Stacks and counted strings

[nidud]

deleted

[jj2007]

Code Select Expand


push ebp
mov ebp,esp
..
leave
or
mov esp,ebp
pop ebp


Added to testbed as npush/npop (without error check). I chose the leave variant because it is much faster - you can try the other one by setting useleave=0 in line 3. Results (cycles for 200 push/pop pairs):

Code Select Expand

Intel(R) Core(TM) i5-2450M CPU @ 2.50GHz (SSE4)

270     cycles for 100 * dpush/dpop with error check
1221    cycles for 100 * rpush/rpop
1214    cycles for 100 * npush/npop
216     cycles for 100 * dpush/dpop without error check

271     cycles for 100 * dpush/dpop with error check
1222    cycles for 100 * rpush/rpop
1191    cycles for 100 * npush/npop
220     cycles for 100 * dpush/dpop without error check

271     cycles for 100 * dpush/dpop with error check
1220    cycles for 100 * rpush/rpop
1224    cycles for 100 * npush/npop
219     cycles for 100 * dpush/dpop without error check

272     cycles for 100 * dpush/dpop with error check
1257    cycles for 100 * rpush/rpop
1200    cycles for 100 * npush/npop
215     cycles for 100 * dpush/dpop without error check

270     cycles for 100 * dpush/dpop with error check
1218    cycles for 100 * rpush/rpop
1205    cycles for 100 * npush/npop
216     cycles for 100 * dpush/dpop without error check

54      bytes for dpush/dpop with error check
25      bytes for rpush/rpop
45      bytes for npush/npop
28      bytes for dpush/dpop without error check

Note also that those versions that push downwards (as in the normal push/pop via esp) cannot simply extend the stack with HeapReAlloc.

[nidud]

deleted

[jj2007]

Also keep in mind that you will have to restore the stack if you make a call to an external proc.

I see ::)

So npush & npop would only be applicable in areas of the program that don't use things like print, Msgbox, InString, len(), lstrcpy, val(), ...?

[nidud]

deleted

[jj2007]

Code Select Expand

  mov edi, AlgoLoops-1 ; loop e.g. 100x
  push ebp
  mov ebp,esp
  mov esp, offset dataStack+1000h ; end of stack area

  align 4
  .Repeat
npush 123
npop eax
npush eax
npop ecx
dec edi
  .Until Sign?
  leave
  ret

Shouldn't that be like that (for reasons of comparability)? See Reply #18 for possible reasons.

Code Select Expand

  mov edi, AlgoLoops-1 ; loop e.g. 100x

  align 4
  .Repeat
        push ebp
        mov ebp,esp
        mov esp, offset dataStack+1000h ; end of stack area
npush 123
npop eax
npush eax
npop ecx
       leave
dec edi
  .Until Sign?
  leave
  ret

[nidud]

deleted

[jj2007]

You were thinking of something like this:

Code Select Expand

dpErrCheck=1 ; do some MasmBasic stuff here..

The code I posted is plain assembler, no MasmBasic in there. But my dpush/dpop allows to use print, InString, len(), lstrcpy, val(), ... and every imaginable WinAPI call in parallel. I doubt that CCurl can implement a FORTH machine without ever using the coder's bread-and-butter functions. Besides, my version is short and fast. But you know what? Assembler is learning by crashing, pardon: doing. Nobody is obliged to use somebody else's code, even if it's shorter and faster. Roll your own, homegrown code is the best :P