MAMEWorld Forums - EmuChat - Re: MAME Opinions

No-one’s going to take kindly to that unless you’re prepared to pay market rates. A decent C++ developer is worth AUD125,000/year. If you want to be able to dictate what even one developer does for a month, you’re going to be up for thousands of dollars. MAME development is only “fun” when you can work on what you want to work on, in proportion to your available time and motivation. If MAME was a day job, most of us would hate it.

Now to be honest, a lot of MAME development isn’t fun for me at all. I don’t spend a lot of time working on stuff I’m interested in. I tend to spend a lot of my time dealing with overheads or fighting the biggest fire. Overheads includes stuff like releases (a pretty busy three days, even with the new, streamlined process), and essential management tasks to keep development moving along smoothly.

There’s a lot of firefighting because MAME is big and decades old. Doing any kind of major rework is a lot of effort. It does pay off, though. A good example is the rate that computer peripherals are added these days. Dragging the codebase into the 21st century has made that a lot easier – life’s a lot simpler for the people working on actual emulation. There’s nowhere near as much boilerplate code and voodoo involved in writing MAME devices now. But getting there was a lot of effort, and it’s slow, thankless work where the best immediate outcome is that you don’t break anything. It’s not until later that you see the benefits come through in terms of more progress in each release.

I’d consider the artwork system changes in the last couple of months to be firefighting as well. It’s an area that’s been plagues with bugs from the very beginning, with major limitations that constantly have to be worked around (full credit to hap for what he manages to pull off with the limited tools he has available). It took me about two years to actually get this stuff done for various reasons, but from a user’s perspective it looks like it just appeared in a couple of months. This is something that probably could’ve been done faster by adding one more developer, but only if you could get them in the same place/timezone to avoid communication breakdown and stepping on each other’s toes. It required some fairly specialised knowledge, including calculus and inexact floating-point maths, as well as understanding how MAME’s render targets and layout elements work. I really appreciate hap testing it and catching a few issues before I did (or before I missed them and they made it into a release).

I’ve worked on some systems in MAME that you might think no-one cares about, like the Osborne 1 and INTELLEC 4, but it’s really important that MAME emulates these things, because nothing else does. The popular systems often have their own emulators and aren’t in any real danger of being forgotten. For a lot of less popular systems, MAME is all there is. I often pick up relatively easy systems that can be emulated just by working off the schematic. I really appreciate that we’ve got people like P-Mack who take on far more complex systems, like InterPro, NEWS, Jazz, etc.

Despite these forums being pretty much dead, MAME is very much alive. There’s more development happening than ever. I mean, that’s why we’ve had to streamline the release process – the old process just wasn’t scaling with the amount happening these days. MAME 0.226 has been a pretty popular release, with thousands of downloads on the first day. MAME has made it through two decades, and I’m going to make sure we can get at least another decade out of it.