We Stand on the Shoulders of Giants
We Stand on the Shoulders of Giants2019-10-01 07:23:46
I believe the advice in this article applies much more broadly than just emulation, but nonetheless I am going to present this article from said vantage point.
When looking to begin a new emulator project, where to begin can be very daunting: is it okay to look at the source code of other emulators? To read notes that explain how to implement various details? Or is that just copying? And how do you make a positive difference in the scene?
When I started on bsnes back in 2004, I did not start from scratch: I started with the combined wisdom and research of dozens of talented individuals that helped to reverse engineer and emulate the Super Ninendo between 1996 and 2004. I had a full eight years of progress at my fingertips: source code, documentation, forum posts, and bugfix changelogs from the work of ZSNES, Snes9X, et al. This, to me, was invaluable.
(I also had six years of my own experience in reverse engineering SNES games for the fan translations I had worked on, which was also a benefit.)
If you search back to the early days of bsnes, you'll find that I was able to catch up to their general level of accuracy and compatibility within only six months, in spite of bsnes being the first major article I had written.
To claim that this represents some extraordinary talent of mine would be facetious: I had a lot of help, and I'm not ashamed of that. I leaned on the folks who came before me, studied their work, asked them questions, and benefited greatly from their patience and assistance.
Folks like anomie, TRAC, etc helped make bsnes a reality. I would not be where I am today without them.
Success Begets Success
After the first six months of development, it was now my turn to improve the state of the art. I developed new techniques to more accurately analyze the cycle timings of the SNES, and I wrote hundreds of test ROMs to suss out countless edge cases and new behaviors.
After relying on pre-existing knowledge, I had to transition to discovering new details myself. This required a hardware setup and knowledge of how to write programs for the SNES. Even though I technically had that from my days of ROM hacking, such a thing is not necessary to begin writing a new emulator, that can come later on. These days, for most major retro systems, there is enough information to create relatively high-quality emulators without ever even owning the original hardware. Case in point: the MiSTer SNES emulation core was written by someone using only bsnes' source code as a reference, having never even owned a Super Nintendo!
Returning the Favor
bsnes has now been under active development for the past fifteen years. But I haven't forgotten my origins. Instead, I've strived to give back. You can see my contributions in pretty much anything in the SNES space today: I have answered questions for and/or donated source code to Snes9X, SNESGT, Mesen-S, the Super Nt, etc.
My source code has always been open source, and where required I've even relicensed it for use in other projects, such as my APU core for Snes9X's non-commercial license.
And just as I was able to catch up to ZSNES and Snes9X before me in only a tiny fraction of the time, modern SNES emulators from 2018 onward have quickly been able to catch up to me.
The Super Nt was created in only nine months, and Mesen-S in only four months.
As an example, Speedy Gonzales for the SNES was an infamous bug. For around 15 years, no emulator could figure out how to run this game. It would deadlock in the middle of stage 6-2 for seemingly no reason.
I spent approximately eighty hours of a two-week period reverse engineering the game, and trying to understand what was happening. It was actually pretty easy to get the game playable: there were several ways of getting around the problem area of code. But only one way was the right way. Implement any of the other ways would mean my emulator had two bugs instead of one. Potentially an incorrect fix would end up breaking a different game in the future.
And so the challenge was ruling out every other possibility through devising a comprehensive and exhausting set of tests, so that I could be sure I had it right.
I eventually reached this conclusion: the game was reading from an unmapped memory address, waiting for a bit to be set that never would be. But by sheer chance, after thousands of scanlines, an Hblank DMA transfer would fetch a value with said bit set, and if the cycles aligned just right, this value would remain on the bus, and the loop condition would terminate.
After this was discovered, I was able to provide the answer to other SNES emulator developers, who could then implement this behavior within a few seconds.
This is no different than all of the bugfixes I gleaned by reading through the old Snes9X changelogs and forum posts about problematic games.
I know it can seem like it, if you view emulation as taking a test and other emulators as having the answer key for the test. And by all means, if it is your ambition to learn how to reverse engineer hardware for the sake of it, you are most welcome to not study what was done before you.
But if your goal is preservation of the original hardware, then it is not in any way "cheating." It is simply being practical and making use of the resources that are available to you.
We have limited time in this world, and the hardware we seek to preserve is not getting any younger, nor any more readily available. Eventually, it will not be possible to pick up working original hardware to analyze. That is already the case for some very old and rare systems, and even more modern ones have become prohibitively expensive and fragile, such as the Atari Jaguar CD.
There's no need to reinvent the wheel: make use of what tools we have, and then strive to build an even better wheel. Any lingering guilt you have can easily be absolved once you give back to the community.
My hope is that the broader community can see emulation as team effort, and not as a competition. It isn't about who did it first, or who did it best. It's about making sure the works of video game studios can live on, and that our great-grandchildren can, if they so choose, look back and see how video gaming began.
No one in emulation is a god or a king. No one is more important than anyone else. We're all in this together, collectively.
Again, you don't have to rely on existing knowledge. You are welcome to forge your own path. But there should be no shame in standing on the shoulders of those who came before you.
Thank you for reading, I hope this will help some of you to get started. I look forward to seeing what you come up with one day soon. Good luck, and don't forget, we're here for you!