I wanted to take another look at the save corruption run since it doesn't include BIOS time and timer emulation is now console accurate.
The first thing i wanted to check is if there are any easy manipulations in CGB in GBA mode for red, since it has console verifiability and the color bios is much shorter then the gb bios.
There are 4 places you can delay to do TID manipulation, from masterjun's old submission here are the values we want for the second byte (the birst one needs to be 0x64): 0x11, 0x13, 0x15, 0x16, 0x17, 0x1A, 0x1B, 0x1D, 0x5A, 0x71, 0x76, 0xCF, 0xD0 and 0xEA. Also in Mrwint's submission we see that 0xBD is possible as well.
Here is a complete list of possiblities with delay values 0-9 for each of the four inputs (starting from the first possible frame each input is selected.)
CGBinGBA CGB GB
0 8 0 9 0x38 0 3 7 5 0x50 0 9 1 2 0x48
0 8 1 8 0x37 0 3 8 4 0x50 0 9 2 1 0x48
0 8 3 6 0x39 0 3 9 3 0x50 0 9 3 0 0x47
1 6 7 6 0x6D 0 4 0 1 0x54 0 9 6 1 0x94
1 6 8 5 0x6D 0 4 0 3 0x78 0 9 7 0 0x95
2 3 3 8 0x6A 0 4 1 0 0x54 1 4 2 9 0x9A
2 3 4 7 0x68 0 4 1 2 0x78 1 4 5 6 0x9A
2 3 5 6 0x67 1 0 0 4 0x65 1 4 6 5 0x99
2 3 7 4 0x67 1 0 1 3 0x65 1 4 8 3 0x9A
2 5 8 1 0x37 1 0 2 2 0x66 1 9 2 7 0xC5
3 5 8 9 0xDA 1 3 1 4 0xE6 1 9 4 5 0xC6
3 5 9 8 0xDB 1 7 8 1 0x65 1 9 7 2 0xC5
3 8 0 9 0x6E 1 7 9 0 0x66 2 0 6 5 0x70
3 8 1 8 0x6F 1 8 5 0 0xF7 2 1 7 7 0xBD
3 8 3 6 0x6F 1 9 1 4 0x07 2 6 2 8 0xBD
3 8 6 3 0x6E 2 4 0 6 0xD2 2 6 5 5 0xBE
3 8 7 2 0x6E 2 4 0 8 0xF8 2 6 7 3 0xBE
3 8 8 1 0x6E 2 4 2 4 0xD2 2 7 8 9 0x94
3 9 2 4 0x42 2 4 2 6 0xF8 2 7 9 8 0x95
3 9 3 3 0x43 2 8 0 6 0x1F 2 8 0 1 0x30
3 9 8 6 0xD6 2 8 1 8 0x58 2 9 0 3 0x6F
4 3 9 5 0xC6 2 8 2 4 0x1F 2 9 1 2 0x6F
4 7 1 8 0xA9 2 8 3 3 0x20 2 9 2 1 0x6F
4 7 9 0 0xA8 2 8 5 1 0x21 2 9 8 9 0x78
6 2 3 8 0x5E 3 3 7 0 0x2A 2 9 9 8 0x78
6 2 4 9 0x82 3 7 7 2 0x91 4 0 8 5 0xB8
6 2 5 6 0x5D 3 7 8 1 0x91 4 6 7 9 0x4E
6 2 7 6 0x81 4 8 5 8 0xC8 4 9 0 0 0x5B
6 2 8 5 0x82 4 8 6 7 0xCA 4 9 2 0 0x80
6 2 9 2 0x5F 4 8 8 5 0xC8 5 5 4 9 0x23
6 9 3 7 0xC1 7 2 5 9 0x9F 5 5 5 8 0x23
6 9 4 6 0xC1 7 4 5 9 0xC9 5 5 7 6 0x21
6 9 6 4 0xC2 7 4 6 8 0xC7 5 5 8 5 0x21
6 9 9 1 0xC2 7 6 6 6 0xC9 5 5 9 4 0x21
8 1 5 8 0xA4 7 6 9 3 0xC9 5 8 6 8 0x5F
8 1 7 6 0xA4 8 0 0 2 0xC1 6 4 7 8 0x37
8 6 7 2 0xA7 8 0 1 1 0xC1 6 6 0 4 0x93
8 6 8 1 0xA6 8 0 2 0 0xC1 6 6 1 3 0x94
8 7 0 3 0x86 8 2 0 3 0xE7 6 6 3 1 0x93
8 7 1 8 0xF2 8 2 2 1 0xE7 6 6 4 0 0x93
9 0 1 8 0x59 8 3 1 8 0xB4 6 8 5 9 0x73
9 1 0 4 0x14 8 3 9 0 0xB7 6 8 7 7 0x72
9 1 2 2 0x14 9 6 2 9 0xDC 7 6 4 2 0xCC
9 3 8 4 0x06 9 6 3 8 0xDD 7 7 6 9 0xCE
9 3 9 3 0x06 9 6 6 5 0xDC 7 7 8 7 0xCE
9 6 7 4 0xDC 8 1 7 9 0x51
9 6 8 3 0xDC 8 1 8 8 0x51
9 8 0 8 0xC8 8 4 7 8 0x60
9 8 1 7 0xC7 8 8 3 6 0x3E
9 8 2 6 0xC7 8 8 6 3 0x3D
9 8 8 0 0xC7 8 9 4 0 0xF1
9 0 7 8 0x3F
9 0 8 7 0x3F
9 5 3 6 0x27
9 5 4 5 0x28
9 5 5 4 0x28
9 8 8 5 0x95
So, none of them work in GBA mode. I'll try again for normal CGB mode, and the for GB mode and update this post with results.
EDIT: done for CGB, no positive results.
EDIT2: So it looks like at least for red, the only viable options are 2 1 7 7 and 2 6 2 8 in GB mode. Not sure if any of these other values have any usefulness. Also I'm not sure how big MrWint's original search space was. technically I guess I'd need to look up to 17 since that's the amount of delay in the current run, but that's beyond the current scope of what I wanted to do.
http://tasvideos.org/userfiles/info/59639073305669766
Here is a resynced run of MrWint's movie to the current BizHawk builds. With a time of 1:21, it's quite a lot longer then the publication would indicate. If the yellow version run also worked, it would be several seconds faster since it uses the CGB instead, so my personal opinion is that yellow should be able to obsolete this one.
This core is shaping up and starting to approach release levels of accuracy.
The biggest problem I'm facing right now is that various homebrew's I test don't look right compared to O2EM, but I'm not confident that O2EM is truly representative of hardware.
At this point if anyone is willing to do some beta testing with some actual retail Odyssey2 games I would appreciate any bug reports.
TASing should be in full working order at this point.
Now that I have this core more in working order I thought I'd make a thread for it. there are still numerous bugs to fix but things are slowly coming together. Some games are now completely playable, but it will probably be another couple of weeks before this core can be declared ready for TAS production.
Small update on Odyssey2. A lot of things are getting cleaned up and into a functional state. I can get in game with controller support now. A lot of things don't work yet but things are progressing at a reasonable pace.
It turns out that this system is quite cumbersome. The PPU is very unconventional and un-intuitive, and the input/output functions of the CPU are used in tricky ways. Good thing there is good documentation or I would be quite lost.