/Zem80

An emulation of the Zilog Z80 written in idiomatic C#

Primary LanguageC#MIT LicenseMIT

Zem80

A simple emulation of the Z80 CPU written in C# and running on .NET 8.0

Version 2.0 now available

Zem80 v2.0 includes a completely redesigned emulator core with improved performance. There are some minor breaking API and event changes, but from the point of view of consuming code, little has changed. Most of the work here was to improve the architecture of the system and tidy up the timing mechanisms.

The emulator is now compiled and built using .NET 8.0, so you will need to have this installed to use it, and your consuming projects must be compiled with .NET 8.0 as well. Since .NET 8.0 is both LTS and about to be superceded by .NET 9.0, I don't feel particularly guilty about this. I am considering a .NET Standard version of the library but this may involve too many compromises. A NativeAOT implementation is still a goal for a potential v3.0.

If you are using the emulator (hi, David!) I strongly suggest you upgrade to v2.0.

Project goals

Most emulators written in most languages are not very representative of those languages. Since emulation is all about twiddling bits and doing things as fast as possible, successful emulators tend to look quite unlike everyday software written on a given platform, and are often pretty incomprehensible.

I'm trying, as far as is possible given the domain, to write idiomatic C# code, with fairly standard OO design principles. I'm aware that I will probably not get the best performance this way. But that's a problem I'll deal with if - and only if - it proves impossible to build an emulation core that can run Z80 software (CP/M, Spectrum, whatever) at a speed comparable to the original hardware.

I have now written a complete Z80 emulation including as much of the undocumented instructions and behaviour as I could find.

I have also added a basic ZX Spectrum emulation, but this is a sample and not intended for actual use as an emulator.

Performance - some thoughts

The amount of CPU required to run the Z80 emulation is significant using the RealTimeClock. This is because the emulation thread has to spin continuously while waiting for clock ticks in order to generate the right events at exactly the right time. This is likely to only be necessary for emulating hardware that is timing-critical, and even then it's probably possible to do it a different way.

To improve this I've added a TimeSlicedClock as an alternative where the emulator will run as fast as possible but after a specified number of Z80 ticks (the 'time slice', which you will need to work out against your timing and your Z80 emulated speed) it will suspend the CPU and wait for a timer (not on the main thread) to elapse and then resumes the CPU. This will resynchronise execution with real time, while avoiding executing any code on the main thread while the timer runs. This allows for instructions to be 'fast-forwarded' as quickly as possible, while still making sure that code like screen updates can run at the proper time. The ZX Spectrum VM sample now uses this technique and this drastically reduces the PC CPU required to run it on my machine (Release build).

The future for this project

Where I think this project can be useful is in perhaps explaining how to do CPU emulation in principle. It's a pretty complex subject that requires a lot of learning to approach and do, and learning from the existing code is difficult because it's often quite... opaque. This emulator is actually quite straightforward - not that you would understand it at first glance, or without knowledge of how the Z80 itself works in quite a lot of detail - and hopefully might give people a useful starting point to build their own emulators. If it does, then my work is done!

I built this thing just to prove to myself that I could. I have a fully-functioning emulator built on a platform that's not generally thought of as one you could build something so low-level on. Take that, C++.

Project status

16/08/2024 - 2.0 release. Mostly a refactoring and redesign release, 2.0 doesn't add any major new features. Many bugs were fixed and timing improved, but timing is still the emulator's achilles heel. If you don't care about absolutely precise timing, you will be fine. The Spectrum sample was updated to work with this version and slightly refactored itself, but is otherwise untouched and remains a fairly poor implementation intended for demo purposes only.

The project moves to .NET 8.0 in this release.

Known issues

  • Interrupt Mode 0 remains essentially untested and may not work properly. I plan to create a virtual machine to test this out.
  • Interrupt Mode 2 appears to be slightly bugged but I'm not 100% sure how or why. I am looking for Spectrum games which are known to use IM2 to test.

The other main component is the ZX Spectrum VM, which has several known issues:

  • Game compatibility is very patchy, many games either don't run or don't run properly
  • Audio in is not supported, so you cannot use the LOAD command. The SAVE command will operate but not successfully (data will be corrupt even if recorded)
  • It's frankly just not very good and is certainly not usable as a real emulator for playing games etc. Download FUSE for that!

Acknowledgements

This project contains materials derived from ZEXALL which is copyright Frank Cringle, with amendments copyright J.G. Harston, and which are distributed under the terms of the ZEXALL license.

This project contains the binary code of the ZX Spectrum 48K ROM which is copyright Amstrad plc and distributed under general permission given to the community to redistribute the code for emulation purposes provided that the code is unchanged.

Portions of this project were created using an amended version of SpectNetIDE (https://github.com/Dotneteer/spectnetide) which is copyright Istvan Novak and other contributors, but this project does not reproduce any code or assets from SpectNetIDE.

Thanks to the wide community of enthusiasts who have documented much of the process of emulating a Z80 and other CPUs in various languages. Without your hard work and contributions, this project would not have been possible.