huhhuh2000
New member
i know the pj64 help says the byte order of a rom doesnt matter, but which, in your opinion loads fastest or is best
best byteswap?
- Thread starter huhhuh2000
- Start date
You really don't need to do this at all and you will see no improvement whatsoever, but if you insist... I'd say the "best" format would be byteswapped (v64) format because it is the most common format. Make sure you make back-ups, as you could potentially ruin your roms in the attempt of converting.
The only instance I know of where byte order would be of concern is when patching. Maybe when using old (read: outdated) emulators as well. The difference in load time, if any, is inconsequential.
AlphaWolf
I prey, not pray.
Saying "byteswapped" is a bit ambiguous, don't you think?...its either big endian (e.g. most risc CPUs), or little endian (e.g. intel CPUs). Byteswap just means you are switching from one endian to the other 
I would figure though, that in the case of a dynarec engine, it would be better to match the endian-ness of the host CPU?
I would figure though, that in the case of a dynarec engine, it would be better to match the endian-ness of the host CPU?
Last edited:
OP
huhhuh2000
New member
- Thread Starter
- #5
another aspect is if u assume that project64 does something like
check to see which format its in
convert to x format
work with x format
namely it is only coded to handle one format and just converts to that format before doing anything else.. so i wonder what the default format is?
..a reminder this discussion is just for the heck of it
check to see which format its in
convert to x format
work with x format
namely it is only coded to handle one format and just converts to that format before doing anything else.. so i wonder what the default format is?
..a reminder this discussion is just for the heck of it
OP
huhhuh2000
New member
- Thread Starter
- #6
yup, heeere we go!
i d/lded the source code for 1.4 to confirm it..
see how case is used 3 times, one for each type.. now for 2 types it converts them and for 1 it leaves it.. so which is which? i added code so a message box would popup saying so
long story short, project64 does not convert the .rom (also .n64?) type, so that type should load a little faster
i d/lded the source code for 1.4 to confirm it..
Code:
void ByteSwapRom (void) {
DWORD count;
SendMessage( hStatusWnd, SB_SETTEXT, 0, (LPARAM)"Byte swapping image" );
switch (*((DWORD *)&ROM[0])) {
case 0x12408037:
for( count = 0 ; count < RomFileSize; count += 4 ) {
ROM[count] ^= ROM[count+2];
ROM[count + 2] ^= ROM[count];
ROM[count] ^= ROM[count+2];
ROM[count + 1] ^= ROM[count + 3];
ROM[count + 3] ^= ROM[count + 1];
ROM[count + 1] ^= ROM[count + 3];
}
break;
case 0x40123780:
for( count = 0 ; count < RomFileSize; count += 4 ) {
ROM[count] ^= ROM[count+3];
ROM[count + 3] ^= ROM[count];
ROM[count] ^= ROM[count+3];
ROM[count + 1] ^= ROM[count + 2];
ROM[count + 2] ^= ROM[count + 1];
ROM[count + 1] ^= ROM[count + 2];
}
break;
case 0x80371240:
break;
default:
DisplayError("ByteSwapRom: %X",ROM[0]);
}
}see how case is used 3 times, one for each type.. now for 2 types it converts them and for 1 it leaves it.. so which is which? i added code so a message box would popup saying so
Code:
void ByteSwapRom (void) {
DWORD count;
SendMessage( hStatusWnd, SB_SETTEXT, 0, (LPARAM)"Byte swapping image" );
switch (*((DWORD *)&ROM[0])) {
case 0x12408037:
for( count = 0 ; count < RomFileSize; count += 4 ) {
ROM[count] ^= ROM[count+2];
ROM[count + 2] ^= ROM[count];
ROM[count] ^= ROM[count+2];
ROM[count + 1] ^= ROM[count + 3];
ROM[count + 3] ^= ROM[count + 1];
ROM[count + 1] ^= ROM[count + 3];
}
MessageBox(NULL,"1","here",MB_OK);
break;
case 0x40123780:
for( count = 0 ; count < RomFileSize; count += 4 ) {
ROM[count] ^= ROM[count+3];
ROM[count + 3] ^= ROM[count];
ROM[count] ^= ROM[count+3];
ROM[count + 1] ^= ROM[count + 2];
ROM[count + 2] ^= ROM[count + 1];
ROM[count + 1] ^= ROM[count + 2];
}
MessageBox(NULL,"2","here",MB_OK);
break;
case 0x80371240:
MessageBox(NULL,"3","here",MB_OK);
break;
default:
DisplayError("ByteSwapRom: %X",ROM[0]);
}
}ShadowPrince
Moderator
AlphaWolf said:Saying "byteswapped" is a bit ambiguous, don't you think?...its either big endian (e.g. most risc CPUs), or little endian (e.g. intel CPUs). Byteswap just means you are switching from one endian to the other
I would figure though, that in the case of a dynarec engine, it would be better to match the endian-ness of the host CPU?
Not totally correct.There are 4 DWORD formats :
Big Endian which mainframe computers using, Byteswapped, Little Endian (mostly PCs) and Wordswapped.
N64 emulators ,at least starting from uhle, disregard rom extension,and determine rom format by the first 4 bytes of the rom :
0x80371240 Big endian (z64)
0x37804012 Byteswapped (v64)
0x40123780 Little endian
0x12408037 Word swapped
When loading a rom,emulator converts it to big endian format.So in this aspect it's better to keep roms in big endian (z64 default),so that loading time will be minimized (by fractions of second in the best case ) .
OP
huhhuh2000
New member
- Thread Starter
- #8
well, not quite-- ive only heard of 3 formats, and pj64 only checks for 3, they are
0x12408037 Byteswapped (v64)
0x40123780 not byteswapped, Big endian (z64)
0x80371240 Word swapped, Little endian (rom n64)
0x12408037 Byteswapped (v64)
0x40123780 not byteswapped, Big endian (z64)
0x80371240 Word swapped, Little endian (rom n64)
ShadowPrince
Moderator
I was refering to byte sequences and not actual DWORD values,as it's different on each machine depending of it's endianess :
instead of 0x80371240 Big endian (z64)
should be 80 37 12 40,if it's what you mean.
instead of 0x80371240 Big endian (z64)
should be 80 37 12 40,if it's what you mean.