2 KB CHR bankswitching inconsistency?

Hi all,

I started my own emulator project, Yane (http://yane.alike.se), in October 2012 with the goal to play SMB1. I thought it was a fun thing to do while being on parental leave. SMB1 has been playable for some time now and I am now adding more mappers.

This is the first post on the NESDev forums, but I have been following the NESDev forums for a while. I have never had to ask a question, since they have all been asked before. Until now. I was recently done (more or less) with MMC3 (#4). But it toke time, since the 2 KB CHR bankswitching never worked. Today I discovered that if I treated the bank index argument (from $8001) as 1 KB indices when doing a 2 KB bankswitch everything worked well. Why is that? Is the documentation of MMC3 (http://wiki.nesdev.com/w/index.php/MMC3) missing something? Or is the 2 KB CHR bankswitching in general just inconsistent?

When I read the ROM file I store CHR in 1 KB banks and a simple mapping scheme is used to map RAM access to the correct 1 KB bank from the ROM file. For all other CHR bankswitching I have to translate the bank index argument to 1 KB banks in the ROM, e.g. bank_index * 4, when doing a 4 KB CHR bankswitch. As I said, with 2 KB CHR bankswitching everything gets messed up if that translation is done.

Anyways, thanks for all the help and awsome test ROMs.

Yours sincerely,
Robin

In the MMC3 writeup, we failed to make overly clear that the 2KB mode discards the LSB. I'll fix thatTepples fixed that.

Thanks for the quick reply! And, that explains why my MMC3 rendered garbage once in a while It all looks fine now.