64 }* %5, i32 0, i32 1
+  store i64 %1, i64* %7, align 8
+  %8 = getelementptr inbounds %struct.data2, %struct.data2* %4, i32 0, i32 0
+  %9 = load i32, i32* %8, align 8
+  %10 = getelementptr inbounds %struct.data2, %struct.data2* %4, i32 0, i32 1
+  %11 = load i64, i64* %10, align 8
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.11, i64 0, i64 0), i32 noundef %9, i64 noundef %11)
+  %13 = bitcast %struct.data2* %3 to i8*
+  %14 = bitcast %struct.data2* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %13, i8* align 8 %14, i64 16, i1 false)
+  %15 = bitcast %struct.data2* %3 to { i32, i64 }*
+  %16 = load { i32, i64 }, { i32, i64 }* %15, align 8
+  ret { i32, i64 } %16
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, i8 } @fn3(i64 %0, i8 %1) #0 {
+  %3 = alloca %struct.data3, align 8
+  %4 = alloca %struct.data3, align 8
+  %5 = bitcast %struct.data3* %4 to { i64, i8 }*
+  %6 = getelementptr inbounds { i64, i8 }, { i64, i8 }* %5, i32 0, i32 0
+  store i64 %0, i64* %6, align 8
+  %7 = getelementptr inbounds { i64, i8 }, { i64, i8 }* %5, i32 0, i32 1
+  store i8 %1, i8* %7, align 8
+  %8 = getelementptr inbounds %struct.data3, %struct.data3* %4, i32 0, i32 0
+  %9 = load i64, i64* %8, align 8
+  %10 = getelementptr inbounds %struct.data3, %struct.data3* %4, i32 0, i32 1
+  %11 = load i8, i8* %10, align 8
+  %12 = sext i8 %11 to i32
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.12, i64 0, i64 0), i64 noundef %9, i32 noundef %12)
+  %14 = bitcast %struct.data3* %3 to i8*
+  %15 = bitcast %struct.data3* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %14, i8* align 8 %15, i64 16, i1 false)
+  %16 = bitcast %struct.data3* %3 to { i64, i8 }*
+  %17 = load { i64, i8 }, { i64, i8 }* %16, align 8
+  ret { i64, i8 } %17
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local float @ff1(float %0) #0 {
+  %2 = alloca %struct.fdata1, align 4
+  %3 = alloca %struct.fdata1, align 4
+  %4 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  store float %0, float* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.13, i64 0, i64 0), double noundef %7)
+  %9 = bitcast %struct.fdata1* %2 to i8*
+  %10 = bitcast %struct.fdata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 4 %10, i64 4, i1 false)
+  %11 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %2, i32 0, i32 0
+  %12 = load float, float* %11, align 4
+  ret float %12
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local double @dd1(double %0) #0 {
+  %2 = alloca %struct.ddata1, align 8
+  %3 = alloca %struct.ddata1, align 8
+  %4 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  store double %0, double* %4, align 8
+  %5 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  %6 = load double, double* %5, align 8
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.14, i64 0, i64 0), double noundef %6)
+  %8 = bitcast %struct.ddata1* %2 to i8*
+  %9 = bitcast %struct.ddata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %8, i8* align 8 %9, i64 8, i1 false)
+  %10 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %2, i32 0, i32 0
+  %11 = load double, double* %10, align 8
+  ret double %11
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { double, double } @dd2(double %0, double %1) #0 {
+  %3 = alloca %struct.ddata2, align 8
+  %4 = alloca %struct.ddata2, align 8
+  %5 = bitcast %struct.ddata2* %4 to { double, double }*
+  %6 = getelementptr inbounds { double, double }, { double, double }* %5, i32 0, i32 0
+  store double %0, double* %6, align 8
+  %7 = getelementptr inbounds { double, double }, { double, double }* %5, i32 0, i32 1
+  store double %1, double* %7, align 8
+  %8 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %4, i32 0, i32 0
+  %9 = load double, double* %8, align 8
+  %10 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %4, i32 0, i32 1
+  %11 = load double, double* %10, align 8
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.15, i64 0, i64 0), double noundef %9, double noundef %11)
+  %13 = bitcast %struct.ddata2* %3 to i8*
+  %14 = bitcast %struct.ddata2* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %13, i8* align 8 %14, i64 16, i1 false)
+  %15 = bitcast %struct.ddata2* %3 to { double, double }*
+  %16 = load { double, double }, { double, double }* %15, align 8
+  ret { double, double } %16
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @dd3(%struct.ddata3* noalias sret(%struct.ddata3) align 8 %0, %struct.ddata3* noundef byval(%struct.ddata3) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 0
+  %4 = load double, double* %3, align 8
+  %5 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 1
+  %6 = load double, double* %5, align 8
+  %7 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 2
+  %8 = load double, double* %7, align 8
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.16, i64 0, i64 0), double noundef %4, double noundef %6, double noundef %8)
+  %10 = bitcast %struct.ddata3* %0 to i8*
+  %11 = bitcast %struct.ddata3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %10, i8* align 8 %11, i64 24, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @ff2i(i64 %0) #0 {
+  %2 = alloca %struct.fdata2i, align 4
+  %3 = alloca %struct.fdata2i, align 4
+  %4 = bitcast %struct.fdata2i* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 1
+  %9 = load i32, i32* %8, align 4
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([13 x i8], [13 x i8]* @.str.17, i64 0, i64 0), double noundef %7, i32 noundef %9)
+  %11 = bitcast %struct.fdata2i* %2 to i8*
+  %12 = bitcast %struct.fdata2i* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %11, i8* align 4 %12, i64 8, i1 false)
+  %13 = bitcast %struct.fdata2i* %2 to i64*
+  %14 = load i64, i64* %13, align 4
+  ret i64 %14
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local <2 x float> @ff2(<2 x float> %0) #3 {
+  %2 = alloca %struct.fdata2, align 4
+  %3 = alloca %struct.fdata2, align 4
+  %4 = bitcast %struct.fdata2* %3 to <2 x float>*
+  store <2 x float> %0, <2 x float>* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.18, i64 0, i64 0), double noundef %7, double noundef %10)
+  %12 = bitcast %struct.fdata2* %2 to i8*
+  %13 = bitcast %struct.fdata2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %12, i8* align 4 %13, i64 8, i1 false)
+  %14 = bitcast %struct.fdata2* %2 to <2 x float>*
+  %15 = load <2 x float>, <2 x float>* %14, align 4
+  ret <2 x float> %15
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { <2 x float>, float } @ff3(<2 x float> %0, float %1) #3 {
+  %3 = alloca %struct.fdata3, align 4
+  %4 = alloca %struct.fdata3, align 4
+  %5 = alloca { <2 x float>, float }, align 4
+  %6 = alloca { <2 x float>, float }, align 8
+  %7 = getelementptr inbounds { <2 x float>, float }, { <2 x float>, float }* %5, i32 0, i32 0
+  store <2 x float> %0, <2 x float>* %7, align 4
+  %8 = getelementptr inbounds { <2 x float>, float }, { <2 x float>, float }* %5, i32 0, i32 1
+  store float %1, float* %8, align 4
+  %9 = bitcast %struct.fdata3* %4 to i8*
+  %10 = bitcast { <2 x float>, float }* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 4 %10, i64 12, i1 false)
+  %11 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %4, i32 0, i32 0
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %4, i32 0, i32 1
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %4, i32 0, i32 2
+  %18 = load float, float* %17, align 4
+  %19 = fpext float %18 to double
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.19, i64 0, i64 0), double noundef %13, double noundef %16, double noundef %19)
+  %21 = bitcast %struct.fdata3* %3 to i8*
+  %22 = bitcast %struct.fdata3* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %21, i8* align 4 %22, i64 12, i1 false)
+  %23 = bitcast { <2 x float>, float }* %6 to i8*
+  %24 = bitcast %struct.fdata3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %23, i8* align 4 %24, i64 12, i1 false)
+  %25 = load { <2 x float>, float }, { <2 x float>, float }* %6, align 8
+  ret { <2 x float>, float } %25
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { <2 x float>, <2 x float> } @ff4(<2 x float> %0, <2 x float> %1) #3 {
+  %3 = alloca %struct.fdata4, align 4
+  %4 = alloca %struct.fdata4, align 4
+  %5 = bitcast %struct.fdata4* %4 to { <2 x float>, <2 x float> }*
+  %6 = getelementptr inbounds { <2 x float>, <2 x float> }, { <2 x float>, <2 x float> }* %5, i32 0, i32 0
+  store <2 x float> %0, <2 x float>* %6, align 4
+  %7 = getelementptr inbounds { <2 x float>, <2 x float> }, { <2 x float>, <2 x float> }* %5, i32 0, i32 1
+  store <2 x float> %1, <2 x float>* %7, align 4
+  %8 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %4, i32 0, i32 0
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %4, i32 0, i32 1
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %4, i32 0, i32 2
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %4, i32 0, i32 3
+  %18 = load float, float* %17, align 4
+  %19 = fpext float %18 to double
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.20, i64 0, i64 0), double noundef %10, double noundef %13, double noundef %16, double noundef %19)
+  %21 = bitcast %struct.fdata4* %3 to i8*
+  %22 = bitcast %struct.fdata4* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %21, i8* align 4 %22, i64 16, i1 false)
+  %23 = bitcast %struct.fdata4* %3 to { <2 x float>, <2 x float> }*
+  %24 = load { <2 x float>, <2 x float> }, { <2 x float>, <2 x float> }* %23, align 4
+  ret { <2 x float>, <2 x float> } %24
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff5(%struct.fdata5* noalias sret(%struct.fdata5) align 4 %0, %struct.fdata5* noundef byval(%struct.fdata5) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 8
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 1
+  %7 = load float, float* %6, align 4
+  %8 = fpext float %7 to double
+  %9 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 2
+  %10 = load float, float* %9, align 8
+  %11 = fpext float %10 to double
+  %12 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 3
+  %13 = load float, float* %12, align 4
+  %14 = fpext float %13 to double
+  %15 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 4
+  %16 = load float, float* %15, align 8
+  %17 = fpext float %16 to double
+  %18 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.21, i64 0, i64 0), double noundef %5, double noundef %8, double noundef %11, double noundef %14, double noundef %17)
+  %19 = bitcast %struct.fdata5* %0 to i8*
+  %20 = bitcast %struct.fdata5* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %19, i8* align 8 %20, i64 20, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, double } @ff2id(i64 %0, double %1) #0 {
+  %3 = alloca %struct.fdata2id, align 8
+  %4 = alloca %struct.fdata2id, align 8
+  %5 = bitcast %struct.fdata2id* %4 to { i64, double }*
+  %6 = getelementptr inbounds { i64, double }, { i64, double }* %5, i32 0, i32 0
+  store i64 %0, i64* %6, align 8
+  %7 = getelementptr inbounds { i64, double }, { i64, double }* %5, i32 0, i32 1
+  store double %1, double* %7, align 8
+  %8 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %4, i32 0, i32 0
+  %9 = load i8, i8* %8, align 8
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %4, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %4, i32 0, i32 2
+  %15 = load double, double* %14, align 8
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.22, i64 0, i64 0), i32 noundef %10, i32 noundef %13, double noundef %15)
+  %17 = bitcast %struct.fdata2id* %3 to i8*
+  %18 = bitcast %struct.fdata2id* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %17, i8* align 8 %18, i64 16, i1 false)
+  %19 = bitcast %struct.fdata2id* %3 to { i64, double }*
+  %20 = load { i64, double }, { i64, double }* %19, align 8
+  ret { i64, double } %20
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, float } @ff7if(i64 %0, float %1) #0 {
+  %3 = alloca %struct.fdata7if, align 4
+  %4 = alloca %struct.fdata7if, align 4
+  %5 = alloca { i64, float }, align 4
+  %6 = alloca { i64, float }, align 8
+  %7 = getelementptr inbounds { i64, float }, { i64, float }* %5, i32 0, i32 0
+  store i64 %0, i64* %7, align 4
+  %8 = getelementptr inbounds { i64, float }, { i64, float }* %5, i32 0, i32 1
+  store float %1, float* %8, align 4
+  %9 = bitcast %struct.fdata7if* %4 to i8*
+  %10 = bitcast { i64, float }* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 4 %10, i64 12, i1 false)
+  %11 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %4, i32 0, i32 0
+  %12 = getelementptr inbounds [7 x i8], [7 x i8]* %11, i64 0, i64 0
+  %13 = load i8, i8* %12, align 4
+  %14 = sext i8 %13 to i32
+  %15 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %4, i32 0, i32 0
+  %16 = getelementptr inbounds [7 x i8], [7 x i8]* %15, i64 0, i64 1
+  %17 = load i8, i8* %16, align 1
+  %18 = sext i8 %17 to i32
+  %19 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %4, i32 0, i32 1
+  %20 = load float, float* %19, align 4
+  %21 = fpext float %20 to double
+  %22 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.23, i64 0, i64 0), i32 noundef %14, i32 noundef %18, double noundef %21)
+  %23 = bitcast %struct.fdata7if* %3 to i8*
+  %24 = bitcast %struct.fdata7if* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %23, i8* align 4 %24, i64 12, i1 false)
+  %25 = bitcast { i64, float }* %6 to i8*
+  %26 = bitcast %struct.fdata7if* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %25, i8* align 4 %26, i64 12, i1 false)
+  %27 = load { i64, float }, { i64, float }* %6, align 8
+  ret { i64, float } %27
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local { i64, <2 x float> } @ff4if(i64 %0, <2 x float> %1) #3 {
+  %3 = alloca %struct.fdata4if, align 4
+  %4 = alloca %struct.fdata4if, align 4
+  %5 = bitcast %struct.fdata4if* %4 to { i64, <2 x float> }*
+  %6 = getelementptr inbounds { i64, <2 x float> }, { i64, <2 x float> }* %5, i32 0, i32 0
+  store i64 %0, i64* %6, align 4
+  %7 = getelementptr inbounds { i64, <2 x float> }, { i64, <2 x float> }* %5, i32 0, i32 1
+  store <2 x float> %1, <2 x float>* %7, align 4
+  %8 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %4, i32 0, i32 0
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %4, i32 0, i32 1
+  %12 = load i8, i8* %11, align 4
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %4, i32 0, i32 2
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %4, i32 0, i32 3
+  %18 = load float, float* %17, align 4
+  %19 = fpext float %18 to double
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str.24, i64 0, i64 0), double noundef %10, i32 noundef %13, double noundef %16, double noundef %19)
+  %21 = bitcast %struct.fdata4if* %3 to i8*
+  %22 = bitcast %struct.fdata4if* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %21, i8* align 4 %22, i64 16, i1 false)
+  %23 = bitcast %struct.fdata4if* %3 to { i64, <2 x float> }*
+  %24 = load { i64, <2 x float> }, { i64, <2 x float> }* %23, align 4
+  ret { i64, <2 x float> } %24
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @demo(%struct.array* noalias sret(%struct.array) align 4 %0, %struct.array* noundef byval(%struct.array) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %4 = getelementptr inbounds [8 x i32], [8 x i32]* %3, i64 0, i64 0
+  %5 = load i32, i32* %4, align 8
+  %6 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %7 = getelementptr inbounds [8 x i32], [8 x i32]* %6, i64 0, i64 1
+  %8 = load i32, i32* %7, align 4
+  %9 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %10 = getelementptr inbounds [8 x i32], [8 x i32]* %9, i64 0, i64 2
+  %11 = load i32, i32* %10, align 8
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.25, i64 0, i64 0), i32 noundef %5, i32 noundef %8, i32 noundef %11)
+  %13 = bitcast %struct.array* %0 to i8*
+  %14 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %13, i8* align 8 %14, i64 32, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @demo2(%struct.array* noalias sret(%struct.array) align 4 %0, i32 noundef %1) #0 {
+  %3 = alloca i32, align 4
+  %4 = alloca i32, align 4
+  store i32 %1, i32* %3, align 4
+  store i32 0, i32* %4, align 4
+  br label %5
+
+5:                                                ; preds = %16, %2
+  %6 = load i32, i32* %4, align 4
+  %7 = icmp slt i32 %6, 8
+  br i1 %7, label %8, label %19
+
+8:                                                ; preds = %5
+  %9 = load i32, i32* %4, align 4
+  %10 = load i32, i32* %3, align 4
+  %11 = add nsw i32 %9, %10
+  %12 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %13 = load i32, i32* %4, align 4
+  %14 = sext i32 %13 to i64
+  %15 = getelementptr inbounds [8 x i32], [8 x i32]* %12, i64 0, i64 %14
+  store i32 %11, i32* %15, align 4
+  br label %16
+
+16:                                               ; preds = %8
+  %17 = load i32, i32* %4, align 4
+  %18 = add nsw i32 %17, 1
+  store i32 %18, i32* %4, align 4
+  br label %5, !llvm.loop !6
+
+19:                                               ; preds = %5
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @callback(void (%struct.array*, %struct.array*, i64, i64, i32)* noundef %0, %struct.array* noundef byval(%struct.array) align 8 %1) #0 {
+  %3 = alloca void (%struct.array*, %struct.array*, i64, i64, i32)*, align 8
+  %4 = alloca %struct.array, align 4
+  %5 = alloca %struct.point, align 4
+  %6 = alloca %struct.point1, align 4
+  %7 = alloca %struct.array, align 8
+  %8 = alloca { i64, i32 }, align 4
+  %9 = alloca %struct.array, align 4
+  store void (%struct.array*, %struct.array*, i64, i64, i32)* %0, void (%struct.array*, %struct.array*, i64, i64, i32)** %3, align 8
+  call void @demo(%struct.array* sret(%struct.array) align 4 %4, %struct.array* noundef byval(%struct.array) align 8 %1)
+  %10 = bitcast %struct.point* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %10, i8* align 4 bitcast (%struct.point* @__const.callback.pt to i8*), i64 8, i1 false)
+  %11 = bitcast %struct.point1* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %11, i8* align 4 bitcast (%struct.point1* @__const.callback.pt1 to i8*), i64 12, i1 false)
+  %12 = load void (%struct.array*, %struct.array*, i64, i64, i32)*, void (%struct.array*, %struct.array*, i64, i64, i32)** %3, align 8
+  %13 = bitcast %struct.point* %5 to i64*
+  %14 = load i64, i64* %13, align 4
+  %15 = bitcast { i64, i32 }* %8 to i8*
+  %16 = bitcast %struct.point1* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %15, i8* align 4 %16, i64 12, i1 false)
+  %17 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %8, i32 0, i32 0
+  %18 = load i64, i64* %17, align 4
+  %19 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %8, i32 0, i32 1
+  %20 = load i32, i32* %19, align 4
+  call void %12(%struct.array* sret(%struct.array) align 4 %7, %struct.array* noundef byval(%struct.array) align 8 %1, i64 %14, i64 %18, i32 %20)
+  call void @demo(%struct.array* sret(%struct.array) align 4 %9, %struct.array* noundef byval(%struct.array) align 8 %7)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @callback1(i64 (%struct.array*, i64, i64, i32)* noundef %0, %struct.array* noundef byval(%struct.array) align 8 %1) #0 {
+  %3 = alloca i64 (%struct.array*, i64, i64, i32)*, align 8
+  %4 = alloca %struct.point, align 4
+  %5 = alloca %struct.point1, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca { i64, i32 }, align 4
+  store i64 (%struct.array*, i64, i64, i32)* %0, i64 (%struct.array*, i64, i64, i32)** %3, align 8
+  %8 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %9 = getelementptr inbounds [8 x i32], [8 x i32]* %8, i64 0, i64 0
+  %10 = load i32, i32* %9, align 8
+  %11 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i64 0, i64 1
+  %13 = load i32, i32* %12, align 4
+  %14 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %15 = getelementptr inbounds [8 x i32], [8 x i32]* %14, i64 0, i64 7
+  %16 = load i32, i32* %15, align 4
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.26, i64 0, i64 0), i32 noundef %10, i32 noundef %13, i32 noundef %16)
+  %18 = bitcast %struct.point* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 bitcast (%struct.point* @__const.callback1.pt to i8*), i64 8, i1 false)
+  %19 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %19, i8* align 4 bitcast (%struct.point1* @__const.callback1.pt1 to i8*), i64 12, i1 false)
+  %20 = load i64 (%struct.array*, i64, i64, i32)*, i64 (%struct.array*, i64, i64, i32)** %3, align 8
+  %21 = bitcast %struct.point* %4 to i64*
+  %22 = load i64, i64* %21, align 4
+  %23 = bitcast { i64, i32 }* %7 to i8*
+  %24 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %23, i8* align 4 %24, i64 12, i1 false)
+  %25 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %7, i32 0, i32 0
+  %26 = load i64, i64* %25, align 4
+  %27 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %7, i32 0, i32 1
+  %28 = load i32, i32* %27, align 4
+  %29 = call i64 %20(%struct.array* noundef byval(%struct.array) align 8 %1, i64 %22, i64 %26, i32 %28)
+  %30 = bitcast %struct.point* %6 to i64*
+  store i64 %29, i64* %30, align 4
+  %31 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %32 = load i32, i32* %31, align 4
+  %33 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %34 = load i32, i32* %33, align 4
+  %35 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.27, i64 0, i64 0), i32 noundef %32, i32 noundef %34)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @mycallback(%struct.array* noundef byval(%struct.array) align 8 %0, i64 %1, i64 %2, i32 %3) #0 {
+  %5 = alloca %struct.point, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca %struct.point1, align 4
+  %8 = alloca { i64, i32 }, align 4
+  %9 = bitcast %struct.point* %6 to i64*
+  store i64 %1, i64* %9, align 4
+  %10 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %8, i32 0, i32 0
+  store i64 %2, i64* %10, align 4
+  %11 = getelementptr inbounds { i64, i32 }, { i64, i32 }* %8, i32 0, i32 1
+  store i32 %3, i32* %11, align 4
+  %12 = bitcast %struct.point1* %7 to i8*
+  %13 = bitcast { i64, i32 }* %8 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %12, i8* align 4 %13, i64 12, i1 false)
+  %14 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %15 = getelementptr inbounds [8 x i32], [8 x i32]* %14, i64 0, i64 0
+  %16 = load i32, i32* %15, align 8
+  %17 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %18 = getelementptr inbounds [8 x i32], [8 x i32]* %17, i64 0, i64 1
+  %19 = load i32, i32* %18, align 4
+  %20 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %21 = getelementptr inbounds [8 x i32], [8 x i32]* %20, i64 0, i64 7
+  %22 = load i32, i32* %21, align 4
+  %23 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([28 x i8], [28 x i8]* @.str.28, i64 0, i64 0), i32 noundef %16, i32 noundef %19, i32 noundef %22)
+  %24 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %25 = load i32, i32* %24, align 4
+  %26 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %27 = load i32, i32* %26, align 4
+  %28 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.29, i64 0, i64 0), i32 noundef %25, i32 noundef %27)
+  %29 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 0
+  %30 = load i32, i32* %29, align 4
+  %31 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 1
+  %32 = load i32, i32* %31, align 4
+  %33 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 2
+  %34 = load i32, i32* %33, align 4
+  %35 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([26 x i8], [26 x i8]* @.str.30, i64 0, i64 0), i32 noundef %30, i32 noundef %32, i32 noundef %34)
+  %36 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  %37 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %38 = load i32, i32* %37, align 4
+  %39 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 0
+  %40 = load i32, i32* %39, align 4
+  %41 = add nsw i32 %38, %40
+  store i32 %41, i32* %36, align 4
+  %42 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  %43 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %44 = load i32, i32* %43, align 4
+  %45 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 1
+  %46 = load i32, i32* %45, align 4
+  %47 = add nsw i32 %44, %46
+  store i32 %47, i32* %42, align 4
+  %48 = bitcast %struct.point* %5 to i64*
+  %49 = load i64, i64* %48, align 4
+  ret i64 %49
+}
 
 attributes #0 = { noinline nounwind optnone uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
-attributes #1 = { argmemonly nofree nounwind willreturn }
+attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
+attributes #2 = { argmemonly nofree nounwind willreturn }
+attributes #3 = { noinline nounwind optnone uwtable "frame-pointer"="all" "min-legal-vector-width"="64" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
 
 !llvm.module.flags = !{!0, !1, !2, !3, !4}
 !llvm.ident = !{!5}
@@ -34,3 +935,5 @@ attributes #1 = { argmemonly nofree nounwind willreturn }
 !3 = !{i32 7, !"uwtable", i32 2}
 !4 = !{i32 7, !"frame-pointer", i32 2}
 !5 = !{!"Apple clang version 14.0.3 (clang-1403.0.22.14.1)"}
+!6 = distinct !{!6, !7}
+!7 = !{!"llvm.loop.mustprogress"}
diff --git a/internal/cabi/_testdata/arch/arm64/demo.ll b/internal/cabi/_testdata/arch/arm64/demo.ll
index 7cfdb107..8220e27b 100644
--- a/internal/cabi/_testdata/arch/arm64/demo.ll
+++ b/internal/cabi/_testdata/arch/arm64/demo.ll
@@ -3,27 +3,888 @@ source_filename = "../../wrap/demo.c"
 target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
 target triple = "aarch64-unknown-linux-gnu"
 
-%struct.st1 = type { i32, i32 }
+%struct.point = type { i32, i32 }
+%struct.point1 = type { i32, i32, i32 }
+%struct.point64 = type { i64, i64 }
+%struct.struct32 = type { i32 }
+%struct.point2 = type { i8, i32, i32 }
+%struct.point3 = type { i8, i8, i8 }
+%struct.point4 = type { i8, i8, i8, i32 }
+%struct.point5 = type { i8, i8, i8, i8, i8 }
+%struct.point6 = type { i8, i8, i8, i8, i8, i32 }
+%struct.point7 = type { i8, i8, i8, i8, i8, i32, i8 }
+%struct.data1 = type { i8, i64 }
+%struct.data2 = type { i32, i64 }
+%struct.data3 = type { i64, i8 }
+%struct.fdata1 = type { float }
+%struct.ddata1 = type { double }
+%struct.ddata2 = type { double, double }
+%struct.ddata3 = type { double, double, double }
+%struct.fdata2i = type { float, i32 }
+%struct.fdata2 = type { float, float }
+%struct.fdata3 = type { float, float, float }
+%struct.fdata4 = type { float, float, float, float }
+%struct.fdata5 = type { float, float, float, float, float }
+%struct.fdata2id = type { i8, i8, double }
+%struct.fdata7if = type { [7 x i8], float }
+%struct.fdata4if = type { float, i8, float, float }
+%struct.array = type { [8 x i32] }
+
+@.str = private unnamed_addr constant [20 x i8] c"point64: %lld %lld\0A\00", align 1
+@.str.1 = private unnamed_addr constant [14 x i8] c"struct32: %d\0A\00", align 1
+@.str.2 = private unnamed_addr constant [14 x i8] c"point: %d %d\0A\00", align 1
+@.str.3 = private unnamed_addr constant [18 x i8] c"point1: %d %d %d\0A\00", align 1
+@.str.4 = private unnamed_addr constant [18 x i8] c"point2: %d %d %d\0A\00", align 1
+@.str.5 = private unnamed_addr constant [18 x i8] c"point3: %d %d %d\0A\00", align 1
+@.str.6 = private unnamed_addr constant [21 x i8] c"point4: %d %d %d %d\0A\00", align 1
+@.str.7 = private unnamed_addr constant [24 x i8] c"point5: %d %d %d %d %d\0A\00", align 1
+@.str.8 = private unnamed_addr constant [27 x i8] c"point6: %d %d %d %d %d %d\0A\00", align 1
+@.str.9 = private unnamed_addr constant [30 x i8] c"point7: %d %d %d %d %d %d %d\0A\00", align 1
+@.str.10 = private unnamed_addr constant [16 x i8] c"data1: %d %lld\0A\00", align 1
+@.str.11 = private unnamed_addr constant [16 x i8] c"data2: %d %lld\0A\00", align 1
+@.str.12 = private unnamed_addr constant [16 x i8] c"data3: %lld %d\0A\00", align 1
+@.str.13 = private unnamed_addr constant [9 x i8] c"ff1: %f\0A\00", align 1
+@.str.14 = private unnamed_addr constant [9 x i8] c"dd1: %f\0A\00", align 1
+@.str.15 = private unnamed_addr constant [12 x i8] c"dd2: %f %f\0A\00", align 1
+@.str.16 = private unnamed_addr constant [15 x i8] c"dd3: %f %f %f\0A\00", align 1
+@.str.17 = private unnamed_addr constant [13 x i8] c"ff2i: %f %d\0A\00", align 1
+@.str.18 = private unnamed_addr constant [12 x i8] c"ff2: %f %f\0A\00", align 1
+@.str.19 = private unnamed_addr constant [15 x i8] c"ff3: %f %f %f\0A\00", align 1
+@.str.20 = private unnamed_addr constant [18 x i8] c"ff4: %f %f %f %f\0A\00", align 1
+@.str.21 = private unnamed_addr constant [21 x i8] c"ff5: %f %f %f %f %f\0A\00", align 1
+@.str.22 = private unnamed_addr constant [15 x i8] c"ff6: %d %d %f\0A\00", align 1
+@.str.23 = private unnamed_addr constant [17 x i8] c"ff7if: %d %d %f\0A\00", align 1
+@.str.24 = private unnamed_addr constant [20 x i8] c"ff4if: %f %d %f %f\0A\00", align 1
+@.str.25 = private unnamed_addr constant [16 x i8] c"demo: %d %d %d\0A\00", align 1
+@__const.callback.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.26 = private unnamed_addr constant [27 x i8] c"callback1 array: %d %d %d\0A\00", align 1
+@__const.callback1.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback1.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.27 = private unnamed_addr constant [22 x i8] c"callback1 ret: %d,%d\0A\00", align 1
+@.str.28 = private unnamed_addr constant [28 x i8] c"mycallback array: %d %d %d\0A\00", align 1
+@.str.29 = private unnamed_addr constant [22 x i8] c"mycallback pt: %d %d\0A\00", align 1
+@.str.30 = private unnamed_addr constant [26 x i8] c"mycallback pt1: %d %d %d\0A\00", align 1
 
 ; Function Attrs: noinline nounwind optnone uwtable
-define dso_local i64 @fn1(i64 %0) #0 {
-  %2 = alloca %struct.st1, align 4
-  %3 = alloca %struct.st1, align 4
-  %4 = bitcast %struct.st1* %3 to i64*
-  store i64 %0, i64* %4, align 4
-  %5 = bitcast %struct.st1* %2 to i8*
-  %6 = bitcast %struct.st1* %3 to i8*
-  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %5, i8* align 4 %6, i64 8, i1 false)
-  %7 = bitcast %struct.st1* %2 to i64*
-  %8 = load i64, i64* %7, align 4
-  ret i64 %8
+define dso_local i32 @demo32(i32 noundef %0) #0 {
+  %2 = alloca i32, align 4
+  store i32 %0, i32* %2, align 4
+  %3 = load i32, i32* %2, align 4
+  %4 = add nsw i32 %3, 100
+  ret i32 %4
 }
 
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @demo64(i64 noundef %0) #0 {
+  %2 = alloca i64, align 8
+  store i64 %0, i64* %2, align 8
+  %3 = load i64, i64* %2, align 8
+  %4 = add nsw i64 %3, 100
+  ret i64 %4
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @pt64([2 x i64] %0) #0 {
+  %2 = alloca %struct.point64, align 8
+  %3 = alloca %struct.point64, align 8
+  %4 = bitcast %struct.point64* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.point64, %struct.point64* %3, i32 0, i32 0
+  %6 = load i64, i64* %5, align 8
+  %7 = getelementptr inbounds %struct.point64, %struct.point64* %3, i32 0, i32 1
+  %8 = load i64, i64* %7, align 8
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str, i64 0, i64 0), i64 noundef %6, i64 noundef %8)
+  %10 = bitcast %struct.point64* %2 to i8*
+  %11 = bitcast %struct.point64* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %10, i8* align 8 %11, i64 16, i1 false)
+  %12 = bitcast %struct.point64* %2 to [2 x i64]*
+  %13 = load [2 x i64], [2 x i64]* %12, align 8
+  ret [2 x i64] %13
+}
+
+declare i32 @printf(i8* noundef, ...) #1
+
 ; Function Attrs: argmemonly nofree nounwind willreturn
-declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #1
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #2
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i32 @demo32s(i64 %0) #0 {
+  %2 = alloca %struct.struct32, align 4
+  %3 = alloca %struct.struct32, align 4
+  %4 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %5 = trunc i64 %0 to i32
+  store i32 %5, i32* %4, align 4
+  %6 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %7 = load i32, i32* %6, align 4
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.1, i64 0, i64 0), i32 noundef %7)
+  %9 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %10 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %11 = load i32, i32* %10, align 4
+  %12 = add nsw i32 %11, 100
+  store i32 %12, i32* %9, align 4
+  %13 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %14 = load i32, i32* %13, align 4
+  ret i32 %14
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @pt(i64 %0) #0 {
+  %2 = alloca %struct.point, align 4
+  %3 = alloca %struct.point, align 4
+  %4 = bitcast %struct.point* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 4
+  %7 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 1
+  %8 = load i32, i32* %7, align 4
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.2, i64 0, i64 0), i32 noundef %6, i32 noundef %8)
+  %10 = bitcast %struct.point* %2 to i8*
+  %11 = bitcast %struct.point* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %10, i8* align 4 %11, i64 8, i1 false)
+  %12 = bitcast %struct.point* %2 to i64*
+  %13 = load i64, i64* %12, align 4
+  ret i64 %13
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @pt1([2 x i64] %0) #0 {
+  %2 = alloca %struct.point1, align 4
+  %3 = alloca %struct.point1, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.point1* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 0
+  %9 = load i32, i32* %8, align 4
+  %10 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 1
+  %11 = load i32, i32* %10, align 4
+  %12 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 2
+  %13 = load i32, i32* %12, align 4
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.3, i64 0, i64 0), i32 noundef %9, i32 noundef %11, i32 noundef %13)
+  %15 = bitcast %struct.point1* %2 to i8*
+  %16 = bitcast %struct.point1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %15, i8* align 4 %16, i64 12, i1 false)
+  %17 = bitcast [2 x i64]* %5 to i8*
+  %18 = bitcast %struct.point1* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %17, i8* align 4 %18, i64 12, i1 false)
+  %19 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %19
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @pt2([2 x i64] %0) #0 {
+  %2 = alloca %struct.point2, align 4
+  %3 = alloca %struct.point2, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.point2* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 4
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 1
+  %12 = load i32, i32* %11, align 4
+  %13 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 2
+  %14 = load i32, i32* %13, align 4
+  %15 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.4, i64 0, i64 0), i32 noundef %10, i32 noundef %12, i32 noundef %14)
+  %16 = bitcast %struct.point2* %2 to i8*
+  %17 = bitcast %struct.point2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %16, i8* align 4 %17, i64 12, i1 false)
+  %18 = bitcast [2 x i64]* %5 to i8*
+  %19 = bitcast %struct.point2* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %18, i8* align 4 %19, i64 12, i1 false)
+  %20 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %20
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i24 @pt3(i64 %0) #0 {
+  %2 = alloca %struct.point3, align 1
+  %3 = alloca %struct.point3, align 1
+  %4 = alloca i64, align 8
+  %5 = alloca i24, align 4
+  store i64 %0, i64* %4, align 8
+  %6 = bitcast %struct.point3* %3 to i8*
+  %7 = bitcast i64* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %6, i8* align 8 %7, i64 3, i1 false)
+  %8 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.5, i64 0, i64 0), i32 noundef %10, i32 noundef %13, i32 noundef %16)
+  %18 = bitcast %struct.point3* %2 to i8*
+  %19 = bitcast %struct.point3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %18, i8* align 1 %19, i64 3, i1 false)
+  %20 = bitcast i24* %5 to i8*
+  %21 = bitcast %struct.point3* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %20, i8* align 1 %21, i64 3, i1 false)
+  %22 = load i24, i24* %5, align 4
+  ret i24 %22
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @pt4(i64 %0) #0 {
+  %2 = alloca %struct.point4, align 4
+  %3 = alloca %struct.point4, align 4
+  %4 = bitcast %struct.point4* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 3
+  %15 = load i32, i32* %14, align 4
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.6, i64 0, i64 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %15)
+  %17 = bitcast %struct.point4* %2 to i8*
+  %18 = bitcast %struct.point4* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %17, i8* align 4 %18, i64 8, i1 false)
+  %19 = bitcast %struct.point4* %2 to i64*
+  %20 = load i64, i64* %19, align 4
+  ret i64 %20
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i40 @pt5(i64 %0) #0 {
+  %2 = alloca %struct.point5, align 1
+  %3 = alloca %struct.point5, align 1
+  %4 = alloca i64, align 8
+  %5 = alloca i40, align 8
+  store i64 %0, i64* %4, align 8
+  %6 = bitcast %struct.point5* %3 to i8*
+  %7 = bitcast i64* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %6, i8* align 8 %7, i64 5, i1 false)
+  %8 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 3
+  %18 = load i8, i8* %17, align 1
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 4
+  %21 = load i8, i8* %20, align 1
+  %22 = zext i8 %21 to i32
+  %23 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([24 x i8], [24 x i8]* @.str.7, i64 0, i64 0), i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %22)
+  %24 = bitcast %struct.point5* %2 to i8*
+  %25 = bitcast %struct.point5* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %24, i8* align 1 %25, i64 5, i1 false)
+  %26 = bitcast i40* %5 to i8*
+  %27 = bitcast %struct.point5* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %26, i8* align 1 %27, i64 5, i1 false)
+  %28 = load i40, i40* %5, align 8
+  ret i40 %28
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @pt6([2 x i64] %0) #0 {
+  %2 = alloca %struct.point6, align 4
+  %3 = alloca %struct.point6, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.point6* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 4
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 2
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 3
+  %18 = load i8, i8* %17, align 1
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 4
+  %21 = load i8, i8* %20, align 4
+  %22 = zext i8 %21 to i32
+  %23 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 5
+  %24 = load i32, i32* %23, align 4
+  %25 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.8, i64 0, i64 0), i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %22, i32 noundef %24)
+  %26 = bitcast %struct.point6* %2 to i8*
+  %27 = bitcast %struct.point6* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %26, i8* align 4 %27, i64 12, i1 false)
+  %28 = bitcast [2 x i64]* %5 to i8*
+  %29 = bitcast %struct.point6* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %28, i8* align 4 %29, i64 12, i1 false)
+  %30 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %30
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @pt7([2 x i64] %0) #0 {
+  %2 = alloca %struct.point7, align 4
+  %3 = alloca %struct.point7, align 4
+  %4 = bitcast %struct.point7* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 4
+  %5 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 3
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 4
+  %18 = load i8, i8* %17, align 4
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 5
+  %21 = load i32, i32* %20, align 4
+  %22 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 6
+  %23 = load i8, i8* %22, align 4
+  %24 = zext i8 %23 to i32
+  %25 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([30 x i8], [30 x i8]* @.str.9, i64 0, i64 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %21, i32 noundef %24)
+  %26 = bitcast %struct.point7* %2 to i8*
+  %27 = bitcast %struct.point7* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %26, i8* align 4 %27, i64 16, i1 false)
+  %28 = bitcast %struct.point7* %2 to [2 x i64]*
+  %29 = load [2 x i64], [2 x i64]* %28, align 4
+  ret [2 x i64] %29
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @fn1([2 x i64] %0) #0 {
+  %2 = alloca %struct.data1, align 8
+  %3 = alloca %struct.data1, align 8
+  %4 = bitcast %struct.data1* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data1, %struct.data1* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 8
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.data1, %struct.data1* %3, i32 0, i32 1
+  %9 = load i64, i64* %8, align 8
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.10, i64 0, i64 0), i32 noundef %7, i64 noundef %9)
+  %11 = bitcast %struct.data1* %2 to i8*
+  %12 = bitcast %struct.data1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %11, i8* align 8 %12, i64 16, i1 false)
+  %13 = bitcast %struct.data1* %2 to [2 x i64]*
+  %14 = load [2 x i64], [2 x i64]* %13, align 8
+  ret [2 x i64] %14
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @fn2([2 x i64] %0) #0 {
+  %2 = alloca %struct.data2, align 8
+  %3 = alloca %struct.data2, align 8
+  %4 = bitcast %struct.data2* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data2, %struct.data2* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 8
+  %7 = getelementptr inbounds %struct.data2, %struct.data2* %3, i32 0, i32 1
+  %8 = load i64, i64* %7, align 8
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.11, i64 0, i64 0), i32 noundef %6, i64 noundef %8)
+  %10 = bitcast %struct.data2* %2 to i8*
+  %11 = bitcast %struct.data2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %10, i8* align 8 %11, i64 16, i1 false)
+  %12 = bitcast %struct.data2* %2 to [2 x i64]*
+  %13 = load [2 x i64], [2 x i64]* %12, align 8
+  ret [2 x i64] %13
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @fn3([2 x i64] %0) #0 {
+  %2 = alloca %struct.data3, align 8
+  %3 = alloca %struct.data3, align 8
+  %4 = bitcast %struct.data3* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data3, %struct.data3* %3, i32 0, i32 0
+  %6 = load i64, i64* %5, align 8
+  %7 = getelementptr inbounds %struct.data3, %struct.data3* %3, i32 0, i32 1
+  %8 = load i8, i8* %7, align 8
+  %9 = zext i8 %8 to i32
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.12, i64 0, i64 0), i64 noundef %6, i32 noundef %9)
+  %11 = bitcast %struct.data3* %2 to i8*
+  %12 = bitcast %struct.data3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %11, i8* align 8 %12, i64 16, i1 false)
+  %13 = bitcast %struct.data3* %2 to [2 x i64]*
+  %14 = load [2 x i64], [2 x i64]* %13, align 8
+  ret [2 x i64] %14
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local %struct.fdata1 @ff1([1 x float] %0) #0 {
+  %2 = alloca %struct.fdata1, align 4
+  %3 = alloca %struct.fdata1, align 4
+  %4 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  %5 = bitcast float* %4 to [1 x float]*
+  store [1 x float] %0, [1 x float]* %5, align 4
+  %6 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  %7 = load float, float* %6, align 4
+  %8 = fpext float %7 to double
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.13, i64 0, i64 0), double noundef %8)
+  %10 = bitcast %struct.fdata1* %2 to i8*
+  %11 = bitcast %struct.fdata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %10, i8* align 4 %11, i64 4, i1 false)
+  %12 = load %struct.fdata1, %struct.fdata1* %2, align 4
+  ret %struct.fdata1 %12
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local %struct.ddata1 @dd1([1 x double] %0) #0 {
+  %2 = alloca %struct.ddata1, align 8
+  %3 = alloca %struct.ddata1, align 8
+  %4 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  %5 = bitcast double* %4 to [1 x double]*
+  store [1 x double] %0, [1 x double]* %5, align 8
+  %6 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  %7 = load double, double* %6, align 8
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.14, i64 0, i64 0), double noundef %7)
+  %9 = bitcast %struct.ddata1* %2 to i8*
+  %10 = bitcast %struct.ddata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %9, i8* align 8 %10, i64 8, i1 false)
+  %11 = load %struct.ddata1, %struct.ddata1* %2, align 8
+  ret %struct.ddata1 %11
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local %struct.ddata2 @dd2([2 x double] %0) #0 {
+  %2 = alloca %struct.ddata2, align 8
+  %3 = alloca %struct.ddata2, align 8
+  %4 = bitcast %struct.ddata2* %3 to [2 x double]*
+  store [2 x double] %0, [2 x double]* %4, align 8
+  %5 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %3, i32 0, i32 0
+  %6 = load double, double* %5, align 8
+  %7 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %3, i32 0, i32 1
+  %8 = load double, double* %7, align 8
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.15, i64 0, i64 0), double noundef %6, double noundef %8)
+  %10 = bitcast %struct.ddata2* %2 to i8*
+  %11 = bitcast %struct.ddata2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %10, i8* align 8 %11, i64 16, i1 false)
+  %12 = load %struct.ddata2, %struct.ddata2* %2, align 8
+  ret %struct.ddata2 %12
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local %struct.ddata3 @dd3([3 x double] %0) #0 {
+  %2 = alloca %struct.ddata3, align 8
+  %3 = alloca %struct.ddata3, align 8
+  %4 = bitcast %struct.ddata3* %3 to [3 x double]*
+  store [3 x double] %0, [3 x double]* %4, align 8
+  %5 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %3, i32 0, i32 0
+  %6 = load double, double* %5, align 8
+  %7 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %3, i32 0, i32 1
+  %8 = load double, double* %7, align 8
+  %9 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %3, i32 0, i32 2
+  %10 = load double, double* %9, align 8
+  %11 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.16, i64 0, i64 0), double noundef %6, double noundef %8, double noundef %10)
+  %12 = bitcast %struct.ddata3* %2 to i8*
+  %13 = bitcast %struct.ddata3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %12, i8* align 8 %13, i64 24, i1 false)
+  %14 = load %struct.ddata3, %struct.ddata3* %2, align 8
+  ret %struct.ddata3 %14
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @ff2i(i64 %0) #0 {
+  %2 = alloca %struct.fdata2i, align 4
+  %3 = alloca %struct.fdata2i, align 4
+  %4 = bitcast %struct.fdata2i* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 1
+  %9 = load i32, i32* %8, align 4
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([13 x i8], [13 x i8]* @.str.17, i64 0, i64 0), double noundef %7, i32 noundef %9)
+  %11 = bitcast %struct.fdata2i* %2 to i8*
+  %12 = bitcast %struct.fdata2i* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %11, i8* align 4 %12, i64 8, i1 false)
+  %13 = bitcast %struct.fdata2i* %2 to i64*
+  %14 = load i64, i64* %13, align 4
+  ret i64 %14
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local %struct.fdata2 @ff2([2 x float] %0) #0 {
+  %2 = alloca %struct.fdata2, align 4
+  %3 = alloca %struct.fdata2, align 4
+  %4 = bitcast %struct.fdata2* %3 to [2 x float]*
+  store [2 x float] %0, [2 x float]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.18, i64 0, i64 0), double noundef %7, double noundef %10)
+  %12 = bitcast %struct.fdata2* %2 to i8*
+  %13 = bitcast %struct.fdata2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %12, i8* align 4 %13, i64 8, i1 false)
+  %14 = load %struct.fdata2, %struct.fdata2* %2, align 4
+  ret %struct.fdata2 %14
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local %struct.fdata3 @ff3([3 x float] %0) #0 {
+  %2 = alloca %struct.fdata3, align 4
+  %3 = alloca %struct.fdata3, align 4
+  %4 = bitcast %struct.fdata3* %3 to [3 x float]*
+  store [3 x float] %0, [3 x float]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.19, i64 0, i64 0), double noundef %7, double noundef %10, double noundef %13)
+  %15 = bitcast %struct.fdata3* %2 to i8*
+  %16 = bitcast %struct.fdata3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %15, i8* align 4 %16, i64 12, i1 false)
+  %17 = load %struct.fdata3, %struct.fdata3* %2, align 4
+  ret %struct.fdata3 %17
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local %struct.fdata4 @ff4([4 x float] %0) #0 {
+  %2 = alloca %struct.fdata4, align 4
+  %3 = alloca %struct.fdata4, align 4
+  %4 = bitcast %struct.fdata4* %3 to [4 x float]*
+  store [4 x float] %0, [4 x float]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.20, i64 0, i64 0), double noundef %7, double noundef %10, double noundef %13, double noundef %16)
+  %18 = bitcast %struct.fdata4* %2 to i8*
+  %19 = bitcast %struct.fdata4* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 16, i1 false)
+  %20 = load %struct.fdata4, %struct.fdata4* %2, align 4
+  ret %struct.fdata4 %20
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff5(%struct.fdata5* noalias sret(%struct.fdata5) align 4 %0, %struct.fdata5* noundef %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 4
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 1
+  %7 = load float, float* %6, align 4
+  %8 = fpext float %7 to double
+  %9 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 2
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 3
+  %13 = load float, float* %12, align 4
+  %14 = fpext float %13 to double
+  %15 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 4
+  %16 = load float, float* %15, align 4
+  %17 = fpext float %16 to double
+  %18 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.21, i64 0, i64 0), double noundef %5, double noundef %8, double noundef %11, double noundef %14, double noundef %17)
+  %19 = bitcast %struct.fdata5* %0 to i8*
+  %20 = bitcast %struct.fdata5* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %19, i8* align 4 %20, i64 20, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @ff2id([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata2id, align 8
+  %3 = alloca %struct.fdata2id, align 8
+  %4 = bitcast %struct.fdata2id* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 8
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 2
+  %12 = load double, double* %11, align 8
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.22, i64 0, i64 0), i32 noundef %7, i32 noundef %10, double noundef %12)
+  %14 = bitcast %struct.fdata2id* %2 to i8*
+  %15 = bitcast %struct.fdata2id* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %14, i8* align 8 %15, i64 16, i1 false)
+  %16 = bitcast %struct.fdata2id* %2 to [2 x i64]*
+  %17 = load [2 x i64], [2 x i64]* %16, align 8
+  ret [2 x i64] %17
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @ff7if([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata7if, align 4
+  %3 = alloca %struct.fdata7if, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.fdata7if* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 0
+  %9 = getelementptr inbounds [7 x i8], [7 x i8]* %8, i64 0, i64 0
+  %10 = load i8, i8* %9, align 4
+  %11 = zext i8 %10 to i32
+  %12 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 0
+  %13 = getelementptr inbounds [7 x i8], [7 x i8]* %12, i64 0, i64 1
+  %14 = load i8, i8* %13, align 1
+  %15 = zext i8 %14 to i32
+  %16 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 1
+  %17 = load float, float* %16, align 4
+  %18 = fpext float %17 to double
+  %19 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.23, i64 0, i64 0), i32 noundef %11, i32 noundef %15, double noundef %18)
+  %20 = bitcast %struct.fdata7if* %2 to i8*
+  %21 = bitcast %struct.fdata7if* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %20, i8* align 4 %21, i64 12, i1 false)
+  %22 = bitcast [2 x i64]* %5 to i8*
+  %23 = bitcast %struct.fdata7if* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %22, i8* align 4 %23, i64 12, i1 false)
+  %24 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %24
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local [2 x i64] @ff4if([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata4if, align 4
+  %3 = alloca %struct.fdata4if, align 4
+  %4 = bitcast %struct.fdata4if* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 4
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str.24, i64 0, i64 0), double noundef %7, i32 noundef %10, double noundef %13, double noundef %16)
+  %18 = bitcast %struct.fdata4if* %2 to i8*
+  %19 = bitcast %struct.fdata4if* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 16, i1 false)
+  %20 = bitcast %struct.fdata4if* %2 to [2 x i64]*
+  %21 = load [2 x i64], [2 x i64]* %20, align 4
+  ret [2 x i64] %21
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @demo(%struct.array* noalias sret(%struct.array) align 4 %0, %struct.array* noundef %1) #0 {
+  %3 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %4 = getelementptr inbounds [8 x i32], [8 x i32]* %3, i64 0, i64 0
+  %5 = load i32, i32* %4, align 4
+  %6 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %7 = getelementptr inbounds [8 x i32], [8 x i32]* %6, i64 0, i64 1
+  %8 = load i32, i32* %7, align 4
+  %9 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %10 = getelementptr inbounds [8 x i32], [8 x i32]* %9, i64 0, i64 2
+  %11 = load i32, i32* %10, align 4
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.25, i64 0, i64 0), i32 noundef %5, i32 noundef %8, i32 noundef %11)
+  %13 = bitcast %struct.array* %0 to i8*
+  %14 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %13, i8* align 4 %14, i64 32, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @demo2(%struct.array* noalias sret(%struct.array) align 4 %0, i32 noundef %1) #0 {
+  %3 = alloca i32, align 4
+  %4 = alloca i32, align 4
+  store i32 %1, i32* %3, align 4
+  store i32 0, i32* %4, align 4
+  br label %5
+
+5:                                                ; preds = %16, %2
+  %6 = load i32, i32* %4, align 4
+  %7 = icmp slt i32 %6, 8
+  br i1 %7, label %8, label %19
+
+8:                                                ; preds = %5
+  %9 = load i32, i32* %4, align 4
+  %10 = load i32, i32* %3, align 4
+  %11 = add nsw i32 %9, %10
+  %12 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %13 = load i32, i32* %4, align 4
+  %14 = sext i32 %13 to i64
+  %15 = getelementptr inbounds [8 x i32], [8 x i32]* %12, i64 0, i64 %14
+  store i32 %11, i32* %15, align 4
+  br label %16
+
+16:                                               ; preds = %8
+  %17 = load i32, i32* %4, align 4
+  %18 = add nsw i32 %17, 1
+  store i32 %18, i32* %4, align 4
+  br label %5, !llvm.loop !10
+
+19:                                               ; preds = %5
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @callback(void (%struct.array*, %struct.array*, i64, [2 x i64])* noundef %0, %struct.array* noundef %1) #0 {
+  %3 = alloca void (%struct.array*, %struct.array*, i64, [2 x i64])*, align 8
+  %4 = alloca %struct.array, align 4
+  %5 = alloca %struct.array, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca %struct.point1, align 4
+  %8 = alloca %struct.array, align 4
+  %9 = alloca %struct.array, align 4
+  %10 = alloca [2 x i64], align 8
+  %11 = alloca %struct.array, align 4
+  %12 = alloca %struct.array, align 4
+  store void (%struct.array*, %struct.array*, i64, [2 x i64])* %0, void (%struct.array*, %struct.array*, i64, [2 x i64])** %3, align 8
+  %13 = bitcast %struct.array* %5 to i8*
+  %14 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %13, i8* align 4 %14, i64 32, i1 false)
+  call void @demo(%struct.array* sret(%struct.array) align 4 %4, %struct.array* noundef %5)
+  %15 = bitcast %struct.point* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %15, i8* align 4 bitcast (%struct.point* @__const.callback.pt to i8*), i64 8, i1 false)
+  %16 = bitcast %struct.point1* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %16, i8* align 4 bitcast (%struct.point1* @__const.callback.pt1 to i8*), i64 12, i1 false)
+  %17 = load void (%struct.array*, %struct.array*, i64, [2 x i64])*, void (%struct.array*, %struct.array*, i64, [2 x i64])** %3, align 8
+  %18 = bitcast %struct.array* %9 to i8*
+  %19 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 32, i1 false)
+  %20 = bitcast %struct.point* %6 to i64*
+  %21 = load i64, i64* %20, align 4
+  %22 = bitcast [2 x i64]* %10 to i8*
+  %23 = bitcast %struct.point1* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %22, i8* align 4 %23, i64 12, i1 false)
+  %24 = load [2 x i64], [2 x i64]* %10, align 8
+  call void %17(%struct.array* sret(%struct.array) align 4 %8, %struct.array* noundef %9, i64 %21, [2 x i64] %24)
+  %25 = bitcast %struct.array* %12 to i8*
+  %26 = bitcast %struct.array* %8 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %25, i8* align 4 %26, i64 32, i1 false)
+  call void @demo(%struct.array* sret(%struct.array) align 4 %11, %struct.array* noundef %12)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @callback1(i64 (%struct.array*, i64, [2 x i64])* noundef %0, %struct.array* noundef %1) #0 {
+  %3 = alloca i64 (%struct.array*, i64, [2 x i64])*, align 8
+  %4 = alloca %struct.point, align 4
+  %5 = alloca %struct.point1, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca %struct.array, align 4
+  %8 = alloca [2 x i64], align 8
+  store i64 (%struct.array*, i64, [2 x i64])* %0, i64 (%struct.array*, i64, [2 x i64])** %3, align 8
+  %9 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %10 = getelementptr inbounds [8 x i32], [8 x i32]* %9, i64 0, i64 0
+  %11 = load i32, i32* %10, align 4
+  %12 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %13 = getelementptr inbounds [8 x i32], [8 x i32]* %12, i64 0, i64 1
+  %14 = load i32, i32* %13, align 4
+  %15 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %16 = getelementptr inbounds [8 x i32], [8 x i32]* %15, i64 0, i64 7
+  %17 = load i32, i32* %16, align 4
+  %18 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.26, i64 0, i64 0), i32 noundef %11, i32 noundef %14, i32 noundef %17)
+  %19 = bitcast %struct.point* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %19, i8* align 4 bitcast (%struct.point* @__const.callback1.pt to i8*), i64 8, i1 false)
+  %20 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %20, i8* align 4 bitcast (%struct.point1* @__const.callback1.pt1 to i8*), i64 12, i1 false)
+  %21 = load i64 (%struct.array*, i64, [2 x i64])*, i64 (%struct.array*, i64, [2 x i64])** %3, align 8
+  %22 = bitcast %struct.array* %7 to i8*
+  %23 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %22, i8* align 4 %23, i64 32, i1 false)
+  %24 = bitcast %struct.point* %4 to i64*
+  %25 = load i64, i64* %24, align 4
+  %26 = bitcast [2 x i64]* %8 to i8*
+  %27 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %26, i8* align 4 %27, i64 12, i1 false)
+  %28 = load [2 x i64], [2 x i64]* %8, align 8
+  %29 = call i64 %21(%struct.array* noundef %7, i64 %25, [2 x i64] %28)
+  %30 = bitcast %struct.point* %6 to i64*
+  store i64 %29, i64* %30, align 4
+  %31 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %32 = load i32, i32* %31, align 4
+  %33 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %34 = load i32, i32* %33, align 4
+  %35 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.27, i64 0, i64 0), i32 noundef %32, i32 noundef %34)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @mycallback(%struct.array* noundef %0, i64 %1, [2 x i64] %2) #0 {
+  %4 = alloca %struct.point, align 4
+  %5 = alloca %struct.point, align 4
+  %6 = alloca %struct.point1, align 4
+  %7 = alloca [2 x i64], align 8
+  %8 = bitcast %struct.point* %5 to i64*
+  store i64 %1, i64* %8, align 4
+  store [2 x i64] %2, [2 x i64]* %7, align 8
+  %9 = bitcast %struct.point1* %6 to i8*
+  %10 = bitcast [2 x i64]* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 8 %10, i64 12, i1 false)
+  %11 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i64 0, i64 0
+  %13 = load i32, i32* %12, align 4
+  %14 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %15 = getelementptr inbounds [8 x i32], [8 x i32]* %14, i64 0, i64 1
+  %16 = load i32, i32* %15, align 4
+  %17 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %18 = getelementptr inbounds [8 x i32], [8 x i32]* %17, i64 0, i64 7
+  %19 = load i32, i32* %18, align 4
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([28 x i8], [28 x i8]* @.str.28, i64 0, i64 0), i32 noundef %13, i32 noundef %16, i32 noundef %19)
+  %21 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  %22 = load i32, i32* %21, align 4
+  %23 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  %24 = load i32, i32* %23, align 4
+  %25 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.29, i64 0, i64 0), i32 noundef %22, i32 noundef %24)
+  %26 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 0
+  %27 = load i32, i32* %26, align 4
+  %28 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 1
+  %29 = load i32, i32* %28, align 4
+  %30 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 2
+  %31 = load i32, i32* %30, align 4
+  %32 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([26 x i8], [26 x i8]* @.str.30, i64 0, i64 0), i32 noundef %27, i32 noundef %29, i32 noundef %31)
+  %33 = getelementptr inbounds %struct.point, %struct.point* %4, i32 0, i32 0
+  %34 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  %35 = load i32, i32* %34, align 4
+  %36 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 0
+  %37 = load i32, i32* %36, align 4
+  %38 = add nsw i32 %35, %37
+  store i32 %38, i32* %33, align 4
+  %39 = getelementptr inbounds %struct.point, %struct.point* %4, i32 0, i32 1
+  %40 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  %41 = load i32, i32* %40, align 4
+  %42 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 1
+  %43 = load i32, i32* %42, align 4
+  %44 = add nsw i32 %41, %43
+  store i32 %44, i32* %39, align 4
+  %45 = bitcast %struct.point* %4 to i64*
+  %46 = load i64, i64* %45, align 4
+  ret i64 %46
+}
 
 attributes #0 = { noinline nounwind optnone uwtable "frame-pointer"="non-leaf" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="generic" "target-features"="+neon,+v8a" }
-attributes #1 = { argmemonly nofree nounwind willreturn }
+attributes #1 = { "frame-pointer"="non-leaf" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="generic" "target-features"="+neon,+v8a" }
+attributes #2 = { argmemonly nofree nounwind willreturn }
 
 !llvm.module.flags = !{!0, !1, !2, !3, !4, !5, !6, !7, !8}
 !llvm.ident = !{!9}
@@ -38,3 +899,5 @@ attributes #1 = { argmemonly nofree nounwind willreturn }
 !7 = !{i32 7, !"uwtable", i32 2}
 !8 = !{i32 7, !"frame-pointer", i32 1}
 !9 = !{!"Apple clang version 14.0.3 (clang-1403.0.22.14.1)"}
+!10 = distinct !{!10, !11}
+!11 = !{!"llvm.loop.mustprogress"}
diff --git a/internal/cabi/_testdata/arch/armv6/demo.ll b/internal/cabi/_testdata/arch/armv6/demo.ll
index cef73bbc..435afa31 100644
--- a/internal/cabi/_testdata/arch/armv6/demo.ll
+++ b/internal/cabi/_testdata/arch/armv6/demo.ll
@@ -3,24 +3,799 @@ source_filename = "../../wrap/demo.c"
 target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
 target triple = "armv6kz-unknown-linux-gnueabihf"
 
-%struct.st1 = type { i32, i32 }
+%struct.point = type { i32, i32 }
+%struct.point1 = type { i32, i32, i32 }
+%struct.point64 = type { i64, i64 }
+%struct.struct32 = type { i32 }
+%struct.point2 = type { i8, i32, i32 }
+%struct.point3 = type { i8, i8, i8 }
+%struct.point4 = type { i8, i8, i8, i32 }
+%struct.point5 = type { i8, i8, i8, i8, i8 }
+%struct.point6 = type { i8, i8, i8, i8, i8, i32 }
+%struct.point7 = type { i8, i8, i8, i8, i8, i32, i8 }
+%struct.data1 = type { i8, i64 }
+%struct.data2 = type { i32, i64 }
+%struct.data3 = type { i64, i8 }
+%struct.fdata1 = type { float }
+%struct.ddata1 = type { double }
+%struct.ddata2 = type { double, double }
+%struct.ddata3 = type { double, double, double }
+%struct.fdata2i = type { float, i32 }
+%struct.fdata2 = type { float, float }
+%struct.fdata3 = type { float, float, float }
+%struct.fdata4 = type { float, float, float, float }
+%struct.fdata5 = type { float, float, float, float, float }
+%struct.fdata2id = type { i8, i8, double }
+%struct.fdata7if = type { [7 x i8], float }
+%struct.fdata4if = type { float, i8, float, float }
+%struct.array = type { [8 x i32] }
+
+@.str = private unnamed_addr constant [20 x i8] c"point64: %lld %lld\0A\00", align 1
+@.str.1 = private unnamed_addr constant [14 x i8] c"struct32: %d\0A\00", align 1
+@.str.2 = private unnamed_addr constant [14 x i8] c"point: %d %d\0A\00", align 1
+@.str.3 = private unnamed_addr constant [18 x i8] c"point1: %d %d %d\0A\00", align 1
+@.str.4 = private unnamed_addr constant [18 x i8] c"point2: %d %d %d\0A\00", align 1
+@.str.5 = private unnamed_addr constant [18 x i8] c"point3: %d %d %d\0A\00", align 1
+@.str.6 = private unnamed_addr constant [21 x i8] c"point4: %d %d %d %d\0A\00", align 1
+@.str.7 = private unnamed_addr constant [24 x i8] c"point5: %d %d %d %d %d\0A\00", align 1
+@.str.8 = private unnamed_addr constant [27 x i8] c"point6: %d %d %d %d %d %d\0A\00", align 1
+@.str.9 = private unnamed_addr constant [30 x i8] c"point7: %d %d %d %d %d %d %d\0A\00", align 1
+@.str.10 = private unnamed_addr constant [16 x i8] c"data1: %d %lld\0A\00", align 1
+@.str.11 = private unnamed_addr constant [16 x i8] c"data2: %d %lld\0A\00", align 1
+@.str.12 = private unnamed_addr constant [16 x i8] c"data3: %lld %d\0A\00", align 1
+@.str.13 = private unnamed_addr constant [9 x i8] c"ff1: %f\0A\00", align 1
+@.str.14 = private unnamed_addr constant [9 x i8] c"dd1: %f\0A\00", align 1
+@.str.15 = private unnamed_addr constant [12 x i8] c"dd2: %f %f\0A\00", align 1
+@.str.16 = private unnamed_addr constant [15 x i8] c"dd3: %f %f %f\0A\00", align 1
+@.str.17 = private unnamed_addr constant [13 x i8] c"ff2i: %f %d\0A\00", align 1
+@.str.18 = private unnamed_addr constant [12 x i8] c"ff2: %f %f\0A\00", align 1
+@.str.19 = private unnamed_addr constant [15 x i8] c"ff3: %f %f %f\0A\00", align 1
+@.str.20 = private unnamed_addr constant [18 x i8] c"ff4: %f %f %f %f\0A\00", align 1
+@.str.21 = private unnamed_addr constant [21 x i8] c"ff5: %f %f %f %f %f\0A\00", align 1
+@.str.22 = private unnamed_addr constant [15 x i8] c"ff6: %d %d %f\0A\00", align 1
+@.str.23 = private unnamed_addr constant [17 x i8] c"ff7if: %d %d %f\0A\00", align 1
+@.str.24 = private unnamed_addr constant [20 x i8] c"ff4if: %f %d %f %f\0A\00", align 1
+@.str.25 = private unnamed_addr constant [16 x i8] c"demo: %d %d %d\0A\00", align 1
+@__const.callback.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.26 = private unnamed_addr constant [27 x i8] c"callback1 array: %d %d %d\0A\00", align 1
+@__const.callback1.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback1.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.27 = private unnamed_addr constant [22 x i8] c"callback1 ret: %d,%d\0A\00", align 1
+@.str.28 = private unnamed_addr constant [28 x i8] c"mycallback array: %d %d %d\0A\00", align 1
+@.str.29 = private unnamed_addr constant [22 x i8] c"mycallback pt: %d %d\0A\00", align 1
+@.str.30 = private unnamed_addr constant [26 x i8] c"mycallback pt1: %d %d %d\0A\00", align 1
 
 ; Function Attrs: noinline nounwind optnone
-define dso_local void @fn1(%struct.st1* noalias sret(%struct.st1) align 4 %0, [2 x i32] %1) #0 {
-  %3 = alloca %struct.st1, align 4
-  %4 = bitcast %struct.st1* %3 to [2 x i32]*
-  store [2 x i32] %1, [2 x i32]* %4, align 4
-  %5 = bitcast %struct.st1* %0 to i8*
-  %6 = bitcast %struct.st1* %3 to i8*
-  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %5, i8* align 4 %6, i32 8, i1 false)
+define dso_local i32 @demo32(i32 noundef %0) #0 {
+  %2 = alloca i32, align 4
+  store i32 %0, i32* %2, align 4
+  %3 = load i32, i32* %2, align 4
+  %4 = add nsw i32 %3, 100
+  ret i32 %4
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @demo64(i64 noundef %0) #0 {
+  %2 = alloca i64, align 8
+  store i64 %0, i64* %2, align 8
+  %3 = load i64, i64* %2, align 8
+  %4 = add nsw i64 %3, 100
+  ret i64 %4
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt64(%struct.point64* noalias sret(%struct.point64) align 8 %0, [2 x i64] %1) #0 {
+  %3 = alloca %struct.point64, align 8
+  %4 = bitcast %struct.point64* %3 to [2 x i64]*
+  store [2 x i64] %1, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.point64, %struct.point64* %3, i32 0, i32 0
+  %6 = load i64, i64* %5, align 8
+  %7 = getelementptr inbounds %struct.point64, %struct.point64* %3, i32 0, i32 1
+  %8 = load i64, i64* %7, align 8
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str, i32 0, i32 0), i64 noundef %6, i64 noundef %8)
+  %10 = bitcast %struct.point64* %0 to i8*
+  %11 = bitcast %struct.point64* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %10, i8* align 8 %11, i32 16, i1 false)
   ret void
 }
 
+declare i32 @printf(i8* noundef, ...) #1
+
 ; Function Attrs: argmemonly nofree nounwind willreturn
-declare void @llvm.memcpy.p0i8.p0i8.i32(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i32, i1 immarg) #1
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i32, i1 immarg) #2
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i32 @demo32s([1 x i32] %0) #0 {
+  %2 = alloca %struct.struct32, align 4
+  %3 = alloca %struct.struct32, align 4
+  %4 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %5 = bitcast i32* %4 to [1 x i32]*
+  store [1 x i32] %0, [1 x i32]* %5, align 4
+  %6 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %7 = load i32, i32* %6, align 4
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.1, i32 0, i32 0), i32 noundef %7)
+  %9 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %10 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %11 = load i32, i32* %10, align 4
+  %12 = add nsw i32 %11, 100
+  store i32 %12, i32* %9, align 4
+  %13 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %14 = load i32, i32* %13, align 4
+  ret i32 %14
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt(%struct.point* noalias sret(%struct.point) align 4 %0, [2 x i32] %1) #0 {
+  %3 = alloca %struct.point, align 4
+  %4 = bitcast %struct.point* %3 to [2 x i32]*
+  store [2 x i32] %1, [2 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 4
+  %7 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 1
+  %8 = load i32, i32* %7, align 4
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.2, i32 0, i32 0), i32 noundef %6, i32 noundef %8)
+  %10 = bitcast %struct.point* %0 to i8*
+  %11 = bitcast %struct.point* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %10, i8* align 4 %11, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt1(%struct.point1* noalias sret(%struct.point1) align 4 %0, [3 x i32] %1) #0 {
+  %3 = alloca %struct.point1, align 4
+  %4 = bitcast %struct.point1* %3 to [3 x i32]*
+  store [3 x i32] %1, [3 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 4
+  %7 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 1
+  %8 = load i32, i32* %7, align 4
+  %9 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 2
+  %10 = load i32, i32* %9, align 4
+  %11 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.3, i32 0, i32 0), i32 noundef %6, i32 noundef %8, i32 noundef %10)
+  %12 = bitcast %struct.point1* %0 to i8*
+  %13 = bitcast %struct.point1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %12, i8* align 4 %13, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt2(%struct.point2* noalias sret(%struct.point2) align 4 %0, [3 x i32] %1) #0 {
+  %3 = alloca %struct.point2, align 4
+  %4 = bitcast %struct.point2* %3 to [3 x i32]*
+  store [3 x i32] %1, [3 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 1
+  %9 = load i32, i32* %8, align 4
+  %10 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 2
+  %11 = load i32, i32* %10, align 4
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.4, i32 0, i32 0), i32 noundef %7, i32 noundef %9, i32 noundef %11)
+  %13 = bitcast %struct.point2* %0 to i8*
+  %14 = bitcast %struct.point2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %13, i8* align 4 %14, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i32 @pt3([1 x i32] %0) #0 {
+  %2 = alloca %struct.point3, align 1
+  %3 = alloca %struct.point3, align 1
+  %4 = alloca [1 x i32], align 4
+  %5 = alloca i32, align 4
+  store [1 x i32] %0, [1 x i32]* %4, align 4
+  %6 = bitcast %struct.point3* %3 to i8*
+  %7 = bitcast [1 x i32]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %6, i8* align 4 %7, i32 3, i1 false)
+  %8 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.5, i32 0, i32 0), i32 noundef %10, i32 noundef %13, i32 noundef %16)
+  %18 = bitcast %struct.point3* %2 to i8*
+  %19 = bitcast %struct.point3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %18, i8* align 1 %19, i32 3, i1 false)
+  %20 = bitcast i32* %5 to i8*
+  %21 = bitcast %struct.point3* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %20, i8* align 1 %21, i32 3, i1 false)
+  %22 = load i32, i32* %5, align 4
+  ret i32 %22
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt4(%struct.point4* noalias sret(%struct.point4) align 4 %0, [2 x i32] %1) #0 {
+  %3 = alloca %struct.point4, align 4
+  %4 = bitcast %struct.point4* %3 to [2 x i32]*
+  store [2 x i32] %1, [2 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 3
+  %15 = load i32, i32* %14, align 4
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.6, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %15)
+  %17 = bitcast %struct.point4* %0 to i8*
+  %18 = bitcast %struct.point4* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %17, i8* align 4 %18, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt5(%struct.point5* noalias sret(%struct.point5) align 1 %0, [2 x i32] %1) #0 {
+  %3 = alloca %struct.point5, align 1
+  %4 = alloca [2 x i32], align 4
+  store [2 x i32] %1, [2 x i32]* %4, align 4
+  %5 = bitcast %struct.point5* %3 to i8*
+  %6 = bitcast [2 x i32]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %5, i8* align 4 %6, i32 5, i1 false)
+  %7 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 0
+  %8 = load i8, i8* %7, align 1
+  %9 = zext i8 %8 to i32
+  %10 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 1
+  %11 = load i8, i8* %10, align 1
+  %12 = zext i8 %11 to i32
+  %13 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 2
+  %14 = load i8, i8* %13, align 1
+  %15 = zext i8 %14 to i32
+  %16 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 3
+  %17 = load i8, i8* %16, align 1
+  %18 = zext i8 %17 to i32
+  %19 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 4
+  %20 = load i8, i8* %19, align 1
+  %21 = zext i8 %20 to i32
+  %22 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([24 x i8], [24 x i8]* @.str.7, i32 0, i32 0), i32 noundef %9, i32 noundef %12, i32 noundef %15, i32 noundef %18, i32 noundef %21)
+  %23 = bitcast %struct.point5* %0 to i8*
+  %24 = bitcast %struct.point5* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %23, i8* align 1 %24, i32 5, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt6(%struct.point6* noalias sret(%struct.point6) align 4 %0, [3 x i32] %1) #0 {
+  %3 = alloca %struct.point6, align 4
+  %4 = bitcast %struct.point6* %3 to [3 x i32]*
+  store [3 x i32] %1, [3 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 3
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 4
+  %18 = load i8, i8* %17, align 4
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 5
+  %21 = load i32, i32* %20, align 4
+  %22 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.8, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %21)
+  %23 = bitcast %struct.point6* %0 to i8*
+  %24 = bitcast %struct.point6* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %23, i8* align 4 %24, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @pt7(%struct.point7* noalias sret(%struct.point7) align 4 %0, [4 x i32] %1) #0 {
+  %3 = alloca %struct.point7, align 4
+  %4 = bitcast %struct.point7* %3 to [4 x i32]*
+  store [4 x i32] %1, [4 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 3
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 4
+  %18 = load i8, i8* %17, align 4
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 5
+  %21 = load i32, i32* %20, align 4
+  %22 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 6
+  %23 = load i8, i8* %22, align 4
+  %24 = zext i8 %23 to i32
+  %25 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([30 x i8], [30 x i8]* @.str.9, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %21, i32 noundef %24)
+  %26 = bitcast %struct.point7* %0 to i8*
+  %27 = bitcast %struct.point7* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %26, i8* align 4 %27, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @fn1(%struct.data1* noalias sret(%struct.data1) align 8 %0, [2 x i64] %1) #0 {
+  %3 = alloca %struct.data1, align 8
+  %4 = bitcast %struct.data1* %3 to [2 x i64]*
+  store [2 x i64] %1, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data1, %struct.data1* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 8
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.data1, %struct.data1* %3, i32 0, i32 1
+  %9 = load i64, i64* %8, align 8
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.10, i32 0, i32 0), i32 noundef %7, i64 noundef %9)
+  %11 = bitcast %struct.data1* %0 to i8*
+  %12 = bitcast %struct.data1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %11, i8* align 8 %12, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @fn2(%struct.data2* noalias sret(%struct.data2) align 8 %0, [2 x i64] %1) #0 {
+  %3 = alloca %struct.data2, align 8
+  %4 = bitcast %struct.data2* %3 to [2 x i64]*
+  store [2 x i64] %1, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data2, %struct.data2* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 8
+  %7 = getelementptr inbounds %struct.data2, %struct.data2* %3, i32 0, i32 1
+  %8 = load i64, i64* %7, align 8
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.11, i32 0, i32 0), i32 noundef %6, i64 noundef %8)
+  %10 = bitcast %struct.data2* %0 to i8*
+  %11 = bitcast %struct.data2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %10, i8* align 8 %11, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @fn3(%struct.data3* noalias sret(%struct.data3) align 8 %0, [2 x i64] %1) #0 {
+  %3 = alloca %struct.data3, align 8
+  %4 = bitcast %struct.data3* %3 to [2 x i64]*
+  store [2 x i64] %1, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data3, %struct.data3* %3, i32 0, i32 0
+  %6 = load i64, i64* %5, align 8
+  %7 = getelementptr inbounds %struct.data3, %struct.data3* %3, i32 0, i32 1
+  %8 = load i8, i8* %7, align 8
+  %9 = zext i8 %8 to i32
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.12, i32 0, i32 0), i64 noundef %6, i32 noundef %9)
+  %11 = bitcast %struct.data3* %0 to i8*
+  %12 = bitcast %struct.data3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %11, i8* align 8 %12, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local %struct.fdata1 @ff1(%struct.fdata1 %0) #0 {
+  %2 = alloca %struct.fdata1, align 4
+  %3 = alloca %struct.fdata1, align 4
+  store %struct.fdata1 %0, %struct.fdata1* %3, align 4
+  %4 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  %5 = load float, float* %4, align 4
+  %6 = fpext float %5 to double
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.13, i32 0, i32 0), double noundef %6)
+  %8 = bitcast %struct.fdata1* %2 to i8*
+  %9 = bitcast %struct.fdata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %8, i8* align 4 %9, i32 4, i1 false)
+  %10 = load %struct.fdata1, %struct.fdata1* %2, align 4
+  ret %struct.fdata1 %10
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local %struct.ddata1 @dd1(%struct.ddata1 %0) #0 {
+  %2 = alloca %struct.ddata1, align 8
+  %3 = alloca %struct.ddata1, align 8
+  store %struct.ddata1 %0, %struct.ddata1* %3, align 8
+  %4 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  %5 = load double, double* %4, align 8
+  %6 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.14, i32 0, i32 0), double noundef %5)
+  %7 = bitcast %struct.ddata1* %2 to i8*
+  %8 = bitcast %struct.ddata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %7, i8* align 8 %8, i32 8, i1 false)
+  %9 = load %struct.ddata1, %struct.ddata1* %2, align 8
+  ret %struct.ddata1 %9
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local %struct.ddata2 @dd2(%struct.ddata2 %0) #0 {
+  %2 = alloca %struct.ddata2, align 8
+  %3 = alloca %struct.ddata2, align 8
+  store %struct.ddata2 %0, %struct.ddata2* %3, align 8
+  %4 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %3, i32 0, i32 0
+  %5 = load double, double* %4, align 8
+  %6 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %3, i32 0, i32 1
+  %7 = load double, double* %6, align 8
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.15, i32 0, i32 0), double noundef %5, double noundef %7)
+  %9 = bitcast %struct.ddata2* %2 to i8*
+  %10 = bitcast %struct.ddata2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %9, i8* align 8 %10, i32 16, i1 false)
+  %11 = load %struct.ddata2, %struct.ddata2* %2, align 8
+  ret %struct.ddata2 %11
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local %struct.ddata3 @dd3(%struct.ddata3 %0) #0 {
+  %2 = alloca %struct.ddata3, align 8
+  %3 = alloca %struct.ddata3, align 8
+  store %struct.ddata3 %0, %struct.ddata3* %3, align 8
+  %4 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %3, i32 0, i32 0
+  %5 = load double, double* %4, align 8
+  %6 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %3, i32 0, i32 1
+  %7 = load double, double* %6, align 8
+  %8 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %3, i32 0, i32 2
+  %9 = load double, double* %8, align 8
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.16, i32 0, i32 0), double noundef %5, double noundef %7, double noundef %9)
+  %11 = bitcast %struct.ddata3* %2 to i8*
+  %12 = bitcast %struct.ddata3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %11, i8* align 8 %12, i32 24, i1 false)
+  %13 = load %struct.ddata3, %struct.ddata3* %2, align 8
+  ret %struct.ddata3 %13
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @ff2i(%struct.fdata2i* noalias sret(%struct.fdata2i) align 4 %0, [2 x i32] %1) #0 {
+  %3 = alloca %struct.fdata2i, align 4
+  %4 = bitcast %struct.fdata2i* %3 to [2 x i32]*
+  store [2 x i32] %1, [2 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 1
+  %9 = load i32, i32* %8, align 4
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([13 x i8], [13 x i8]* @.str.17, i32 0, i32 0), double noundef %7, i32 noundef %9)
+  %11 = bitcast %struct.fdata2i* %0 to i8*
+  %12 = bitcast %struct.fdata2i* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %11, i8* align 4 %12, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local %struct.fdata2 @ff2(%struct.fdata2 %0) #0 {
+  %2 = alloca %struct.fdata2, align 4
+  %3 = alloca %struct.fdata2, align 4
+  store %struct.fdata2 %0, %struct.fdata2* %3, align 4
+  %4 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 0
+  %5 = load float, float* %4, align 4
+  %6 = fpext float %5 to double
+  %7 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 1
+  %8 = load float, float* %7, align 4
+  %9 = fpext float %8 to double
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.18, i32 0, i32 0), double noundef %6, double noundef %9)
+  %11 = bitcast %struct.fdata2* %2 to i8*
+  %12 = bitcast %struct.fdata2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %11, i8* align 4 %12, i32 8, i1 false)
+  %13 = load %struct.fdata2, %struct.fdata2* %2, align 4
+  ret %struct.fdata2 %13
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local %struct.fdata3 @ff3(%struct.fdata3 %0) #0 {
+  %2 = alloca %struct.fdata3, align 4
+  %3 = alloca %struct.fdata3, align 4
+  store %struct.fdata3 %0, %struct.fdata3* %3, align 4
+  %4 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 0
+  %5 = load float, float* %4, align 4
+  %6 = fpext float %5 to double
+  %7 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 1
+  %8 = load float, float* %7, align 4
+  %9 = fpext float %8 to double
+  %10 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 2
+  %11 = load float, float* %10, align 4
+  %12 = fpext float %11 to double
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.19, i32 0, i32 0), double noundef %6, double noundef %9, double noundef %12)
+  %14 = bitcast %struct.fdata3* %2 to i8*
+  %15 = bitcast %struct.fdata3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %14, i8* align 4 %15, i32 12, i1 false)
+  %16 = load %struct.fdata3, %struct.fdata3* %2, align 4
+  ret %struct.fdata3 %16
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local %struct.fdata4 @ff4(%struct.fdata4 %0) #0 {
+  %2 = alloca %struct.fdata4, align 4
+  %3 = alloca %struct.fdata4, align 4
+  store %struct.fdata4 %0, %struct.fdata4* %3, align 4
+  %4 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 0
+  %5 = load float, float* %4, align 4
+  %6 = fpext float %5 to double
+  %7 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 1
+  %8 = load float, float* %7, align 4
+  %9 = fpext float %8 to double
+  %10 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 2
+  %11 = load float, float* %10, align 4
+  %12 = fpext float %11 to double
+  %13 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 3
+  %14 = load float, float* %13, align 4
+  %15 = fpext float %14 to double
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.20, i32 0, i32 0), double noundef %6, double noundef %9, double noundef %12, double noundef %15)
+  %17 = bitcast %struct.fdata4* %2 to i8*
+  %18 = bitcast %struct.fdata4* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %17, i8* align 4 %18, i32 16, i1 false)
+  %19 = load %struct.fdata4, %struct.fdata4* %2, align 4
+  ret %struct.fdata4 %19
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @ff5(%struct.fdata5* noalias sret(%struct.fdata5) align 4 %0, [5 x i32] %1) #0 {
+  %3 = alloca %struct.fdata5, align 4
+  %4 = bitcast %struct.fdata5* %3 to [5 x i32]*
+  store [5 x i32] %1, [5 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %3, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %3, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %3, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %3, i32 0, i32 4
+  %18 = load float, float* %17, align 4
+  %19 = fpext float %18 to double
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.21, i32 0, i32 0), double noundef %7, double noundef %10, double noundef %13, double noundef %16, double noundef %19)
+  %21 = bitcast %struct.fdata5* %0 to i8*
+  %22 = bitcast %struct.fdata5* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %21, i8* align 4 %22, i32 20, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @ff2id(%struct.fdata2id* noalias sret(%struct.fdata2id) align 8 %0, [2 x i64] %1) #0 {
+  %3 = alloca %struct.fdata2id, align 8
+  %4 = bitcast %struct.fdata2id* %3 to [2 x i64]*
+  store [2 x i64] %1, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 8
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 2
+  %12 = load double, double* %11, align 8
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.22, i32 0, i32 0), i32 noundef %7, i32 noundef %10, double noundef %12)
+  %14 = bitcast %struct.fdata2id* %0 to i8*
+  %15 = bitcast %struct.fdata2id* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %14, i8* align 8 %15, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @ff7if(%struct.fdata7if* noalias sret(%struct.fdata7if) align 4 %0, [3 x i32] %1) #0 {
+  %3 = alloca %struct.fdata7if, align 4
+  %4 = bitcast %struct.fdata7if* %3 to [3 x i32]*
+  store [3 x i32] %1, [3 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 0
+  %6 = getelementptr inbounds [7 x i8], [7 x i8]* %5, i32 0, i32 0
+  %7 = load i8, i8* %6, align 4
+  %8 = zext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 0
+  %10 = getelementptr inbounds [7 x i8], [7 x i8]* %9, i32 0, i32 1
+  %11 = load i8, i8* %10, align 1
+  %12 = zext i8 %11 to i32
+  %13 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 1
+  %14 = load float, float* %13, align 4
+  %15 = fpext float %14 to double
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.23, i32 0, i32 0), i32 noundef %8, i32 noundef %12, double noundef %15)
+  %17 = bitcast %struct.fdata7if* %0 to i8*
+  %18 = bitcast %struct.fdata7if* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %17, i8* align 4 %18, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @ff4if(%struct.fdata4if* noalias sret(%struct.fdata4if) align 4 %0, [4 x i32] %1) #0 {
+  %3 = alloca %struct.fdata4if, align 4
+  %4 = bitcast %struct.fdata4if* %3 to [4 x i32]*
+  store [4 x i32] %1, [4 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 4
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str.24, i32 0, i32 0), double noundef %7, i32 noundef %10, double noundef %13, double noundef %16)
+  %18 = bitcast %struct.fdata4if* %0 to i8*
+  %19 = bitcast %struct.fdata4if* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %18, i8* align 4 %19, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @demo(%struct.array* noalias sret(%struct.array) align 4 %0, [8 x i32] %1) #0 {
+  %3 = alloca %struct.array, align 4
+  %4 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  store [8 x i32] %1, [8 x i32]* %4, align 4
+  %5 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %6 = getelementptr inbounds [8 x i32], [8 x i32]* %5, i32 0, i32 0
+  %7 = load i32, i32* %6, align 4
+  %8 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %9 = getelementptr inbounds [8 x i32], [8 x i32]* %8, i32 0, i32 1
+  %10 = load i32, i32* %9, align 4
+  %11 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i32 0, i32 2
+  %13 = load i32, i32* %12, align 4
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.25, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13)
+  %15 = bitcast %struct.array* %0 to i8*
+  %16 = bitcast %struct.array* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %15, i8* align 4 %16, i32 32, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @demo2(%struct.array* noalias sret(%struct.array) align 4 %0, i32 noundef %1) #0 {
+  %3 = alloca i32, align 4
+  %4 = alloca i32, align 4
+  store i32 %1, i32* %3, align 4
+  store i32 0, i32* %4, align 4
+  br label %5
+
+5:                                                ; preds = %15, %2
+  %6 = load i32, i32* %4, align 4
+  %7 = icmp slt i32 %6, 8
+  br i1 %7, label %8, label %18
+
+8:                                                ; preds = %5
+  %9 = load i32, i32* %4, align 4
+  %10 = load i32, i32* %3, align 4
+  %11 = add nsw i32 %9, %10
+  %12 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %13 = load i32, i32* %4, align 4
+  %14 = getelementptr inbounds [8 x i32], [8 x i32]* %12, i32 0, i32 %13
+  store i32 %11, i32* %14, align 4
+  br label %15
+
+15:                                               ; preds = %8
+  %16 = load i32, i32* %4, align 4
+  %17 = add nsw i32 %16, 1
+  store i32 %17, i32* %4, align 4
+  br label %5, !llvm.loop !10
+
+18:                                               ; preds = %5
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @callback(void (%struct.array*, [8 x i32], [2 x i32], [3 x i32])* noundef %0, [8 x i32] %1) #0 {
+  %3 = alloca %struct.array, align 4
+  %4 = alloca void (%struct.array*, [8 x i32], [2 x i32], [3 x i32])*, align 4
+  %5 = alloca %struct.array, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca %struct.point1, align 4
+  %8 = alloca %struct.array, align 4
+  %9 = alloca %struct.array, align 4
+  %10 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  store [8 x i32] %1, [8 x i32]* %10, align 4
+  store void (%struct.array*, [8 x i32], [2 x i32], [3 x i32])* %0, void (%struct.array*, [8 x i32], [2 x i32], [3 x i32])** %4, align 4
+  %11 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %12 = load [8 x i32], [8 x i32]* %11, align 4
+  call void @demo(%struct.array* sret(%struct.array) align 4 %5, [8 x i32] %12)
+  %13 = bitcast %struct.point* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %13, i8* align 4 bitcast (%struct.point* @__const.callback.pt to i8*), i32 8, i1 false)
+  %14 = bitcast %struct.point1* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %14, i8* align 4 bitcast (%struct.point1* @__const.callback.pt1 to i8*), i32 12, i1 false)
+  %15 = load void (%struct.array*, [8 x i32], [2 x i32], [3 x i32])*, void (%struct.array*, [8 x i32], [2 x i32], [3 x i32])** %4, align 4
+  %16 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %17 = load [8 x i32], [8 x i32]* %16, align 4
+  %18 = bitcast %struct.point* %6 to [2 x i32]*
+  %19 = load [2 x i32], [2 x i32]* %18, align 4
+  %20 = bitcast %struct.point1* %7 to [3 x i32]*
+  %21 = load [3 x i32], [3 x i32]* %20, align 4
+  call void %15(%struct.array* sret(%struct.array) align 4 %8, [8 x i32] %17, [2 x i32] %19, [3 x i32] %21)
+  %22 = getelementptr inbounds %struct.array, %struct.array* %8, i32 0, i32 0
+  %23 = load [8 x i32], [8 x i32]* %22, align 4
+  call void @demo(%struct.array* sret(%struct.array) align 4 %9, [8 x i32] %23)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @callback1(void (%struct.point*, [8 x i32], [2 x i32], [3 x i32])* noundef %0, [8 x i32] %1) #0 {
+  %3 = alloca %struct.array, align 4
+  %4 = alloca void (%struct.point*, [8 x i32], [2 x i32], [3 x i32])*, align 4
+  %5 = alloca %struct.point, align 4
+  %6 = alloca %struct.point1, align 4
+  %7 = alloca %struct.point, align 4
+  %8 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  store [8 x i32] %1, [8 x i32]* %8, align 4
+  store void (%struct.point*, [8 x i32], [2 x i32], [3 x i32])* %0, void (%struct.point*, [8 x i32], [2 x i32], [3 x i32])** %4, align 4
+  %9 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %10 = getelementptr inbounds [8 x i32], [8 x i32]* %9, i32 0, i32 0
+  %11 = load i32, i32* %10, align 4
+  %12 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %13 = getelementptr inbounds [8 x i32], [8 x i32]* %12, i32 0, i32 1
+  %14 = load i32, i32* %13, align 4
+  %15 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %16 = getelementptr inbounds [8 x i32], [8 x i32]* %15, i32 0, i32 7
+  %17 = load i32, i32* %16, align 4
+  %18 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.26, i32 0, i32 0), i32 noundef %11, i32 noundef %14, i32 noundef %17)
+  %19 = bitcast %struct.point* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %19, i8* align 4 bitcast (%struct.point* @__const.callback1.pt to i8*), i32 8, i1 false)
+  %20 = bitcast %struct.point1* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %20, i8* align 4 bitcast (%struct.point1* @__const.callback1.pt1 to i8*), i32 12, i1 false)
+  %21 = load void (%struct.point*, [8 x i32], [2 x i32], [3 x i32])*, void (%struct.point*, [8 x i32], [2 x i32], [3 x i32])** %4, align 4
+  %22 = getelementptr inbounds %struct.array, %struct.array* %3, i32 0, i32 0
+  %23 = load [8 x i32], [8 x i32]* %22, align 4
+  %24 = bitcast %struct.point* %5 to [2 x i32]*
+  %25 = load [2 x i32], [2 x i32]* %24, align 4
+  %26 = bitcast %struct.point1* %6 to [3 x i32]*
+  %27 = load [3 x i32], [3 x i32]* %26, align 4
+  call void %21(%struct.point* sret(%struct.point) align 4 %7, [8 x i32] %23, [2 x i32] %25, [3 x i32] %27)
+  %28 = getelementptr inbounds %struct.point, %struct.point* %7, i32 0, i32 0
+  %29 = load i32, i32* %28, align 4
+  %30 = getelementptr inbounds %struct.point, %struct.point* %7, i32 0, i32 1
+  %31 = load i32, i32* %30, align 4
+  %32 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.27, i32 0, i32 0), i32 noundef %29, i32 noundef %31)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @mycallback(%struct.point* noalias sret(%struct.point) align 4 %0, [8 x i32] %1, [2 x i32] %2, [3 x i32] %3) #0 {
+  %5 = alloca %struct.array, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca %struct.point1, align 4
+  %8 = getelementptr inbounds %struct.array, %struct.array* %5, i32 0, i32 0
+  store [8 x i32] %1, [8 x i32]* %8, align 4
+  %9 = bitcast %struct.point* %6 to [2 x i32]*
+  store [2 x i32] %2, [2 x i32]* %9, align 4
+  %10 = bitcast %struct.point1* %7 to [3 x i32]*
+  store [3 x i32] %3, [3 x i32]* %10, align 4
+  %11 = getelementptr inbounds %struct.array, %struct.array* %5, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i32 0, i32 0
+  %13 = load i32, i32* %12, align 4
+  %14 = getelementptr inbounds %struct.array, %struct.array* %5, i32 0, i32 0
+  %15 = getelementptr inbounds [8 x i32], [8 x i32]* %14, i32 0, i32 1
+  %16 = load i32, i32* %15, align 4
+  %17 = getelementptr inbounds %struct.array, %struct.array* %5, i32 0, i32 0
+  %18 = getelementptr inbounds [8 x i32], [8 x i32]* %17, i32 0, i32 7
+  %19 = load i32, i32* %18, align 4
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([28 x i8], [28 x i8]* @.str.28, i32 0, i32 0), i32 noundef %13, i32 noundef %16, i32 noundef %19)
+  %21 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %22 = load i32, i32* %21, align 4
+  %23 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %24 = load i32, i32* %23, align 4
+  %25 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.29, i32 0, i32 0), i32 noundef %22, i32 noundef %24)
+  %26 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 0
+  %27 = load i32, i32* %26, align 4
+  %28 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 1
+  %29 = load i32, i32* %28, align 4
+  %30 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 2
+  %31 = load i32, i32* %30, align 4
+  %32 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([26 x i8], [26 x i8]* @.str.30, i32 0, i32 0), i32 noundef %27, i32 noundef %29, i32 noundef %31)
+  %33 = getelementptr inbounds %struct.point, %struct.point* %0, i32 0, i32 0
+  %34 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %35 = load i32, i32* %34, align 4
+  %36 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 0
+  %37 = load i32, i32* %36, align 4
+  %38 = add nsw i32 %35, %37
+  store i32 %38, i32* %33, align 4
+  %39 = getelementptr inbounds %struct.point, %struct.point* %0, i32 0, i32 1
+  %40 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %41 = load i32, i32* %40, align 4
+  %42 = getelementptr inbounds %struct.point1, %struct.point1* %7, i32 0, i32 1
+  %43 = load i32, i32* %42, align 4
+  %44 = add nsw i32 %41, %43
+  store i32 %44, i32* %39, align 4
+  ret void
+}
 
 attributes #0 = { noinline nounwind optnone "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="arm1176jzf-s" "target-features"="+armv6kz,+dsp,+fp64,+strict-align,+vfp2,+vfp2sp,-aes,-d32,-fp-armv8,-fp-armv8d16,-fp-armv8d16sp,-fp-armv8sp,-fp16,-fp16fml,-fullfp16,-neon,-sha2,-thumb-mode,-vfp3,-vfp3d16,-vfp3d16sp,-vfp3sp,-vfp4,-vfp4d16,-vfp4d16sp,-vfp4sp" }
-attributes #1 = { argmemonly nofree nounwind willreturn }
+attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="arm1176jzf-s" "target-features"="+armv6kz,+dsp,+fp64,+strict-align,+vfp2,+vfp2sp,-aes,-d32,-fp-armv8,-fp-armv8d16,-fp-armv8d16sp,-fp-armv8sp,-fp16,-fp16fml,-fullfp16,-neon,-sha2,-thumb-mode,-vfp3,-vfp3d16,-vfp3d16sp,-vfp3sp,-vfp4,-vfp4d16,-vfp4d16sp,-vfp4sp" }
+attributes #2 = { argmemonly nofree nounwind willreturn }
 
 !llvm.module.flags = !{!0, !1, !2, !3, !4, !5, !6, !7, !8}
 !llvm.ident = !{!9}
@@ -35,3 +810,5 @@ attributes #1 = { argmemonly nofree nounwind willreturn }
 !7 = !{i32 7, !"PIE Level", i32 2}
 !8 = !{i32 7, !"frame-pointer", i32 2}
 !9 = !{!"Apple clang version 14.0.3 (clang-1403.0.22.14.1)"}
+!10 = distinct !{!10, !11}
+!11 = !{!"llvm.loop.mustprogress"}
diff --git a/internal/cabi/_testdata/arch/i386/demo.ll b/internal/cabi/_testdata/arch/i386/demo.ll
index 1b5b84fe..09d01f16 100644
--- a/internal/cabi/_testdata/arch/i386/demo.ll
+++ b/internal/cabi/_testdata/arch/i386/demo.ll
@@ -3,29 +3,845 @@ source_filename = "../../wrap/demo.c"
 target datalayout = "e-m:e-p:32:32-p270:32:32-p271:32:32-p272:64:64-f64:32:64-f80:32-n8:16:32-S128"
 target triple = "i386-unknown-linux-gnu"
 
-%struct.st1 = type { i32, i32 }
+%struct.point = type { i32, i32 }
+%struct.point1 = type { i32, i32, i32 }
+%struct.point64 = type { i64, i64 }
+%struct.struct32 = type { i32 }
+%struct.point2 = type { i8, i32, i32 }
+%struct.point3 = type { i8, i8, i8 }
+%struct.point4 = type { i8, i8, i8, i32 }
+%struct.point5 = type { i8, i8, i8, i8, i8 }
+%struct.point6 = type { i8, i8, i8, i8, i8, i32 }
+%struct.point7 = type { i8, i8, i8, i8, i8, i32, i8 }
+%struct.data1 = type { i8, i64 }
+%struct.data2 = type { i32, i64 }
+%struct.data3 = type { i64, i8 }
+%struct.fdata1 = type { float }
+%struct.ddata1 = type { double }
+%struct.ddata2 = type { double, double }
+%struct.ddata3 = type { double, double, double }
+%struct.fdata2i = type { float, i32 }
+%struct.fdata2 = type { float, float }
+%struct.fdata3 = type { float, float, float }
+%struct.fdata4 = type { float, float, float, float }
+%struct.fdata5 = type { float, float, float, float, float }
+%struct.fdata2id = type { i8, i8, double }
+%struct.fdata7if = type { [7 x i8], float }
+%struct.fdata4if = type { float, i8, float, float }
+%struct.array = type { [8 x i32] }
+
+@.str = private unnamed_addr constant [20 x i8] c"point64: %lld %lld\0A\00", align 1
+@.str.1 = private unnamed_addr constant [14 x i8] c"struct32: %d\0A\00", align 1
+@.str.2 = private unnamed_addr constant [14 x i8] c"point: %d %d\0A\00", align 1
+@.str.3 = private unnamed_addr constant [18 x i8] c"point1: %d %d %d\0A\00", align 1
+@.str.4 = private unnamed_addr constant [18 x i8] c"point2: %d %d %d\0A\00", align 1
+@.str.5 = private unnamed_addr constant [18 x i8] c"point3: %d %d %d\0A\00", align 1
+@.str.6 = private unnamed_addr constant [21 x i8] c"point4: %d %d %d %d\0A\00", align 1
+@.str.7 = private unnamed_addr constant [24 x i8] c"point5: %d %d %d %d %d\0A\00", align 1
+@.str.8 = private unnamed_addr constant [27 x i8] c"point6: %d %d %d %d %d %d\0A\00", align 1
+@.str.9 = private unnamed_addr constant [30 x i8] c"point7: %d %d %d %d %d %d %d\0A\00", align 1
+@.str.10 = private unnamed_addr constant [16 x i8] c"data1: %d %lld\0A\00", align 1
+@.str.11 = private unnamed_addr constant [16 x i8] c"data2: %d %lld\0A\00", align 1
+@.str.12 = private unnamed_addr constant [16 x i8] c"data3: %lld %d\0A\00", align 1
+@.str.13 = private unnamed_addr constant [9 x i8] c"ff1: %f\0A\00", align 1
+@.str.14 = private unnamed_addr constant [9 x i8] c"dd1: %f\0A\00", align 1
+@.str.15 = private unnamed_addr constant [12 x i8] c"dd2: %f %f\0A\00", align 1
+@.str.16 = private unnamed_addr constant [15 x i8] c"dd3: %f %f %f\0A\00", align 1
+@.str.17 = private unnamed_addr constant [13 x i8] c"ff2i: %f %d\0A\00", align 1
+@.str.18 = private unnamed_addr constant [12 x i8] c"ff2: %f %f\0A\00", align 1
+@.str.19 = private unnamed_addr constant [15 x i8] c"ff3: %f %f %f\0A\00", align 1
+@.str.20 = private unnamed_addr constant [18 x i8] c"ff4: %f %f %f %f\0A\00", align 1
+@.str.21 = private unnamed_addr constant [21 x i8] c"ff5: %f %f %f %f %f\0A\00", align 1
+@.str.22 = private unnamed_addr constant [15 x i8] c"ff6: %d %d %f\0A\00", align 1
+@.str.23 = private unnamed_addr constant [17 x i8] c"ff7if: %d %d %f\0A\00", align 1
+@.str.24 = private unnamed_addr constant [20 x i8] c"ff4if: %f %d %f %f\0A\00", align 1
+@.str.25 = private unnamed_addr constant [16 x i8] c"demo: %d %d %d\0A\00", align 1
+@__const.callback.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.26 = private unnamed_addr constant [27 x i8] c"callback1 array: %d %d %d\0A\00", align 1
+@__const.callback1.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback1.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.27 = private unnamed_addr constant [22 x i8] c"callback1 ret: %d,%d\0A\00", align 1
+@.str.28 = private unnamed_addr constant [28 x i8] c"mycallback array: %d %d %d\0A\00", align 1
+@.str.29 = private unnamed_addr constant [22 x i8] c"mycallback pt: %d %d\0A\00", align 1
+@.str.30 = private unnamed_addr constant [26 x i8] c"mycallback pt1: %d %d %d\0A\00", align 1
 
 ; Function Attrs: noinline nounwind optnone uwtable
-define dso_local void @fn1(%struct.st1* noalias sret(%struct.st1) align 4 %0, i32 %1, i32 %2) #0 {
+define dso_local i32 @demo32(i32 noundef %0) #0 {
+  %2 = alloca i32, align 4
+  store i32 %0, i32* %2, align 4
+  %3 = load i32, i32* %2, align 4
+  %4 = add nsw i32 %3, 100
+  ret i32 %4
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local i64 @demo64(i64 noundef %0) #0 {
+  %2 = alloca i64, align 8
+  store i64 %0, i64* %2, align 8
+  %3 = load i64, i64* %2, align 8
+  %4 = add nsw i64 %3, 100
+  ret i64 %4
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt64(%struct.point64* noalias sret(%struct.point64) align 4 %0, i64 %1, i64 %2) #0 {
   %4 = alloca i8*, align 4
-  %5 = alloca %struct.st1, align 4
-  %6 = bitcast %struct.st1* %0 to i8*
+  %5 = alloca %struct.point64, align 4
+  %6 = bitcast %struct.point64* %0 to i8*
   store i8* %6, i8** %4, align 4
-  %7 = getelementptr inbounds %struct.st1, %struct.st1* %5, i32 0, i32 0
-  store i32 %1, i32* %7, align 4
-  %8 = getelementptr inbounds %struct.st1, %struct.st1* %5, i32 0, i32 1
-  store i32 %2, i32* %8, align 4
-  %9 = bitcast %struct.st1* %0 to i8*
-  %10 = bitcast %struct.st1* %5 to i8*
-  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %9, i8* align 4 %10, i32 8, i1 false)
+  %7 = getelementptr inbounds %struct.point64, %struct.point64* %5, i32 0, i32 0
+  store i64 %1, i64* %7, align 4
+  %8 = getelementptr inbounds %struct.point64, %struct.point64* %5, i32 0, i32 1
+  store i64 %2, i64* %8, align 4
+  %9 = getelementptr inbounds %struct.point64, %struct.point64* %5, i32 0, i32 0
+  %10 = load i64, i64* %9, align 4
+  %11 = getelementptr inbounds %struct.point64, %struct.point64* %5, i32 0, i32 1
+  %12 = load i64, i64* %11, align 4
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str, i32 0, i32 0), i64 noundef %10, i64 noundef %12)
+  %14 = bitcast %struct.point64* %0 to i8*
+  %15 = bitcast %struct.point64* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %14, i8* align 4 %15, i32 16, i1 false)
   ret void
 }
 
+declare i32 @printf(i8* noundef, ...) #1
+
 ; Function Attrs: argmemonly nofree nounwind willreturn
-declare void @llvm.memcpy.p0i8.p0i8.i32(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i32, i1 immarg) #1
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i32, i1 immarg) #2
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @demo32s(%struct.struct32* noalias sret(%struct.struct32) align 4 %0, i32 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = alloca %struct.struct32, align 4
+  %5 = bitcast %struct.struct32* %0 to i8*
+  store i8* %5, i8** %3, align 4
+  %6 = getelementptr inbounds %struct.struct32, %struct.struct32* %4, i32 0, i32 0
+  store i32 %1, i32* %6, align 4
+  %7 = getelementptr inbounds %struct.struct32, %struct.struct32* %4, i32 0, i32 0
+  %8 = load i32, i32* %7, align 4
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.1, i32 0, i32 0), i32 noundef %8)
+  %10 = getelementptr inbounds %struct.struct32, %struct.struct32* %0, i32 0, i32 0
+  %11 = getelementptr inbounds %struct.struct32, %struct.struct32* %4, i32 0, i32 0
+  %12 = load i32, i32* %11, align 4
+  %13 = add nsw i32 %12, 100
+  store i32 %13, i32* %10, align 4
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt(%struct.point* noalias sret(%struct.point) align 4 %0, i32 %1, i32 %2) #0 {
+  %4 = alloca i8*, align 4
+  %5 = alloca %struct.point, align 4
+  %6 = bitcast %struct.point* %0 to i8*
+  store i8* %6, i8** %4, align 4
+  %7 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  store i32 %1, i32* %7, align 4
+  %8 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  store i32 %2, i32* %8, align 4
+  %9 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  %10 = load i32, i32* %9, align 4
+  %11 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  %12 = load i32, i32* %11, align 4
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.2, i32 0, i32 0), i32 noundef %10, i32 noundef %12)
+  %14 = bitcast %struct.point* %0 to i8*
+  %15 = bitcast %struct.point* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %14, i8* align 4 %15, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt1(%struct.point1* noalias sret(%struct.point1) align 4 %0, i32 %1, i32 %2, i32 %3) #0 {
+  %5 = alloca i8*, align 4
+  %6 = alloca %struct.point1, align 4
+  %7 = bitcast %struct.point1* %0 to i8*
+  store i8* %7, i8** %5, align 4
+  %8 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 0
+  store i32 %1, i32* %8, align 4
+  %9 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 1
+  store i32 %2, i32* %9, align 4
+  %10 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 2
+  store i32 %3, i32* %10, align 4
+  %11 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 0
+  %12 = load i32, i32* %11, align 4
+  %13 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 1
+  %14 = load i32, i32* %13, align 4
+  %15 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 2
+  %16 = load i32, i32* %15, align 4
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.3, i32 0, i32 0), i32 noundef %12, i32 noundef %14, i32 noundef %16)
+  %18 = bitcast %struct.point1* %0 to i8*
+  %19 = bitcast %struct.point1* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %18, i8* align 4 %19, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt2(%struct.point2* noalias sret(%struct.point2) align 4 %0, %struct.point2* noundef byval(%struct.point2) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.point2* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.point2, %struct.point2* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point2, %struct.point2* %1, i32 0, i32 1
+  %9 = load i32, i32* %8, align 4
+  %10 = getelementptr inbounds %struct.point2, %struct.point2* %1, i32 0, i32 2
+  %11 = load i32, i32* %10, align 4
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.4, i32 0, i32 0), i32 noundef %7, i32 noundef %9, i32 noundef %11)
+  %13 = bitcast %struct.point2* %0 to i8*
+  %14 = bitcast %struct.point2* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %13, i8* align 4 %14, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt3(%struct.point3* noalias sret(%struct.point3) align 1 %0, %struct.point3* noundef byval(%struct.point3) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.point3* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.point3, %struct.point3* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point3, %struct.point3* %1, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point3, %struct.point3* %1, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = sext i8 %12 to i32
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.5, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13)
+  %15 = bitcast %struct.point3* %0 to i8*
+  %16 = bitcast %struct.point3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %15, i8* align 4 %16, i32 3, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt4(%struct.point4* noalias sret(%struct.point4) align 4 %0, %struct.point4* noundef byval(%struct.point4) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.point4* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 3
+  %15 = load i32, i32* %14, align 4
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.6, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %15)
+  %17 = bitcast %struct.point4* %0 to i8*
+  %18 = bitcast %struct.point4* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %17, i8* align 4 %18, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt5(%struct.point5* noalias sret(%struct.point5) align 1 %0, %struct.point5* noundef byval(%struct.point5) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.point5* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 3
+  %15 = load i8, i8* %14, align 1
+  %16 = sext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 4
+  %18 = load i8, i8* %17, align 4
+  %19 = sext i8 %18 to i32
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([24 x i8], [24 x i8]* @.str.7, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19)
+  %21 = bitcast %struct.point5* %0 to i8*
+  %22 = bitcast %struct.point5* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %21, i8* align 4 %22, i32 5, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt6(%struct.point6* noalias sret(%struct.point6) align 4 %0, %struct.point6* noundef byval(%struct.point6) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.point6* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 3
+  %15 = load i8, i8* %14, align 1
+  %16 = sext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 4
+  %18 = load i8, i8* %17, align 4
+  %19 = sext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 5
+  %21 = load i32, i32* %20, align 4
+  %22 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.8, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %21)
+  %23 = bitcast %struct.point6* %0 to i8*
+  %24 = bitcast %struct.point6* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %23, i8* align 4 %24, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @pt7(%struct.point7* noalias sret(%struct.point7) align 4 %0, %struct.point7* noundef byval(%struct.point7) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.point7* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = sext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 3
+  %15 = load i8, i8* %14, align 1
+  %16 = sext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 4
+  %18 = load i8, i8* %17, align 4
+  %19 = sext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 5
+  %21 = load i32, i32* %20, align 4
+  %22 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 6
+  %23 = load i8, i8* %22, align 4
+  %24 = sext i8 %23 to i32
+  %25 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([30 x i8], [30 x i8]* @.str.9, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13, i32 noundef %16, i32 noundef %19, i32 noundef %21, i32 noundef %24)
+  %26 = bitcast %struct.point7* %0 to i8*
+  %27 = bitcast %struct.point7* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %26, i8* align 4 %27, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @fn1(%struct.data1* noalias sret(%struct.data1) align 4 %0, %struct.data1* noundef byval(%struct.data1) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.data1* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.data1, %struct.data1* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.data1, %struct.data1* %1, i32 0, i32 1
+  %9 = load i64, i64* %8, align 4
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.10, i32 0, i32 0), i32 noundef %7, i64 noundef %9)
+  %11 = bitcast %struct.data1* %0 to i8*
+  %12 = bitcast %struct.data1* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %11, i8* align 4 %12, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @fn2(%struct.data2* noalias sret(%struct.data2) align 4 %0, i32 %1, i64 %2) #0 {
+  %4 = alloca i8*, align 4
+  %5 = alloca %struct.data2, align 4
+  %6 = bitcast %struct.data2* %0 to i8*
+  store i8* %6, i8** %4, align 4
+  %7 = getelementptr inbounds %struct.data2, %struct.data2* %5, i32 0, i32 0
+  store i32 %1, i32* %7, align 4
+  %8 = getelementptr inbounds %struct.data2, %struct.data2* %5, i32 0, i32 1
+  store i64 %2, i64* %8, align 4
+  %9 = getelementptr inbounds %struct.data2, %struct.data2* %5, i32 0, i32 0
+  %10 = load i32, i32* %9, align 4
+  %11 = getelementptr inbounds %struct.data2, %struct.data2* %5, i32 0, i32 1
+  %12 = load i64, i64* %11, align 4
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.11, i32 0, i32 0), i32 noundef %10, i64 noundef %12)
+  %14 = bitcast %struct.data2* %0 to i8*
+  %15 = bitcast %struct.data2* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %14, i8* align 4 %15, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @fn3(%struct.data3* noalias sret(%struct.data3) align 4 %0, %struct.data3* noundef byval(%struct.data3) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.data3* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.data3, %struct.data3* %1, i32 0, i32 0
+  %6 = load i64, i64* %5, align 4
+  %7 = getelementptr inbounds %struct.data3, %struct.data3* %1, i32 0, i32 1
+  %8 = load i8, i8* %7, align 4
+  %9 = sext i8 %8 to i32
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.12, i32 0, i32 0), i64 noundef %6, i32 noundef %9)
+  %11 = bitcast %struct.data3* %0 to i8*
+  %12 = bitcast %struct.data3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %11, i8* align 4 %12, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff1(%struct.fdata1* noalias sret(%struct.fdata1) align 4 %0, float %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = alloca %struct.fdata1, align 4
+  %5 = bitcast %struct.fdata1* %0 to i8*
+  store i8* %5, i8** %3, align 4
+  %6 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %4, i32 0, i32 0
+  store float %1, float* %6, align 4
+  %7 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %4, i32 0, i32 0
+  %8 = load float, float* %7, align 4
+  %9 = fpext float %8 to double
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.13, i32 0, i32 0), double noundef %9)
+  %11 = bitcast %struct.fdata1* %0 to i8*
+  %12 = bitcast %struct.fdata1* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %11, i8* align 4 %12, i32 4, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @dd1(%struct.ddata1* noalias sret(%struct.ddata1) align 4 %0, double %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = alloca %struct.ddata1, align 4
+  %5 = bitcast %struct.ddata1* %0 to i8*
+  store i8* %5, i8** %3, align 4
+  %6 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %4, i32 0, i32 0
+  store double %1, double* %6, align 4
+  %7 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %4, i32 0, i32 0
+  %8 = load double, double* %7, align 4
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.14, i32 0, i32 0), double noundef %8)
+  %10 = bitcast %struct.ddata1* %0 to i8*
+  %11 = bitcast %struct.ddata1* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %10, i8* align 4 %11, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @dd2(%struct.ddata2* noalias sret(%struct.ddata2) align 4 %0, double %1, double %2) #0 {
+  %4 = alloca i8*, align 4
+  %5 = alloca %struct.ddata2, align 4
+  %6 = bitcast %struct.ddata2* %0 to i8*
+  store i8* %6, i8** %4, align 4
+  %7 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %5, i32 0, i32 0
+  store double %1, double* %7, align 4
+  %8 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %5, i32 0, i32 1
+  store double %2, double* %8, align 4
+  %9 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %5, i32 0, i32 0
+  %10 = load double, double* %9, align 4
+  %11 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %5, i32 0, i32 1
+  %12 = load double, double* %11, align 4
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.15, i32 0, i32 0), double noundef %10, double noundef %12)
+  %14 = bitcast %struct.ddata2* %0 to i8*
+  %15 = bitcast %struct.ddata2* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %14, i8* align 4 %15, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @dd3(%struct.ddata3* noalias sret(%struct.ddata3) align 4 %0, %struct.ddata3* noundef byval(%struct.ddata3) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.ddata3* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 0
+  %6 = load double, double* %5, align 4
+  %7 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 1
+  %8 = load double, double* %7, align 4
+  %9 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 2
+  %10 = load double, double* %9, align 4
+  %11 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.16, i32 0, i32 0), double noundef %6, double noundef %8, double noundef %10)
+  %12 = bitcast %struct.ddata3* %0 to i8*
+  %13 = bitcast %struct.ddata3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %12, i8* align 4 %13, i32 24, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff2i(%struct.fdata2i* noalias sret(%struct.fdata2i) align 4 %0, float %1, i32 %2) #0 {
+  %4 = alloca i8*, align 4
+  %5 = alloca %struct.fdata2i, align 4
+  %6 = bitcast %struct.fdata2i* %0 to i8*
+  store i8* %6, i8** %4, align 4
+  %7 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %5, i32 0, i32 0
+  store float %1, float* %7, align 4
+  %8 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %5, i32 0, i32 1
+  store i32 %2, i32* %8, align 4
+  %9 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %5, i32 0, i32 0
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %5, i32 0, i32 1
+  %13 = load i32, i32* %12, align 4
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([13 x i8], [13 x i8]* @.str.17, i32 0, i32 0), double noundef %11, i32 noundef %13)
+  %15 = bitcast %struct.fdata2i* %0 to i8*
+  %16 = bitcast %struct.fdata2i* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %15, i8* align 4 %16, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff2(%struct.fdata2* noalias sret(%struct.fdata2) align 4 %0, float %1, float %2) #0 {
+  %4 = alloca i8*, align 4
+  %5 = alloca %struct.fdata2, align 4
+  %6 = bitcast %struct.fdata2* %0 to i8*
+  store i8* %6, i8** %4, align 4
+  %7 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %5, i32 0, i32 0
+  store float %1, float* %7, align 4
+  %8 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %5, i32 0, i32 1
+  store float %2, float* %8, align 4
+  %9 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %5, i32 0, i32 0
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %5, i32 0, i32 1
+  %13 = load float, float* %12, align 4
+  %14 = fpext float %13 to double
+  %15 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.18, i32 0, i32 0), double noundef %11, double noundef %14)
+  %16 = bitcast %struct.fdata2* %0 to i8*
+  %17 = bitcast %struct.fdata2* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %16, i8* align 4 %17, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff3(%struct.fdata3* noalias sret(%struct.fdata3) align 4 %0, float %1, float %2, float %3) #0 {
+  %5 = alloca i8*, align 4
+  %6 = alloca %struct.fdata3, align 4
+  %7 = bitcast %struct.fdata3* %0 to i8*
+  store i8* %7, i8** %5, align 4
+  %8 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %6, i32 0, i32 0
+  store float %1, float* %8, align 4
+  %9 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %6, i32 0, i32 1
+  store float %2, float* %9, align 4
+  %10 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %6, i32 0, i32 2
+  store float %3, float* %10, align 4
+  %11 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %6, i32 0, i32 0
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %6, i32 0, i32 1
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %6, i32 0, i32 2
+  %18 = load float, float* %17, align 4
+  %19 = fpext float %18 to double
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.19, i32 0, i32 0), double noundef %13, double noundef %16, double noundef %19)
+  %21 = bitcast %struct.fdata3* %0 to i8*
+  %22 = bitcast %struct.fdata3* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %21, i8* align 4 %22, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff4(%struct.fdata4* noalias sret(%struct.fdata4) align 4 %0, float %1, float %2, float %3, float %4) #0 {
+  %6 = alloca i8*, align 4
+  %7 = alloca %struct.fdata4, align 4
+  %8 = bitcast %struct.fdata4* %0 to i8*
+  store i8* %8, i8** %6, align 4
+  %9 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 0
+  store float %1, float* %9, align 4
+  %10 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 1
+  store float %2, float* %10, align 4
+  %11 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 2
+  store float %3, float* %11, align 4
+  %12 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 3
+  store float %4, float* %12, align 4
+  %13 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 0
+  %14 = load float, float* %13, align 4
+  %15 = fpext float %14 to double
+  %16 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 1
+  %17 = load float, float* %16, align 4
+  %18 = fpext float %17 to double
+  %19 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 2
+  %20 = load float, float* %19, align 4
+  %21 = fpext float %20 to double
+  %22 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %7, i32 0, i32 3
+  %23 = load float, float* %22, align 4
+  %24 = fpext float %23 to double
+  %25 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.20, i32 0, i32 0), double noundef %15, double noundef %18, double noundef %21, double noundef %24)
+  %26 = bitcast %struct.fdata4* %0 to i8*
+  %27 = bitcast %struct.fdata4* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %26, i8* align 4 %27, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff5(%struct.fdata5* noalias sret(%struct.fdata5) align 4 %0, %struct.fdata5* noundef byval(%struct.fdata5) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.fdata5* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 4
+  %18 = load float, float* %17, align 4
+  %19 = fpext float %18 to double
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.21, i32 0, i32 0), double noundef %7, double noundef %10, double noundef %13, double noundef %16, double noundef %19)
+  %21 = bitcast %struct.fdata5* %0 to i8*
+  %22 = bitcast %struct.fdata5* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %21, i8* align 4 %22, i32 20, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff2id(%struct.fdata2id* noalias sret(%struct.fdata2id) align 4 %0, %struct.fdata2id* noundef byval(%struct.fdata2id) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.fdata2id* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %1, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = sext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %1, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %1, i32 0, i32 2
+  %12 = load double, double* %11, align 4
+  %13 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.22, i32 0, i32 0), i32 noundef %7, i32 noundef %10, double noundef %12)
+  %14 = bitcast %struct.fdata2id* %0 to i8*
+  %15 = bitcast %struct.fdata2id* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %14, i8* align 4 %15, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff7if(%struct.fdata7if* noalias sret(%struct.fdata7if) align 4 %0, %struct.fdata7if* noundef byval(%struct.fdata7if) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.fdata7if* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %1, i32 0, i32 0
+  %6 = getelementptr inbounds [7 x i8], [7 x i8]* %5, i32 0, i32 0
+  %7 = load i8, i8* %6, align 4
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %1, i32 0, i32 0
+  %10 = getelementptr inbounds [7 x i8], [7 x i8]* %9, i32 0, i32 1
+  %11 = load i8, i8* %10, align 1
+  %12 = sext i8 %11 to i32
+  %13 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %1, i32 0, i32 1
+  %14 = load float, float* %13, align 4
+  %15 = fpext float %14 to double
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.23, i32 0, i32 0), i32 noundef %8, i32 noundef %12, double noundef %15)
+  %17 = bitcast %struct.fdata7if* %0 to i8*
+  %18 = bitcast %struct.fdata7if* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %17, i8* align 4 %18, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @ff4if(%struct.fdata4if* noalias sret(%struct.fdata4if) align 4 %0, %struct.fdata4if* noundef byval(%struct.fdata4if) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.fdata4if* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 1
+  %9 = load i8, i8* %8, align 4
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str.24, i32 0, i32 0), double noundef %7, i32 noundef %10, double noundef %13, double noundef %16)
+  %18 = bitcast %struct.fdata4if* %0 to i8*
+  %19 = bitcast %struct.fdata4if* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %18, i8* align 4 %19, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @demo(%struct.array* noalias sret(%struct.array) align 4 %0, %struct.array* noundef byval(%struct.array) align 4 %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = bitcast %struct.array* %0 to i8*
+  store i8* %4, i8** %3, align 4
+  %5 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %6 = getelementptr inbounds [8 x i32], [8 x i32]* %5, i32 0, i32 0
+  %7 = load i32, i32* %6, align 4
+  %8 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %9 = getelementptr inbounds [8 x i32], [8 x i32]* %8, i32 0, i32 1
+  %10 = load i32, i32* %9, align 4
+  %11 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i32 0, i32 2
+  %13 = load i32, i32* %12, align 4
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.25, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13)
+  %15 = bitcast %struct.array* %0 to i8*
+  %16 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %15, i8* align 4 %16, i32 32, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @demo2(%struct.array* noalias sret(%struct.array) align 4 %0, i32 noundef %1) #0 {
+  %3 = alloca i8*, align 4
+  %4 = alloca i32, align 4
+  %5 = alloca i32, align 4
+  %6 = bitcast %struct.array* %0 to i8*
+  store i8* %6, i8** %3, align 4
+  store i32 %1, i32* %4, align 4
+  store i32 0, i32* %5, align 4
+  br label %7
+
+7:                                                ; preds = %17, %2
+  %8 = load i32, i32* %5, align 4
+  %9 = icmp slt i32 %8, 8
+  br i1 %9, label %10, label %20
+
+10:                                               ; preds = %7
+  %11 = load i32, i32* %5, align 4
+  %12 = load i32, i32* %4, align 4
+  %13 = add nsw i32 %11, %12
+  %14 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %15 = load i32, i32* %5, align 4
+  %16 = getelementptr inbounds [8 x i32], [8 x i32]* %14, i32 0, i32 %15
+  store i32 %13, i32* %16, align 4
+  br label %17
+
+17:                                               ; preds = %10
+  %18 = load i32, i32* %5, align 4
+  %19 = add nsw i32 %18, 1
+  store i32 %19, i32* %5, align 4
+  br label %7, !llvm.loop !7
+
+20:                                               ; preds = %7
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @callback(void (%struct.array*, %struct.array*, i32, i32, i32, i32, i32)* noundef %0, %struct.array* noundef byval(%struct.array) align 4 %1) #0 {
+  %3 = alloca void (%struct.array*, %struct.array*, i32, i32, i32, i32, i32)*, align 4
+  %4 = alloca %struct.array, align 4
+  %5 = alloca %struct.point, align 4
+  %6 = alloca %struct.point1, align 4
+  %7 = alloca %struct.array, align 4
+  %8 = alloca %struct.array, align 4
+  store void (%struct.array*, %struct.array*, i32, i32, i32, i32, i32)* %0, void (%struct.array*, %struct.array*, i32, i32, i32, i32, i32)** %3, align 4
+  call void @demo(%struct.array* sret(%struct.array) align 4 %4, %struct.array* noundef byval(%struct.array) align 4 %1)
+  %9 = bitcast %struct.point* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %9, i8* align 4 bitcast (%struct.point* @__const.callback.pt to i8*), i32 8, i1 false)
+  %10 = bitcast %struct.point1* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %10, i8* align 4 bitcast (%struct.point1* @__const.callback.pt1 to i8*), i32 12, i1 false)
+  %11 = load void (%struct.array*, %struct.array*, i32, i32, i32, i32, i32)*, void (%struct.array*, %struct.array*, i32, i32, i32, i32, i32)** %3, align 4
+  %12 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  %13 = load i32, i32* %12, align 4
+  %14 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  %15 = load i32, i32* %14, align 4
+  %16 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 0
+  %17 = load i32, i32* %16, align 4
+  %18 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 1
+  %19 = load i32, i32* %18, align 4
+  %20 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 2
+  %21 = load i32, i32* %20, align 4
+  call void %11(%struct.array* sret(%struct.array) align 4 %7, %struct.array* noundef byval(%struct.array) align 4 %1, i32 %13, i32 %15, i32 %17, i32 %19, i32 %21)
+  call void @demo(%struct.array* sret(%struct.array) align 4 %8, %struct.array* noundef byval(%struct.array) align 4 %7)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @callback1(void (%struct.point*, %struct.array*, i32, i32, i32, i32, i32)* noundef %0, %struct.array* noundef byval(%struct.array) align 4 %1) #0 {
+  %3 = alloca void (%struct.point*, %struct.array*, i32, i32, i32, i32, i32)*, align 4
+  %4 = alloca %struct.point, align 4
+  %5 = alloca %struct.point1, align 4
+  %6 = alloca %struct.point, align 4
+  store void (%struct.point*, %struct.array*, i32, i32, i32, i32, i32)* %0, void (%struct.point*, %struct.array*, i32, i32, i32, i32, i32)** %3, align 4
+  %7 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %8 = getelementptr inbounds [8 x i32], [8 x i32]* %7, i32 0, i32 0
+  %9 = load i32, i32* %8, align 4
+  %10 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %11 = getelementptr inbounds [8 x i32], [8 x i32]* %10, i32 0, i32 1
+  %12 = load i32, i32* %11, align 4
+  %13 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %14 = getelementptr inbounds [8 x i32], [8 x i32]* %13, i32 0, i32 7
+  %15 = load i32, i32* %14, align 4
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.26, i32 0, i32 0), i32 noundef %9, i32 noundef %12, i32 noundef %15)
+  %17 = bitcast %struct.point* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %17, i8* align 4 bitcast (%struct.point* @__const.callback1.pt to i8*), i32 8, i1 false)
+  %18 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %18, i8* align 4 bitcast (%struct.point1* @__const.callback1.pt1 to i8*), i32 12, i1 false)
+  %19 = load void (%struct.point*, %struct.array*, i32, i32, i32, i32, i32)*, void (%struct.point*, %struct.array*, i32, i32, i32, i32, i32)** %3, align 4
+  %20 = getelementptr inbounds %struct.point, %struct.point* %4, i32 0, i32 0
+  %21 = load i32, i32* %20, align 4
+  %22 = getelementptr inbounds %struct.point, %struct.point* %4, i32 0, i32 1
+  %23 = load i32, i32* %22, align 4
+  %24 = getelementptr inbounds %struct.point1, %struct.point1* %5, i32 0, i32 0
+  %25 = load i32, i32* %24, align 4
+  %26 = getelementptr inbounds %struct.point1, %struct.point1* %5, i32 0, i32 1
+  %27 = load i32, i32* %26, align 4
+  %28 = getelementptr inbounds %struct.point1, %struct.point1* %5, i32 0, i32 2
+  %29 = load i32, i32* %28, align 4
+  call void %19(%struct.point* sret(%struct.point) align 4 %6, %struct.array* noundef byval(%struct.array) align 4 %1, i32 %21, i32 %23, i32 %25, i32 %27, i32 %29)
+  %30 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %31 = load i32, i32* %30, align 4
+  %32 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %33 = load i32, i32* %32, align 4
+  %34 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.27, i32 0, i32 0), i32 noundef %31, i32 noundef %33)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone uwtable
+define dso_local void @mycallback(%struct.point* noalias sret(%struct.point) align 4 %0, %struct.array* noundef byval(%struct.array) align 4 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6) #0 {
+  %8 = alloca i8*, align 4
+  %9 = alloca %struct.point, align 4
+  %10 = alloca %struct.point1, align 4
+  %11 = bitcast %struct.point* %0 to i8*
+  store i8* %11, i8** %8, align 4
+  %12 = getelementptr inbounds %struct.point, %struct.point* %9, i32 0, i32 0
+  store i32 %2, i32* %12, align 4
+  %13 = getelementptr inbounds %struct.point, %struct.point* %9, i32 0, i32 1
+  store i32 %3, i32* %13, align 4
+  %14 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 0
+  store i32 %4, i32* %14, align 4
+  %15 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 1
+  store i32 %5, i32* %15, align 4
+  %16 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 2
+  store i32 %6, i32* %16, align 4
+  %17 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %18 = getelementptr inbounds [8 x i32], [8 x i32]* %17, i32 0, i32 0
+  %19 = load i32, i32* %18, align 4
+  %20 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %21 = getelementptr inbounds [8 x i32], [8 x i32]* %20, i32 0, i32 1
+  %22 = load i32, i32* %21, align 4
+  %23 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %24 = getelementptr inbounds [8 x i32], [8 x i32]* %23, i32 0, i32 7
+  %25 = load i32, i32* %24, align 4
+  %26 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([28 x i8], [28 x i8]* @.str.28, i32 0, i32 0), i32 noundef %19, i32 noundef %22, i32 noundef %25)
+  %27 = getelementptr inbounds %struct.point, %struct.point* %9, i32 0, i32 0
+  %28 = load i32, i32* %27, align 4
+  %29 = getelementptr inbounds %struct.point, %struct.point* %9, i32 0, i32 1
+  %30 = load i32, i32* %29, align 4
+  %31 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.29, i32 0, i32 0), i32 noundef %28, i32 noundef %30)
+  %32 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 0
+  %33 = load i32, i32* %32, align 4
+  %34 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 1
+  %35 = load i32, i32* %34, align 4
+  %36 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 2
+  %37 = load i32, i32* %36, align 4
+  %38 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([26 x i8], [26 x i8]* @.str.30, i32 0, i32 0), i32 noundef %33, i32 noundef %35, i32 noundef %37)
+  %39 = getelementptr inbounds %struct.point, %struct.point* %0, i32 0, i32 0
+  %40 = getelementptr inbounds %struct.point, %struct.point* %9, i32 0, i32 0
+  %41 = load i32, i32* %40, align 4
+  %42 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 0
+  %43 = load i32, i32* %42, align 4
+  %44 = add nsw i32 %41, %43
+  store i32 %44, i32* %39, align 4
+  %45 = getelementptr inbounds %struct.point, %struct.point* %0, i32 0, i32 1
+  %46 = getelementptr inbounds %struct.point, %struct.point* %9, i32 0, i32 1
+  %47 = load i32, i32* %46, align 4
+  %48 = getelementptr inbounds %struct.point1, %struct.point1* %10, i32 0, i32 1
+  %49 = load i32, i32* %48, align 4
+  %50 = add nsw i32 %47, %49
+  store i32 %50, i32* %45, align 4
+  ret void
+}
 
 attributes #0 = { noinline nounwind optnone uwtable "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="pentium4" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
-attributes #1 = { argmemonly nofree nounwind willreturn }
+attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="pentium4" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" }
+attributes #2 = { argmemonly nofree nounwind willreturn }
 
 !llvm.module.flags = !{!0, !1, !2, !3, !4, !5}
 !llvm.ident = !{!6}
@@ -37,3 +853,5 @@ attributes #1 = { argmemonly nofree nounwind willreturn }
 !4 = !{i32 7, !"uwtable", i32 2}
 !5 = !{i32 7, !"frame-pointer", i32 2}
 !6 = !{!"Apple clang version 14.0.3 (clang-1403.0.22.14.1)"}
+!7 = distinct !{!7, !8}
+!8 = !{!"llvm.loop.mustprogress"}
diff --git a/internal/cabi/_testdata/arch/riscv64/demo.ll b/internal/cabi/_testdata/arch/riscv64/demo.ll
index c022c427..b1baae61 100644
--- a/internal/cabi/_testdata/arch/riscv64/demo.ll
+++ b/internal/cabi/_testdata/arch/riscv64/demo.ll
@@ -3,27 +3,916 @@ source_filename = "../../wrap/demo.c"
 target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n64-S128"
 target triple = "riscv64-unknown-unknown-elf"
 
-%struct.st1 = type { i32, i32 }
+%struct.point = type { i32, i32 }
+%struct.point1 = type { i32, i32, i32 }
+%struct.point64 = type { i64, i64 }
+%struct.struct32 = type { i32 }
+%struct.point2 = type { i8, i32, i32 }
+%struct.point3 = type { i8, i8, i8 }
+%struct.point4 = type { i8, i8, i8, i32 }
+%struct.point5 = type { i8, i8, i8, i8, i8 }
+%struct.point6 = type { i8, i8, i8, i8, i8, i32 }
+%struct.point7 = type { i8, i8, i8, i8, i8, i32, i8 }
+%struct.data1 = type { i8, i64 }
+%struct.data2 = type { i32, i64 }
+%struct.data3 = type { i64, i8 }
+%struct.fdata1 = type { float }
+%struct.ddata1 = type { double }
+%struct.ddata2 = type { double, double }
+%struct.ddata3 = type { double, double, double }
+%struct.fdata2i = type { float, i32 }
+%struct.fdata2 = type { float, float }
+%struct.fdata3 = type { float, float, float }
+%struct.fdata4 = type { float, float, float, float }
+%struct.fdata5 = type { float, float, float, float, float }
+%struct.fdata2id = type { i8, i8, double }
+%struct.fdata7if = type { [7 x i8], float }
+%struct.fdata4if = type { float, i8, float, float }
+%struct.array = type { [8 x i32] }
+
+@.str = private unnamed_addr constant [20 x i8] c"point64: %lld %lld\0A\00", align 1
+@.str.1 = private unnamed_addr constant [14 x i8] c"struct32: %d\0A\00", align 1
+@.str.2 = private unnamed_addr constant [14 x i8] c"point: %d %d\0A\00", align 1
+@.str.3 = private unnamed_addr constant [18 x i8] c"point1: %d %d %d\0A\00", align 1
+@.str.4 = private unnamed_addr constant [18 x i8] c"point2: %d %d %d\0A\00", align 1
+@.str.5 = private unnamed_addr constant [18 x i8] c"point3: %d %d %d\0A\00", align 1
+@.str.6 = private unnamed_addr constant [21 x i8] c"point4: %d %d %d %d\0A\00", align 1
+@.str.7 = private unnamed_addr constant [24 x i8] c"point5: %d %d %d %d %d\0A\00", align 1
+@.str.8 = private unnamed_addr constant [27 x i8] c"point6: %d %d %d %d %d %d\0A\00", align 1
+@.str.9 = private unnamed_addr constant [30 x i8] c"point7: %d %d %d %d %d %d %d\0A\00", align 1
+@.str.10 = private unnamed_addr constant [16 x i8] c"data1: %d %lld\0A\00", align 1
+@.str.11 = private unnamed_addr constant [16 x i8] c"data2: %d %lld\0A\00", align 1
+@.str.12 = private unnamed_addr constant [16 x i8] c"data3: %lld %d\0A\00", align 1
+@.str.13 = private unnamed_addr constant [9 x i8] c"ff1: %f\0A\00", align 1
+@.str.14 = private unnamed_addr constant [9 x i8] c"dd1: %f\0A\00", align 1
+@.str.15 = private unnamed_addr constant [12 x i8] c"dd2: %f %f\0A\00", align 1
+@.str.16 = private unnamed_addr constant [15 x i8] c"dd3: %f %f %f\0A\00", align 1
+@.str.17 = private unnamed_addr constant [13 x i8] c"ff2i: %f %d\0A\00", align 1
+@.str.18 = private unnamed_addr constant [12 x i8] c"ff2: %f %f\0A\00", align 1
+@.str.19 = private unnamed_addr constant [15 x i8] c"ff3: %f %f %f\0A\00", align 1
+@.str.20 = private unnamed_addr constant [18 x i8] c"ff4: %f %f %f %f\0A\00", align 1
+@.str.21 = private unnamed_addr constant [21 x i8] c"ff5: %f %f %f %f %f\0A\00", align 1
+@.str.22 = private unnamed_addr constant [15 x i8] c"ff6: %d %d %f\0A\00", align 1
+@.str.23 = private unnamed_addr constant [17 x i8] c"ff7if: %d %d %f\0A\00", align 1
+@.str.24 = private unnamed_addr constant [20 x i8] c"ff4if: %f %d %f %f\0A\00", align 1
+@.str.25 = private unnamed_addr constant [16 x i8] c"demo: %d %d %d\0A\00", align 1
+@__const.callback.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.26 = private unnamed_addr constant [27 x i8] c"callback1 array: %d %d %d\0A\00", align 1
+@__const.callback1.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback1.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.27 = private unnamed_addr constant [22 x i8] c"callback1 ret: %d,%d\0A\00", align 1
+@.str.28 = private unnamed_addr constant [28 x i8] c"mycallback array: %d %d %d\0A\00", align 1
+@.str.29 = private unnamed_addr constant [22 x i8] c"mycallback pt: %d %d\0A\00", align 1
+@.str.30 = private unnamed_addr constant [26 x i8] c"mycallback pt1: %d %d %d\0A\00", align 1
 
 ; Function Attrs: noinline nounwind optnone
-define dso_local i64 @fn1(i64 %0) #0 {
-  %2 = alloca %struct.st1, align 4
-  %3 = alloca %struct.st1, align 4
-  %4 = bitcast %struct.st1* %3 to i64*
-  store i64 %0, i64* %4, align 4
-  %5 = bitcast %struct.st1* %2 to i8*
-  %6 = bitcast %struct.st1* %3 to i8*
-  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %5, i8* align 4 %6, i64 8, i1 false)
-  %7 = bitcast %struct.st1* %2 to i64*
-  %8 = load i64, i64* %7, align 4
-  ret i64 %8
+define dso_local signext i32 @demo32(i32 noundef signext %0) #0 {
+  %2 = alloca i32, align 4
+  store i32 %0, i32* %2, align 4
+  %3 = load i32, i32* %2, align 4
+  %4 = add nsw i32 %3, 100
+  ret i32 %4
 }
 
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @demo64(i64 noundef %0) #0 {
+  %2 = alloca i64, align 8
+  store i64 %0, i64* %2, align 8
+  %3 = load i64, i64* %2, align 8
+  %4 = add nsw i64 %3, 100
+  ret i64 %4
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @pt64([2 x i64] %0) #0 {
+  %2 = alloca %struct.point64, align 8
+  %3 = alloca %struct.point64, align 8
+  %4 = bitcast %struct.point64* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.point64, %struct.point64* %3, i32 0, i32 0
+  %6 = load i64, i64* %5, align 8
+  %7 = getelementptr inbounds %struct.point64, %struct.point64* %3, i32 0, i32 1
+  %8 = load i64, i64* %7, align 8
+  %9 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str, i64 0, i64 0), i64 noundef %6, i64 noundef %8)
+  %10 = bitcast %struct.point64* %2 to i8*
+  %11 = bitcast %struct.point64* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %10, i8* align 8 %11, i64 16, i1 false)
+  %12 = bitcast %struct.point64* %2 to [2 x i64]*
+  %13 = load [2 x i64], [2 x i64]* %12, align 8
+  ret [2 x i64] %13
+}
+
+declare dso_local signext i32 @printf(i8* noundef, ...) #1
+
 ; Function Attrs: argmemonly nofree nounwind willreturn
-declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #1
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #2
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @demo32s(i64 %0) #0 {
+  %2 = alloca %struct.struct32, align 4
+  %3 = alloca %struct.struct32, align 4
+  %4 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %5 = trunc i64 %0 to i32
+  store i32 %5, i32* %4, align 4
+  %6 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %7 = load i32, i32* %6, align 4
+  %8 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.1, i64 0, i64 0), i32 noundef signext %7)
+  %9 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %10 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %11 = load i32, i32* %10, align 4
+  %12 = add nsw i32 %11, 100
+  store i32 %12, i32* %9, align 4
+  %13 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %14 = load i32, i32* %13, align 4
+  %15 = zext i32 %14 to i64
+  ret i64 %15
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @pt(i64 %0) #0 {
+  %2 = alloca %struct.point, align 4
+  %3 = alloca %struct.point, align 4
+  %4 = bitcast %struct.point* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 4
+  %7 = getelementptr inbounds %struct.point, %struct.point* %3, i32 0, i32 1
+  %8 = load i32, i32* %7, align 4
+  %9 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.2, i64 0, i64 0), i32 noundef signext %6, i32 noundef signext %8)
+  %10 = bitcast %struct.point* %2 to i8*
+  %11 = bitcast %struct.point* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %10, i8* align 4 %11, i64 8, i1 false)
+  %12 = bitcast %struct.point* %2 to i64*
+  %13 = load i64, i64* %12, align 4
+  ret i64 %13
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @pt1([2 x i64] %0) #0 {
+  %2 = alloca %struct.point1, align 4
+  %3 = alloca %struct.point1, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.point1* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 0
+  %9 = load i32, i32* %8, align 4
+  %10 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 1
+  %11 = load i32, i32* %10, align 4
+  %12 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 2
+  %13 = load i32, i32* %12, align 4
+  %14 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.3, i64 0, i64 0), i32 noundef signext %9, i32 noundef signext %11, i32 noundef signext %13)
+  %15 = bitcast %struct.point1* %2 to i8*
+  %16 = bitcast %struct.point1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %15, i8* align 4 %16, i64 12, i1 false)
+  %17 = bitcast [2 x i64]* %5 to i8*
+  %18 = bitcast %struct.point1* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %17, i8* align 4 %18, i64 12, i1 false)
+  %19 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %19
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @pt2([2 x i64] %0) #0 {
+  %2 = alloca %struct.point2, align 4
+  %3 = alloca %struct.point2, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.point2* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 4
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 1
+  %12 = load i32, i32* %11, align 4
+  %13 = getelementptr inbounds %struct.point2, %struct.point2* %3, i32 0, i32 2
+  %14 = load i32, i32* %13, align 4
+  %15 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.4, i64 0, i64 0), i32 noundef signext %10, i32 noundef signext %12, i32 noundef signext %14)
+  %16 = bitcast %struct.point2* %2 to i8*
+  %17 = bitcast %struct.point2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %16, i8* align 4 %17, i64 12, i1 false)
+  %18 = bitcast [2 x i64]* %5 to i8*
+  %19 = bitcast %struct.point2* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %18, i8* align 4 %19, i64 12, i1 false)
+  %20 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %20
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @pt3(i64 %0) #0 {
+  %2 = alloca %struct.point3, align 1
+  %3 = alloca %struct.point3, align 1
+  %4 = alloca i64, align 8
+  %5 = alloca i64, align 8
+  store i64 %0, i64* %4, align 8
+  %6 = bitcast %struct.point3* %3 to i8*
+  %7 = bitcast i64* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %6, i8* align 8 %7, i64 3, i1 false)
+  %8 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point3, %struct.point3* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.5, i64 0, i64 0), i32 noundef signext %10, i32 noundef signext %13, i32 noundef signext %16)
+  %18 = bitcast %struct.point3* %2 to i8*
+  %19 = bitcast %struct.point3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %18, i8* align 1 %19, i64 3, i1 false)
+  %20 = bitcast i64* %5 to i8*
+  %21 = bitcast %struct.point3* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %20, i8* align 1 %21, i64 3, i1 false)
+  %22 = load i64, i64* %5, align 8
+  ret i64 %22
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @pt4(i64 %0) #0 {
+  %2 = alloca %struct.point4, align 4
+  %3 = alloca %struct.point4, align 4
+  %4 = bitcast %struct.point4* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point4, %struct.point4* %3, i32 0, i32 3
+  %15 = load i32, i32* %14, align 4
+  %16 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.6, i64 0, i64 0), i32 noundef signext %7, i32 noundef signext %10, i32 noundef signext %13, i32 noundef signext %15)
+  %17 = bitcast %struct.point4* %2 to i8*
+  %18 = bitcast %struct.point4* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %17, i8* align 4 %18, i64 8, i1 false)
+  %19 = bitcast %struct.point4* %2 to i64*
+  %20 = load i64, i64* %19, align 4
+  ret i64 %20
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @pt5(i64 %0) #0 {
+  %2 = alloca %struct.point5, align 1
+  %3 = alloca %struct.point5, align 1
+  %4 = alloca i64, align 8
+  %5 = alloca i64, align 8
+  store i64 %0, i64* %4, align 8
+  %6 = bitcast %struct.point5* %3 to i8*
+  %7 = bitcast i64* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %6, i8* align 8 %7, i64 5, i1 false)
+  %8 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 3
+  %18 = load i8, i8* %17, align 1
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point5, %struct.point5* %3, i32 0, i32 4
+  %21 = load i8, i8* %20, align 1
+  %22 = zext i8 %21 to i32
+  %23 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([24 x i8], [24 x i8]* @.str.7, i64 0, i64 0), i32 noundef signext %10, i32 noundef signext %13, i32 noundef signext %16, i32 noundef signext %19, i32 noundef signext %22)
+  %24 = bitcast %struct.point5* %2 to i8*
+  %25 = bitcast %struct.point5* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 1 %24, i8* align 1 %25, i64 5, i1 false)
+  %26 = bitcast i64* %5 to i8*
+  %27 = bitcast %struct.point5* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %26, i8* align 1 %27, i64 5, i1 false)
+  %28 = load i64, i64* %5, align 8
+  ret i64 %28
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @pt6([2 x i64] %0) #0 {
+  %2 = alloca %struct.point6, align 4
+  %3 = alloca %struct.point6, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.point6* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 0
+  %9 = load i8, i8* %8, align 4
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 1
+  %12 = load i8, i8* %11, align 1
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 2
+  %15 = load i8, i8* %14, align 2
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 3
+  %18 = load i8, i8* %17, align 1
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 4
+  %21 = load i8, i8* %20, align 4
+  %22 = zext i8 %21 to i32
+  %23 = getelementptr inbounds %struct.point6, %struct.point6* %3, i32 0, i32 5
+  %24 = load i32, i32* %23, align 4
+  %25 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.8, i64 0, i64 0), i32 noundef signext %10, i32 noundef signext %13, i32 noundef signext %16, i32 noundef signext %19, i32 noundef signext %22, i32 noundef signext %24)
+  %26 = bitcast %struct.point6* %2 to i8*
+  %27 = bitcast %struct.point6* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %26, i8* align 4 %27, i64 12, i1 false)
+  %28 = bitcast [2 x i64]* %5 to i8*
+  %29 = bitcast %struct.point6* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %28, i8* align 4 %29, i64 12, i1 false)
+  %30 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %30
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @pt7([2 x i64] %0) #0 {
+  %2 = alloca %struct.point7, align 4
+  %3 = alloca %struct.point7, align 4
+  %4 = bitcast %struct.point7* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 4
+  %5 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 4
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 2
+  %12 = load i8, i8* %11, align 2
+  %13 = zext i8 %12 to i32
+  %14 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 3
+  %15 = load i8, i8* %14, align 1
+  %16 = zext i8 %15 to i32
+  %17 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 4
+  %18 = load i8, i8* %17, align 4
+  %19 = zext i8 %18 to i32
+  %20 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 5
+  %21 = load i32, i32* %20, align 4
+  %22 = getelementptr inbounds %struct.point7, %struct.point7* %3, i32 0, i32 6
+  %23 = load i8, i8* %22, align 4
+  %24 = zext i8 %23 to i32
+  %25 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([30 x i8], [30 x i8]* @.str.9, i64 0, i64 0), i32 noundef signext %7, i32 noundef signext %10, i32 noundef signext %13, i32 noundef signext %16, i32 noundef signext %19, i32 noundef signext %21, i32 noundef signext %24)
+  %26 = bitcast %struct.point7* %2 to i8*
+  %27 = bitcast %struct.point7* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %26, i8* align 4 %27, i64 16, i1 false)
+  %28 = bitcast %struct.point7* %2 to [2 x i64]*
+  %29 = load [2 x i64], [2 x i64]* %28, align 4
+  ret [2 x i64] %29
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @fn1([2 x i64] %0) #0 {
+  %2 = alloca %struct.data1, align 8
+  %3 = alloca %struct.data1, align 8
+  %4 = bitcast %struct.data1* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data1, %struct.data1* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 8
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.data1, %struct.data1* %3, i32 0, i32 1
+  %9 = load i64, i64* %8, align 8
+  %10 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.10, i64 0, i64 0), i32 noundef signext %7, i64 noundef %9)
+  %11 = bitcast %struct.data1* %2 to i8*
+  %12 = bitcast %struct.data1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %11, i8* align 8 %12, i64 16, i1 false)
+  %13 = bitcast %struct.data1* %2 to [2 x i64]*
+  %14 = load [2 x i64], [2 x i64]* %13, align 8
+  ret [2 x i64] %14
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @fn2([2 x i64] %0) #0 {
+  %2 = alloca %struct.data2, align 8
+  %3 = alloca %struct.data2, align 8
+  %4 = bitcast %struct.data2* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data2, %struct.data2* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 8
+  %7 = getelementptr inbounds %struct.data2, %struct.data2* %3, i32 0, i32 1
+  %8 = load i64, i64* %7, align 8
+  %9 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.11, i64 0, i64 0), i32 noundef signext %6, i64 noundef %8)
+  %10 = bitcast %struct.data2* %2 to i8*
+  %11 = bitcast %struct.data2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %10, i8* align 8 %11, i64 16, i1 false)
+  %12 = bitcast %struct.data2* %2 to [2 x i64]*
+  %13 = load [2 x i64], [2 x i64]* %12, align 8
+  ret [2 x i64] %13
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @fn3([2 x i64] %0) #0 {
+  %2 = alloca %struct.data3, align 8
+  %3 = alloca %struct.data3, align 8
+  %4 = bitcast %struct.data3* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.data3, %struct.data3* %3, i32 0, i32 0
+  %6 = load i64, i64* %5, align 8
+  %7 = getelementptr inbounds %struct.data3, %struct.data3* %3, i32 0, i32 1
+  %8 = load i8, i8* %7, align 8
+  %9 = zext i8 %8 to i32
+  %10 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.12, i64 0, i64 0), i64 noundef %6, i32 noundef signext %9)
+  %11 = bitcast %struct.data3* %2 to i8*
+  %12 = bitcast %struct.data3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %11, i8* align 8 %12, i64 16, i1 false)
+  %13 = bitcast %struct.data3* %2 to [2 x i64]*
+  %14 = load [2 x i64], [2 x i64]* %13, align 8
+  ret [2 x i64] %14
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @ff1(i64 %0) #0 {
+  %2 = alloca %struct.fdata1, align 4
+  %3 = alloca %struct.fdata1, align 4
+  %4 = alloca i64, align 8
+  %5 = alloca i64, align 8
+  %6 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  store i64 %0, i64* %4, align 8
+  %7 = bitcast float* %6 to i8*
+  %8 = bitcast i64* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %7, i8* align 8 %8, i64 4, i1 false)
+  %9 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.13, i64 0, i64 0), double noundef %11)
+  %13 = bitcast %struct.fdata1* %2 to i8*
+  %14 = bitcast %struct.fdata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %13, i8* align 4 %14, i64 4, i1 false)
+  %15 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %2, i32 0, i32 0
+  %16 = bitcast i64* %5 to i8*
+  %17 = bitcast float* %15 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %16, i8* align 4 %17, i64 4, i1 false)
+  %18 = load i64, i64* %5, align 8
+  ret i64 %18
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @dd1(i64 %0) #0 {
+  %2 = alloca %struct.ddata1, align 8
+  %3 = alloca %struct.ddata1, align 8
+  %4 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  %5 = bitcast double* %4 to i64*
+  store i64 %0, i64* %5, align 8
+  %6 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  %7 = load double, double* %6, align 8
+  %8 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.14, i64 0, i64 0), double noundef %7)
+  %9 = bitcast %struct.ddata1* %2 to i8*
+  %10 = bitcast %struct.ddata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %9, i8* align 8 %10, i64 8, i1 false)
+  %11 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %2, i32 0, i32 0
+  %12 = bitcast double* %11 to i64*
+  %13 = load i64, i64* %12, align 8
+  ret i64 %13
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @dd2([2 x i64] %0) #0 {
+  %2 = alloca %struct.ddata2, align 8
+  %3 = alloca %struct.ddata2, align 8
+  %4 = bitcast %struct.ddata2* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %3, i32 0, i32 0
+  %6 = load double, double* %5, align 8
+  %7 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %3, i32 0, i32 1
+  %8 = load double, double* %7, align 8
+  %9 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.15, i64 0, i64 0), double noundef %6, double noundef %8)
+  %10 = bitcast %struct.ddata2* %2 to i8*
+  %11 = bitcast %struct.ddata2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %10, i8* align 8 %11, i64 16, i1 false)
+  %12 = bitcast %struct.ddata2* %2 to [2 x i64]*
+  %13 = load [2 x i64], [2 x i64]* %12, align 8
+  ret [2 x i64] %13
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @dd3(%struct.ddata3* noalias sret(%struct.ddata3) align 8 %0, %struct.ddata3* noundef %1) #0 {
+  %3 = alloca i8*, align 8
+  %4 = bitcast %struct.ddata3* %0 to i8*
+  store i8* %4, i8** %3, align 8
+  %5 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 0
+  %6 = load double, double* %5, align 8
+  %7 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 1
+  %8 = load double, double* %7, align 8
+  %9 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 2
+  %10 = load double, double* %9, align 8
+  %11 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.16, i64 0, i64 0), double noundef %6, double noundef %8, double noundef %10)
+  %12 = bitcast %struct.ddata3* %0 to i8*
+  %13 = bitcast %struct.ddata3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %12, i8* align 8 %13, i64 24, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @ff2i(i64 %0) #0 {
+  %2 = alloca %struct.fdata2i, align 4
+  %3 = alloca %struct.fdata2i, align 4
+  %4 = bitcast %struct.fdata2i* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %3, i32 0, i32 1
+  %9 = load i32, i32* %8, align 4
+  %10 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([13 x i8], [13 x i8]* @.str.17, i64 0, i64 0), double noundef %7, i32 noundef signext %9)
+  %11 = bitcast %struct.fdata2i* %2 to i8*
+  %12 = bitcast %struct.fdata2i* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %11, i8* align 4 %12, i64 8, i1 false)
+  %13 = bitcast %struct.fdata2i* %2 to i64*
+  %14 = load i64, i64* %13, align 4
+  ret i64 %14
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @ff2(i64 %0) #0 {
+  %2 = alloca %struct.fdata2, align 4
+  %3 = alloca %struct.fdata2, align 4
+  %4 = bitcast %struct.fdata2* %3 to i64*
+  store i64 %0, i64* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %3, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.18, i64 0, i64 0), double noundef %7, double noundef %10)
+  %12 = bitcast %struct.fdata2* %2 to i8*
+  %13 = bitcast %struct.fdata2* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %12, i8* align 4 %13, i64 8, i1 false)
+  %14 = bitcast %struct.fdata2* %2 to i64*
+  %15 = load i64, i64* %14, align 4
+  ret i64 %15
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @ff3([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata3, align 4
+  %3 = alloca %struct.fdata3, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.fdata3* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 0
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 1
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %3, i32 0, i32 2
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.19, i64 0, i64 0), double noundef %10, double noundef %13, double noundef %16)
+  %18 = bitcast %struct.fdata3* %2 to i8*
+  %19 = bitcast %struct.fdata3* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 12, i1 false)
+  %20 = bitcast [2 x i64]* %5 to i8*
+  %21 = bitcast %struct.fdata3* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %20, i8* align 4 %21, i64 12, i1 false)
+  %22 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %22
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @ff4([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata4, align 4
+  %3 = alloca %struct.fdata4, align 4
+  %4 = bitcast %struct.fdata4* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %3, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.20, i64 0, i64 0), double noundef %7, double noundef %10, double noundef %13, double noundef %16)
+  %18 = bitcast %struct.fdata4* %2 to i8*
+  %19 = bitcast %struct.fdata4* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 16, i1 false)
+  %20 = bitcast %struct.fdata4* %2 to [2 x i64]*
+  %21 = load [2 x i64], [2 x i64]* %20, align 4
+  ret [2 x i64] %21
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @ff5(%struct.fdata5* noalias sret(%struct.fdata5) align 4 %0, %struct.fdata5* noundef %1) #0 {
+  %3 = alloca i8*, align 8
+  %4 = bitcast %struct.fdata5* %0 to i8*
+  store i8* %4, i8** %3, align 8
+  %5 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 1
+  %9 = load float, float* %8, align 4
+  %10 = fpext float %9 to double
+  %11 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 4
+  %18 = load float, float* %17, align 4
+  %19 = fpext float %18 to double
+  %20 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.21, i64 0, i64 0), double noundef %7, double noundef %10, double noundef %13, double noundef %16, double noundef %19)
+  %21 = bitcast %struct.fdata5* %0 to i8*
+  %22 = bitcast %struct.fdata5* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %21, i8* align 4 %22, i64 20, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @ff2id([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata2id, align 8
+  %3 = alloca %struct.fdata2id, align 8
+  %4 = bitcast %struct.fdata2id* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %5 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 0
+  %6 = load i8, i8* %5, align 8
+  %7 = zext i8 %6 to i32
+  %8 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %3, i32 0, i32 2
+  %12 = load double, double* %11, align 8
+  %13 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.22, i64 0, i64 0), i32 noundef signext %7, i32 noundef signext %10, double noundef %12)
+  %14 = bitcast %struct.fdata2id* %2 to i8*
+  %15 = bitcast %struct.fdata2id* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %14, i8* align 8 %15, i64 16, i1 false)
+  %16 = bitcast %struct.fdata2id* %2 to [2 x i64]*
+  %17 = load [2 x i64], [2 x i64]* %16, align 8
+  ret [2 x i64] %17
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @ff7if([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata7if, align 4
+  %3 = alloca %struct.fdata7if, align 4
+  %4 = alloca [2 x i64], align 8
+  %5 = alloca [2 x i64], align 8
+  store [2 x i64] %0, [2 x i64]* %4, align 8
+  %6 = bitcast %struct.fdata7if* %3 to i8*
+  %7 = bitcast [2 x i64]* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %6, i8* align 8 %7, i64 12, i1 false)
+  %8 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 0
+  %9 = getelementptr inbounds [7 x i8], [7 x i8]* %8, i64 0, i64 0
+  %10 = load i8, i8* %9, align 4
+  %11 = zext i8 %10 to i32
+  %12 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 0
+  %13 = getelementptr inbounds [7 x i8], [7 x i8]* %12, i64 0, i64 1
+  %14 = load i8, i8* %13, align 1
+  %15 = zext i8 %14 to i32
+  %16 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %3, i32 0, i32 1
+  %17 = load float, float* %16, align 4
+  %18 = fpext float %17 to double
+  %19 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.23, i64 0, i64 0), i32 noundef signext %11, i32 noundef signext %15, double noundef %18)
+  %20 = bitcast %struct.fdata7if* %2 to i8*
+  %21 = bitcast %struct.fdata7if* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %20, i8* align 4 %21, i64 12, i1 false)
+  %22 = bitcast [2 x i64]* %5 to i8*
+  %23 = bitcast %struct.fdata7if* %2 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %22, i8* align 4 %23, i64 12, i1 false)
+  %24 = load [2 x i64], [2 x i64]* %5, align 8
+  ret [2 x i64] %24
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local [2 x i64] @ff4if([2 x i64] %0) #0 {
+  %2 = alloca %struct.fdata4if, align 4
+  %3 = alloca %struct.fdata4if, align 4
+  %4 = bitcast %struct.fdata4if* %3 to [2 x i64]*
+  store [2 x i64] %0, [2 x i64]* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 1
+  %9 = load i8, i8* %8, align 4
+  %10 = zext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 2
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %3, i32 0, i32 3
+  %15 = load float, float* %14, align 4
+  %16 = fpext float %15 to double
+  %17 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str.24, i64 0, i64 0), double noundef %7, i32 noundef signext %10, double noundef %13, double noundef %16)
+  %18 = bitcast %struct.fdata4if* %2 to i8*
+  %19 = bitcast %struct.fdata4if* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 16, i1 false)
+  %20 = bitcast %struct.fdata4if* %2 to [2 x i64]*
+  %21 = load [2 x i64], [2 x i64]* %20, align 4
+  ret [2 x i64] %21
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @demo(%struct.array* noalias sret(%struct.array) align 4 %0, %struct.array* noundef %1) #0 {
+  %3 = alloca i8*, align 8
+  %4 = bitcast %struct.array* %0 to i8*
+  store i8* %4, i8** %3, align 8
+  %5 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %6 = getelementptr inbounds [8 x i32], [8 x i32]* %5, i64 0, i64 0
+  %7 = load i32, i32* %6, align 4
+  %8 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %9 = getelementptr inbounds [8 x i32], [8 x i32]* %8, i64 0, i64 1
+  %10 = load i32, i32* %9, align 4
+  %11 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i64 0, i64 2
+  %13 = load i32, i32* %12, align 4
+  %14 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.25, i64 0, i64 0), i32 noundef signext %7, i32 noundef signext %10, i32 noundef signext %13)
+  %15 = bitcast %struct.array* %0 to i8*
+  %16 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %15, i8* align 4 %16, i64 32, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @demo2(%struct.array* noalias sret(%struct.array) align 4 %0, i32 noundef signext %1) #0 {
+  %3 = alloca i8*, align 8
+  %4 = alloca i32, align 4
+  %5 = alloca i32, align 4
+  %6 = bitcast %struct.array* %0 to i8*
+  store i8* %6, i8** %3, align 8
+  store i32 %1, i32* %4, align 4
+  store i32 0, i32* %5, align 4
+  br label %7
+
+7:                                                ; preds = %18, %2
+  %8 = load i32, i32* %5, align 4
+  %9 = icmp slt i32 %8, 8
+  br i1 %9, label %10, label %21
+
+10:                                               ; preds = %7
+  %11 = load i32, i32* %5, align 4
+  %12 = load i32, i32* %4, align 4
+  %13 = add nsw i32 %11, %12
+  %14 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %15 = load i32, i32* %5, align 4
+  %16 = sext i32 %15 to i64
+  %17 = getelementptr inbounds [8 x i32], [8 x i32]* %14, i64 0, i64 %16
+  store i32 %13, i32* %17, align 4
+  br label %18
+
+18:                                               ; preds = %10
+  %19 = load i32, i32* %5, align 4
+  %20 = add nsw i32 %19, 1
+  store i32 %20, i32* %5, align 4
+  br label %7, !llvm.loop !5
+
+21:                                               ; preds = %7
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @callback(void (%struct.array*, %struct.array*, i64, [2 x i64])* noundef %0, %struct.array* noundef %1) #0 {
+  %3 = alloca void (%struct.array*, %struct.array*, i64, [2 x i64])*, align 8
+  %4 = alloca %struct.array, align 4
+  %5 = alloca %struct.array, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca %struct.point1, align 4
+  %8 = alloca %struct.array, align 4
+  %9 = alloca %struct.array, align 4
+  %10 = alloca [2 x i64], align 8
+  %11 = alloca %struct.array, align 4
+  %12 = alloca %struct.array, align 4
+  store void (%struct.array*, %struct.array*, i64, [2 x i64])* %0, void (%struct.array*, %struct.array*, i64, [2 x i64])** %3, align 8
+  %13 = bitcast %struct.array* %5 to i8*
+  %14 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %13, i8* align 4 %14, i64 32, i1 false)
+  call void @demo(%struct.array* sret(%struct.array) align 4 %4, %struct.array* noundef %5)
+  %15 = bitcast %struct.point* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %15, i8* align 4 bitcast (%struct.point* @__const.callback.pt to i8*), i64 8, i1 false)
+  %16 = bitcast %struct.point1* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %16, i8* align 4 bitcast (%struct.point1* @__const.callback.pt1 to i8*), i64 12, i1 false)
+  %17 = load void (%struct.array*, %struct.array*, i64, [2 x i64])*, void (%struct.array*, %struct.array*, i64, [2 x i64])** %3, align 8
+  %18 = bitcast %struct.array* %9 to i8*
+  %19 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %18, i8* align 4 %19, i64 32, i1 false)
+  %20 = bitcast %struct.point* %6 to i64*
+  %21 = load i64, i64* %20, align 4
+  %22 = bitcast [2 x i64]* %10 to i8*
+  %23 = bitcast %struct.point1* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %22, i8* align 4 %23, i64 12, i1 false)
+  %24 = load [2 x i64], [2 x i64]* %10, align 8
+  call void %17(%struct.array* sret(%struct.array) align 4 %8, %struct.array* noundef %9, i64 %21, [2 x i64] %24)
+  %25 = bitcast %struct.array* %12 to i8*
+  %26 = bitcast %struct.array* %8 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %25, i8* align 4 %26, i64 32, i1 false)
+  call void @demo(%struct.array* sret(%struct.array) align 4 %11, %struct.array* noundef %12)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local void @callback1(i64 (%struct.array*, i64, [2 x i64])* noundef %0, %struct.array* noundef %1) #0 {
+  %3 = alloca i64 (%struct.array*, i64, [2 x i64])*, align 8
+  %4 = alloca %struct.point, align 4
+  %5 = alloca %struct.point1, align 4
+  %6 = alloca %struct.point, align 4
+  %7 = alloca %struct.array, align 4
+  %8 = alloca [2 x i64], align 8
+  store i64 (%struct.array*, i64, [2 x i64])* %0, i64 (%struct.array*, i64, [2 x i64])** %3, align 8
+  %9 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %10 = getelementptr inbounds [8 x i32], [8 x i32]* %9, i64 0, i64 0
+  %11 = load i32, i32* %10, align 4
+  %12 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %13 = getelementptr inbounds [8 x i32], [8 x i32]* %12, i64 0, i64 1
+  %14 = load i32, i32* %13, align 4
+  %15 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %16 = getelementptr inbounds [8 x i32], [8 x i32]* %15, i64 0, i64 7
+  %17 = load i32, i32* %16, align 4
+  %18 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.26, i64 0, i64 0), i32 noundef signext %11, i32 noundef signext %14, i32 noundef signext %17)
+  %19 = bitcast %struct.point* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %19, i8* align 4 bitcast (%struct.point* @__const.callback1.pt to i8*), i64 8, i1 false)
+  %20 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %20, i8* align 4 bitcast (%struct.point1* @__const.callback1.pt1 to i8*), i64 12, i1 false)
+  %21 = load i64 (%struct.array*, i64, [2 x i64])*, i64 (%struct.array*, i64, [2 x i64])** %3, align 8
+  %22 = bitcast %struct.array* %7 to i8*
+  %23 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %22, i8* align 4 %23, i64 32, i1 false)
+  %24 = bitcast %struct.point* %4 to i64*
+  %25 = load i64, i64* %24, align 4
+  %26 = bitcast [2 x i64]* %8 to i8*
+  %27 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %26, i8* align 4 %27, i64 12, i1 false)
+  %28 = load [2 x i64], [2 x i64]* %8, align 8
+  %29 = call i64 %21(%struct.array* noundef %7, i64 %25, [2 x i64] %28)
+  %30 = bitcast %struct.point* %6 to i64*
+  store i64 %29, i64* %30, align 4
+  %31 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %32 = load i32, i32* %31, align 4
+  %33 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %34 = load i32, i32* %33, align 4
+  %35 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.27, i64 0, i64 0), i32 noundef signext %32, i32 noundef signext %34)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define dso_local i64 @mycallback(%struct.array* noundef %0, i64 %1, [2 x i64] %2) #0 {
+  %4 = alloca %struct.point, align 4
+  %5 = alloca %struct.point, align 4
+  %6 = alloca %struct.point1, align 4
+  %7 = alloca [2 x i64], align 8
+  %8 = bitcast %struct.point* %5 to i64*
+  store i64 %1, i64* %8, align 4
+  store [2 x i64] %2, [2 x i64]* %7, align 8
+  %9 = bitcast %struct.point1* %6 to i8*
+  %10 = bitcast [2 x i64]* %7 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %9, i8* align 8 %10, i64 12, i1 false)
+  %11 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i64 0, i64 0
+  %13 = load i32, i32* %12, align 4
+  %14 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %15 = getelementptr inbounds [8 x i32], [8 x i32]* %14, i64 0, i64 1
+  %16 = load i32, i32* %15, align 4
+  %17 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %18 = getelementptr inbounds [8 x i32], [8 x i32]* %17, i64 0, i64 7
+  %19 = load i32, i32* %18, align 4
+  %20 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([28 x i8], [28 x i8]* @.str.28, i64 0, i64 0), i32 noundef signext %13, i32 noundef signext %16, i32 noundef signext %19)
+  %21 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  %22 = load i32, i32* %21, align 4
+  %23 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  %24 = load i32, i32* %23, align 4
+  %25 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.29, i64 0, i64 0), i32 noundef signext %22, i32 noundef signext %24)
+  %26 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 0
+  %27 = load i32, i32* %26, align 4
+  %28 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 1
+  %29 = load i32, i32* %28, align 4
+  %30 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 2
+  %31 = load i32, i32* %30, align 4
+  %32 = call signext i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([26 x i8], [26 x i8]* @.str.30, i64 0, i64 0), i32 noundef signext %27, i32 noundef signext %29, i32 noundef signext %31)
+  %33 = getelementptr inbounds %struct.point, %struct.point* %4, i32 0, i32 0
+  %34 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 0
+  %35 = load i32, i32* %34, align 4
+  %36 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 0
+  %37 = load i32, i32* %36, align 4
+  %38 = add nsw i32 %35, %37
+  store i32 %38, i32* %33, align 4
+  %39 = getelementptr inbounds %struct.point, %struct.point* %4, i32 0, i32 1
+  %40 = getelementptr inbounds %struct.point, %struct.point* %5, i32 0, i32 1
+  %41 = load i32, i32* %40, align 4
+  %42 = getelementptr inbounds %struct.point1, %struct.point1* %6, i32 0, i32 1
+  %43 = load i32, i32* %42, align 4
+  %44 = add nsw i32 %41, %43
+  store i32 %44, i32* %39, align 4
+  %45 = bitcast %struct.point* %4 to i64*
+  %46 = load i64, i64* %45, align 4
+  ret i64 %46
+}
 
 attributes #0 = { noinline nounwind optnone "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+a,+c,+m,+relax,-save-restore" }
-attributes #1 = { argmemonly nofree nounwind willreturn }
+attributes #1 = { "frame-pointer"="all" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-features"="+64bit,+a,+c,+m,+relax,-save-restore" }
+attributes #2 = { argmemonly nofree nounwind willreturn }
 
 !llvm.module.flags = !{!0, !1, !2, !3}
 !llvm.ident = !{!4}
@@ -33,3 +922,5 @@ attributes #1 = { argmemonly nofree nounwind willreturn }
 !2 = !{i32 7, !"frame-pointer", i32 2}
 !3 = !{i32 1, !"SmallDataLimit", i32 8}
 !4 = !{!"Apple clang version 14.0.3 (clang-1403.0.22.14.1)"}
+!5 = distinct !{!5, !6}
+!6 = !{!"llvm.loop.mustprogress"}
diff --git a/internal/cabi/_testdata/arch/wasm32/demo.ll b/internal/cabi/_testdata/arch/wasm32/demo.ll
index 9826f6c5..f44eaee3 100644
--- a/internal/cabi/_testdata/arch/wasm32/demo.ll
+++ b/internal/cabi/_testdata/arch/wasm32/demo.ll
@@ -3,24 +3,690 @@ source_filename = "../../wrap/demo.c"
 target datalayout = "e-m:e-p:32:32-p10:8:8-p20:8:8-i64:64-f128:64-n32:64-S128-ni:1:10:20"
 target triple = "wasm32-unknown-emscripten"
 
-%struct.st1 = type { i32, i32 }
+%struct.point = type { i32, i32 }
+%struct.point1 = type { i32, i32, i32 }
+%struct.point64 = type { i64, i64 }
+%struct.struct32 = type { i32 }
+%struct.point2 = type { i8, i32, i32 }
+%struct.point3 = type { i8, i8, i8 }
+%struct.point4 = type { i8, i8, i8, i32 }
+%struct.point5 = type { i8, i8, i8, i8, i8 }
+%struct.point6 = type { i8, i8, i8, i8, i8, i32 }
+%struct.point7 = type { i8, i8, i8, i8, i8, i32, i8 }
+%struct.data1 = type { i8, i64 }
+%struct.data2 = type { i32, i64 }
+%struct.data3 = type { i64, i8 }
+%struct.fdata1 = type { float }
+%struct.ddata1 = type { double }
+%struct.ddata2 = type { double, double }
+%struct.ddata3 = type { double, double, double }
+%struct.fdata2i = type { float, i32 }
+%struct.fdata2 = type { float, float }
+%struct.fdata3 = type { float, float, float }
+%struct.fdata4 = type { float, float, float, float }
+%struct.fdata5 = type { float, float, float, float, float }
+%struct.fdata2id = type { i8, i8, double }
+%struct.fdata7if = type { [7 x i8], float }
+%struct.fdata4if = type { float, i8, float, float }
+%struct.array = type { [8 x i32] }
+
+@.str = private unnamed_addr constant [20 x i8] c"point64: %lld %lld\0A\00", align 1
+@.str.1 = private unnamed_addr constant [14 x i8] c"struct32: %d\0A\00", align 1
+@.str.2 = private unnamed_addr constant [14 x i8] c"point: %d %d\0A\00", align 1
+@.str.3 = private unnamed_addr constant [18 x i8] c"point1: %d %d %d\0A\00", align 1
+@.str.4 = private unnamed_addr constant [18 x i8] c"point2: %d %d %d\0A\00", align 1
+@.str.5 = private unnamed_addr constant [18 x i8] c"point3: %d %d %d\0A\00", align 1
+@.str.6 = private unnamed_addr constant [21 x i8] c"point4: %d %d %d %d\0A\00", align 1
+@.str.7 = private unnamed_addr constant [24 x i8] c"point5: %d %d %d %d %d\0A\00", align 1
+@.str.8 = private unnamed_addr constant [27 x i8] c"point6: %d %d %d %d %d %d\0A\00", align 1
+@.str.9 = private unnamed_addr constant [30 x i8] c"point7: %d %d %d %d %d %d %d\0A\00", align 1
+@.str.10 = private unnamed_addr constant [16 x i8] c"data1: %d %lld\0A\00", align 1
+@.str.11 = private unnamed_addr constant [16 x i8] c"data2: %d %lld\0A\00", align 1
+@.str.12 = private unnamed_addr constant [16 x i8] c"data3: %lld %d\0A\00", align 1
+@.str.13 = private unnamed_addr constant [9 x i8] c"ff1: %f\0A\00", align 1
+@.str.14 = private unnamed_addr constant [9 x i8] c"dd1: %f\0A\00", align 1
+@.str.15 = private unnamed_addr constant [12 x i8] c"dd2: %f %f\0A\00", align 1
+@.str.16 = private unnamed_addr constant [15 x i8] c"dd3: %f %f %f\0A\00", align 1
+@.str.17 = private unnamed_addr constant [13 x i8] c"ff2i: %f %d\0A\00", align 1
+@.str.18 = private unnamed_addr constant [12 x i8] c"ff2: %f %f\0A\00", align 1
+@.str.19 = private unnamed_addr constant [15 x i8] c"ff3: %f %f %f\0A\00", align 1
+@.str.20 = private unnamed_addr constant [18 x i8] c"ff4: %f %f %f %f\0A\00", align 1
+@.str.21 = private unnamed_addr constant [21 x i8] c"ff5: %f %f %f %f %f\0A\00", align 1
+@.str.22 = private unnamed_addr constant [15 x i8] c"ff6: %d %d %f\0A\00", align 1
+@.str.23 = private unnamed_addr constant [17 x i8] c"ff7if: %d %d %f\0A\00", align 1
+@.str.24 = private unnamed_addr constant [20 x i8] c"ff4if: %f %d %f %f\0A\00", align 1
+@.str.25 = private unnamed_addr constant [16 x i8] c"demo: %d %d %d\0A\00", align 1
+@__const.callback.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.26 = private unnamed_addr constant [27 x i8] c"callback1 array: %d %d %d\0A\00", align 1
+@__const.callback1.pt = private unnamed_addr constant %struct.point { i32 1, i32 2 }, align 4
+@__const.callback1.pt1 = private unnamed_addr constant %struct.point1 { i32 1, i32 2, i32 3 }, align 4
+@.str.27 = private unnamed_addr constant [22 x i8] c"callback1 ret: %d,%d\0A\00", align 1
+@.str.28 = private unnamed_addr constant [28 x i8] c"mycallback array: %d %d %d\0A\00", align 1
+@.str.29 = private unnamed_addr constant [22 x i8] c"mycallback pt: %d %d\0A\00", align 1
+@.str.30 = private unnamed_addr constant [26 x i8] c"mycallback pt1: %d %d %d\0A\00", align 1
 
 ; Function Attrs: noinline nounwind optnone
-define hidden void @fn1(%struct.st1* noalias sret(%struct.st1) align 4 %0, %struct.st1* noundef byval(%struct.st1) align 4 %1) #0 {
-  %3 = bitcast %struct.st1* %0 to i8*
-  %4 = bitcast %struct.st1* %1 to i8*
-  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %3, i8* align 4 %4, i32 8, i1 false)
+define hidden i32 @demo32(i32 noundef %0) #0 {
+  %2 = alloca i32, align 4
+  store i32 %0, i32* %2, align 4
+  %3 = load i32, i32* %2, align 4
+  %4 = add nsw i32 %3, 100
+  ret i32 %4
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden i64 @demo64(i64 noundef %0) #0 {
+  %2 = alloca i64, align 8
+  store i64 %0, i64* %2, align 8
+  %3 = load i64, i64* %2, align 8
+  %4 = add nsw i64 %3, 100
+  ret i64 %4
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt64(%struct.point64* noalias sret(%struct.point64) align 8 %0, %struct.point64* noundef byval(%struct.point64) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.point64, %struct.point64* %1, i32 0, i32 0
+  %4 = load i64, i64* %3, align 8
+  %5 = getelementptr inbounds %struct.point64, %struct.point64* %1, i32 0, i32 1
+  %6 = load i64, i64* %5, align 8
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str, i32 0, i32 0), i64 noundef %4, i64 noundef %6)
+  %8 = bitcast %struct.point64* %0 to i8*
+  %9 = bitcast %struct.point64* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %8, i8* align 8 %9, i32 16, i1 false)
   ret void
 }
 
+declare i32 @printf(i8* noundef, ...) #1
+
 ; Function Attrs: argmemonly nofree nounwind willreturn
-declare void @llvm.memcpy.p0i8.p0i8.i32(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i32, i1 immarg) #1
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i32, i1 immarg) #2
+
+; Function Attrs: noinline nounwind optnone
+define hidden i32 @demo32s(i32 %0) #0 {
+  %2 = alloca %struct.struct32, align 4
+  %3 = alloca %struct.struct32, align 4
+  %4 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  store i32 %0, i32* %4, align 4
+  %5 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %6 = load i32, i32* %5, align 4
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.1, i32 0, i32 0), i32 noundef %6)
+  %8 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %9 = getelementptr inbounds %struct.struct32, %struct.struct32* %3, i32 0, i32 0
+  %10 = load i32, i32* %9, align 4
+  %11 = add nsw i32 %10, 100
+  store i32 %11, i32* %8, align 4
+  %12 = getelementptr inbounds %struct.struct32, %struct.struct32* %2, i32 0, i32 0
+  %13 = load i32, i32* %12, align 4
+  ret i32 %13
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt(%struct.point* noalias sret(%struct.point) align 4 %0, %struct.point* noundef byval(%struct.point) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.point, %struct.point* %1, i32 0, i32 0
+  %4 = load i32, i32* %3, align 4
+  %5 = getelementptr inbounds %struct.point, %struct.point* %1, i32 0, i32 1
+  %6 = load i32, i32* %5, align 4
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([14 x i8], [14 x i8]* @.str.2, i32 0, i32 0), i32 noundef %4, i32 noundef %6)
+  %8 = bitcast %struct.point* %0 to i8*
+  %9 = bitcast %struct.point* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %8, i8* align 4 %9, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt1(%struct.point1* noalias sret(%struct.point1) align 4 %0, %struct.point1* noundef byval(%struct.point1) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.point1, %struct.point1* %1, i32 0, i32 0
+  %4 = load i32, i32* %3, align 4
+  %5 = getelementptr inbounds %struct.point1, %struct.point1* %1, i32 0, i32 1
+  %6 = load i32, i32* %5, align 4
+  %7 = getelementptr inbounds %struct.point1, %struct.point1* %1, i32 0, i32 2
+  %8 = load i32, i32* %7, align 4
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.3, i32 0, i32 0), i32 noundef %4, i32 noundef %6, i32 noundef %8)
+  %10 = bitcast %struct.point1* %0 to i8*
+  %11 = bitcast %struct.point1* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %10, i8* align 4 %11, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt2(%struct.point2* noalias sret(%struct.point2) align 4 %0, %struct.point2* noundef byval(%struct.point2) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.point2, %struct.point2* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 4
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.point2, %struct.point2* %1, i32 0, i32 1
+  %7 = load i32, i32* %6, align 4
+  %8 = getelementptr inbounds %struct.point2, %struct.point2* %1, i32 0, i32 2
+  %9 = load i32, i32* %8, align 4
+  %10 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.4, i32 0, i32 0), i32 noundef %5, i32 noundef %7, i32 noundef %9)
+  %11 = bitcast %struct.point2* %0 to i8*
+  %12 = bitcast %struct.point2* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %11, i8* align 4 %12, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt3(%struct.point3* noalias sret(%struct.point3) align 1 %0, %struct.point3* noundef byval(%struct.point3) align 1 %1) #0 {
+  %3 = getelementptr inbounds %struct.point3, %struct.point3* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 1
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.point3, %struct.point3* %1, i32 0, i32 1
+  %7 = load i8, i8* %6, align 1
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.point3, %struct.point3* %1, i32 0, i32 2
+  %10 = load i8, i8* %9, align 1
+  %11 = sext i8 %10 to i32
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.5, i32 0, i32 0), i32 noundef %5, i32 noundef %8, i32 noundef %11)
+  %13 = bitcast %struct.point3* %0 to i8*
+  %14 = bitcast %struct.point3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %13, i8* align 1 %14, i32 3, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt4(%struct.point4* noalias sret(%struct.point4) align 4 %0, %struct.point4* noundef byval(%struct.point4) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 4
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 1
+  %7 = load i8, i8* %6, align 1
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 2
+  %10 = load i8, i8* %9, align 2
+  %11 = sext i8 %10 to i32
+  %12 = getelementptr inbounds %struct.point4, %struct.point4* %1, i32 0, i32 3
+  %13 = load i32, i32* %12, align 4
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.6, i32 0, i32 0), i32 noundef %5, i32 noundef %8, i32 noundef %11, i32 noundef %13)
+  %15 = bitcast %struct.point4* %0 to i8*
+  %16 = bitcast %struct.point4* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %15, i8* align 4 %16, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt5(%struct.point5* noalias sret(%struct.point5) align 1 %0, %struct.point5* noundef byval(%struct.point5) align 1 %1) #0 {
+  %3 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 1
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 1
+  %7 = load i8, i8* %6, align 1
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 2
+  %10 = load i8, i8* %9, align 1
+  %11 = sext i8 %10 to i32
+  %12 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 3
+  %13 = load i8, i8* %12, align 1
+  %14 = sext i8 %13 to i32
+  %15 = getelementptr inbounds %struct.point5, %struct.point5* %1, i32 0, i32 4
+  %16 = load i8, i8* %15, align 1
+  %17 = sext i8 %16 to i32
+  %18 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([24 x i8], [24 x i8]* @.str.7, i32 0, i32 0), i32 noundef %5, i32 noundef %8, i32 noundef %11, i32 noundef %14, i32 noundef %17)
+  %19 = bitcast %struct.point5* %0 to i8*
+  %20 = bitcast %struct.point5* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %19, i8* align 1 %20, i32 5, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt6(%struct.point6* noalias sret(%struct.point6) align 4 %0, %struct.point6* noundef byval(%struct.point6) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 4
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 1
+  %7 = load i8, i8* %6, align 1
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 2
+  %10 = load i8, i8* %9, align 2
+  %11 = sext i8 %10 to i32
+  %12 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 3
+  %13 = load i8, i8* %12, align 1
+  %14 = sext i8 %13 to i32
+  %15 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 4
+  %16 = load i8, i8* %15, align 4
+  %17 = sext i8 %16 to i32
+  %18 = getelementptr inbounds %struct.point6, %struct.point6* %1, i32 0, i32 5
+  %19 = load i32, i32* %18, align 4
+  %20 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.8, i32 0, i32 0), i32 noundef %5, i32 noundef %8, i32 noundef %11, i32 noundef %14, i32 noundef %17, i32 noundef %19)
+  %21 = bitcast %struct.point6* %0 to i8*
+  %22 = bitcast %struct.point6* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %21, i8* align 4 %22, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @pt7(%struct.point7* noalias sret(%struct.point7) align 4 %0, %struct.point7* noundef byval(%struct.point7) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 4
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 1
+  %7 = load i8, i8* %6, align 1
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 2
+  %10 = load i8, i8* %9, align 2
+  %11 = sext i8 %10 to i32
+  %12 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 3
+  %13 = load i8, i8* %12, align 1
+  %14 = sext i8 %13 to i32
+  %15 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 4
+  %16 = load i8, i8* %15, align 4
+  %17 = sext i8 %16 to i32
+  %18 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 5
+  %19 = load i32, i32* %18, align 4
+  %20 = getelementptr inbounds %struct.point7, %struct.point7* %1, i32 0, i32 6
+  %21 = load i8, i8* %20, align 4
+  %22 = sext i8 %21 to i32
+  %23 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([30 x i8], [30 x i8]* @.str.9, i32 0, i32 0), i32 noundef %5, i32 noundef %8, i32 noundef %11, i32 noundef %14, i32 noundef %17, i32 noundef %19, i32 noundef %22)
+  %24 = bitcast %struct.point7* %0 to i8*
+  %25 = bitcast %struct.point7* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %24, i8* align 4 %25, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @fn1(%struct.data1* noalias sret(%struct.data1) align 8 %0, %struct.data1* noundef byval(%struct.data1) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.data1, %struct.data1* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 8
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.data1, %struct.data1* %1, i32 0, i32 1
+  %7 = load i64, i64* %6, align 8
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.10, i32 0, i32 0), i32 noundef %5, i64 noundef %7)
+  %9 = bitcast %struct.data1* %0 to i8*
+  %10 = bitcast %struct.data1* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %9, i8* align 8 %10, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @fn2(%struct.data2* noalias sret(%struct.data2) align 8 %0, %struct.data2* noundef byval(%struct.data2) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.data2, %struct.data2* %1, i32 0, i32 0
+  %4 = load i32, i32* %3, align 8
+  %5 = getelementptr inbounds %struct.data2, %struct.data2* %1, i32 0, i32 1
+  %6 = load i64, i64* %5, align 8
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.11, i32 0, i32 0), i32 noundef %4, i64 noundef %6)
+  %8 = bitcast %struct.data2* %0 to i8*
+  %9 = bitcast %struct.data2* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %8, i8* align 8 %9, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @fn3(%struct.data3* noalias sret(%struct.data3) align 8 %0, %struct.data3* noundef byval(%struct.data3) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.data3, %struct.data3* %1, i32 0, i32 0
+  %4 = load i64, i64* %3, align 8
+  %5 = getelementptr inbounds %struct.data3, %struct.data3* %1, i32 0, i32 1
+  %6 = load i8, i8* %5, align 8
+  %7 = sext i8 %6 to i32
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.12, i32 0, i32 0), i64 noundef %4, i32 noundef %7)
+  %9 = bitcast %struct.data3* %0 to i8*
+  %10 = bitcast %struct.data3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %9, i8* align 8 %10, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden float @ff1(float %0) #0 {
+  %2 = alloca %struct.fdata1, align 4
+  %3 = alloca %struct.fdata1, align 4
+  %4 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  store float %0, float* %4, align 4
+  %5 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %3, i32 0, i32 0
+  %6 = load float, float* %5, align 4
+  %7 = fpext float %6 to double
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.13, i32 0, i32 0), double noundef %7)
+  %9 = bitcast %struct.fdata1* %2 to i8*
+  %10 = bitcast %struct.fdata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %9, i8* align 4 %10, i32 4, i1 false)
+  %11 = getelementptr inbounds %struct.fdata1, %struct.fdata1* %2, i32 0, i32 0
+  %12 = load float, float* %11, align 4
+  ret float %12
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden double @dd1(double %0) #0 {
+  %2 = alloca %struct.ddata1, align 8
+  %3 = alloca %struct.ddata1, align 8
+  %4 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  store double %0, double* %4, align 8
+  %5 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %3, i32 0, i32 0
+  %6 = load double, double* %5, align 8
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([9 x i8], [9 x i8]* @.str.14, i32 0, i32 0), double noundef %6)
+  %8 = bitcast %struct.ddata1* %2 to i8*
+  %9 = bitcast %struct.ddata1* %3 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %8, i8* align 8 %9, i32 8, i1 false)
+  %10 = getelementptr inbounds %struct.ddata1, %struct.ddata1* %2, i32 0, i32 0
+  %11 = load double, double* %10, align 8
+  ret double %11
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @dd2(%struct.ddata2* noalias sret(%struct.ddata2) align 8 %0, %struct.ddata2* noundef byval(%struct.ddata2) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %1, i32 0, i32 0
+  %4 = load double, double* %3, align 8
+  %5 = getelementptr inbounds %struct.ddata2, %struct.ddata2* %1, i32 0, i32 1
+  %6 = load double, double* %5, align 8
+  %7 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.15, i32 0, i32 0), double noundef %4, double noundef %6)
+  %8 = bitcast %struct.ddata2* %0 to i8*
+  %9 = bitcast %struct.ddata2* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %8, i8* align 8 %9, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @dd3(%struct.ddata3* noalias sret(%struct.ddata3) align 8 %0, %struct.ddata3* noundef byval(%struct.ddata3) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 0
+  %4 = load double, double* %3, align 8
+  %5 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 1
+  %6 = load double, double* %5, align 8
+  %7 = getelementptr inbounds %struct.ddata3, %struct.ddata3* %1, i32 0, i32 2
+  %8 = load double, double* %7, align 8
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.16, i32 0, i32 0), double noundef %4, double noundef %6, double noundef %8)
+  %10 = bitcast %struct.ddata3* %0 to i8*
+  %11 = bitcast %struct.ddata3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %10, i8* align 8 %11, i32 24, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff2i(%struct.fdata2i* noalias sret(%struct.fdata2i) align 4 %0, %struct.fdata2i* noundef byval(%struct.fdata2i) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 4
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata2i, %struct.fdata2i* %1, i32 0, i32 1
+  %7 = load i32, i32* %6, align 4
+  %8 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([13 x i8], [13 x i8]* @.str.17, i32 0, i32 0), double noundef %5, i32 noundef %7)
+  %9 = bitcast %struct.fdata2i* %0 to i8*
+  %10 = bitcast %struct.fdata2i* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %9, i8* align 4 %10, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff2(%struct.fdata2* noalias sret(%struct.fdata2) align 4 %0, %struct.fdata2* noundef byval(%struct.fdata2) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 4
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata2, %struct.fdata2* %1, i32 0, i32 1
+  %7 = load float, float* %6, align 4
+  %8 = fpext float %7 to double
+  %9 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.18, i32 0, i32 0), double noundef %5, double noundef %8)
+  %10 = bitcast %struct.fdata2* %0 to i8*
+  %11 = bitcast %struct.fdata2* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %10, i8* align 4 %11, i32 8, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff3(%struct.fdata3* noalias sret(%struct.fdata3) align 4 %0, %struct.fdata3* noundef byval(%struct.fdata3) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 4
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %1, i32 0, i32 1
+  %7 = load float, float* %6, align 4
+  %8 = fpext float %7 to double
+  %9 = getelementptr inbounds %struct.fdata3, %struct.fdata3* %1, i32 0, i32 2
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.19, i32 0, i32 0), double noundef %5, double noundef %8, double noundef %11)
+  %13 = bitcast %struct.fdata3* %0 to i8*
+  %14 = bitcast %struct.fdata3* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %13, i8* align 4 %14, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff4(%struct.fdata4* noalias sret(%struct.fdata4) align 4 %0, %struct.fdata4* noundef byval(%struct.fdata4) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 4
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %1, i32 0, i32 1
+  %7 = load float, float* %6, align 4
+  %8 = fpext float %7 to double
+  %9 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %1, i32 0, i32 2
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = getelementptr inbounds %struct.fdata4, %struct.fdata4* %1, i32 0, i32 3
+  %13 = load float, float* %12, align 4
+  %14 = fpext float %13 to double
+  %15 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([18 x i8], [18 x i8]* @.str.20, i32 0, i32 0), double noundef %5, double noundef %8, double noundef %11, double noundef %14)
+  %16 = bitcast %struct.fdata4* %0 to i8*
+  %17 = bitcast %struct.fdata4* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %16, i8* align 4 %17, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff5(%struct.fdata5* noalias sret(%struct.fdata5) align 4 %0, %struct.fdata5* noundef byval(%struct.fdata5) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 4
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 1
+  %7 = load float, float* %6, align 4
+  %8 = fpext float %7 to double
+  %9 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 2
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 3
+  %13 = load float, float* %12, align 4
+  %14 = fpext float %13 to double
+  %15 = getelementptr inbounds %struct.fdata5, %struct.fdata5* %1, i32 0, i32 4
+  %16 = load float, float* %15, align 4
+  %17 = fpext float %16 to double
+  %18 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([21 x i8], [21 x i8]* @.str.21, i32 0, i32 0), double noundef %5, double noundef %8, double noundef %11, double noundef %14, double noundef %17)
+  %19 = bitcast %struct.fdata5* %0 to i8*
+  %20 = bitcast %struct.fdata5* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %19, i8* align 4 %20, i32 20, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff2id(%struct.fdata2id* noalias sret(%struct.fdata2id) align 8 %0, %struct.fdata2id* noundef byval(%struct.fdata2id) align 8 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %1, i32 0, i32 0
+  %4 = load i8, i8* %3, align 8
+  %5 = sext i8 %4 to i32
+  %6 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %1, i32 0, i32 1
+  %7 = load i8, i8* %6, align 1
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.fdata2id, %struct.fdata2id* %1, i32 0, i32 2
+  %10 = load double, double* %9, align 8
+  %11 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([15 x i8], [15 x i8]* @.str.22, i32 0, i32 0), i32 noundef %5, i32 noundef %8, double noundef %10)
+  %12 = bitcast %struct.fdata2id* %0 to i8*
+  %13 = bitcast %struct.fdata2id* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 8 %12, i8* align 8 %13, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff7if(%struct.fdata7if* noalias sret(%struct.fdata7if) align 4 %0, %struct.fdata7if* noundef byval(%struct.fdata7if) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %1, i32 0, i32 0
+  %4 = getelementptr inbounds [7 x i8], [7 x i8]* %3, i32 0, i32 0
+  %5 = load i8, i8* %4, align 4
+  %6 = sext i8 %5 to i32
+  %7 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %1, i32 0, i32 0
+  %8 = getelementptr inbounds [7 x i8], [7 x i8]* %7, i32 0, i32 1
+  %9 = load i8, i8* %8, align 1
+  %10 = sext i8 %9 to i32
+  %11 = getelementptr inbounds %struct.fdata7if, %struct.fdata7if* %1, i32 0, i32 1
+  %12 = load float, float* %11, align 4
+  %13 = fpext float %12 to double
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.23, i32 0, i32 0), i32 noundef %6, i32 noundef %10, double noundef %13)
+  %15 = bitcast %struct.fdata7if* %0 to i8*
+  %16 = bitcast %struct.fdata7if* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %15, i8* align 4 %16, i32 12, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @ff4if(%struct.fdata4if* noalias sret(%struct.fdata4if) align 4 %0, %struct.fdata4if* noundef byval(%struct.fdata4if) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 0
+  %4 = load float, float* %3, align 4
+  %5 = fpext float %4 to double
+  %6 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 1
+  %7 = load i8, i8* %6, align 4
+  %8 = sext i8 %7 to i32
+  %9 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 2
+  %10 = load float, float* %9, align 4
+  %11 = fpext float %10 to double
+  %12 = getelementptr inbounds %struct.fdata4if, %struct.fdata4if* %1, i32 0, i32 3
+  %13 = load float, float* %12, align 4
+  %14 = fpext float %13 to double
+  %15 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([20 x i8], [20 x i8]* @.str.24, i32 0, i32 0), double noundef %5, i32 noundef %8, double noundef %11, double noundef %14)
+  %16 = bitcast %struct.fdata4if* %0 to i8*
+  %17 = bitcast %struct.fdata4if* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %16, i8* align 4 %17, i32 16, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @demo(%struct.array* noalias sret(%struct.array) align 4 %0, %struct.array* noundef byval(%struct.array) align 4 %1) #0 {
+  %3 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %4 = getelementptr inbounds [8 x i32], [8 x i32]* %3, i32 0, i32 0
+  %5 = load i32, i32* %4, align 4
+  %6 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %7 = getelementptr inbounds [8 x i32], [8 x i32]* %6, i32 0, i32 1
+  %8 = load i32, i32* %7, align 4
+  %9 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %10 = getelementptr inbounds [8 x i32], [8 x i32]* %9, i32 0, i32 2
+  %11 = load i32, i32* %10, align 4
+  %12 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([16 x i8], [16 x i8]* @.str.25, i32 0, i32 0), i32 noundef %5, i32 noundef %8, i32 noundef %11)
+  %13 = bitcast %struct.array* %0 to i8*
+  %14 = bitcast %struct.array* %1 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %13, i8* align 4 %14, i32 32, i1 false)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @demo2(%struct.array* noalias sret(%struct.array) align 4 %0, i32 noundef %1) #0 {
+  %3 = alloca i32, align 4
+  %4 = alloca i32, align 4
+  store i32 %1, i32* %3, align 4
+  store i32 0, i32* %4, align 4
+  br label %5
+
+5:                                                ; preds = %15, %2
+  %6 = load i32, i32* %4, align 4
+  %7 = icmp slt i32 %6, 8
+  br i1 %7, label %8, label %18
+
+8:                                                ; preds = %5
+  %9 = load i32, i32* %4, align 4
+  %10 = load i32, i32* %3, align 4
+  %11 = add nsw i32 %9, %10
+  %12 = getelementptr inbounds %struct.array, %struct.array* %0, i32 0, i32 0
+  %13 = load i32, i32* %4, align 4
+  %14 = getelementptr inbounds [8 x i32], [8 x i32]* %12, i32 0, i32 %13
+  store i32 %11, i32* %14, align 4
+  br label %15
+
+15:                                               ; preds = %8
+  %16 = load i32, i32* %4, align 4
+  %17 = add nsw i32 %16, 1
+  store i32 %17, i32* %4, align 4
+  br label %5, !llvm.loop !2
+
+18:                                               ; preds = %5
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @callback(void (%struct.array*, %struct.array*, %struct.point*, %struct.point1*)* noundef %0, %struct.array* noundef byval(%struct.array) align 4 %1) #0 {
+  %3 = alloca void (%struct.array*, %struct.array*, %struct.point*, %struct.point1*)*, align 4
+  %4 = alloca %struct.array, align 4
+  %5 = alloca %struct.point, align 4
+  %6 = alloca %struct.point1, align 4
+  %7 = alloca %struct.array, align 4
+  %8 = alloca %struct.array, align 4
+  store void (%struct.array*, %struct.array*, %struct.point*, %struct.point1*)* %0, void (%struct.array*, %struct.array*, %struct.point*, %struct.point1*)** %3, align 4
+  call void @demo(%struct.array* sret(%struct.array) align 4 %4, %struct.array* noundef byval(%struct.array) align 4 %1)
+  %9 = bitcast %struct.point* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %9, i8* align 4 bitcast (%struct.point* @__const.callback.pt to i8*), i32 8, i1 false)
+  %10 = bitcast %struct.point1* %6 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %10, i8* align 4 bitcast (%struct.point1* @__const.callback.pt1 to i8*), i32 12, i1 false)
+  %11 = load void (%struct.array*, %struct.array*, %struct.point*, %struct.point1*)*, void (%struct.array*, %struct.array*, %struct.point*, %struct.point1*)** %3, align 4
+  call void %11(%struct.array* sret(%struct.array) align 4 %7, %struct.array* noundef byval(%struct.array) align 4 %1, %struct.point* noundef byval(%struct.point) align 4 %5, %struct.point1* noundef byval(%struct.point1) align 4 %6)
+  call void @demo(%struct.array* sret(%struct.array) align 4 %8, %struct.array* noundef byval(%struct.array) align 4 %7)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @callback1(void (%struct.point*, %struct.array*, %struct.point*, %struct.point1*)* noundef %0, %struct.array* noundef byval(%struct.array) align 4 %1) #0 {
+  %3 = alloca void (%struct.point*, %struct.array*, %struct.point*, %struct.point1*)*, align 4
+  %4 = alloca %struct.point, align 4
+  %5 = alloca %struct.point1, align 4
+  %6 = alloca %struct.point, align 4
+  store void (%struct.point*, %struct.array*, %struct.point*, %struct.point1*)* %0, void (%struct.point*, %struct.array*, %struct.point*, %struct.point1*)** %3, align 4
+  %7 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %8 = getelementptr inbounds [8 x i32], [8 x i32]* %7, i32 0, i32 0
+  %9 = load i32, i32* %8, align 4
+  %10 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %11 = getelementptr inbounds [8 x i32], [8 x i32]* %10, i32 0, i32 1
+  %12 = load i32, i32* %11, align 4
+  %13 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %14 = getelementptr inbounds [8 x i32], [8 x i32]* %13, i32 0, i32 7
+  %15 = load i32, i32* %14, align 4
+  %16 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([27 x i8], [27 x i8]* @.str.26, i32 0, i32 0), i32 noundef %9, i32 noundef %12, i32 noundef %15)
+  %17 = bitcast %struct.point* %4 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %17, i8* align 4 bitcast (%struct.point* @__const.callback1.pt to i8*), i32 8, i1 false)
+  %18 = bitcast %struct.point1* %5 to i8*
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %18, i8* align 4 bitcast (%struct.point1* @__const.callback1.pt1 to i8*), i32 12, i1 false)
+  %19 = load void (%struct.point*, %struct.array*, %struct.point*, %struct.point1*)*, void (%struct.point*, %struct.array*, %struct.point*, %struct.point1*)** %3, align 4
+  call void %19(%struct.point* sret(%struct.point) align 4 %6, %struct.array* noundef byval(%struct.array) align 4 %1, %struct.point* noundef byval(%struct.point) align 4 %4, %struct.point1* noundef byval(%struct.point1) align 4 %5)
+  %20 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 0
+  %21 = load i32, i32* %20, align 4
+  %22 = getelementptr inbounds %struct.point, %struct.point* %6, i32 0, i32 1
+  %23 = load i32, i32* %22, align 4
+  %24 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.27, i32 0, i32 0), i32 noundef %21, i32 noundef %23)
+  ret void
+}
+
+; Function Attrs: noinline nounwind optnone
+define hidden void @mycallback(%struct.point* noalias sret(%struct.point) align 4 %0, %struct.array* noundef byval(%struct.array) align 4 %1, %struct.point* noundef byval(%struct.point) align 4 %2, %struct.point1* noundef byval(%struct.point1) align 4 %3) #0 {
+  %5 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %6 = getelementptr inbounds [8 x i32], [8 x i32]* %5, i32 0, i32 0
+  %7 = load i32, i32* %6, align 4
+  %8 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %9 = getelementptr inbounds [8 x i32], [8 x i32]* %8, i32 0, i32 1
+  %10 = load i32, i32* %9, align 4
+  %11 = getelementptr inbounds %struct.array, %struct.array* %1, i32 0, i32 0
+  %12 = getelementptr inbounds [8 x i32], [8 x i32]* %11, i32 0, i32 7
+  %13 = load i32, i32* %12, align 4
+  %14 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([28 x i8], [28 x i8]* @.str.28, i32 0, i32 0), i32 noundef %7, i32 noundef %10, i32 noundef %13)
+  %15 = getelementptr inbounds %struct.point, %struct.point* %2, i32 0, i32 0
+  %16 = load i32, i32* %15, align 4
+  %17 = getelementptr inbounds %struct.point, %struct.point* %2, i32 0, i32 1
+  %18 = load i32, i32* %17, align 4
+  %19 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([22 x i8], [22 x i8]* @.str.29, i32 0, i32 0), i32 noundef %16, i32 noundef %18)
+  %20 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 0
+  %21 = load i32, i32* %20, align 4
+  %22 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 1
+  %23 = load i32, i32* %22, align 4
+  %24 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 2
+  %25 = load i32, i32* %24, align 4
+  %26 = call i32 (i8*, ...) @printf(i8* noundef getelementptr inbounds ([26 x i8], [26 x i8]* @.str.30, i32 0, i32 0), i32 noundef %21, i32 noundef %23, i32 noundef %25)
+  %27 = getelementptr inbounds %struct.point, %struct.point* %0, i32 0, i32 0
+  %28 = getelementptr inbounds %struct.point, %struct.point* %2, i32 0, i32 0
+  %29 = load i32, i32* %28, align 4
+  %30 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 0
+  %31 = load i32, i32* %30, align 4
+  %32 = add nsw i32 %29, %31
+  store i32 %32, i32* %27, align 4
+  %33 = getelementptr inbounds %struct.point, %struct.point* %0, i32 0, i32 1
+  %34 = getelementptr inbounds %struct.point, %struct.point* %2, i32 0, i32 1
+  %35 = load i32, i32* %34, align 4
+  %36 = getelementptr inbounds %struct.point1, %struct.point1* %3, i32 0, i32 1
+  %37 = load i32, i32* %36, align 4
+  %38 = add nsw i32 %35, %37
+  store i32 %38, i32* %33, align 4
+  ret void
+}
 
 attributes #0 = { noinline nounwind optnone "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="generic" }
-attributes #1 = { argmemonly nofree nounwind willreturn }
+attributes #1 = { "frame-pointer"="none" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="generic" }
+attributes #2 = { argmemonly nofree nounwind willreturn }
 
 !llvm.module.flags = !{!0}
 !llvm.ident = !{!1}
 
 !0 = !{i32 1, !"wchar_size", i32 4}
 !1 = !{!"Apple clang version 14.0.3 (clang-1403.0.22.14.1)"}
+!2 = distinct !{!2, !3}
+!3 = !{!"llvm.loop.mustprogress"}
diff --git a/internal/cabi/cabi_test.go b/internal/cabi/cabi_test.go
index 4f821702..7754fd5b 100644
--- a/internal/cabi/cabi_test.go
+++ b/internal/cabi/cabi_test.go
@@ -6,6 +6,7 @@ package cabi_test
 import (
 	"os"
 	"path/filepath"
+	"runtime"
 	"strings"
 	"testing"
 
@@ -21,6 +22,14 @@ var (
 	archDir = []string{"amd64", "arm64", "riscv64", "armv6"}
 )
 
+func init() {
+	// crosscompile
+	if runtime.GOOS == "darwin" {
+		archs = append(archs, "wasm")
+		archDir = append(archDir, "wasm32")
+	}
+}
+
 func TestBuild(t *testing.T) {
 	for _, mode := range modes {
 		for _, arch := range archs {
@@ -28,6 +37,9 @@ func TestBuild(t *testing.T) {
 			conf.AbiMode = mode
 			conf.Goarch = arch
 			conf.Goos = "linux"
+			if arch == "wasm" {
+				conf.Goos = "wasip1"
+			}
 			_, err := build.Do([]string{"./_testdata/demo/demo.go"}, conf)
 			if err != nil {
 				t.Fatalf("build error: %v-%v %v", arch, mode, err)
@@ -62,6 +74,9 @@ func testArch(t *testing.T, arch string, archDir string, files []string) {
 	conf.AbiMode = cabi.ModeAllFunc
 	conf.Goarch = arch
 	conf.Goos = "linux"
+	if arch == "wasm" {
+		conf.Goos = "wasip1"
+	}
 	for _, file := range files {
 		pkgs, err := build.Do([]string{filepath.Join("./_testdata/demo", file)}, conf)
 		if err != nil {
@@ -97,7 +112,10 @@ func testModule(t *testing.T, ctx context, td llvm.TargetData, m llvm.Module, c
 		fn = llvm.NextFunction(fn)
 	}
 	for _, fn := range fns {
+		// check c linkname
 		testFunc(t, ctx, td, m.NamedFunction(fn.Name()), fn)
+		// check go
+		testFunc(t, ctx, td, m.NamedFunction("command-line-arguments."+fn.Name()), fn)
 	}
 }