atol, atodw and negative numbers

[jj2007]

I seem to have difficulties to interpret the MasmLib help correctly. When I pass a string starting with a minus sign to atol and atodw, I get rubbish:

Code Select Expand

Positive, atodw:        12345
Positive, atol:         12345
Negative, atodw:        25312345
Negative, atol:         -112345

Any explanation for this?

Thanks,
jj

Code Select Expand

include \masm32\include\masm32rt.inc

.code
StrPos db "12345", 0
StrNeg db "-12345", 0

start:
print "Positive, atodw: ", 9
invoke atodw, offset StrPos
print str$(eax), 13, 10
print "Positive, atol:  ", 9
invoke atol, offset StrPos
print str$(eax), 13, 10

print "Negative, atodw: ", 9
invoke atodw, offset StrNeg
print str$(eax), 13, 10
print "Negative, atol:  ", 9
invoke atol, offset StrNeg
print str$(eax), 13, 10

getkey
exit

end start

[Jimg]

Running your code unchanged, atol works properly for me.
Clearly, atodw is not meant for signed strings, and should be documented as such.

[drizz]

atodw proc:

Code Select Expand

   cmp al, 2D
   jne proceed


should be 2DH or
cmp al, '-'

[jj2007]

Running your code unchanged, atol works properly for me.

I get an extra 1:
Negative, atol:         -112345

My atol.asm has 1327 bytes, 13.08.2006

EDIT: Ze bug is here:

atol proc lpSrc:DWORD
    xor eax, eax                ; zero EAX
    mov edx, [esp+4]
    movzx ecx, BYTE PTR [edx]
    add edx, 1

    cmp ecx, "-"                ; test for sign
    jne lbl0
    add eax, 1                  ; set EAX if sign
   
   if 1   ; corrected
      movzx ecx, BYTE PTR [edx+1]
      add edx, 2
   else   ; original
      movzx ecx, BYTE PTR [edx]
      add edx, 1
   endif

  lbl0:
    push eax                    ; store sign on stack
...

[Jimg]

Mine is 1445 bytes from 28 may 2008.

Time to update.

It still works and looks like-

Code Select Expand

atol proc lpSrc:DWORD

    xor eax, eax                ; zero EAX
    mov edx, [esp+4]
    movzx ecx, BYTE PTR [edx]
    add edx, 1
    cmp ecx, "-"                ; test for sign
    jne lbl0
    add eax, 1                  ; set EAX if sign
    movzx ecx, BYTE PTR [edx]
    add edx, 1

  lbl0:
    push eax                    ; store sign on stack
    xor eax, eax                ; so eax*10 will be 0 for first digit

  lbl1:
    sub ecx, 48
    jc  lbl2
    lea eax, [eax+eax*4]        ; mul eax by 5
    lea eax, [ecx+eax*2]        ; mul eax by 2 and add digit value
    movzx ecx, BYTE PTR [edx]   ; get next digit
    add edx, 1
    jmp lbl1

  lbl2:
    pop ecx                     ; retrieve sign
    test ecx, ecx
    jnz lbl3
    ret 4

  lbl3:
    neg eax                     ; negative return value is sign set
    ret 4

atol endp


[Jimg]

Good catch Drizz, (and jj).  Hard to believe it's been around this long with that error.

[PBrennick]

Very nice catch, JJ. I think it needs to be fixed via the next service pack. I guess not many people have been passing negative numbers to that algo else it may have been caught sooner. You are a very meticulous individual.

Paul

[mitchi]

These are my results here :

Positive, atodw:        12345
Positive, atol:         12345
Negative, atodw:        25312345
Negative, atol:         -12345

[jj2007]

I found a newer version on my other machine...

Just for fun:

Code Select Expand

Intel(R) Celeron(R) M CPU        420  @ 1.60GHz (SSE3)

31      cycles for atodjj, offset StrNeg
45      cycles for atol,   offset StrNeg

31      cycles for atodjj, offset StrNeg
45      cycles for atol,   offset StrNeg

31      cycles for atodjj, offset StrPos
57      cycles for atol,   offset StrPos
45      cycles for atodw,  offset StrPos

31      cycles for atodjj, offset StrPos
57      cycles for atol,   offset StrPos
43      cycles for atodw,  offset StrPos

Positive, atodw:        123456789
Positive, atol:         123456789
Positive, atodjj:       123456789
Negative, atol:         -123456789
Negative, atodjj:       -123456789

Code sizes:
atodjj:         58
atol:           61
a2dw:           71

EDIT: A bit of shuffling, an unrolled loop - and voilà, it's speeding up :bg

[attachment deleted by admin]

[lingo]

New results:  :lol

Code Select Expand


Intel(R) Core(TM)2 Duo CPU     E8500  @ 3.16GHz (SSE4)

22      cycles for atodjj, offset StrNeg
18      cycles for a2dLingo, offset StrNeg
40      cycles for atol,   offset StrNeg

22      cycles for atodjj, offset StrNeg
18      cycles for a2dLingo, offset StrNeg
40      cycles for atol,   offset StrNeg

22      cycles for atodjj, offset StrPos
18      cycles for a2dLingo, offset StrPos
38      cycles for atol,   offset StrPos
39      cycles for atodw,  offset StrPos

22      cycles for atodjj, offset StrPos
18      cycles for a2dLingo, offset StrPos
38      cycles for atol,   offset StrPos
42      cycles for atodw,  offset StrPos

Positive, atodw:         123456789
Positive, a2dLingo:      123456789
Positive, atol:          123456789
Positive, atodjj:        123456789
Negative, atol:         -123456789
Negative, atodjj:       -123456789
Negative, a2dLingo:     -123456789

Code sizes:
atodjj:         58
a2dLingo:      113
atol:           61
a2dw:           71
[attachment deleted by admin]

[jj2007]

Not bad, my young friend, although it's a bit bloated, and on my archaic CPU it's on average a cycle slower :green

:U

Code Select Expand

              Intel(R) Pentium(R) 4 CPU 3.40GHz (SSE3)

57      cycles for atodjj, offset StrNeg
54      cycles for a2dLingo, offset StrNeg
52      cycles for atol,   offset StrNeg

53      cycles for atodjj, offset StrNeg
54      cycles for a2dLingo, offset StrNeg
52      cycles for atol,   offset StrNeg

51      cycles for atodjj, offset StrPos
53      cycles for a2dLingo, offset StrPos
52      cycles for atol,   offset StrPos
59      cycles for atodw,  offset StrPos

51      cycles for atodjj, offset StrPos
55      cycles for a2dLingo, offset StrPos
52      cycles for atol,   offset StrPos
59      cycles for atodw,  offset StrPos

Positive, atodw:        123456789
Positive, a2dLingo:     123456789
Positive, atol:         123456789
Positive, atodjj:       123456789
Negative, atol:         -123456789
Negative, atodjj:       -123456789
Negative, a2dLingo:     -123456789

Code sizes:
atodjj:         58
a2dLingo:       113
atol:           61
a2dw:           71

[MichaelW]

I split this topic and put the parts that definitely do not belong in the Campus here:

http://www.masm32.com/board/index.php?topic=11280.0