#include <windows.h> #include <stdio.h> #include <time.h> #include <stdlib.h> int main (/* daviddr 091129 */) { const int W = 80, H = 24, L = 12; const float PL = 2.5; float pos[W], step[W]; int x, y, c; char s[] = " "; srand (time(0)); SetConsoleTitle (L"Art"); HDC hdc = GetDC (FindWindow (0, L"Art")); SetBkColor (hdc,0); for (x=0; x<W; ++x) { pos [x] = rand() % int(H*PL); step[x] = float (rand() % 7 + 5) / 5; } for (;;Sleep(120)) for (x=0; x<W; ++x ) { for (y=0; y<H; ++y) { c = int (pos[x]) - y; c = ((0<c && c<L)? 255-c*255/L: 0); *s = 33 + rand() % 93; SetTextColor (hdc, c>225? 0xFFFFFF: RGB(0,c,0)); TextOutA (hdc, x*8, y*16, s, 1); } pos[x] += step[x]; if (pos[x] >= H*PL) { pos[x] = 0; step[x] = float (rand() % 7 + 5) / 5; } } }截圖：
float InvSqrt (float x) { float xhalf = 0.5f*x; int i = *(int*)&x; i = 0x5f3759df - (i>>1); x = *(float*)&i; x = x*(1.5f - xhalf*x*x); return x; }Haskell 著名示例：
qsort [] = [] qsort (x:xs) = qsort (filter (< x) xs) ++ [x] ++ qsort (filter (>= x) xs)談最佳化時常提及的 Duff's device，雖然只是 loop unrolling~
void strcpy (to, from, count) register char *to, *from; register int count; { int n = (count + 7) / 8; switch (count % 8) { case 0: do { *to = *from++; case 7: *to = *from++; case 6: *to = *from++; case 5: *to = *from++; case 4: *to = *from++; case 3: *to = *from++; case 2: *to = *from++; case 1: *to = *from++; } while (--n > 0); } }Evil C 裡的另一例：
switch (c&3) while ((c-=4)>=0) { foo(); case 3: foo(); case 2: foo(); case 1: foo(); case 0: }theycallhimtom post。給定 adjacency matrix，計算各個最短路徑：
for (int i = 0; i < adj.length; i++) for (int j = 0; j < adj.length; j++) for (int k = 0; k < adj.length; k++) adj[j][k] = min (adj[j][k], adj[j][i] + adj[i][k]);位元倒置：
unsigned int reverse (register unsigned int x) { x = ((x & 0xaaaaaaaa) >> 1) | ((x & 0x55555555) << 1); x = ((x & 0xcccccccc) >> 2) | ((x & 0x33333333) << 2); x = ((x & 0xf0f0f0f0) >> 4) | ((x & 0x0f0f0f0f) << 4); x = ((x & 0xff00ff00) >> 8) | ((x & 0x00ff00ff) << 8); return (x >> 16) | (x << 16); }ob-C Mandelbrot:
main(){float A,B,P,Q,X,Y,d;int i,D=80,n=3120;for(X=-2.,Y=-1.5,d=6./D;B=2*A
*B+Y,A=P-Q+X,n;((P=A*A)+(Q=B*B)>4||++i>D)&&putchar(*((n--%D?X+=d/2,i<D?i%11
:11:(X=-2.0,Y+=d,12))+"Mandelbrot! \n"))&&(A=B=P=Q=i=0));}
Another one#include <unistd.h>; float o=0.075,h=1.5,T,r,O,l,I;int _,L=80,s=3200;main(){for(;s%L|| (h-=o,T= -2),s;4 -(r=O*O)<(l=I*I)|++ _==L&&write(1,(--s%L?_<L?--_ %6:6:7)+"World! \n",1)&&(O=I=l=_=r=0,T+=o /2))O=I*2*O+h,I=l+T-r;}Ken Perlin 更短的..
main(k){float i,j,r,x,y=-16;while(puts(""),y++<15)for(x
=0;x++<84;putchar(" .:-;!/>)|&IH%*#"[k&15]))for(i=k=r=0;
j=r*r-i*i-2+x/25,i=2*r*i+y/10,j*j+i*i<11&&k++<111;r=j);}
FStruct s = {0};
或
FStruct s = {};
void search (char* query) { http://www.google.com/search?q=query return; }然後... Peter Mortensen 這樣寫到：
Fred* f = new(ram) Fred(); http://www.parashift.com/c++-faq-lite/dtors.html#faq-11.10 f->~Fred();同樣是 Peter Mortensen，身份變換 :-)
class A {}; struct B { A a; operator A&() { return a; } }; void func(A a) { } int main() { A a, c; B b; a = c; func (b); //yeah baby a = b; //gotta love this }Daniel Martin，Perl 運算式解析
sub e{($_)=@_;$n='( *-?[.\d]++ *)'; s:\($n\)|$n(.)$n:($1,0,$2*$4,$2+$4,0,$2-$4)[7&ord$3]//$2/$4:e?e($_):$_}gnibbler，Golfscript 文字模式 Sierpinski’s Triangle:
' /\ /__\ '4/{).+: ;.{ \ ++}%\{.+}%+~ ]}@~(*n*
惡作劇用：<td align="right" style="text-align: left">也是惡作劇用，fork bomb：
:(){ :|:& };:
放在 source code 最上方.. 然後執行它://&>/dev/null;x="${0%.*}";[ ! "$x" -ot "$0" ]||(rm -f "$x";cc -o "$x" "$0")&&"$x" $*;exit
In Quake 2 src:double normals[][] = { #include "normals.txt" };Multi-character constants:
enum state { stopped = 'STOP', running = 'RUN!', waiting = 'WAIT', };三元運算：
(a? b: c) = (b<c ? b: c);Patrick post，同樣是三元運算 (q = b ? x : y)
bool b; int x, y; int q = -!!b&x|-!b&y;算 PI:
double fldpi () { double pi; asm ("fldpi" : "=t" (pi)); return pi; }Static linked lists：
struct llist { int a; struct llist* next;}; #define cons(x,y) (struct llist[]) {{x,y}} struct llist *list = cons(1, cons(2, cons(3, cons(4, NULL))));整數求和:
#define sum(...) \ _sum(sizeof((int []){ __VA_ARGS__ })\ / sizeof(int), (int []){ __VA_ARGS__ }) int _sum (size_t count, int values[]) { int s = 0; while (count--) s += values[count]; return s; } int main () {printf("%i", sum (1,2,3));}上例的變形：
void f (int len, int *n, char c) {...} #define SIZEOF(a) sizeof(a)/sizeof*(a) #define NOPAREN(...) __VA_ARGS__ #define F1(arr,c) {int v[]={NOPAREN arr}; f(SIZEOF(v),v,c);} #define F2(c,...) {int v[]=__VA_ARGS__; f(SIZEOF(v),v,c);}再一種應用：
#include <algorithm>; #define MAX(x, ...) ({\ typeof(x) ra[] = {(x), __VA_ARGS__}; \ *std::max_element (&ra[0], &ra[sizeof ra/sizeof 0[ra]]); \ }) int main() { int i = 12; std::cout << MAX (i+1,1,3,6,i); }Johannes Schaub 寫的：
template <template From, template To> union union_cast { From from; To to; union_cast (From from): from(from) { } To getTo() const { return to; } };這是真的嗎 :-)
const MyClass& x = MyClass(); // temporary exists as long as x is in scopeRube 累計遞增:
def foo(n) lambda { |i| n += i } end測驗題，不用空白來定義結構。By Bojan Resnik：
typedef/**/struct{const/**/char/**/c;unsigned/**/u;}nospace;vividos 寫的：
int Func() try { // if success return 0; } catch(...) { return -1; }goto 和 dtor 的關係：
struct A { ~A() {cout<<"dtor";} }; int main() { { A a; goto L; } L: cout<<"After L"; }Evil C：
int logical_xor (int a, int b) { return !a != !b; }From Canadian Mind Products:
char *p; switch (n) { case 1: p = "one"; if (0) case 2: p = "two"; if (0) case 3: p = "three"; printf ("%s", p); break; }自動初始化 a 中元素為 0（那麼，本來會做嗎？）:
class C { int a[10]; C(): a() {} };template bitfields，by Kaz Dragon:
template <size_t X, size_t Y> struct bitfield { char left : X; char right : Y; };出自 Comptrol：
template <typename T> class Creator { friend void appear() { // a new function ::appear(), but it doesn't … // exist until Creator is instantiated } }; Creator<void> miracle; // ::appear() is created at this point Creator<double> oops; // ERROR: ::appear() is created a second time! template <typename T> class Creator { friend void feed(Creator<T>*){ // every T generates a different … // function ::feed() } }; Creator<void> one; // generates ::feed(Creator<void>*) Creator<double> two; // generates ::feed(Creator<double>*)用於 non-tmpl 的類 boost::identity。
template<typename T> struct id { typedef T type; }; // void (*f)(); // same id<void()>::type *f; // void (*f(void(*p)()))(int); // same id<void(int)>::type *f(id<void()>::type *p); // int (*p)[2] = new int[10][2]; // same id<int[2]>::type *p = new int[10][2]; // void (C::*p)(int) = 0; // same id<void(int)>::type C::*p = 0;原載於 Sumant Tambe 的 Blog:
template<class T> // (a) a base template void f( T ) { std::cout << "f(T)\n"; } template<> void f<>(int*) { // (b) an explicit specialization std::cout << "f(int *) specilization\n"; } template<class T> // (c) another, overloads (a) void f( T* ) { std::cout << "f(T *)\n"; } template<> void f<>(int*) { // (d) another identical explicit specialization std::cout << "f(int *) another specilization\n"; }用複雜方法做簡單的事? 出自 Marcus Lindblom:
template<int> class foo; template<> struct foo<0> { int* get() { return array; } int* array; }; template<int i> struct foo: public foo<i-1> { int* get() { return array + 1; } };
#include <ole2.h> ..... CLSID clsid; IDispatch pApp; CoInitialize (0); CLSIDFromProgID (L"Excel.Application", &clsid); CoCreateInstance (clsid, 0, CLSCTX_LOCAL_SERVER, IID_IDispatch, (void**)&pApp);(2) 用 wgunplot
template <typename T> inline void swap (T& a, T& b) { T t(a); a=b; b=t; }
#define SWAP(T,a,b) {T t=a; a=b; b=t;}
驗證：int main () { float a=-88, b=57; SWAP (float, a, b); printf ("\n a=%f b=%f", a, b); char *c="alcohol", *d="water"; SWAP (char*, c, d); printf ("\n c=%s d=%s", c, d); }
void swap_int (int* a, int* b)
{
int c = *a; *a = *b; *b = c;
}
不過上面的寫法在 compiler 未最佳化前，void swap2 (void* a, void* b) { int t = *(int*)a; *(int*)a = *(int*)b; *(int*)b = t; }
int t = *(int*)a; mov eax,dword ptr [a]
mov ecx,dword ptr [eax]
mov dword ptr [t],ecx
*(int*)a = *(int*)b; mov eax,dword ptr [a]
mov ecx,dword ptr [ b ]
mov edx,dword ptr [ecx]
mov dword ptr [eax],edx
*(int*)b = t; mov eax,dword ptr [ b ]
mov ecx,dword ptr [t]
mov dword ptr [eax],ecx
驗證：int main () { float a=-88.5, b=57.12; swap2 (&a, &b); printf ("\n a=%f b=%f", a, b); char c='5', d='9'; swap2 (&c, &d); printf ("\n c=%c d=%c", c, d); }
__asm {
mov eax, a
mov ebx, b
mov b , eax
mov a , ebx
}比較短，但慢些的寫法：
_asm {
mov eax, a
xchg eax, b
mov a , eax
}同樣短，但更慢：
_asm {
xchg a, eax
xchg b, eax
xchg a, eax
}比較慢，又額外浪費堆疊空間的寫法是：
_asm{
push a
push b
pop a
pop b
}或是浮點數專用的：
_asm{
fld a
fld b
fstp a
fstp b
}
是否存在不使用額外變數，而能將變數值正確交換的方法呢？
a ^= b ^= a ^= b; (1)
a -= b = (a+=b) - b; (2)
a -= b += a -= b = -b; (3)
a = a+b, b = a-b, a = a-b; (4)
a *= b, b = a/b, a /= b; (5)
(2)(3)(4) 其實是相同的邏輯。 //原始陣列 int A[] = { ... }; int B[] = { ... }; //用指標代表某個陣列 int * a = A; int * b = B; //交換所代表的陣列 int *t = a; a = b; b = t;
void swap (void** a, void** b) { void* t = *a; *a = *b; *b = t; } void show (void* a, int n) { int i, *p = (int*) a; //此處強制定成 int for (i=0; i<n; ++i) printf (" %d", p[i]); puts (""); } int main() { const int NUM = 8; int i; int A[NUM] = {1,2,3,4,5,6,7,8}; int B[NUM] = {11,12,13,14,15,16,17,18}; void* a = A; void* b = B; show (a, NUM); show (b, NUM); swap (&a, &b); show (a, NUM); show (b, NUM); }更一般化地，我們可用 Widget 模式包覆在非基本類型上，
void show (int* a, int n) { for (int i=0; i<n; ++i) printf (" %d", a[i]); puts (""); } void left_rotate (int* a, int n) { int t = a[0]; memmove (a, a+1, (n-1) * sizeof(int)); a[n-1] = t; } int main() { const int NUM = 8; int i=3, a[NUM] = {1,2,3,4,5,6,7,8}; while (i--) left_rotate (a, NUM), show (a, NUM); }
這是個 nano-second 精度的計時器，所做的僅是計算 MSR 差值。
在 multi-core processors 下，
各 CPU 的 TSC 同步率若不高，計時結果也會有誤差，
故建構 TimeProfile 時，便設法讓行程在單一 CPU 上跑。
設計上很簡便，但存在一些缺點：
1. 當他前幾次執行時，算出來的時間間隔會比較長。
因為 CPU 要花時間將算法的相關指令與資料 load 到 L2 甚至 L1 裡，
並調整諸如分支預測等機制的最佳模式。
2. Process 會因 context switch 被切出去，但 MSR 仍持續累加，
造成計時誤差。為避免此情形，可在檢測算法前，
將 process 與其中的 main thread 之 priority 提高，
待算完後再行回復：
DWORD oldProcess = GetPriorityClass (GetCurrentProcess());
DWORD oldThread = GetThreadPriority (GetCurrentThread());
SetPriorityClass (GetCurrentProcess(), REALTIME_PRIORITY_CLASS);
SetThreadPriority (GetCurrentThread(), THREAD_PRIORITY_TIME_CRITICAL);
...........[Algorithm]...........
SetThreadPriority (GetCurrentThread(), oldThread);
SetPriorityClass (GetCurrentProcess(), oldProcess);
但這步驟可能會讓其他 thread 無法順暢執行，
為保持 TimeProfile 的「輕鬆」簡便，便不加上這層考量。
3. 現今許多 CPU 為了省電節能，當處於閒置狀態會降低主頻，
如： Intel 的 EIST、AMD 的 C&Q，
故 TimeProfile 中額外定義了 set_freq，
以便在「需要」時重新讀取 CPU 頻率。
4. 需注意：真正的計時類別尚有許多細節待考慮，
本程式僅是個以簡便為主的粗糙設計。
#include <stdio.h> #include <windows.h> class TimeProfile { unsigned __int64 Hz; unsigned __int64 cycle() //讀取 64-bit 的 { //Model Specific Register (MSR) _asm { xor eax, eax _emit 0x0F //使用 cpuid 讓 CPU 不採無序執行指令 _emit 0xA2 //讓接下來的指令照流水線運行 _emit 0x0F //注意: 586 後的 IA32 才支援 RDTSC _emit 0x31 //結果置於 EDX:EAX } } public: void set_freq() { LARGE_INTEGER freq; QueryPerformanceFrequency (&freq); Hz = freq.QuadPart; } TimeProfile() { set_freq(); //若 multi-core 同步率高, 可不用加下面那行 SetThreadAffinityMask (GetCurrentThread(), 1); } double time; void start() {time = (double) cycle();} void end() {time = double (cycle()-time) /Hz;} };測試碼：
double cal_pi (int i = 50) { double pi = 2; for (; i>0; i--) pi = pi * double(i)/(2*i+1) + 2; return pi; } int main (/* daviddr 081211 */) { TimeProfile tp; double pi; for (int i=1; i<52; i+=5) { tp.start(); pi = cal_pi (i); tp.end(); printf ("\n[%2d] PI = %.15f Time: %.9f", i, pi, tp.time); } return system ("pause"); }從結果可以觀察到：第一次執行時，
[ 1] PI = 2.666666666666667 Time: 0.000004494
[ 6] PI = 3.132156732156732 Time: 0.000001445
[11] PI = 3.141358472520136 Time: 0.000001507
[16] PI = 3.141586396037061 Time: 0.000001674
[21] PI = 3.141592479958224 Time: 0.000001756
[26] PI = 3.141592648659952 Time: 0.000001881
[31] PI = 3.141592653447636 Time: 0.000002006
[36] PI = 3.141592653585648 Time: 0.000002130
[41] PI = 3.141592653589671 Time: 0.000002255
[46] PI = 3.141592653589790 Time: 0.000002379
[51] PI = 3.141592653589793 Time: 0.000002504
