久久精品国产99国产精品,精品久久久久成人码免费动漫,久久精品国产亚洲av影院

intern

Snape — Sat, 01 Sep 2012 17:16:00 GMT

1.软�gp4, vs, vim, cscope.
fix step�Q?br />先reproduce
再确定出现问题的函数�Q�行�?br />分析原因
fix

vs中远�E�调试，指定ip�Q�port�Q?同时在远�E�主��Z��要同意远�E�调试，接受指定user。要保证code与exe文�g一致�?br />讑֮�断点�Q�观察debug输入�Q�与预想的有什么不同。观察bt�Q�thread stack 扑ֈ�出现问题的函数或者行受��?br />
p4 中的check out, commit, diff, opened, edit�Q?sync命��o和作用�?br />submit step�Q�merge changlist (from branch) (to branch)
check diff
resolve
make (so make sure the code is right)
submit (add comments)
set the bug state in Web

vim:熟悉了各�U�命令�?br />
cscope�Q�ctag�Q�在linux中浏览代码很有用�Q�可以找到function definition, struct definition, calling function, called function.

在一个目录中建立cscope database�Q?可以再主目录建立�Q�之后只在这里��用cscope�Q�就能找到全部引用�?br />
linux�Q?lib 文�g的��用�?�?a �?o的lib文�g�?br />
2.code style.
文�g命名要有层次感。例如，snape_webserver_msg.c, snape_webserver_thread.c, snape_webserver_log.c, snape_client_msg.c,,,
函数命名也要有层�ơ感�Q?例如�Q�snape_webserver_msg_set_connection(), snape_webserver_thread_create(),snape_client_request_get_pic().
变量命名也要有层�ơ感和意义�?br />出错处理的专门函数�?br />debug level�Q�debug�Q�info, basic, webserver, client,,,

3.函数的定义与��览�?br />call graph definition.
函数名字最好能够显�C�出函数调用的graph�?br />使用hash table 保存大量�c�d��的数据�?br />使用内容的md5作�ؓ�q�个内容的id�Q�就可以完美配合hash table�?br />
4.thread�Q?process
windows 中， object , event, cs
cs同步速度较快�Q�但是��用cs�Ҏ��q�入deadlock状态。因为在�{�待�q�入cs�Ӟ��无法讑֮��时倹{�?br />互斥对象与内核对象属于内核对象，利用内核对象�q�行�U�程同步�Q�同步速度较慢�Q�但�q�种方式可在多个�q�程的各个对象之间进行同步�?br />event分�ؓ人工重置与自动重�|�事�Ӟ��两者在�l�节上，不同�?br />
5.cache implementation
client端可以��用类似os中，cache �?虚拟内存的方法。b_modified表示是否�l�过更改�Q�从而是否需要更新。每�ơ只��L��新的东西�?br />而在server端，可以使用内容的MD5判断是否需要处理来到的内容�Q�可以用多层�ơ的MD5来更加去improve performance。例如用��M��的md5�Q�和各个部分的md5.
先判断��M��MD5�Q�如果改变，在去��查部分md5.如果没有改变�Q�就整个都可以丢弃，不去处理�?br />design is very important. 代码的执行过�E�要完全依赖design�?br />
6.select function
异步select�Q?可以让线�E�不去busy waiting�?br />但是如果需要传输内容，需要trigger signal�Q�让它不用等到timeout�Q�就能发送数据�?br />
7.timestamp/timeout/ ts_diff
可以用ts_diff��d��步时间�?br />使用timeout来防止�Q意操作的无限制等�?br />使用timestamp来决定哪一个更新。不要用counter。（无法同步�Q�无法控制单一性）

8.xml api
在传送数据方面，可以使用xml

9.performace/memory usage
mm tools.

Snape 2012-09-02 01:16 发表评论

Snape — Mon, 26 Mar 2012 17:20:00 GMT

�Ҏ��1:

1 #include
2 using namespace std;
3
4 //by snape 2012-3-25
5 //method 1: drawback is we need calculate the one-dimensional index to access the 2D array
6 int main()
7 {
8     int rowSize, colSize, totalSize;
9     int index, i,j;
10
11     cout<<"Enter the row and column size for your 2D array!"<12     cin>>rowSize>>colSize;
13
14     totalSize=rowSize*colSize;
15     int *pArray;
16     pArray=new int[totalSize];
17
18     //file  the array with integers from 0 to totalsize
19     //file across the rows, moving down the colums
20
21     int arrayValue=0;
22     for(i=0; i//outer-loop traverse down the "rows"
23     {
24         for(j=0; j25         {
26             //caculate array index
27             //index=rowSize*j+i;        //both index=rowSize*j+i; and index=colSize*i+j; are ok
28             index=colSize*i+j;            //but if index=rowsize*i+j; or index=colSize*j+i; then there will be a bug.
29             pArray[index]=arrayValue;    // i like index=colSize*i+j; since the arrange of 2D is according to rows
30             cout<<"index="<31             cout<<"i(row)="<32             ++arrayValue;
33         }
34     }
35
36     //output the array
37     for(int k=0; k38     {
39         cout<40     }
41     cout<<"The End"<42     delete [] pArray;
43     return 0;
44 }
45
46

�Ҏ��2�Q?/span>

1 //by snape
2 //method 2: better than method 1, but call new twice
3 int main()
4 {
5     int rowSize, colSize, totalSize;
6     int i, j;
7     cout<<"Enter the row and column size for your 2D array"< 8     cin>>rowSize>>colSize;
9
10     totalSize=rowSize*colSize;
11
12     int *pArray; //pointer to an integer
13     int **pPointerArray; //pointer to an integer pointer
14
15     pArray=new int[totalSize];    //memory for totalSize integers
16     pPointerArray=new int*[rowSize];    //memory for rowSize # of int pointers
17
18     //fill the pointer array with the pArray[i][0] address
19     for(i=0; i20         pPointerArray[i]=pArray+i*colSize;    //place the address into the pointer
21
22     //now fill the pArray by using the pPointerArray to access elements
23     int arrayValue=0;
24     for(i=0; i25     {
26         for(j=0; j27         {
28             pPointerArray[i][j]=arrayValue;    //cool
29             cout<<"i(row)="<30             ++arrayValue;
31         }
32     }
33
34         //output the array
35     for(int k=0; k36     {
37         cout<38     }
39     cout<<"The End!"<40     delete [] pArray;
41     delete [] pPointerArray;
42     return 0;
43 }

�Ҏ��3�Q?/span>

1 // by snape
2 //method 3: better than method 2. just malloc once and the memory is contiguous block. the best
3 int **my2DAlloc(int rowSize, int colSize)
4 {
5     int i;
6     int header=rowSize * sizeof(int *);
7     int data=rowSize * colSize * sizeof(int);
8     int **rowptr=(int **)malloc(header+data);    //malloc memory for both data and pointerArray(the header)
9
10     if(rowptr==NULL)
11         return NULL;
12     int *buf=(int *)(rowptr+rowSize);    //buf: the pointer to the first data
13     for(i=0; i//assign the address of each row to pointerArray(the header)
14         rowptr[i]=buf+i*colSize;
15
16     return rowptr;
17 }
18
19 int main()
20 {
21     cout<<"Enter the row and column size for your 2D array"<22     int rowSize, colSize;
23     cin>>rowSize>>colSize;
24     int **p=my2DAlloc(rowSize, colSize);
25
26     //assign values
27     int i, j, arrayValue=0;
28     for(i=0; i29         for(j=0; j30             p[i][j]=arrayValue++;
31
32     //output values
33     for(i=0; i34         for(j=0; j35             cout<36
37     free((void *)p);
38 }

�Ҏ��3�Q�我感觉最好，只调用一�ơmalloc�Q?�I�间�q�箋�Q�释攑ֆ�存也比较方便�?/span>

大家有什么想法的�Ƣ迎交流

Snape 2012-03-27 01:20 发表评论

C 边界寚w��的malloc和free函数实现

Snape — Mon, 26 Mar 2012 17:19:00 GMT

MSVC 下面有：

1
2 void * _aligned_malloc(
3 size_t size,
4 size_t alignment
5 );

Unix/Linux下面有：

1 int posix_memalign(void **memptr, size_t alignment, size_t size);

自己实现�Q?br />

1 #include
2 #include
3 using namespace std;
4
5 //by snape 2012-3-25
6
7 void *aligned_malloc(size_t required_bytes, size_t alignment)
8 {
9     void *p1;    //orignal block
10     void **p2;    //aligned block
11
12     int offset=alignment-1+sizeof(void*);
13     if( (p1=(void*)malloc(required_bytes+offset))==NULL )
14         return NULL;
15     //cast void* to void** to let p1 store into heap in the form of pointer
16     p2=(void**)( ((size_t)(p1)+offset) & ~(alignment-1) );    //add offset to p1 and then back to the aligned address
17     p2[-1]=p1;
18     return p2;
19 }
20
21 void aigned_free(void *p2)
22 {
23     void* p1=((void**)p2)[-1];    //get the p1
24     free(p1);
25 }
26
27
28
29 int main()
30 {
31     int size=10, align=16;
32     int *p=(int *)aligned_malloc(size, align);
33
34     cout<<"The start address="<35
36     aigned_free(p);
37 }

上面实现很考验�Ҏ��针的掌握。原理不是很难�?br />
Line12�Q�分配内存的时候，加上alignment-1个空��_��p��保证能够扑ֈ�和alignment寚w��的开始地址。多出来的sizeof�Q�void*�Q�是��Z��保存malloc最初返回的指针p1而分配的一个空�?br />
Line 17: 保存原始的malloc�q�回的地址p1到对齐地址的前一个空�?/div>

Snape 2012-03-27 01:19 发表评论