diff --git a/Makefile b/Makefile
index a48e5b2a8..c672c313e 100644
--- a/Makefile
+++ b/Makefile
@@ -22,23 +22,20 @@ SRC_FILE		:= native/native.c
 CPU_avx			:= amd64
 CPU_avx2		:= amd64
 CPU_sse  		:= amd64
-CPU_neon  		:= arm64
 
-TMPL_amd64		:= fastint_test fastfloat_test native_test recover_amd64_test
-TMPL_arm64		:= fastint_test fastfloat_test native_test recover_arm64_test native_export_arm64
+TMPL_avx		:= fastint_amd64_test fastfloat_amd64_test native_amd64_test recover_amd64_test
+TMPL_avx2		:= fastint_amd64_test fastfloat_amd64_test native_amd64_test recover_amd64_test
+TMPL_sse 		:= fastint_amd64_test fastfloat_amd64_test native_amd64_test recover_amd64_test
 
-ASM2ASM_amd64	:= tools/asm2asm/asm2asm.py
-ASM2ASM_arm64	:= tools/asm2arm/arm.py
-
-ASM_FLAGS_amd64 := -r
-ASM_FLAGS_arm64 := 
+CFLAGS_avx		:= -msse -mno-sse4 -mavx -mpclmul -mno-avx2 -mstack-alignment=0 -DUSE_AVX=1 -DUSE_AVX2=0
+CFLAGS_avx2		:= -msse -mno-sse4 -mavx -mpclmul -mavx2 -mstack-alignment=0 -DUSE_AVX=1 -DUSE_AVX2=1 
+CFLAGS_sse		:= -msse -mno-sse4 -mno-avx -mno-avx2 -mpclmul
 
-ASM_OUT_amd64   := native_text_amd64.go
-ASM_OUT_arm64   := native_arm64.s
-
-CC		        := clang
+CC_amd64		:= clang
+ASM2ASM_amd64	:= tools/asm2asm/asm2asm.py
 
 CFLAGS			:= -mno-red-zone
+CFLAGS			+= -target x86_64-apple-macos11
 CFLAGS			+= -fno-asynchronous-unwind-tables
 CFLAGS			+= -fno-builtin
 CFLAGS			+= -fno-exceptions
@@ -46,11 +43,7 @@ CFLAGS			+= -fno-rtti
 CFLAGS			+= -fno-stack-protector
 CFLAGS			+= -nostdlib
 CFLAGS			+= -O3
-# CFLAGS			+= -Wall -Werror
-CFLAGS_avx		:= -target x86_64-apple-macos11 -msse -mno-sse4 -mavx -mpclmul -mno-avx2 -mstack-alignment=0 -DUSE_AVX=1 -DUSE_AVX2=0
-CFLAGS_avx2		:= -target x86_64-apple-macos11 -msse -mno-sse4 -mavx -mpclmul -mavx2 -mstack-alignment=0 -DUSE_AVX=1 -DUSE_AVX2=1 
-CFLAGS_sse		:= -target x86_64-apple-macos11 -msse -mno-sse4 -mno-avx -mno-avx2 -mpclmul
-CFLAGS_neon		:= -target aarch64-apple-macos11 -march=armv8-a+simd -Itools/simde/simde
+CFLAGS			+= -Wall -Werror
 
 NATIVE_SRC		:= $(wildcard native/*.h)
 NATIVE_SRC		+= $(wildcard native/*.c)
@@ -70,12 +63,19 @@ ${@dest}: ${@tmpl}
 endef
 
 define build_arch
-$(eval @cpu		:= $(value CPU_$(1)))
-$(eval @deps	:= $(foreach tmpl,$(value TMPL_${@cpu}),${OUT_DIR}/$(1)/${tmpl}.go))
-$(eval @asmin	:= ${TMP_DIR}/$(1)/native.s)
-$(eval @asmout	:= ${OUT_DIR}/$(1)/$(value ASM_OUT_${@cpu}))
-$(eval @stubin	:= ${OUT_DIR}/native_${@cpu}.tmpl)
-$(eval @stubout	:= ${OUT_DIR}/$(1)/native_${@cpu}.go)
+	$(eval @cpu		:= $(value CPU_$(1)))
+	$(eval @deps	:= $(foreach tmpl,$(value TMPL_$(1)),${OUT_DIR}/$(1)/${tmpl}.go))
+	$(eval @asmin	:= ${TMP_DIR}/$(1)/native.s)
+	$(eval @asmout	:= ${OUT_DIR}/$(1)/native_text_${@cpu}.go)
+	$(eval @stubin	:= ${OUT_DIR}/native_${@cpu}.tmpl)
+	$(eval @stubout	:= ${OUT_DIR}/$(1)/native_${@cpu}.go)
+
+$(1): ${@asmout} ${@deps}
+
+${@asmout}: ${@stubout} ${NATIVE_SRC}
+	mkdir -p ${TMP_DIR}/$(1)
+	$${CC_${@cpu}} $${CFLAGS} $${CFLAGS_$(1)} -S -o ${TMP_DIR}/$(1)/native.s ${SRC_FILE}
+	python3 $${ASM2ASM_${@cpu}} -r ${@stubout} ${TMP_DIR}/$(1)/native.s
 
 $(eval $(call 	\
 	build_tmpl,	\
@@ -84,16 +84,9 @@ $(eval $(call 	\
 	${@stubout}	\
 ))
 
-$(1): ${@asmout} ${@deps}
-
-${@asmout}: ${@stubout} ${NATIVE_SRC}
-	mkdir -p ${TMP_DIR}/$(1)
-	$${CC} $${CFLAGS} $${CFLAGS_$(1)} -S -o ${TMP_DIR}/$(1)/native.s ${SRC_FILE}
-	python3 $${ASM2ASM_${@cpu}} $${ASM_FLAGS_${@cpu}} ${@stubout} ${TMP_DIR}/$(1)/native.s
-
 $(foreach 							\
 	tmpl,							\
-	$(value TMPL_${@cpu}),			\
+	$(value TMPL_$(1)),				\
 	$(eval $(call 					\
 		build_tmpl,					\
 		$(1),						\
diff --git a/internal/decoder/decoder_test.go b/internal/decoder/decoder_test.go
index 120d8b590..745eb38eb 100644
--- a/internal/decoder/decoder_test.go
+++ b/internal/decoder/decoder_test.go
@@ -400,3 +400,163 @@ func TestDecoder_MapWithIndirectElement(t *testing.T) {
     require.NoError(t, err)
     assert.Equal(t, [129]byte{1, 2, 3, 4, 5}, v[""].A)
 }
+
+func BenchmarkDecoder_Generic_Sonic(b *testing.B) {
+    var w interface{}
+    _, _ = decode(TwitterJson, &w, true)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    for i := 0; i < b.N; i++ {
+        var v interface{}
+        _, _ = decode(TwitterJson, &v, true)
+    }
+}
+
+func BenchmarkDecoder_Generic_Sonic_Fast(b *testing.B) {
+    var w interface{}
+    _, _ = decode(TwitterJson, &w, false)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    for i := 0; i < b.N; i++ {
+        var v interface{}
+        _, _ = decode(TwitterJson, &v, false)
+    }
+}
+
+func BenchmarkDecoder_Generic_StdLib(b *testing.B) {
+    var w interface{}
+    m := []byte(TwitterJson)
+    _ = json.Unmarshal(m, &w)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    for i := 0; i < b.N; i++ {
+        var v interface{}
+        _ = json.Unmarshal(m, &v)
+    }
+}
+
+func BenchmarkDecoder_Binding_Sonic(b *testing.B) {
+    var w TwitterStruct
+    _, _ = decode(TwitterJson, &w, true)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    for i := 0; i < b.N; i++ {
+        var v TwitterStruct
+        _, _ = decode(TwitterJson, &v, true)
+    }
+}
+
+func BenchmarkDecoder_Binding_Sonic_Fast(b *testing.B) {
+    var w TwitterStruct
+    _, _ = decode(TwitterJson, &w, false)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    for i := 0; i < b.N; i++ {
+        var v TwitterStruct
+        _, _ = decode(TwitterJson, &v, false)
+    }
+}
+
+func BenchmarkDecoder_Binding_StdLib(b *testing.B) {
+    var w TwitterStruct
+    m := []byte(TwitterJson)
+    _ = json.Unmarshal(m, &w)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    for i := 0; i < b.N; i++ {
+        var v TwitterStruct
+        _ = json.Unmarshal(m, &v)
+    }
+}
+
+func BenchmarkDecoder_Parallel_Generic_Sonic(b *testing.B) {
+    var w interface{}
+    _, _ = decode(TwitterJson, &w, true)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    b.RunParallel(func(pb *testing.PB) {
+        for pb.Next() {
+            var v interface{}
+            _, _ = decode(TwitterJson, &v, true)
+        }
+    })
+}
+
+func BenchmarkDecoder_Parallel_Generic_Sonic_Fast(b *testing.B) {
+    var w interface{}
+    _, _ = decode(TwitterJson, &w, false)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    b.RunParallel(func(pb *testing.PB) {
+        for pb.Next() {
+            var v interface{}
+            _, _ = decode(TwitterJson, &v, false)
+        }
+    })
+}
+
+func BenchmarkDecoder_Parallel_Generic_StdLib(b *testing.B) {
+    var w interface{}
+    m := []byte(TwitterJson)
+    _ = json.Unmarshal(m, &w)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    b.RunParallel(func(pb *testing.PB) {
+        for pb.Next() {
+            var v interface{}
+            _ = json.Unmarshal(m, &v)
+        }
+    })
+}
+
+func BenchmarkDecoder_Parallel_Binding_Sonic(b *testing.B) {
+    var w TwitterStruct
+    _, _ = decode(TwitterJson, &w, true)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    b.RunParallel(func(pb *testing.PB) {
+        for pb.Next() {
+            var v TwitterStruct
+            _, _ = decode(TwitterJson, &v, true)
+        }
+    })
+}
+
+func BenchmarkDecoder_Parallel_Binding_Sonic_Fast(b *testing.B) {
+    var w TwitterStruct
+    _, _ = decode(TwitterJson, &w, false)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    b.RunParallel(func(pb *testing.PB) {
+        for pb.Next() {
+            var v TwitterStruct
+            _, _ = decode(TwitterJson, &v, false)
+        }
+    })
+}
+
+func BenchmarkDecoder_Parallel_Binding_StdLib(b *testing.B) {
+    var w TwitterStruct
+    m := []byte(TwitterJson)
+    _ = json.Unmarshal(m, &w)
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    b.RunParallel(func(pb *testing.PB) {
+        for pb.Next() {
+            var v TwitterStruct
+            _ = json.Unmarshal(m, &v)
+        }
+    })
+}
+
+func BenchmarkSkip_Sonic(b *testing.B) {
+    var data = rt.Str2Mem(TwitterJson)
+    if ret, _ := Skip(data); ret < 0 {
+        b.Fatal()
+    }
+    b.SetBytes(int64(len(TwitterJson)))
+    b.ResetTimer()
+    for i:=0; i<b.N; i++ {
+        _, _ = Skip(data)
+    }
+}