feat: compile native with new asm2asm

Makefile

@@ -21,15 +21,15 @@ SRC_FILE		:= native/native.c
 
 CPU_avx			:= amd64
 CPU_avx2		:= amd64
-CPU_sse		    := amd64
+CPU_sse  		:= amd64
 
-TMPL_avx		:= fastint_amd64_test fastfloat_amd64_test native_amd64_test native_export_amd64
-TMPL_avx2		:= fastint_amd64_test fastfloat_amd64_test native_amd64_test native_export_amd64
-TMPL_sse		:= fastint_amd64_test fastfloat_amd64_test native_amd64_test native_export_amd64
+TMPL_avx		:= native_amd64_test
+TMPL_avx2		:= native_amd64_test
+TMPL_sse 		:= native_amd64_test
 
-CFLAGS_avx		:= -msse -mno-sse4 -mavx -mno-avx2 -DUSE_AVX=1 -DUSE_AVX2=0
-CFLAGS_avx2		:= -msse -mno-sse4 -mavx -mavx2    -DUSE_AVX=0 -DUSE_AVX2=1
-CFLAGS_sse		:= -msse -mno-sse4 -mno-avx -mno-avx2
+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
 
 CC_amd64		:= clang
 ASM2ASM_amd64	:= tools/asm2asm/asm2asm.py
@@ -43,6 +43,7 @@ CFLAGS			+= -fno-rtti
 CFLAGS			+= -fno-stack-protector
 CFLAGS			+= -nostdlib
 CFLAGS			+= -O3
+CFLAGS			+= -Wall
 
 NATIVE_SRC		:= $(wildcard native/*.h)
 NATIVE_SRC		+= $(wildcard native/*.c)
@@ -65,7 +66,7 @@ define build_arch
 	$(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_${@cpu}.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)
 
@@ -74,8 +75,7 @@ $(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}} ${@asmout} ${TMP_DIR}/$(1)/native.s
-	asmfmt -w ${@asmout}
+	python3 $${ASM2ASM_${@cpu}} -r ${@stubout} ${TMP_DIR}/$(1)/native.s
 
 $(eval $(call 	\
 	build_tmpl,	\
@@ -99,11 +99,13 @@ endef
 all: ${ARCH}
 
 clean:
-	rm -vfr ${TMP_DIR}/{sse,avx,avx2}
-	rm -vfr ${OUT_DIR}/{sse,avx,avx2}
+	for arch in ${ARCH}; do \
+		rm -vfr ${TMP_DIR}/$${arch}; \
+		rm -vfr ${OUT_DIR}/$${arch}; \
+	done
 
 $(foreach 								\
 	arch,								\
 	${ARCH},							\
 	$(eval $(call build_arch,${arch}))	\
-)
+)

go.mod

@@ -4,14 +4,13 @@ go 1.16
 
 require (
 	github.com/apache/thrift v0.13.0
-	github.com/bytedance/sonic v1.8.8
-	github.com/chenzhuoyu/base64x v0.0.0-20221115062448-fe3a3abad311
+	github.com/bytedance/sonic v1.10.2
+	github.com/chenzhuoyu/base64x v0.0.0-20230717121745-296ad89f973d
 	github.com/cloudwego/kitex v0.6.1
 	github.com/cloudwego/thriftgo v0.2.11
 	github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc
 	github.com/fatih/structtag v1.2.0
 	github.com/iancoleman/strcase v0.2.0
 	github.com/klauspost/cpuid/v2 v2.2.4
 	github.com/stretchr/testify v1.8.2
-
 )

go.sum

@@ -20,14 +20,19 @@ github.com/bytedance/gopkg v0.0.0-20230531144706-a12972768317 h1:SReMVmTCeJ5Nf0h
 github.com/bytedance/gopkg v0.0.0-20230531144706-a12972768317/go.mod h1:FtQG3YbQG9L/91pbKSw787yBQPutC+457AvDW77fgUQ=
 github.com/bytedance/mockey v1.2.0/go.mod h1:+Jm/fzWZAuhEDrPXVjDf/jLM2BlLXJkwk94zf2JZ3X4=
 github.com/bytedance/sonic v1.5.0/go.mod h1:ED5hyg4y6t3/9Ku1R6dU/4KyJ48DZ4jPhfY1O2AihPM=
-github.com/bytedance/sonic v1.8.8 h1:Kj4AYbZSeENfyXicsYppYKO0K2YWab+i2UTSY7Ukz9Q=
 github.com/bytedance/sonic v1.8.8/go.mod h1:i736AoUSYt75HyZLoJW9ERYxcy6eaN6h4BZXU064P/U=
+github.com/bytedance/sonic v1.10.0-rc/go.mod h1:ElCzW+ufi8qKqNW0FY314xriJhyJhuoJ3gFZdAHF7NM=
+github.com/bytedance/sonic v1.10.2 h1:GQebETVBxYB7JGWJtLBi07OVzWwt+8dWA00gEVW2ZFE=
+github.com/bytedance/sonic v1.10.2/go.mod h1:iZcSUejdk5aukTND/Eu/ivjQuEL0Cu9/rf50Hi0u/g4=
 github.com/census-instrumentation/opencensus-proto v0.2.1/go.mod h1:f6KPmirojxKA12rnyqOA5BBL4O983OfeGPqjHWSTneU=
 github.com/chenzhuoyu/base64x v0.0.0-20211019084208-fb5309c8db06/go.mod h1:DH46F32mSOjUmXrMHnKwZdA8wcEefY7UVqBKYGjpdQY=
-github.com/chenzhuoyu/base64x v0.0.0-20221115062448-fe3a3abad311 h1:qSGYFH7+jGhDF8vLC+iwCD4WpbV1EBDSzWkJODFLams=
 github.com/chenzhuoyu/base64x v0.0.0-20221115062448-fe3a3abad311/go.mod h1:b583jCggY9gE99b6G5LEC39OIiVsWj+R97kbl5odCEk=
+github.com/chenzhuoyu/base64x v0.0.0-20230717121745-296ad89f973d h1:77cEq6EriyTZ0g/qfRdp61a3Uu/AWrgIq2s0ClJV1g0=
+github.com/chenzhuoyu/base64x v0.0.0-20230717121745-296ad89f973d/go.mod h1:8EPpVsBuRksnlj1mLy4AWzRNQYxauNi62uWcE3to6eA=
 github.com/chenzhuoyu/iasm v0.0.0-20220818063314-28c361dae733/go.mod h1:wOQ0nsbeOLa2awv8bUYFW/EHXbjQMlZ10fAlXDB2sz8=
 github.com/chenzhuoyu/iasm v0.0.0-20230222070914-0b1b64b0e762/go.mod h1:Xjy2NpN3h7aUqeqM+woSuuvxmIe6+DDsiNLIrkAmYog=
+github.com/chenzhuoyu/iasm v0.9.0 h1:9fhXjVzq5hUy2gkhhgHl95zG2cEAhw9OSGs8toWWAwo=
+github.com/chenzhuoyu/iasm v0.9.0/go.mod h1:Xjy2NpN3h7aUqeqM+woSuuvxmIe6+DDsiNLIrkAmYog=
 github.com/choleraehyq/pid v0.0.13/go.mod h1:uhzeFgxJZWQsZulelVQZwdASxQ9TIPZYL4TPkQMtL/U=
 github.com/choleraehyq/pid v0.0.15/go.mod h1:uhzeFgxJZWQsZulelVQZwdASxQ9TIPZYL4TPkQMtL/U=
 github.com/choleraehyq/pid v0.0.16 h1:1/714sMH9IBlE/aK6xM0acTagGKSzpiR0bDt7l0cG7o=