[C++] War3 key algorithms

[Eibro]

C++ implementation of war3 key algorithms reversed from install.exe. This was an independant effort; it has nothing to do with the other war3 key code floating around. Email questions/comments to [email protected]

Get it at http://cs.smu.ca/~e_brooks/war3/
Future updates (if any) will be posted there.

The IDB is there for IDA 4.5 if anyone is interested in looking at the reversal. Also, an idc script (maketable.idc) is there that was used to format the key tables into an array.

[Skywing]

It's not really a C++ implementation if it's mostly inline assembler :p

[Eibro]

It's not really a C++ implementation if it's mostly inline assembler :p

Figured someone would say that. If I get some time (and interest) i'll convert the remaining parts (see the readme).

[Skywing]

It's not really a C++ implementation if it's mostly inline assembler :p

Figured someone would say that. If I get some time (and interest) i'll convert the remaining parts (see the readme).

BTW, there are C++ equivalents for what you call multPass -- it's 64-bit math, though.

[Eibro]

It's not really a C++ implementation if it's mostly inline assembler :p

Figured someone would say that. If I get some time (and interest) i'll convert the remaining parts (see the readme).

BTW, there are C++ equivalents for what you call multPass -- it's 64-bit math, though.

Yeah, I figured as much. Though I couldn't get the compiler to emit the correct instructons (see post in the assembly forum)

[iago]

when we did it, Maddox was stuck on that line for quite awhile.  __int64 or long long works in C++ (VS and gcc respectively), and long works in Java.

[Arta]

I used LARGE_INTEGER, which works quite nicely. Translates easily from the assembly. That is, if we're talking about the same bit

[Eibro]

Ah, that is how I did it the first time through, but it wasn't generating the correct instructions so I scrapped it. I guess now that that problem is solved I should roll back to my previous code.

Code Select Expand

union LargeInt {
struct {
unsigned long low;
unsigned long high;
};
int64 val;
};   */
BTW, we're talking about the code inside what I called 'tableMult' (.text:0041D050)

[Arta]

I can't find anything there in my notes, I may have used TFT though.

[Maddox]

It's not really a C++ implementation if it's mostly inline assembler :p

Looks more like a C implementation.

This is the way I ended up doing that specific function.

Code Select Expand


#ifdef WIN32
typedef unsigned __int64 ULONGLONG;
#else
typedef unsigned long long ULONGLONG;
#endif

DWORD Mult(DWORD Rounds, DWORD Mul, DWORD* BufA, DWORD* BufB, DWORD DecodedByte)
{
while(Rounds--)
{
ULONGLONG edxeax = (ULONGLONG)BufA[Rounds] * (ULONGLONG)Mul;
BufB[Rounds] = DecodedByte + (DWORD)edxeax;
DecodedByte = (DWORD)(edxeax >> 32);
}

return DecodedByte;
}


[Skywing]

Recommend looking at the UInt32x32To64 macro, and the code it generates (for future reference).

[Maddox]

Recommend looking at the UInt32x32To64 macro, and the code it generates (for future reference).

Except that is native only to Windows.

[Skywing]

Yes.  However, you were reverse engineering a Windows program.

[Zakath]

Haha...this is what I wound up with when I did my port back to C++ of the code that iago released:

Code Select Expand

void Mult( int nRounds, int nMulX, LPDWORD lpdwBufferA, LPDWORD lpdwBufferB, DWORD dwDecodedByte ) {
while ( nRounds ) {
unsigned __int64 edxeax = (unsigned __int64)((unsigned __int64)lpdwBufferA[ nRounds - 1 ] * (DWORD)nMulX);
lpdwBufferB[ --nRounds ] = dwDecodedByte + (DWORD)edxeax;
dwDecodedByte = (DWORD)(edxeax >> 32);
}
}

I would guess he called the function slightly differently than you do, since he's using different offsets into the buffers.

[Eibro]

It's not really a C++ implementation if it's mostly inline assembler :p

Looks more like a C implementation.

This is the way I ended up doing that specific function.

Code Select Expand


#ifdef WIN32
typedef unsigned __int64 ULONGLONG;
#else
typedef unsigned long long ULONGLONG;
#endif

DWORD Mult(DWORD Rounds, DWORD Mul, DWORD* BufA, DWORD* BufB, DWORD DecodedByte)
{
while(Rounds--)
{
ULONGLONG edxeax = (ULONGLONG)BufA[Rounds] * (ULONGLONG)Mul;
BufB[Rounds] = DecodedByte + (DWORD)edxeax;
DecodedByte = (DWORD)(edxeax >> 32);
}

return DecodedByte;
}


I tried to make it easy for C programmers to use.

I would guess he called the function slightly differently than you do, since he's using different offsets into the buffers.

No, look closer. He indexes the buffer by rounds - 1 (rounds--) while I index as buflen - i - 1 (i++)