MAMEDev probably has some things, I can check around for you.

UNDOCUMENTED 9801/9821 Vol.2 http://www.webtech.co.jp/company/doc/undocumented_mem/
upd7220 datasheet http://bitsavers.trailing-edge.com/pdf/nec/uP … anual_Dec85.pdf
Neko Project 2 https://github.com/nonakap/xnp2
MAME https://github.com/mamedev/mame/blob/master/s … vers/pc9801.cpp

... or crazyc can beat me to it. Good one, crazyc 😉

Excellent! I hope this shows up on Google search for anyone else interested in PC-98 I/O ports and hardware.

Now to see if I can figure out enough Japanese to read that documentation... 😀

I learned more than anyone rational would want to know about that machine while working on the mess emulation. If you have any questions, just ask. BTW google translate does a marginally understandable job on the undocumented_mem files.

crazyc wrote:

I learned more than anyone rational would want to know about that machine while working on the mess emulation. If you have any questions, just ask. BTW google translate does a marginally understandable job on the undocumented_mem files.

I appreciate it. I'm putting in the framework now for DOSLIB to compile versions of it's libraries for PC-98 systems. I understand some of the hardware (like the PIT timer and IDE) is sort of the same interface, just moved around to different I/O ports, spaced two I/O ports apart for some reason and interleaved even/odd numbers between devices. Weird.

The intent is for DOSLIB to open up cross-development to PC-98 systems the same way DOSLIB opens up IBM PC/XT/AT cross-development now. I will be writing code to run in PC-98 emulation, and hopefully, run on actual PC-98 hardware.

https://github.com/joncampbell123/doslib

I understand some of the hardware (like the PIT timer and IDE) is sort of the same interface, just moved around to different I/O ports, spaced two I/O ports apart for some reason and interleaved even/odd numbers between devices. Weird.

There is a sane reason to interleave the devices like that. Even the original pc9801 had an 8086 so it has 16 data lines which is a pain to use with 8bit devices. That way they could use A0 as a chip select and route D0-7 to one device and D8-15 to the other.

crazyc wrote:

I understand some of the hardware (like the PIT timer and IDE) is sort of the same interface, just moved around to different I/O ports, spaced two I/O ports apart for some reason and interleaved even/odd numbers between devices. Weird.

There is a sane reason to interleave the devices like that. Even the original pc9801 had an 8086 so it has 16 data lines which is a pain to use with 8bit devices. That way they could use A0 as a chip select and route D0-7 to one device and D8-15 to the other.

So what happens if you do a 16-bit I/O write, do you talk to two chips at once? Or does it still decode A0 as chip select and let the chip take in all 16 bits? Or... is that undefined behavior?

I've never used an actual machine but probably the high chip uses ~BHE for it's select and the low uses ~A0.

crazyc, a quick question, why does NT 4.0 here suffer from the BSOD 0x7b (second parameter is 0xc0000034) in general? is it related to the DMA or the IDE code? (not on MAME as it's already solved)

And btw, I'm using this fork of NP21 and win2000 PC98 works there, but not NT 4.0 or 3.5x: https://sites.google.com/site/np21win/home

Does np21w have PCI IDE emulation? If not than it wouldn't be DMA as PC98 ISA (acutally CBUS but the differences aren't important) IDE, like the AT, lacks any useful DMA support. The problem might not even be IDE though, could be a CPU emulation issue.

well, I can post a screenshot and yes and no, PCI is partially supported on np21w, though not for IDE.
np21w, like the standard mainline one, has a defective FPU, but I don't think it's the problem.

If you post a screenshot, I can tell you if there's anything obviously wrong but not anything deeper.

nvm, got NT 3.51 working on Neko Project 21/W (this fork) through the use of the PC9821 BIOS and ITF and PC9821CX2 ide bios.

I know I haven't spoken in this topic for awhile but I just finished adding basic PC-98 emulation to DOSBox-X.

In the latest commit (not the latest Win32 binary) you can set machine=pc98 to enable it.

You will also need ANEX86.BMP for the character ROM to display the full character set.

So far I can run:

Edge
DGA
Tetris (PC-98 verson)
Pacman (PC-98 version)
Rude Breaker

and of course:

Touhou Project (2 through 5).

Github issue tracker for this progress:

https://github.com/joncampbell123/dosbox-x/issues/106

I just finished adding more PC-98 emulation. I've decided to start releasing Windows binaries again, though Linux will still be my primary dev target.

I also borrowed the FM synthesis emulation from Neko Project II and ported it to DOSBox-X, so your favorite games should have some music and sound effects. MIDI emulation was adapted as well, so you can use your Windows MIDI device or Linux with Fluidsynth to hear it if the game uses MIDI playback.

Enjoy.

https://github.com/joncampbell123/dosbox-x/commits/master