VS2010中内联汇编入门--cdecl

水深无声 2023-10-11 10:30 19阅读 0赞

在vs2010中进行内联汇编时，注意在被调函数中，编译器自动在栈中为局部变量预留一个区域，如下面的反汇编代码所示，ebp-0xc0，并且初始化为0xcccccccc

int __cdecl mse_sse(unsigned char * a, unsigned char * b, int len){
    00B213B0  push        ebp  
    00B213B1  mov         ebp,esp  
    00B213B3  sub         esp,0C0h  
    00B213B9  push        ebx  
    00B213BA  push        esi  
    00B213BB  push        edi  
    00B213BC  lea         edi,[ebp-0C0h]  
    00B213C2  mov         ecx,30h  
    00B213C7  mov         eax,0CCCCCCCCh  
    00B213CC  rep stos    dword ptr es:[edi]

参数传递：

在cdecl调用下，参数由上层调用函数进行入栈，在内联汇编模式下，访问参数要用中括号加参数名称的方式获取参数

mov  esi,[a];//a
    00B213CE  mov         esi,dword ptr [a]  
    		mov  edi,[b];//b
    00B213D1  mov         edi,dword ptr [b]  
    		mov  ecx,[len];//len
    00B213D4  mov         ecx,dword ptr [len]

不要使用下面的方式：

mov  esi,[ebp+8];//a
    		mov  edi,[ebp+8+4];//b
    		mov  ecx,[ebp+8+8];//len

这种方式在DEBUG下是正确的，但是在Release下会报错  
  
下面是一个求MSE的小函数

int __cdecl mse_sse(unsigned char * a, unsigned char * b, int len){
    	__asm{
    		/*
    		mov  esi,[ebp+8];//a
    		mov  edi,[ebp+8+4];//b
    		mov  ecx,[ebp+8+8];//len
    		*/
    		mov  esi,[a];//a
    		mov  edi,[b];//b
    		mov  ecx,[len];//len
    		
    		pxor   xmm7,xmm7;//128bit 0
    		movdqu xmm0,[esi];//a[0-7]
    		punpcklbw xmm0,xmm7;//8bit->16bit
    		movdqu xmm1,[edi];//b[0-7]
    		punpcklbw xmm1,xmm7;//8bit->16bit
    		psubw  xmm0,xmm1;//a-b
    		pabsw  xmm0,xmm0;	 //|a-b|
    		//movdqu [esi],xmm0
    		pmullw xmm0,xmm0;//|a-b|^2
    		phaddw xmm0,xmm0;//0,1->0
    		phaddw xmm0,xmm0;//0,1,2,3->0
    		phaddw xmm0,xmm0;//0,...,7->0
    		
    		movd    eax,xmm0;
    	}
    }

完整的代码：

//fiel:main.c
    #include<stdlib.h>
    #include<stdio.h>
    #include"mem_align.h"
    #include "asm.h"
    struct sByte{
    	char c;
    	int  i;
    } aa;
    
    int main(int argc, char * argv[])
    {
    	int size =30;
    	int align_size =16;
    	//unsigned char * pVar_mem=(unsigned char *)(malloc(size+align_size));\
    	//unsigned char * pVar = pVar_mem+(align_size-(int)pVar_mem&0xf);
    	align_mem_get(pBuffer,30,16);
    	struct sByte bb;
    	unsigned char a[32]={0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7}; 
    	unsigned char b[32]={0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7};
    	int len = 8;
    	int mse=mse_sse_v2(a,b,len);
    	printf("mse = : %d \n",mse);
    
    	mse=mse_sse(a,b,len);
    	printf("mse = : %d \n",mse);
    	printf("%d \n",sizeof(aa));
    	printf("%d \n",sizeof(bb));
    	printf("align_mem_get pBuffer : 0x%08x \n",(int)pBuffer);
    	printf("align_mem_get pBuffer : 0x%08x mod 16 = %d \n",(int)pBuffer,(int)pBuffer%16);
    	align_mem_free(pBuffer);
    	return 0;
    }

/*
    FILE: mem_align.h
    */
    #ifndef _MEM_AQLIGN_H_
    #define _MEM_AQLIGN_H_
    
    #define align_mem_get(pVar,size,align_size)\
    	unsigned char * pVar_mem=(unsigned char *)(malloc(size+align_size));\
    	unsigned char * pVar = pVar_mem+(align_size-(int)pVar_mem&0xf)
    
    #define align_mem_free(pVar) \
    	free(pVar_mem);  \
    	pVar = 0
    #endif

// asm.h
    #ifndef _ASM_H_
    #define _ASM_H_
    #include <xmmintrin.h>//SSE
    #include <emmintrin.h>//SSE2
    #include <pmmintrin.h>//SSE3
    #include <tmmintrin.h>//SSSE3
    #include <smmintrin.h>//SSE4.1
    #include <nmmintrin.h>//SSE4.2
    #include <immintrin.h>//AVX,AVX2,FMA
    
    //__declspec(naked) static int mse_sse(unsigned char * a, unsigned char * b, int len);
    int __cdecl mse_sse(unsigned char * a, unsigned char * b, int len);
    int __cdecl  mse_sse_v2(unsigned char * a, unsigned char * b, int len);
    #endif

//file:asm.c
    #include "asm.h"
    //8x8 16bit unsigned short 
    //int __stdcall mse_sse(unsigned char * a, unsigned char * b, int len){
    //__declspec(naked) static int mse_sse(unsigned char * a, unsigned char * b, int len){
    int __cdecl mse_sse(unsigned char * a, unsigned char * b, int len){
    	__asm{
    		/*
    		mov  esi,[ebp+8];//a
    		mov  edi,[ebp+8+4];//b
    		mov  ecx,[ebp+8+8];//len
    		*/
    		mov  esi,[a];//a
    		mov  edi,[b];//b
    		mov  ecx,[len];//len
    		
    		pxor   xmm7,xmm7;//128bit 0
    		movdqu xmm0,[esi];//a[0-7]
    		punpcklbw xmm0,xmm7;//8bit->16bit
    		movdqu xmm1,[edi];//b[0-7]
    		punpcklbw xmm1,xmm7;//8bit->16bit
    		psubw  xmm0,xmm1;//a-b
    		pabsw  xmm0,xmm0;	 //|a-b|
    		//movdqu [esi],xmm0
    		pmullw xmm0,xmm0;//|a-b|^2
    		phaddw xmm0,xmm0;//0,1->0
    		phaddw xmm0,xmm0;//0,1,2,3->0
    		phaddw xmm0,xmm0;//0,...,7->0
    		
    		movd    eax,xmm0;
    	}
    }
    int __cdecl mse_sse_v2(unsigned char * a, unsigned char * b, int len){
    	int x = (int)a;
    	return len;
    }

【资料】比较好的说明： [ http://blog.csdn.net/kzh313561014/article/details/7283765][http_blog.csdn.net_kzh313561014_article_details_7283765]

【GGC 下的 at&t 汇编】[http://blog.csdn.net/zhangqingsup/article/details/6704253][http_blog.csdn.net_zhangqingsup_article_details_6704253]

[http_blog.csdn.net_kzh313561014_article_details_7283765]: http://blog.csdn.net/kzh313561014/article/details/7283765
[http_blog.csdn.net_zhangqingsup_article_details_6704253]: http://blog.csdn.net/zhangqingsup/article/details/6704253