How to calc with div & mul?

[LCF-AT]

Hi guys,

sorry to create a new topic but I got a little stupid problem with some calculation and I don't get it work.

Problem: I have created a slidebar and did set the maximum range to 100 (dec) = used as 0-100 %.The problem now are the calculations.So I remember I have done this in the past with a progressbar & filesize (downloader) and there is was working but now I don't get it work anymore.

<0100739D>

	MOV EDI,0x60 ; full lenght
	NOP

@L00000001:
	MOV EAX,EDI
	MOV EDX,0x0
	MOV ECX,0x64 ; 100 dec
	DIV ECX
	CMP EDX,0x0
	JE SHORT @L00000002
	DEC EDI
	JMP SHORT @L00000001

@L00000002:
	MOV EDI,EAX
	MOV ESI,0x30 ; is lenght
	NOP

@L00000003:
	MOV EAX,ESI
	MOV EDX,0x0
	DIV ECX
	CMP EDX,0x0
	JE SHORT @L00000004
	DEC ESI
	JMP SHORT @L00000003

@L00000004:
	MOV ESI,EAX
	IMUL ESI,ESI,0x64 ; 100 dec
	MOV EAX,ESI
	MOV ECX,EDI
	MOV EDX,0x0
	DIV ECX
	NOP
	NOP

Normaly I should get 96 % from that calculation to use this value to set the position of slidebar but I get 0.So I do anything wrong so maybe you can help how to make it correctly and easy too if possible.

Thank you

[Kurapica]

Maybe because you are doing integer divisions ?

The progress is less than 100 and when you divide the current progress on 100 it will return 0.

maybe you should use a different approach or try with floating points instructions.

http://stackoverflow.com/questions/8804770/how-to-divide-floating-point-number-in-x86-assembly

[CodeExplorer]

0x60 = full lenght = 100%

0x30 = 50%

0x60                    ... 100 %
current progress ... ??? % (X)

X = (current_progress * 100) / 0x060

So for example:
if progress = 0x30 the result should be 50:
so we do:
x = (0x30 * 100) / 0x060
x = (48 dec * 100) / 96
x = 4800 / 96 = 50 %

So do: multiply current lenght with 100 (64h) and divide the result with full lenght
 

"Normaly I should get 96 % from that calculation to use this value to set the position of slidebar but I get 0"
if full lenght = 0x60 and you are position 0x30 you should get 50% from where 96 % ???

 

[CodeExplorer]

From Kurapica post:
"The progress is less than 100 and when you divide the current progress on 100 it will return 0. "
Whell you supose to multiply by 100 not divide!

 

 

[LCF-AT]

Hi guys and thanks for the answers.

Ok now I wrote this....

mov eax, 30 ; current
mov ecx, 64 ; 100
mov edx, 0
mul ecx

mov ecx, 60 ; full length
mov edx, 0
div ecx

.....and I got 50 & in dec.Oh my so simple and I did spend yesterday few hours and didn't got it work.Now it works and slider does move correctly so far.Thanks again very much and sorry for asking about it.

greetz

[LCF-AT]

Hi again,

have another small question.Can anyone tell me how to calc size bytes to QWORD = int64?Lets say I do alloc any section with a size of 30000 hex bytes so how I have to calc to get this size into QWORD = int64 pushed to stack?

greetz

[CodeExplorer]

So you need the size in qword (int 64) of 30000 hex bytes;
An qword = 8 bytes (from what I know)
So we do 30000 hex bytes / 8 bytes, so we do 30000 / 8
probable the result value needs to be runded so if reminder of divide operation (30000 / 8) is not 0 increment the result with 1
 

[LCF-AT]

Hi again and thanks so far.

So have I to work with any QWORD commands maybe?FISTP or so?Problem is I wanna use a API and there I have to use QWORD into stack.

datalength : QWORD

DataLength [in] Length of that file 

There is also a other API...

function BassVideo_GetLength(handle: DWORD): Double; stdcall;

If successful, then the channel's length in seconds is returned, else -1 is returned

...so here I can see Double = also QWORD (2 DWORDs ) right?Now I did debug a file what used this API and right after this API call I see that operation commands...

			sub esp,8
			FISTP QWORD PTR SS:[ESP]

...also I am not sure whether it wants the filesize as lenght or seconds.

function BassVideo_StreamCreateFileMem(data :Pointer; datalength : QWORD; MediaType, Flags, BassFlags : DWORD; CallBackProc : VIDEOPROC; user : Pointer): DWORD; stdcall;

Unfortunately there is no example file using this API above which I could check.

 

greetz

[LCF-AT]

Hi guys,

I have again another questions about calculations with values getting higher than -1h.All in all I have really trouble to work with float and not float together and now I thought there would be any other easier solution which I could maybe use to prevent massiv calculation etc.My question is: Are there maybe some API of any module out there which I could use to handle math operations (+/-/*/div/percent) only by strings?Lets say I have 2 ASCII strings and wanna add them and as results I will get also a ASCII string back.
 

1234567890 ; String 1
445454     ; String 2

1234567890+445454=1235013344

Invoke CalcAPIxy,String1,String2,addition,Buffer
=
Buffer "1235013344"

Maybe you know any helpfully APIs which I could use if there some who can do such string operations etc maybe any msvcrt API or so.Would be nice.

greetz

[T-rad]

I don't know of any API the will do math directly on / using ASCII string but  you could do something like this...

invoke sscanf, addr String1, chr$("%d"), addr buf1    ;"1234567890"
invoke sscanf, addr String2, chr$("%d"), addr buf2    ;"445454"

mov eax, buf1
mov edx, buf2
add eax, edx

invoke sprintf, Buffer, chr$("%d"), eax            ; "1235013344"

 

[LCF-AT]

Hi,

no it dosen't work like I expected.You only can handle values till max -1h on that way like you did post above.I also don't whether there are any API which I could use to do calc operations only with dez ASCII strings.

But thanks for trying to help T-rad

greetz

[T-rad]

no problem.

Just a note. When using chr$("%d") you are limited to the max value of an integer so you would change it according to your variable

chr$("%d")    signed decimal or integer
chr$("%i")    signed decimal or integer
chr$("%u")    unsigned decimal or integer
chr$("%ld")    signed long decimal or integer
chr$("%lu")    unsigned long decimal or integer
chr$("%hd")    short
chr$("%l")    long
chr$("%f")    float
chr$("%lf")    double
chr$("%Lf")    long double
chr$("%lld")    signed long long decimal or integer
chr$("%I64d")    signed long long decimal or integer
chr$("%llu")    unsigned long long decimal or integer
chr$("%I64u")    unsigned long long decimal or integer
chr$("%llx")    long long decimal or integer in hex format
chr$("%I64x")    long long decimal or integer in hex format

for example... if you were to do something using a long double like this

String1        db "12345678909876", 0
String2        db "445454.123", 0

invoke sscanf, addr String1, chr$("%Lf"), addr buf1
invoke sscanf, addr String2, chr$("%Lf"), addr buf2
fld [buf1]
fld [buf2]
fadd
fstp [buf1]
invoke sprintf, serial, chr$("%Lf"), buf1

the result would yield "12345679355330.123000"

or an unsigned 64 bit integer

String1        db "1234567890987654", 0
String2        db "445454", 0

invoke sscanf, addr String1, chr$("%llu"), addr buf1
invoke sscanf, addr String2, chr$("%llu"), addr buf2
fld [buf1]
fld [buf2]
fadd
fst [buf1]
invoke sprintf, serial, chr$("%llu"), buf1

the result would yield "1234567891433108"

[LCF-AT]

Hi and thanks for the infos so far.The example for unsigned 64 bit integer does not work for me.Maybe I wrote anything wrong.

00401C60  PUSH 0x40A960
00401C65  PUSH 0x4054DD                     ; ASCII "%llu"
00401C6A  PUSH 0x4054C5                     ; ASCII "1234567890987654"
00401C6F  CALL DWORD PTR DS:[0x404058]      ; msvcrt.sscanf
00401C75  ADD ESP,0xC
00401C78  PUSH 0x40A964
00401C7D  PUSH 0x4054E4                     ; ASCII "%llu"
00401C82  PUSH 0x4054D6                     ; ASCII "445454"
00401C87  CALL DWORD PTR DS:[0x404058]      ; msvcrt.sscanf
00401C8D  ADD ESP,0xC
00401C90  FLD DWORD PTR DS:[0x40A960]
00401C96  FLD DWORD PTR DS:[0x40A964]
00401C9C  FADDP ST(1),ST
00401C9E  FST DWORD PTR DS:[0x40A960]
00401CA4  PUSH DWORD PTR DS:[0x40A960]
00401CAA  PUSH 0x4054EC                     ; ASCII "%llu"
00401CAF  PUSH 0x40C978                     ; ASCII "1016588934"
00401CB4  CALL DWORD PTR DS:[0x40405C]      ; msvcrt.sprintf
00401CBA  ADD ESP,0xC

1016588934

1016588934 is the results I get.

.data
String1        db "1234567890987654", 0
String2        db "445454", 0
.data?
buf1 dd ?
buf2 dd ?
serial dd 256 dup (?)

invoke crt_sscanf, addr String1, chr$("%llu"), addr buf1
invoke crt_sscanf, addr String2, chr$("%llu"), addr buf2
fld [buf1]
fld [buf2]
fadd
fst [buf1]
invoke crt_sprintf, addr serial, chr$("%llu"), buf1

greetz

[T-rad]

dd is a dword (4 bytes) and not big enough to hold the the size of the int64. you should declare it as dq or qword (8 bytes)

Like this

.data?

buf1 dq ?
buf2 dq ?

 

 

 

Edited January 6, 2016 by T-rad

[LCF-AT]

Ok I did but still get not the right result back.

00401C60  PUSH 0x40A968
00401C65  PUSH 0x4054DD                   ; ASCII "%llu"
00401C6A  PUSH 0x4054C5                   ; ASCII "1234567890987654"
00401C6F  CALL DWORD PTR DS:[0x404058]    ; msvcrt.sscanf
00401C75  ADD ESP,0xC
00401C78  PUSH 0x40A970
00401C7D  PUSH 0x4054E4                   ; ASCII "%llu"
00401C82  PUSH 0x4054D6                   ; ASCII "445454"
00401C87  CALL DWORD PTR DS:[0x404058]    ; msvcrt.sscanf
00401C8D  ADD ESP,0xC
00401C90  FLD QWORD PTR DS:[0x40A968]
00401C96  FLD QWORD PTR DS:[0x40A970]
00401C9C  FADDP ST(1),ST
00401C9E  FST QWORD PTR DS:[0x40A968]
00401CA4  PUSH DWORD PTR DS:[0x40A96C]
00401CAA  PUSH DWORD PTR DS:[0x40A968]
00401CB0  PUSH 0x4054EC                   ; ASCII "%llu"
00401CB5  PUSH 0x40C988                   ; ASCII "1017034388"
00401CBA  CALL DWORD PTR DS:[0x40405C]    ; msvcrt.sprintf
00401CC0  ADD ESP,0x10

1017034388

0040A968  3C9EB694
0040A96C  00000000
0040A970  0006CC0E
0040A974  00000000

greetz

[kao]

For some reason, you have all zeroes at 0040A96C. It's incorrect, it should have value "000462d5". Quick look at the disassembly doesn't show anything particularly wrong, so I can only suggest that you debug it instruction-by-instruction.

For reference, here's the code produced by VS2010. I can see 2 main differences. First, it uses FADD QWORD PTR instead of FLD/FADDP. Second, it uses FLD/FSTP to prepare arguments for sprintf.

	sscanf("1234567890987654", "%llu", &a);
00D61269 68 30 62 D6 00       push        offset a (0D66230h)  
00D6126E 68 58 47 D6 00       push        offset string "%llu" (0D64758h)  
00D61273 68 44 47 D6 00       push        offset string "1234567890987654" (0D64744h)  
00D61278 FF 15 3C 72 D6 00    call        dword ptr [__imp__sscanf (0D6723Ch)]  
00D6127E 83 C4 0C             add         esp,0Ch  
	sscanf("445454", "%llu", &b);
00D61281 68 20 62 D6 00       push        offset b (0D66220h)  
00D61286 68 58 47 D6 00       push        offset string "%llu" (0D64758h)  
00D6128B 68 3C 47 D6 00       push        offset string "445454" (0D6473Ch)  
00D61290 FF 15 3C 72 D6 00    call        dword ptr [__imp__sscanf (0D6723Ch)]  
00D61296 83 C4 0C             add         esp,0Ch  
	c = a + b;
00D61299 DD 05 30 62 D6 00    fld         qword ptr [a (0D66230h)]  
00D6129F DC 05 20 62 D6 00    fadd        qword ptr [b (0D66220h)]  
00D612A5 DD 1D 28 62 D6 00    fstp        qword ptr [c (0D66228h)]  
	sprintf((char *)&result, "%llu", c);
00D612AB 83 EC 08             sub         esp,8  
00D612AE DD 05 28 62 D6 00    fld         qword ptr [c (0D66228h)]  
00D612B4 DD 1C 24             fstp        qword ptr [esp]  
00D612B7 68 58 47 D6 00       push        offset string "%llu" (0D64758h)  
00D612BC 68 20 61 D6 00       push        offset result (0D66120h)  
00D612C1 FF 15 40 72 D6 00    call        dword ptr [__imp__sprintf (0D67240h)]  
00D612C7 83 C4 10             add         esp,10h  

 

[T-rad]

my rip looks the same but everything is working here. I have attached my src so you can have a look.

sscanf app.rar

0040102A   PUSH Keygen.00403034
0040102F   PUSH Keygen.00403018                                       ; |format = "%llu"
00401034   PUSH Keygen.00403000                                       ; |s = "1234567890987654"
00401039   CALL DWORD PTR DS:[<&msvcrt.sscanf>]                       ; \sscanf
0040103F   ADD ESP,0C
00401042   PUSH Keygen.0040303C
00401047   PUSH Keygen.0040301D                                       ; |format = "%llu"
0040104C   PUSH Keygen.00403011                                       ; |s = "445454"
00401051   CALL DWORD PTR DS:[<&msvcrt.sscanf>]                       ; \sscanf
00401057   ADD ESP,0C
0040105A   FLD QWORD PTR DS:[403034]
00401060   FLD QWORD PTR DS:[40303C]
00401066   FADDP ST(1),ST
00401068   FST QWORD PTR DS:[403034]
0040106E   PUSH DWORD PTR DS:[403038]
00401074   PUSH DWORD PTR DS:[403034]                                 ; |
0040107A   PUSH Keygen.00403022                                       ; |format = "%llu"
0040107F   PUSH Keygen.00403044                                       ; |s = Keygen.00403044
00401084   CALL DWORD PTR DS:[<&msvcrt.sprintf>]                      ; \sprintf
0040108A   ADD ESP,10

 

[LCF-AT]

Hhmm.Thanks for the file / source but I get same as before too.See pic.

[ragdog]

 

Fpu for a simply Qword Addition?
Crazy coding

[simple]

None of the sample codes here have error checking. The return value of sscanf will indicate success/failure.

Besides, strtol(), strtoul(), etc (family of functions) are specifically designed for string/numerical conversions & probably the cleanest way.

Assuming you are using a MS compiler, atoi(), atof(), atoi64() etc are probably the easiest but they aren't cross platform. Personally, sscanf() would be my last choice for this, it has many odd things about it.

[kao]

2 hours ago, ragdog said:

Fpu for a simply Qword Addition?
Crazy coding

That's LCF-AT, always having "unconventional" ideas.  I'm already used to it..

[ragdog]

Quote

That's LCF-AT, always having "unconventional" ideas.  I'm already used to it..

Hehe but with Fpu is a really bad solution  No error checking no Overflow check.

And is not Optimized

[simple]

LCF, I had exact same problem as u w/Trad code, big number was getting cut off. For me, problem was %llu wasn't introduced till C99 or later and ur msvcrt is still running on C89. To fix this problem change the format string for arguments to sscanf & sprintf from "%llu" to C89 standard "%I64d" and it should work fine. When using masm w/libc functions probably best to use only C89. I still prefer other functions for this though.

edit - also to add 445454 all u have to do is use add instruction

Edited January 6, 2016 by simple
add

[LCF-AT]

Hi guys,

yes maybe I am the one who has unconventional ideas and also looking for unconventional solutions too. All in all I wanna find special solutions (simple of course if possible) to prevent unnecessary headache.Ok it works now with the hint of simple to change the format.Thanks again so far guys.

greetz

[T-rad]

question... whats so wrong with FPU for the addition

i can use 3-ish lines of code with FPU

fld [buf1]
fadd [buf2]
fst [buf1]

or 6 lines without it.

MOV EAX,DWORD PTR DS:[buf1]
MOV EDX,DWORD PTR DS:[buf1+4]
ADD EAX,DWORD PTR DS:[buf2]
ADC EDX,DWORD PTR DS:[buf2+4]

MOV DWORD PTR DS:[buf1],EAX
MOV DWORD PTR DS:[buf1+4],EDX

Please note that this is a genuine question. I normally code in C/C++ so i don't code in ASM much and would like to learn the proper do's and dont's.