{ Just for fun, I sat down last night and generated an inline asm version of the C code I posted that permuted character strings. This version uses about 90 clock cycles/call on a 486 and 60 on a Pentium, i.e. one microsecond/call on my P5-60: (The average time is very nearly independent of the length of the string.) From: terjem@hda.hydro.com (Terje Mathisen) } {$R-,S-} Program Permutate; Function Permute(var a; alen : word): Boolean; Assembler; asm push ds { BX -> beginning of array } lds bx,[a] mov di,[alen] { Array length } xor cx,cx { Return value = False } lea si,[di+bx-2]{ SI -> one character before last } lea di,[di+bx-1]{ DI -> last position in a[] } @loop1: cmp si,bx { Check if we've gotten to the beginning of the array!} jb @reverse { Yes, so no permutation found, return reversed array!} mov ax,[si] { Get another pair of bytes from the end } dec si cmp al,ah jae @loop1 { Loop until a[si] < a[si+1] } { We have found a pair of bytes where the first is less than the second, which means that there exists at least one more permutation of the a[] array. } mov bx,di { BX -> Last byte in a[] } inc si { SI was decremented one extra time, adjust back } mov cl,True { Return value = True } { Find the last byte which is > al (a[si]) } @loop2: mov dl,[bx] dec bx cmp dl,al jbe @loop2 { Swap the two positions found: } mov [si],dl mov [bx+1],al jmp @reverse { Reverse the rest of the array! } @loop3: mov al,[si] mov dl,[di] mov [di],al dec di mov [si],dl @reverse: inc si cmp si,di jb @loop3 mov ax,cx pop ds end; var test, org : String; n : LongInt; begin org := ParamStr(1); test := org; n := 0; if ParamCount > 1 then begin {Verbose version } repeat WriteLn(n:10,' ',test); Inc(n); until not Permute(test[1],Length(test)); WriteLn(n:10,' ',test); end else begin repeat Inc(n); until not Permute(test[1],Length(test)); WriteLn(n,' permutations of ',org,' found!'); end; end.