Apple //e emulator, primarily for the Teensy 4.1
Go to file
2019-02-22 01:01:48 -05:00
apple tie disk spin-down to CPU cycles 2019-02-22 01:01:48 -05:00
linuxfb tie disk spin-down to CPU cycles 2019-02-22 01:01:48 -05:00
nix incremental cleanup and prep for dynamic track loading (instead of full track preloading) 2019-02-20 22:06:55 -05:00
schematics
sdl tie disk spin-down to CPU cycles 2019-02-22 01:01:48 -05:00
teensy tie disk spin-down to CPU cycles 2019-02-22 01:01:48 -05:00
tests test abstraction 2018-06-15 23:13:06 -04:00
util added script to side-load a binary right in to a running SDL build via the debugger 2019-02-12 21:24:40 -05:00
.gitignore
bios-font.h
bios.cpp abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
bios.h finish abstracting file extension filter 2018-02-07 14:23:52 -05:00
cpu.cpp woz uses built-in file management system 2019-02-20 11:51:55 -05:00
cpu.h abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
filemanager.h incremental cleanup and prep for dynamic track loading (instead of full track preloading) 2019-02-20 22:06:55 -05:00
globals.cpp abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
globals.h abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
images.cpp
images.h
lcg.cpp
lcg.h
LRingBuffer.cpp
LRingBuffer.h
Makefile Real-time dropout only makes sense on the Teensy; don't force it on SDL 2019-02-20 17:41:04 -05:00
mmu.h
physicaldisplay.h abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
physicalkeyboard.h
physicalpaddles.h
physicalprinter.h
physicalspeaker.h incremental cleanup and prep for dynamic track loading (instead of full track preloading) 2019-02-20 22:06:55 -05:00
prefsstore.h abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
README.md updated for full 65c02 and linuxfb 2018-06-16 07:48:22 -04:00
vm.h abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
vmdisplay.h abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00
vmkeyboard.h
vmram.cpp
vmram.h
vmui.h abstract drawing so devices may have full or half resolution; rebuild BIOS interface 2018-02-17 20:44:04 -05:00

Aiie!

Aiie! is an Apple //e emulator, written ground-up for the Teensy 3.6.

The name comes from a game I used to play on the Apple //e back around 1986 - Ali Baba and the Forty Thieves, published by Quality Software in 1981.

http://crpgaddict.blogspot.com/2013/07/game-103-ali-baba-and-forty-thieves-1981.html

When characters in the game did damage to each other, they exclaimed something like "HAH! JUST A SCRATCH!" or "AAARGH!" or "OH MA, I THINK ITS MY TIME" [sic]. One of these exclamations was "AIIYEEEEE!!"

Build log:

https://hackaday.io/project/19925-aiie-an-embedded-apple-e-emulator

Getting the ROMs

As with many emulators, you have to go get the ROMs yourself. I've got the ROMs that I dumped out of my Apple //e. You can probably get yours a lot easier.

There are four files that you'll need:

The MD5 sums of those files are:

  • 003a780b461c96ae3e72861ed0f4d3d9 apple2e.rom
  • 2020aa1413ff77fe29353f3ee72dc295 disk.rom
  • 5902996f16dc78fc013f6e1db14805b3 parallel.rom
  • e91f379957d87aa0af0c7255f6ee6ba0 HDDRVR.BIN

From those, the appropriate headers will be automatically generated by "make roms" (or any other target that relies on the ROMs).

Building (for the Teensy)

The directory 'teensy' contains 'teensy.ino' - the Arduino development environment project file. You'll need to open that up and compile from within.

However.

I built this on a Mac, and I used a lot of symlinks because of limitations in the Arduino IDE. There's no reason that shouldn't work under Linux, but I have absolutely no idea what Windows will make of it. I would expect trouble. No, I won't accept pull requests that remove the symlinks and replace them with the bare files. Sorry.

Also, you'll have to build the ROM headers (above) with 'make roms' before you can build the Teensy .ino file.

If anyone knows how to make the Arduino development environment do any form of scripting that could be used to generate those headers, I'd gladly adopt that instead of forcing folks to run the Perl script via Makefile. And if you have a better way of dealing with subfolders of code, with the Teensy-specific code segregated as it is, I'm all ears...

I compile this with optimization set to "Faster" for the Teensy 3.6 at 180MHz. There's no need to overclock the CPU -- but it does give better video performance, all the way up to 240MHz, but still not perfect. Do as you see fit :)

Environment and Libraries

I built this with arduino 1.8.5 and TeensyDuino 1.40.

  https://www.pjrc.com/teensy/td_download.html

These libraries I'm using right from Teensy's environment: TimerOne; SPI; EEPROM; Time; Keypad.

I'm also using these libraries that don't come with TeensyDuino:

SdFat

SD card support - accelerated for the Teensy, and with long filename support.

 https://github.com/greiman/SdFat

Running (on the Teensy)

The reset/menu button brings up a BIOS menu:

Resume
Reset
Cold Reboot
Drop to Monitor
Display: RGB
Debug: off
Insert/Eject Disk 1
Insert/Eject Disk 2
Insert/Eject HD 1
Insert/Eject HD 2
Volume +
Volume -
Suspend
Restore
Prioritize Display/Audio

Reset

This is the same as control-reset on the actual hardware. If you want to execute the Apple //e self-test, then hold down the two joystick buttons; hit the reset/menu key; and select "Reset".

Cold Reboot

This resets much of the hardware to a default state and forces a reboot. (You can get the self-test using this, too.)

Drop to Monitor

"Drop to Monitor" tries fairly hard to get you back to a monitor prompt. Useful for debugging, probably not for much else.

Display

"Display" has four values, and they're only really implemented for text and hi-res modes (not for lo-res modes). To describe them, I have to talk about the details of the Apple II display system.

In hires modes, the Apple II can only display certain colors in certain horizontal pixel columns. Because of how the composite video out works, the color "carries over" from one pixel to its neighbor; multiple pixels turned on in a row makes them all white. Which means that, if you're trying to display a picture in hires mode, you get color artifacts on the edges of white areas.

The Apple Color Composite Monitor had a button on it that turned on "Monochrome" mode, with the finer resolution necessary to display the pixels without the color cast. So its two display modes would be the ones I call "NTSC-like" and "Black and White."

There are two other video modes. The "RGB" mode (the default, because it's my preference) shows the color pixels as they're actually drawn in to memory. That means there can't be a solid field of, for example, orange; there can only be vertical stripes of orange with black between them.

The last mode is "Monochrome" which looks like the original "Monitor II", a black-and-green display.

Debug

This has several settings:

off
Show FPS
Show mem free
Show paddles
Show PC
Show cycles
Show battery
Show time

... these are all fairly self-explanatory.

Insert/Eject Disk1/2 HD1/2

Fairly self-explanatory. Disks may be .dsk, .po, or .nib images (although .nib images aren't very heavily tested, particularly for write support). Hard drives are raw 32MB files, whose filenames must end in .img.

Suspend and Restore

The Teensy can be fully suspended and restored - including what disks are inserted. It's a full VM hibernation. It currently writes to a file named "suspend.vm" in the root of the MicroSD card. (I would like to be able to select from multiple suspend/restore files eventually. It wouldn't be terribly hard; it's just that the BIOS interface is very limited.)

Prioritize Display/Audio

Any Apple emulator needs real-time support for audio. The hardware is very direct: when a particular memory location ($C030) is read from or written to, the speaker's state is toggled. By doing this at specific times, the Apple generates a square wave - and this is the basis of the built-in audio. Obviously, then, the CPU has to be running at a specific frequency and the updates to the speaker have to come at exactly the right time.

The Mac version of AiiE runs the CPU a few cycles fast; and then while it's waiting for the CPU to catch up, it's slowly toggling the speaker at the right time. This works well because the actual computer it's running on is so much faster than the CPU it's emulating.

The Teensy verison can't afford all that overhead. It runs the virtual CPU in batches of 24-ish cycles, unless it updates the speaker; in which case, the virtual CPU returns control to the main program early. This was important while the Teensy code was queueing... but the queueing has since been removed, and its virtual MMU directly toggles the speaker as the virtual CPU is running. (This isn't quite as accurate as the Mac verison but it's close enough; uses less RAM; and uses less CPU.)

While all that's running, the display is also updating from the main loop. The trobule is that it takes quite a lot of time to update the display. Much longer than one CPU cycle. Much longer than hundreds of CPU cycles. Which means that one of two bad things happens.

Either (a) the CPU can run and update the video while you're still pushing that video to the screen - which means you can wind up with (e.g.) half a cloud in Skyfox at the top of the screen, and nothing immediately below it, because the video is "tearing" (getting partial results that change over time as it scans down the screen); or (b) you block the CPU from running while updating the display, which means the CPU runs in fits and starts. Since the audio is directly coupled to the CPU that also means the audio will be very messed up.

So this BIOS option lets you choose whether you want good audio or good video.

Building (on a Mac)

While this isn't the purpose of the emulator, it is functional, and is my first test target for most of the work. With MacOS 10.11.6 and Homebrew, you can build and run it like this:

<pre>
  $ make 
  $ ./aiie-sdl /path/to/disk.dsk
</pre>

As the name implies, this requires that SDL is installed and in /usr/local/lib. I've done that with Homebrew like this

<pre>
  $ brew install sdl2
</pre>

Building (on Linux)

I've been experimenting with Aiie running under a handmade OS on a Raspberry Pi Zero W; the hardware is decent, and cheap. I just don't want Linux in the way. So I built JOSS (see [https://hackaday.io/project/19925-aiie-an-embedded-apple-e-emulator/log/87286-entry-18-pi-zero-w-and-joss](my Hackaday page about JOSS)).

Well, performance under JOSS is poor, so I built a Linux framebuffer wrapper for Aiie so that I can do performance testing on the Zero W, directly between JOSS and Linux.

$ make linuxfb
$ ./linuxfb

Mockingboard

Mockingboard support is slowly taking shape, based on the schematic in the Apple II Documentation Project:

https://mirrors.apple2.org.za/Apple%20II%20Documentation%20Project/Interface%20Cards/Audio/Sweet%20Microsystems%20Mockingboard/Schematics/Mockingboard%20Schematic.gif

I'm not sure the Teensy 3.6 has enough horsepower to run this, so I'm not sure it will ever come to be. It struggles to refresh the display and run the speaker at full tilt; but maybe that's primarily a function of the throughput to the display, and adding the Mockingboard support won't require further sacrifice...

VM

The virtual machine architecture is broken in half - the virtual and physical pieces. There's the root VM object (vm.h), which ties together the MMU, virtual keyboard, and virtual display.

Then there are the physical interfaces, which aren't as well organized. They exist as globals in globals.cpp:

	   FileManager *g_filemanager = NULL;
	   PhysicalDisplay *g_display = NULL;
	   PhysicalKeyboard *g_keyboard = NULL;
	   PhysicalSpeaker *g_speaker = NULL;
	   PhysicalPaddles *g_paddles = NULL;

There are the two globals that point to the VM and the virtual CPU:

	   Cpu *g_cpu = NULL;
	   VM *g_vm = NULL;

And there are two global configuration values that probably belong in some sort of Prefs class:

	   int16_t g_volume;
	   uint8_t g_displayType;

CPU

The CPU is a full and complete 65C02. It supports all of the 65C02 documented and undocumented opcodes:

   http://www.oxyron.de/html/opcodes02.html

The timing of the CPU is close to, but not quite, correct. It doesn't count cycles due to page boundary crossings during branch instructions, for example. (See the "cycle count footnotes" in cpu.cpp.)

The CPU passes all of the the 6502 tests from here, including the undocumented ADC and SBC handling of Decimal mode and the overflow flag:

https://github.com/Klaus2m5/6502_65C02_functional_tests

Doing a

$ make test

will build the test harness and execute the three tests that encompass all others. (There are more tests in the tests/ directory - only 3 of them are truly unique.) Two of the tests should emit

All tests successful!

while the third says

Test complete. Result: passed