bsnes WIP :: Emu-Land.net

Новости > Показать отдельно

bsnes WIP

Ice, 01:47, 14-10-2006

byuu, автор эмулятора bsnes, сообщает о прогрессе своего эмулятора консоли Nintendo Super NES/Famicom:

10/11/2006

CPU core timing bugfix, other improvements

Well, this is certainly embarassing. I had S-CPU condition 4 wrong (see the GTE65816 PDF for more info on this condition). So, as it turns out, the true reason for the Jumbo Ozaki no Hole in One bug was actually because bsnes was running too fast, not because it was running too slowly as I previously suspected. It seems shameful to have bsnes referred to as such an accurate emulator with such a basic mistake, but eh... can't catch them all I suppose. And it only gets more accurate with time, right? I'm going to start writing CPU opcode timing test ROMs to verify the accuracy of the rest of the opcodes now as well. Anyway, this fix was actually a few separate fixes. Writes now always return true for this condition, and reads can now be true when regs.p.x != 0 and page wrapping occurs, and reads are always true when regs.p.x == 0.
I also added in the CPU version differences for DRAM refresh and HDMA init timing. You'll need to compile the emulator yourself if you want to set the CPU version to something other than 1 for now. I may change that later, possibly let you change that in the config file or something. The HDMA init timing needs a bit of work to be perfect again, but it's getting there little by little.
Lastly, I reverted the Street Racer fix from a while back. I'm pretty sure HDMA starts at HCLOCK=1106, so I'm going with my instinct here. Once I get the last little bits ironed out with HDMA init timing, I can verify this on hardware properly and then go from there.

Сообщение Для того, чтобы иметь возможность писать комментарии,
пожалуйста, войдите или зарегистрируйтесь.

Вход