!14741 [MS][LITE][Develop]fix assembly headers and declaration style

From: @lx0095
Reviewed-by: @zhanghaibo5,@zhang_xue_tong
Signed-off-by: @zhang_xue_tong

mindspore/lite/nnacl/assembly/arm32/ConvDw3x3Int8BorderPixel.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Int8BorderPixel
-#ifndef __APPLE__
-.type ConvDw3x3Int8BorderPixel, %function
-#endif
 
 // void ConvDw3x3Int8BorderPixel(int8_t *dst, const int8_t *src, const int16_t *weight, const int32_t *bias, size_t height,
 //                               size_t width, size_t in_kh_step, size_t in_kw_step, size_t channel, size_t in_zp, size_t out_zp,
@@ -116,4 +111,3 @@ asm_function ConvDw3x3Int8BorderPixel
     vpop {q4-q7}
     pop {r4-r8, r9-r12, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/ConvDwFp32Border.S

@@ -3,10 +3,6 @@
 
 .text
 .align 5
-.global ConvDwFp32Border
-#ifndef __APPLE__
-.type ConvDwFp32Border, %function
-#endif
 
 // void ConvDwFp32Border(float *dst, const float *src, const float *weight, const float *bias, size_t height, size_t width,
 //                       size_t in_kh_step, size_t in_kw_step, size_t kernel_w, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm32/ConvDwFp32Center.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp32Center
-#ifndef __APPLE__
-.type ConvDwFp32Center, %function
-#endif
 
 // void ConvDwFp32Center(float *dst, const float *src, const float *weight, const float *bias, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step, size_t in_sw_step,
@@ -164,4 +159,3 @@ LoopWEnd:
     vpop {q4-q7}
     pop {r0-r8, r10, r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/ConvDwFp32Row.S

@@ -3,10 +3,6 @@
 
 .text
 .align 5
-.global ConvDwFp32Row
-#ifndef __APPLE__
-.type ConvDwFp32Row, %function
-#endif
 
 // voidConvDwFp32Row(float* output_ptr, const float* input_ptr, const float* filter_ptr,
 //                   size_t num_pixels, size_t input_channel, size_t input_step)

mindspore/lite/nnacl/assembly/arm32/ConvDwInt8Center.S

@@ -1,13 +1,9 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8Center
-#ifndef __APPLE__
-.type ConvDwInt8Center, %function
-#endif
+
 // void DepthwiseCenterInt8(int8_t *dst, const int8_t *src, const int16_t *weight, const int32_t *bias, int height,
 //                          int width, int kernel_h, int kernel_w, int out_h_step, int block_channel, int in_sh_step,
 //                          int in_sw_step, int in_kh_step, int in_kw_step, int8_t *in_zp, int32_t *out_zp,
@@ -277,4 +273,3 @@ asm_function ConvDwInt8Center
         vpop {q4-q7}
         pop {r0-r8, r10, r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/ConvDwInt8PostAlign4.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8PostAlign4
-#ifndef __APPLE__
-.type ConvDwInt8PostAlign4, %function
-#endif
 
 // void ConvDwInt8PostAlign4(int8_t *dst, int32_t *buffer, int num_pixels, int32_t output_zp, int32_t out_multiplier,
 //                           int32_t left_shift, int32_t right_shift, int32_t acc_min, int32_t acc_max);
@@ -108,4 +103,3 @@ asm_function ConvDwInt8PostAlign4
         bx lr
 
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/ConvDwInt8PostAlign4PerChannel.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8PostAlign4PerChannel
-#ifndef __APPLE__
-.type ConvDwInt8PostAlign4PerChannel, %function
-#endif
 
 // void ConvDwInt8PostAlign4PerChannel(int8_t *dst, int32_t *buffer, int channel4, int32_t output_zp, int32_t *out_multiplier,
 //                                     int32_t *left_shift, int32_t *right_shift, int32_t acc_min, int32_t acc_max);
@@ -111,4 +106,3 @@ asm_function ConvDwInt8PostAlign4PerChannel
         bx lr
 
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/ConvDwInt8Row.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8Row
-#ifndef __APPLE__
-.type ConvDwInt8Row, %function
-#endif
 
 // void ConvDwInt8Row(int32_t *output_ptr, const int8_t *input_ptr, const int16_t *weight_ptr, int num_pixels,
 //                    int output_channel, int input_step, int8_t input_zp)
@@ -132,4 +127,3 @@ asm_function ConvDwInt8Row
     vpop {q4-q7}
     pop {r4-r8, r9-r12, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/DeconvDwFp32Center.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwFp32Center
-#ifndef __APPLE__
-.type DeconvDwFp32Center, %function
-#endif
 
 // void DeconvDwFp32Center(float *dst, const float *src, const float *weight, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step,
@@ -67,4 +62,3 @@ asm_function DeconvDwFp32Center
 
     pop {r0-r8, r10, r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/DeconvDwInt8Center.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwInt8Center
-#ifndef __APPLE__
-.type DeconvDwInt8Center, %function
-#endif
 
 // void DeconvDwInt8Center(int32_t *dst, const int16_t *src, const int16_t *weight, size_t height, size_t width,
 //                         size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step,
@@ -67,4 +62,3 @@ asm_function DeconvDwInt8Center
 
     pop {r0-r8, r10, r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/DeconvDwInt8Post.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwInt8Post
-#ifndef __APPLE__
-.type DeconvDwInt8Post, %function
-#endif
 
 // void DeconvDwInt8Post(int8_t *dst, int32_t *output_buffer, const int32_t *bias, int block_channel, int pixel_nums,
 //                       int out_multiplier, int left_shift, int right_shift, int32_t out_zp, int32_t acc_min,
@@ -72,4 +67,3 @@ asm_function DeconvDwInt8Post
         bx lr
 
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/IndirectGemmInt16to32_8x4.S

@@ -3,10 +3,6 @@
 
 .text
 .align 5
-.global IndirectGemmInt16to32_8x4
-#ifndef __APPLE__
-.type IndirectGemmInt16to32_8x4, %function
-#endif
 
 // void IndirectGemmInt16to32_8x4(int *output, short *input, short *weight, size_t kszie, size_t ic8, size_t oc4, size_t offset);
 // r0: output, r1: input, r2: weight, r3: kszie, r4: ic8, r5: oc4, r6: offset

mindspore/lite/nnacl/assembly/arm32/IndirectGemmInt8_2x4.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global IndirectGemmInt8_2x4
-#ifndef __APPLE__
-.type IndirectGemmInt8_2x4, %function
-#endif
 
 // void IndirectGemmInt8_2x4(int8_t *output, int8_t *input, int8_t *weight, int32_t *bias, size_t ksize, size_t ic4,
 // size_t oc, size_t offset, int32_t *input_sum, size_t act_min, size_t act_max, size_t out_zp, int32_t *out_multiplier,
@@ -294,4 +289,3 @@ LoopOcEnd:
     vpop {q4-q7}
     pop {r4-r8, r10, r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/MatVecMulFp32.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global MatVecMulFp32
-#ifndef __APPLE__
-.type MatVecMulFp32, %function
-#endif
 
 // void MatVecMulFp32(const float *a, const float *b, float *c, const float *bias, int act_type, int depth, int col)
 // r0: a
@@ -183,4 +178,3 @@ End:
   sub sp, sp, #52
   pop {r0-r8, r9, r10, r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/MatmulFp32.S

@@ -1,11 +1,8 @@
 #ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulFloatNeon32
-#ifndef __APPLE__
-    .type MatmulFloatNeon32, %function
-#endif
+
+.text
+.align 5
 
 // void MatmulFloatNeon32(const float *a, const float *b, float *c, const float *bias, int act_type, int depth
 //                        int row, int col, size_t stride, size_t writeNhwc, size_t WriteWino)

mindspore/lite/nnacl/assembly/arm32/MatmulFp32Opt.S

@@ -1,11 +1,8 @@
 #ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulFloatNeon32Opt
-#ifndef __APPLE__
-    .type MatmulFloatNeon32Opt, %function
-#endif
+
+.text
+.align 5
 
 // void MatmulFloatNeon32Opt(const float *a, const float *b, float *c, const float *bias, int act_type, int depth
 //                        int row, int col, size_t stride, size_t writeMode)

mindspore/lite/nnacl/assembly/arm32/MatmulFp32Opt12x4.S

@@ -1,11 +1,8 @@
 #ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulFloatNeon32Opt12x4
-#ifndef __APPLE__
-    .type MatmulFloatNeon32Opt12x4, %function
-#endif
+
+.text
+.align 5
 
 // void MatmulFloatNeon32Opt12x4(const float *a, const float *b, float *c, const float *bias, int act_type, int depth
 //                               int row, int col, size_t stride, size_t writeMode)

mindspore/lite/nnacl/assembly/arm32/MatmulInt8.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global MatmulInt8Neon32
-#ifndef __APPLE__
-.type MatmulInt8Neon32, %function
-#endif
 
 //void MatmulInt8Neon32(const int8_t *a, const int8_t *b, int8_t *dst, int row, int col, int deep16, 
 //                      const int *input_sums, const int *weight_bias, int act_min, int act_max, int out_zp,
@@ -286,4 +281,3 @@ End1:
   vpop {q4-q7}
   pop {r0-r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/MatmulInt8Opt.S

@@ -1,13 +1,8 @@
-#ifdef __arm__
-#ifndef __aarch64__
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global MatmulInt8Opt
-#ifndef __APPLE__
-.type MatmulInt8Opt, %function
-#endif
 
 //void MatmulInt8Neon32Opt(const int8_t *a, const int8_t *b, int8_t *dst, int row, int col, int deep16, 
 //                         const int *input_sums, const int *weight_bias, int act_min, int act_max, int out_zp,
@@ -288,4 +283,3 @@ LoopRowEnd:
     vpop {q4-q7}
     pop {r0-r8, r10, r11, pc}
 #endif
-#endif

mindspore/lite/nnacl/assembly/arm32/MatmulWinogradFp32.S

@@ -3,10 +3,6 @@
 
 .text
 .align 5
-.global MatrixMultiplyWinograd
-#ifndef __APPLE__
-.type MatrixMultiplyWinograd, %function
-#endif
 
 // MatrixMultiplyWinograd(float *matix_a, float *matrix_b, float *matrix_c, int m, int k, int n, int in_channel, int c4_channel)
     // r0: matrix_a, r1: matrix_b, r2: matrix_c, r3: m, r4: k, r5: n, r6: in_channel, r7: c4_channel * 4

mindspore/lite/nnacl/assembly/arm32/PostFuncBiasReluC4.S

@@ -1,12 +1,8 @@
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-//.p2align 5,,15
-.global PostFuncBiasReluC4
-#ifndef __APPLE__
-.type PostFuncBiasReluC4, %function
-#endif
 
 asm_function PostFuncBiasReluC4
   push {r4-r8, r10, r11, lr}
@@ -234,3 +230,4 @@ Loop_C1_3_Write:
 End:
   sub sp, sp, #32
   pop {r4-r8, r10, r11, pc}
+#endif

mindspore/lite/nnacl/assembly/arm32/PostFuncBiasReluC8.S

@@ -3,11 +3,6 @@
 
 .text
 .align 5
-//.p2align 5,,15
-.global PostFuncBiasReluC8
-#ifndef __APPLE__
-.type PostFuncBiasReluC8, %function
-#endif
 
 //void PostFuncBiasReluC8(float *dst, const float *src, const float *bias, size_t oc8div,size_t oc8mod
 //                        size_t plane_size, size_t stride, int relu_type);

mindspore/lite/nnacl/assembly/arm32/PreSum4x16Int8Peroc.S

@@ -1,12 +1,8 @@
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global PreSum4x16Int8Peroc
-#ifndef __APPLE__
-.type PreSum4x16Int8Peroc, %function
-#endif
-
 
 //void PreSum4x16Int8Peroc(const int8_t *src, int32_t *sum, int32_t *zp, size_t hw4, size_t ic16, int32_t oc_div2,
 //                            size_t oc_res2, size_t stride);
@@ -129,3 +125,4 @@ End:
   sub sp, sp, #100
   vpop {q4-q7}
   pop {r4-r11, pc}
+#endif

mindspore/lite/nnacl/assembly/arm32/PreSum4x16Int8Pert.S

@@ -1,12 +1,8 @@
+#ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global PreSum4x16Int8Pert
-#ifndef __APPLE__
-.type PreSum4x16Int8Pert, %function
-#endif
-
 
 // void PreSum4x16Int8Pert(const int8_t *src, int32_t *sum, size_t row4, size_t col16, int32_t filter_zp);
 
@@ -80,3 +76,4 @@ End:
   sub sp, sp, #96
   vpop {q4-q7}
   pop {r4-r8, r10, r11, pc}
+#endif

mindspore/lite/nnacl/assembly/arm32/TiledC4MatmulFp32.S

@@ -1,11 +1,8 @@
 #ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global TiledC4MatmulFp32
-#ifndef __APPLE__
-    .type TiledC4MatmulFp32, %function
-#endif
+
+.text
+.align 5
 
 asm_function TiledC4MatmulFp32
 //void TiledC4MatmulFp32(float* dst, const float* src, const float* weight, size_t cal_num, size_t ic4, size_t oc4)

mindspore/lite/nnacl/assembly/arm32/WinogradTransLeft.S

@@ -1,12 +1,8 @@
 #ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    .global WinogradTransLeft
-#ifndef __APPLE__
-    .type WinogradTransLeft, %function
-#endif
+.text
+.align 5
 
 //void WinogradTransLeft(const float* S, const float* B, float* M, size_t w, size_t h, size_t k, size_t length);
 //x0: S

mindspore/lite/nnacl/assembly/arm32/WinogradTransRight.S

@@ -1,12 +1,8 @@
 #ifdef ENABLE_ARM32
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    .global WinogradTransRight
-#ifndef __APPLE__
-    .type WinogradTransRight, %function
-#endif
+.text
+.align 5
 
 //void WinogradTransRight(const float* S, const float* B, float* M, size_t w, size_t h, size_t k, size_t length);
 //x0: S

mindspore/lite/nnacl/assembly/arm64/AdderFp32.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global AdderFloatNeon64
-#ifndef __APPLE__
-    .type AdderFloatNeon64, %function
-#endif
+
+.text
+.align 5
 
 // void AdderFloatNeon64(const float *a, const float *b, float *c, const float *bias, int act_type, int depth
 //                        int row, int col, size_t stride)

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Fp32Corner.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Corner
-#ifndef __APPLE__
-.type ConvDw3x3Corner, %function
-#endif
 
 // void ConvDw3x3Corner(float *dst, const float *src, const float *weight, const float *bias, int in_kh_step,
 //                      int in_kw_step, int channel, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Fp32Horizontal.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Horizontal
-#ifndef __APPLE__
-.type ConvDw3x3Horizontal, %function
-#endif
 
 // void ConvDw3x3Horizontal(float *dst, const float *src, const float *weight, const float *bias, int in_kh_step,
 //                          int in_kw_step, int channel, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Fp32Stride1.S

@@ -1,13 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Stride1
-#ifndef __APPLE__
-.type ConvDw3x3Stride1, %function
-#endif
-
 
 // void ConvDw3x3Stride1(float *output, const float *buffer, const float *weight, const float *bias, int col_size,
 //                       int row_size, int channel, int output_h, int output_w, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Fp32Stride2.S

@@ -1,13 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Stride2
-#ifndef __APPLE__
-.type ConvDw3x3Stride2, %function
-#endif
-
 
 // void ConvDw3x3Stride2(float *output, const float *buffer, const float *weight, const float *bias, int col_size,
 //                       int row_size, int channel, int output_h, int output_w, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Fp32Vertical.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Vertical
-#ifndef __APPLE__
-.type ConvDw3x3Vertical, %function
-#endif
 
 // void ConvDw3x3Vertical(float *dst, const float *src, const float *weight, const float *bias, int in_kh_step,
 //                        int in_kw_step, int channel, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Int8.S

@@ -1,13 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Int8Neon64
-#ifndef __APPLE__
-.type ConvDw3x3Int8Neon64, %function
-#endif
-
 
 // void ConvDw3x3Int8Neon64(int8_t *output, const int8_t *input, const int16_t *weight, const int32_t *bias, int input_col_size,
 //                          int input_row_size, int channel, int output_h, int output_w, int8_t in_zp, int32_t out_zp,

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Int8Corner.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Int8Corner
-#ifndef __APPLE__
-.type ConvDw3x3Int8Corner, %function
-#endif
 
 // void ConvDw3x3Int8Corner(int8_t *dst, const int8_t *src, const int16_t *weight, const int32_t *bias, size_t in_kh_step,
 //                          size_t in_kw_step, size_t channel, size_t in_zp, size_t out_zp, int32_t *out_multiplier,

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Int8Horizontal.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Int8Horizontal
-#ifndef __APPLE__
-.type ConvDw3x3Int8Horizontal, %function
-#endif
 
 // void ConvDw3x3Int8Horizontal(int8_t *dst, const int8_t *src, const int16_t *weight, const int32_t *bias, size_t in_kh_step,
 //                              size_t in_kw_step, size_t channel, size_t in_zp, size_t out_zp, int32_t *out_multiplier,

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Int8Stride2.S

@@ -1,13 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Int8Stride2
-#ifndef __APPLE__
-.type ConvDw3x3Int8Stride2, %function
-#endif
-
 
 // void ConvDw3x3Int8Stride2(int8_t *output, const int8_t *input, const int16_t *weight, const int32_t *bias, int input_col_size,
 //                           int input_row_size, int channel, int output_h, int output_w, int8_t in_zp, int32_t out_zp,

mindspore/lite/nnacl/assembly/arm64/ConvDw3x3Int8Vertical.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDw3x3Int8Vertical
-#ifndef __APPLE__
-.type ConvDw3x3Int8Vertical, %function
-#endif
 
 // void ConvDw3x3Int8Vertical(int8_t *dst, const int8_t *src, const int16_t *weight, const int32_t *bias, size_t in_kh_step,
 //                            size_t in_kw_step, size_t channel, size_t in_zp, size_t out_zp, int32_t *out_multiplier,

mindspore/lite/nnacl/assembly/arm64/ConvDwFp32Border.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp32Border
-#ifndef __APPLE__
-.type ConvDwFp32Border, %function
-#endif
 
 // void ConvDwFp32Border(float *dst, const float *src, const float *weight, const float *bias, size_t height, size_t width,
 //                       size_t in_kh_step, size_t in_kw_step, size_t kernel_w, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDwFp32Center.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp32Center
-#ifndef __APPLE__
-.type ConvDwFp32Center, %function
-#endif
 
 // void ConvDwFp32Center(float *dst, const float *src, const float *weight, const float *bias, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step, size_t in_sw_step,

mindspore/lite/nnacl/assembly/arm64/ConvDwFp32Indirect3x3.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp32Indirect3x3
-#ifndef __APPLE__
-.type ConvDwFp32Indirect3x3, %function
-#endif
 
 // void ConvDwFp32Indirect3x3(float *output, float **input, const float *weights, const float *bias, int channels, int output_width,
 //                            size_t input_stride, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDwFp32Indirect5x5.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp32Indirect5x5
-#ifndef __APPLE__
-.type ConvDwFp32Indirect5x5, %function
-#endif
 
 // void ConvDwFp32Indirect5x5(float *output, float **input, const float *weights, const float *bias, int channels, int output_width,
 //                            size_t input_stride, size_t relu, size_t relu6)

mindspore/lite/nnacl/assembly/arm64/ConvDwFp32Row.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp32Row
-#ifndef __APPLE__
-.type ConvDwFp32Row, %function
-#endif
 
 // void ConvDwFp32Row(float* output_ptr, const float* input_ptr,const float* filter_ptr,
 //                    size_t num_pixels, size_t input_channel, size_t input_step)

mindspore/lite/nnacl/assembly/arm64/ConvDwInt8Center.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8Center
-#ifndef __APPLE__
-.type ConvDwInt8Center, %function
-#endif
 
 // void ConvDwInt8Center(int8_t *dst, const int8_t *src, const int16_t *weight, const int32_t *bias, size_t height,
 //                       size_t width, size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel,

mindspore/lite/nnacl/assembly/arm64/ConvDwInt8PostAlign4.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8PostAlign4
-#ifndef __APPLE__
-.type ConvDwInt8PostAlign4, %function
-#endif
 
 // void ConvDwInt8PostAlign4(int8_t *dst, int32_t *buffer, int num_pixels, int32_t output_zp, int32_t out_multiplier,
 //                           int32_t left_shift, int32_t right_shift, int32_t acc_min, int32_t acc_max);

mindspore/lite/nnacl/assembly/arm64/ConvDwInt8PostAlign4PerChannel.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8PostAlign4PerChannel
-#ifndef __APPLE__
-.type ConvDwInt8PostAlign4PerChannel, %function
-#endif
 
 // void ConvDwInt8PostAlign4PerChannel(int8_t *dst, int32_t *buffer, int channel4, int32_t output_zp, int32_t *out_multiplier,
 //                                     int32_t *left_shift, int32_t *right_shift, int32_t acc_min, int32_t acc_max);

mindspore/lite/nnacl/assembly/arm64/ConvDwInt8Row.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwInt8Row
-#ifndef __APPLE__
-.type ConvDwInt8Row, %function
-#endif
 
 // void ConvDwInt8Row(int32_t *output_ptr, const int8_t *input_ptr, const int16_t *weight_ptr, int num_pixels,
 //                    int output_channel, int input_step, int8_t input_zp)

mindspore/lite/nnacl/assembly/arm64/ConvFp32Center.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvSwFp32Center
-#ifndef __APPLE__
-.type ConvSwFp32Center, %function
-#endif
 
 // void ConvSwFp32Center(float *dst, const float *src, const float *weight, const float *bias, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t ic4, size_t in_sh_step,

mindspore/lite/nnacl/assembly/arm64/DeconvDwFp32Border.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwFp32Border
-#ifndef __APPLE__
-.type DeconvDwFp32Border, %function
-#endif
 
 // void DeconvDwFp32Border(float *dst, const float *src, const float *weight, size_t height, size_t width,
 //                         size_t in_kh_step, size_t in_kw_step, size_t kernel_w)

mindspore/lite/nnacl/assembly/arm64/DeconvDwFp32Center.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwFp32Center
-#ifndef __APPLE__
-.type DeconvDwFp32Center, %function
-#endif
 
 // void DeconvDwFp32Center(float *dst, const float *src, const float *weight, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step, size_t in_sw_step,

mindspore/lite/nnacl/assembly/arm64/DeconvDwInt8Center.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwInt8Center
-#ifndef __APPLE__
-.type DeconvDwInt8Center, %function
-#endif
 
 // void DeconvDwInt8Center(int32_t *dst, const int16_t *src, const int16_t *weight, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step, size_t in_sw_step,

mindspore/lite/nnacl/assembly/arm64/DeconvDwInt8Post.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwInt8Post
-#ifndef __APPLE__
-.type DeconvDwInt8Post, %function
-#endif
 
 // void DeconvDwInt8Post(int8_t *dst, int32_t *output_buffer, const int32_t *bias, int block_channel, int pixel_nums,
 //                       int out_multiplier, int left_shift, int right_shift, int32_t out_zp, int32_t acc_min,

mindspore/lite/nnacl/assembly/arm64/IndirectGemmInt16to32_8x4.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global IndirectGemmInt16to32_8x4
-#ifndef __APPLE__
-.type IndirectGemmInt16to32_8x4, %function
-#endif
 
 // void IndirectGemmInt16to32_8x4(int *output, short *input, short *weight, size_t ksize, size_t ic8, size_t oc4, size_t offset);
 // x0: output, x1: input, x2: weight, x3: ksize, x4: ic8, x5: oc4, x6: offset

mindspore/lite/nnacl/assembly/arm64/MatVecMulFp32.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatVecMulFp32
-#ifndef __APPLE__
-    .type MatVecMulFp32, %function
-#endif
+    
+.text
+.align 5
 
 // void MatVecMulFp32(const float *a, const float *b, float *c, const float *bias, int act_type, int depth, int col)
 // x0: a

mindspore/lite/nnacl/assembly/arm64/MatmulFp32.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulFloatNeon64
-#ifndef __APPLE__
-    .type MatmulFloatNeon64, %function
-#endif
+
+.text
+.align 5
 
 // void MatmulFloatNeon64(const float *a, const float *b, float *c, const float *bias, int act_type, int depth
 //                        int row, int col, size_t stride, size_t writeNhwc, size_t WriteWino)

mindspore/lite/nnacl/assembly/arm64/MatmulFp32Opt.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulFloatNeon64Opt
-#ifndef __APPLE__
-    .type MatmulFloatNeon64Opt, %function
-#endif
+
+.text
+.align 5
 
 // void MatmulFloatNeon64(const float *a, const float *b, float *c, const float *bias, int act_type, int depth
 //                        int row, int col, size_t stride, size_t writeMode)

mindspore/lite/nnacl/assembly/arm64/MatmulInt8.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulInt8Neon64
-#ifndef __APPLE__
-    .type MatmulInt8Neon64, %function
-#endif
+
+.text
+.align 5
 
 //void MatmulInt8Neon64(const int8_t *a, const int8_t *b, int8_t *dst, int row4, int col4, int deep16, const int *a_sums,
 //                      const int *bias, int act_min, int act_max, int out_zp, int32_t *multiplier, int32_t *left_shift,

mindspore/lite/nnacl/assembly/arm64/MatmulInt8Opt.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulInt8Opt
-#ifndef __APPLE__
-    .type MatmulInt8Opt, %function
-#endif
+
+.text
+.align 5
 
 //void MatmulInt8Opt(const int8_t *a, const int8_t *b, int8_t *dst, int row, int col, int deep16, const int *a_sums,
 //                   const int *bias, int act_min, int act_max, int out_zp, int32_t *multiplier, int32_t *left_shift,

mindspore/lite/nnacl/assembly/arm64/MatmulR4Int8.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatMulR4Int8Neon64
-#ifndef __APPLE__
-    .type MatMulR4Int8Neon64, %function
-#endif
+
+.text
+.align 5
 
 //void MatMulR4Int8Neon64(const int8_t *a, const int8_t *b, int32_t *dst, int row4, int col4, int deep16,
 //                        const int *input_sum, const int *bias)

mindspore/lite/nnacl/assembly/arm64/MatmulWinogradFp32.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global MatrixMultiplyWinograd
-#ifndef __APPLE__
-.type MatrixMultiplyWinograd, %function
-#endif
 
 // MatrixMultiplyWinograd(float *matix_a, float *matrix_b, float *matrix_c, int m, int k, int n, int in_channel, int c4_channel)
                // x0: matrix_a, x1: matrix_b, x2: matrix_c, x3: m, x4: k, x5: n, x6: in_channel, x7: c4_channel

mindspore/lite/nnacl/assembly/arm64/PostFuncBiasReluC4.S

@@ -1,13 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    //.p2align 5,,15
-    .global PostFuncBiasReluC4
-#ifndef __APPLE__
-    .type PostFuncBiasReluC4, %function
-#endif
+.text
+.align 5
 
 //void PostFuncBiasReluC4(float *dst, const float *src, const float *bias, size_t oc4div, size_t oc4mod,
 //                        size_t plane_size, size_t plane_stride, size_t relu_type);

mindspore/lite/nnacl/assembly/arm64/PostFuncBiasReluC8.S

@@ -1,13 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    //.p2align 5,,15
-    .global PostFuncBiasReluC8
-#ifndef __APPLE__
-    .type PostFuncBiasReluC8, %function
-#endif
+.text
+.align 5
 
 //void PostFuncBiasReluC8(float *dst, const float *src, const float *bias, size_t oc8div,size_t oc8mod
 //                        size_t plane_size, size_t stride, int relu_type);

mindspore/lite/nnacl/assembly/arm64/PostFuncInt8C4Neon64.S

@@ -1,14 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    //.p2align 5,,15
-    .global PostFuncInt8C4Neon64
-#ifndef __APPLE__
-    .type PostFuncInt8C4Neon64, %function
-#endif
-
+.text
+.align 5
 
 //void PostFuncInt8C4Neon64(const int32_t *in, const int32_t *bias, int8_t *out, size_t oc4div, size_t oc4res,
 //                          size_t plane, size_t stride, int32_t multiplier, int32_t left_shift, int32_t right_shift,

mindspore/lite/nnacl/assembly/arm64/PreSum4x16Int8Peroc.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    //.p2align 5,,15
-    .global PreSum4x16Int8Peroc
-#ifndef __APPLE__
-    .type PreSum4x16Int8Peroc, %function
-#endif
+
+.text
+.align 5
 
 //void PreSum4x16Int8Peroc(const int8_t *src, int32_t *sum, int32_t *zp, size_t hw4, size_t ic16, int32_t oc_div4,
 //                            size_t oc_res4, size_t stride);

mindspore/lite/nnacl/assembly/arm64/PreSum4x16Int8Pert.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    //.p2align 5,,15
-    .global PreSum4x16Int8Pert
-#ifndef __APPLE__
-    .type PreSum4x16Int8Pert, %function
-#endif
+
+.text
+.align 5
 
 // void PreSum4x16Int8Pert(const int8_t *src, int32_t *dst, size_t row4, size_t col16, int32_t filter_zp);
 

mindspore/lite/nnacl/assembly/arm64/TiledC4MatmulFp32.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    .global TiledC4MatmulFp32
-#ifndef __APPLE__
-    .type TiledC4MatmulFp32, %function
-#endif
+.text
+.align 5
 
 asm_function TiledC4MatmulFp32
 //void TiledC4MatmulFp32(float* dst, const float* src, const float* weight, size_t ic4, size_t cal_num, size_t oc4)

mindspore/lite/nnacl/assembly/arm64/WinogradTransLeft.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    .global WinogradTransLeft
-#ifndef __APPLE__
-    .type WinogradTransLeft, %function
-#endif
+.text
+.align 5
 
 asm_function WinogradTransLeft
 //void WinogradTransLeft(const float* S, const float* B, float* M, size_t w, size_t h, size_t k, size_t length);

mindspore/lite/nnacl/assembly/arm64/WinogradTransRight.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    .global WinogradTransRight
-#ifndef __APPLE__
-    .type WinogradTransRight, %function
-#endif
+.text
+.align 5
 
 asm_function WinogradTransRight
 //void WinogradTransRight(const float* S, const float* B, float* M, size_t w, size_t h, size_t k, size_t length);

mindspore/lite/nnacl/assembly/avx/ConvDwFp32Avx3x3.S

@@ -2,12 +2,6 @@
 #include "nnacl/assembly_global.h"
 .text
 .align 4
-.global ConvDwFp32Avx3x3
-#ifndef __APPLE__
-#ifndef WIN32
-.type ConvDwFp32Avx3x3, %function
-#endif
-#endif
 
 // void ConvDwFp32Avx3x3(float *output, float **input, const float *weights, const float *bias, size_t channels, size_t output_width,
 //                       size_t input_stride, size_t relum, szie_t relu6)

mindspore/lite/nnacl/assembly/avx/ConvDwFp32BorderAvx.S

@@ -1,18 +1,14 @@
 #ifdef ENABLE_AVX
-    .text
-    .align 4
-    .global ConvDwFp32Border
-#ifndef __APPLE__
-#ifndef WIN32
-    .type ConvDwFp32Border, %function
-#endif
-#endif
+#include "nnacl/assembly_global.h"
+
+.text
+.align 4
 
 // void ConvDwFp32Border(float *dst, const float *src, const float *weight, const float *bias, size_t height,
 //                       size_t width, size_t in_kh_step, size_t in_kw_step, size_t kernel_w, size_t relu,
 //                       size_t relu6);
 
-ConvDwFp32Border:
+asm_function ConvDwFp32Border
     pushq %r15
     pushq %r14
     pushq %r13

mindspore/lite/nnacl/assembly/avx/ConvDwFp32RowAvx.S

@@ -1,12 +1,8 @@
 #ifdef ENABLE_AVX
-    .text
-    .align 4
-    .global ConvDwFp32Row
-#ifndef __APPLE__
-#ifndef WIN32
-    .type ConvDwFp32Row, %function
-#endif
-#endif
+#include "nnacl/assembly_global.h"
+
+.text
+.align 4
 
 // void ConvDwFp32Row(float *output_ptr, const float *input_tmp, const float *weight_ptr, size_t num_pixels,
 //                    size_t output_channel, size_t input_step);
@@ -26,7 +22,7 @@
 // 40: output_channel
 // 48: input_step
 
-ConvDwFp32Row:
+asm_function ConvDwFp32Row
     pushq %r15
     pushq %r14
     pushq %r13

mindspore/lite/nnacl/assembly/avx/MatmulAvx.S

@@ -1,13 +1,8 @@
 #ifdef ENABLE_AVX
 #include "nnacl/assembly_global.h"
-    .text
-    .align 4
-    .global MatmulFloatAvxOpt
-#ifndef __APPLE__
-#ifndef WIN32
-    .type MatmulFloatAvxOpt, %function
-#endif
-#endif
+
+.text
+.align 4
 
 // void MatmulFloatAvxOpt(const float *a, const float *b, float *c, const float *bias, int act_type, int depth
 //                        int row, int col, size_t stride, size_t writeMode)

mindspore/lite/nnacl/assembly/fp16/ConvDwFp16Border.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp16Border
-#ifndef __APPLE__
-.type ConvDwFp16Border, %function
-#endif
 
 // void ConvDwFp16Border(float16_t *dst, const float16_t *src, const float16_t *weight, const float16_t *bias,
 //                       size_t height, size_t width, size_t in_kh_step, size_t in_kw_step, size_t kernel_w, size_t relu,

mindspore/lite/nnacl/assembly/fp16/ConvDwFp16Center.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp16Center
-#ifndef __APPLE__
-.type ConvDwFp16Center, %function
-#endif
 
 // void ConvDwFp16Center(float16_t *dst, const float16_t *src, const float16_t *weight, const float16_t *bias, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step, size_t in_sw_step,

mindspore/lite/nnacl/assembly/fp16/ConvDwFp16Row.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global ConvDwFp16Row
-#ifndef __APPLE__
-.type ConvDwFp16Row, %function
-#endif
 
 // void ConvDwFp16Row(float16_t* output_ptr, const float16_t* input_ptr,const float16_t* filter_ptr,
 //                    size_t num_pixels, size_t input_channel, size_t input_step)

mindspore/lite/nnacl/assembly/fp16/DeconvDwFp16Border.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwFp16Border
-#ifndef __APPLE__
-.type DeconvDwFp16Border, %function
-#endif
 
 // void DeconvDwFp16Border(float *dst, const float *src, const float *weight, size_t height, size_t width,
 //                         size_t in_kh_step, size_t in_kw_step, size_t kernel_w)

mindspore/lite/nnacl/assembly/fp16/DeconvDwFp16Center.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global DeconvDwFp16Center
-#ifndef __APPLE__
-.type DeconvDwFp16Center, %function
-#endif
 
 // void DeconvDwFp16Center(float16_t *dst, const float16_t *src, const float16_t *weight, size_t height, size_t width,
 //                      size_t kernel_h, size_t kernel_w, size_t out_h_step, size_t block_channel, size_t in_sh_step, size_t in_sw_step,

mindspore/lite/nnacl/assembly/fp16/Float16ToFloat32.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global Float16ToFloat32
-#ifndef __APPLE__
-.type Float16ToFloat32, %function
-#endif
 
 // void Float16ToFloat32(const float16_t *input, float *output, int number);
 // x0: input, x1: output, x2: number

mindspore/lite/nnacl/assembly/fp16/Float32ToFloat16.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global Float32ToFloat16
-#ifndef __APPLE__
-.type Float32ToFloat16, %function
-#endif
 
 // void Float32ToFloat16(const float *input, float16_t output, int number);
 // x0: input, x1: output, x2: number

mindspore/lite/nnacl/assembly/fp16/IndirectGemmFp16_16x8.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global IndirectGemmFp16_16x8
-#ifndef __APPLE__
-.type IndirectGemmFp16_16x8, %function
-#endif
 
 // void IndirectGemmFp16_16x8(float16_t *output, float16_t *input, float16_t *weight, float16_t *bias,
 //     size_t step, size_t ic4, size_t oc8, size_t offset, size_t mode, size_t writeC4, size_t relu, size_t relu6);

mindspore/lite/nnacl/assembly/fp16/MatVecMulFp16.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatVecMulFp16Neon64
-#ifndef __APPLE__
-    .type MatVecMulFp16Neon64, %function
-#endif
+
+.text
+.align 5
 
 // void MatVecMulFp16Neon64(const float16_t *a, const float16_t *b, float16_t *c, const float16_t *bias, int act_type, int depth, int col)
 // x0: a

mindspore/lite/nnacl/assembly/fp16/MatmulFp16.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulFp16Neon64
-#ifndef __APPLE__
-    .type MatmulFp16Neon64, %function
-#endif
+
+.text
+.align 5
 
 // void MatmulFp16Neon64(const float16_t *a, const float16_t *b, float16_t *c, const float16_t *bias, int act_type,
 //                        int depth, int row, int col, int stride, bool write_nhwc)

mindspore/lite/nnacl/assembly/fp16/MatmulFp16Opt.S

@@ -1,11 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulFp16Neon64Opt
-#ifndef __APPLE__
-    .type MatmulFp16Neon64Opt, %function
-#endif
+
+.text
+.align 5
 
 // void MatmulFp16Neon64Opt(const float16_t *a, const float16_t *b, float16_t *c, const float16_t *bias, int act_type,
 //                          int depth, int row, int col, size_t stride, size_t writeMode)

mindspore/lite/nnacl/assembly/fp16/MatmulWinogradFp16.S

@@ -1,12 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global MatrixMultiplyWinogradFp16
-#ifndef __APPLE__
-.type MatrixMultiplyWinogradFp16, %function
-#endif
 
 // MatrixMultiplyWinogradFp16(float16_t *matix_a, float16_t *matrix_b, float16_t *matrix_c, int m, int k, int n, int in_channel)
     // x0: matrix_a, x1: matrix_b, x2: matrix_c, x3: m, x4: k, x5: n, x6: in_channel

mindspore/lite/nnacl/assembly/fp16/PostFuncBiasReluC4Fp16.S

@@ -1,12 +1,8 @@
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    //.p2align 5,,15
-    .global PostFuncBiasReluC4Fp16
-#ifndef __APPLE__
-    .type PostFuncBiasReluC4Fp16, %function
-#endif
+.text
+.align 5
 
 //void PostFuncBiasReluC4Fp16(float16_t *dst, const float16_t *src, const float16_t *bias, size_t oc4div, size_t oc4mod,
 //                            size_t plane_size, size_t plane_stride, size_t relu_type);
@@ -278,3 +274,5 @@ Loop_C1_3_Write:
 
 End:
   ret
+
+#endif

mindspore/lite/nnacl/assembly/fp16/PostFuncBiasReluC8Fp16.S

@@ -1,13 +1,8 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-    .text
-    .align 5
-    //.p2align 5,,15
-    .global PostFuncBiasReluC8Fp16
-#ifndef __APPLE__
-    .type PostFuncBiasReluC8Fp16, %function
-#endif
+.text
+.align 5
 
 //void PostFuncBiasReluC8Fp16(float *dst, const float *src, const float *bias, size_t oc8div,size_t oc8mod
 //                            size_t plane_size, size_t stride, int relu_type);

mindspore/lite/nnacl/assembly/fp16/TiledC4MatmulFp16.S

@@ -1,11 +1,8 @@
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
 .text
 .align 5
-.global TiledC4MatmulFp16
-#ifndef __APPLE__
-.type TiledC4MatmulFp16, %function
-#endif
 
 asm_function TiledC4MatmulFp16
 
@@ -258,3 +255,4 @@ LoopOcEnd:
     ld1 {v12.4s, v13.4s, v14.4s, v15.4s}, [sp], #64
     ret
 
+#endif

mindspore/lite/nnacl/assembly/fp16/WinogradTransLeftFp16.S

@@ -1,11 +1,8 @@
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-  .text
-  .align 5
-  .global WinogradTransLeftFp16
-#ifndef __APPLE__
-  .type WinogradTransLeftFp16, %function
-#endif
+.text
+.align 5
 
 asm_function WinogradTransLeftFp16
 
@@ -135,3 +132,5 @@ LoopH:
     sub sp, sp, #16
     ldp x19, x20, [sp], #16
     ret
+
+#endif

mindspore/lite/nnacl/assembly/fp16/WinogradTransRightFp16.S

@@ -1,11 +1,8 @@
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
 
-  .text
-  .align 5
-  .global WinogradTransRightFp16
-#ifndef __APPLE__
-  .type WinogradTransRightFp16, %function
-#endif
+.text
+.align 5
 
 asm_function WinogradTransRightFp16
 
@@ -138,4 +135,6 @@ LoopH:
     sub sp, sp, #16
     ldp x19, x20, [sp], #16
 
-    ret
+    ret
+
+#endif

mindspore/lite/nnacl/assembly/opt/MatmulDpInt8.S

@@ -1,11 +1,7 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulInt8DpNeon64
-#ifndef __APPLE__
-    .type MatmulInt8DpNeon64, %function
-#endif
+.text
+.align 5
 
 //void MatmulInt8DpNeon64(const int8_t *a, const int8_t *b, int8_t *dst, int row8, int col8, int deep4, 
 //                      const int *a_sums, const int *bias, int act_min, int act_max, int out_zp,

mindspore/lite/nnacl/assembly/opt/MatmulDpInt8Opt.S

@@ -1,11 +1,7 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatmulInt8DpOpt
-#ifndef __APPLE__
-    .type MatmulInt8DpOpt, %function
-#endif
+.text
+.align 5
 
 //void MatmulInt8DpOpt(const int8_t *a, const int8_t *b, int8_t *dst, int row, int col, int deep4, const int *a_sums,
 //                     const int *bias, int act_min, int act_max, int out_zp, int32_t *multiplier, int32_t *left_shift,

mindspore/lite/nnacl/assembly/opt/MatmulOptR4Int8.S

@@ -1,11 +1,7 @@
-#ifdef __aarch64__
+#ifdef ENABLE_ARM64
 #include "nnacl/assembly_global.h"
-    .text
-    .align 5
-    .global MatMulOptR4Int8Neon64
-#ifndef __APPLE__
-    .type MatMulOptR4Int8Neon64, %function
-#endif
+.text
+.align 5
 
 //void MatMulOptR4Int8Neon64(const int8_t *a, const int8_t *b, int *dst, int row4, int col4, int deep16,
 //                          const int *input_sum, const int *bias)

mindspore/lite/nnacl/assembly_global.h

@@ -16,18 +16,21 @@
 #ifndef MINDSPORE_LITE_NNACL_ASSEMBLY_GLOBAL_H
 #define MINDSPORE_LITE_NNACL_ASSEMBLY_GLOBAL_H
 
+// clang-format off
 .macro asm_function fname
 #ifdef __APPLE__
-  .globl _\fname;
-_\fname :
+.globl _\fname
+_\fname:
 #else
-  .global \fname;
-#ifdef __ELE__
-.hidden \fname;
-.type \fname, % function;
+.global \fname
+#ifdef __ELF__
+.hidden \fname
+.type \fname, %function
 #endif
-\fname :
+\fname:
 #endif
-      .endm
+.endm
+
+// clang-format on
 
 #endif  // MINDSPORE_LITE_NNACL_ASSEMBLY_GLOBAL_H