|
|
|
|
I finally tracked down Commodore's old inventory of "round tuits" and swiped one. In other words, I got a round tuit - and finally put UAE on some computers at work.
The U in UAE originally stood for Unusable - the early versions didn't boot at all, the author was writing it as an intellectual exercise - and then became the Unix Amiga Emulator, primarily because the original author did all his development on Linux. These days, it's up to version 0.6.9, and the U has been taken to stand for everything from Ubiquitous to Ultimate. I prefer Universal myself - it's been ported to everything under the sun, including Be, Acorn, Wintel, NeXT, every UNIX, Power Mac, and even the Amiga itself. And yes, it's finally robust enough to run itself - thus being the only emulator of which I'm aware for ANY platform that can theoretically host itself recursively. But is it still "useless?" Read on.
UAE for Power Mac is still a "beta port" - which means the conversion to Power Mac has introduced bugs above and beyond the bugs in the code itself. It lacks a number of features - file systems, serial port, etc. It also lacks some speed, even on the 120MHz Power Computing "slowpoke" I tested it on. Lack of file system emulation meant it was nearly useless for what I wanted it to do - so off to the other office I went.
At Hoosier On-Line, we have these funky things called Monorails. They're supercheap PC clones, essentially designed for kiosk applications, data entry, presentations, and the like. They look eerily Mac-like, but then, if Apple made PC clones like these, they'd be white and cost twice as much. Anyway, they're black plastic, with a passive-matrix LCD color 640x480 screen. One of these babies with 16MB RAM and an AMD K5-75 (Pentium 75-class) will run you $899. They're not powerhouses by any stretch - but they'll run a Web browser passably, and they play a decent game of Freecell. The screens suck. They're like laptops that just couldn't make the grade - they're not conventional desktops, but they require AC power and aren't small enough to be laptops.
Anyway, I finally had enough of Freecell. On went UAE. I had been warned it would be tricky to set up, but it actually didn't take that long - I fed it a ROM image (39.106 from my 1200, I won't be using the Monorail and my Amiga at the same time), the ADFs (floppy images) of the entire six-disk Workbench 3.0 Install set, told it to run in 640x480x256, turned off aspect correction, turned on clever centering in both directions, frame rate 1, and went to town.
One problem I did encounter almost immediately: it doesn't like Windows 95. It runs under it, but once you switch to another task, UAE becomes "unrecoverable". I'm running the MS-DOS version; I don't think these Monorails are beefy enough for the Win32/DirectX version. And the Win95 incompatibility seems directly tied to how Win95 handles 256-color screens for DOS apps.
BUT. The results were astounding. Last time I ran UAE, it was on a Power Mac 6100/66, and it was about four versions ago. This is a very different animal. The last time, it took eight minutes to bring up a Workbench that was so slow simply moving the mouse pointer produced a "striping" effect. Now, however, on that AMD K5-75, it delivers approximately the performance of a 5MHz 68000-based A500.
This thing's incredible. It doesn't take long for it to hit the insides of the PC's limitations, for instance it will only accept about 7MB RAM for the emulation, despite 16MB physically installed in the machine. (7MB? Yeah - 2 chip, 4 fast, and 1 "bogo". Bogo? That's the "slow RAM" A500's had up at $C00000, the 512K that lived on the clock card. UAE can apparently shoehorn an entire meg in there - great since the 68000 emulation only does the old 10MB memory map, this ups that limit to 11MB.) The graphics mode I chose, 640x480x256 with no dithering, does NOT accurately represent all 4096 colors of the Amiga - which means, for example, MagicWB is worthless on it, those color shades simply aren't there.
But it does deliver rather quick floppy emulation, it can use hardfiles, period, and it can do this nifty "emulated filesystem" trick that I adore. The emulated filesystem, for the uninitiated, is the ability to mount entire drives or directories on the native computer as AmigaDOS volumes. In this case, I gave it a Windows 95 folder all to itself. It's perfect for sharing files between OSes - for example, I hopped off to Aminet with Internet Explorer, downloaded some LHA archives, and tossed them in that shared volume so the Amiga "side" could play with 'em. It worked a treat.
I timed it cold-starting: from hitting enter on "Run UAE" to the busy pointer going away, 27 seconds booting from the hardfile with a clean OS install and only a couple of commodities. I was highly impressed. My Amiga at home can do that in 18 seconds - but for an emulation, that's wicked fast.
I have not used it to run games or demos. In fact, I have yet to put the emulation into HAM mode. My first priority has been getting the thing tricked out with the usual complement of commodities and hacks - MagicMenu, CycletoMenu, a decent text editor, ToolsDaemon. Then, since I didn't have the apps yet that I wanted to run, I tinkered.
I installed VisualPrefs, which is something I wouldn't dare try on the system at home. It does some exceedingly vicious things to the system - sorta like SysIHack on steroids - it does cause the emulation to lock up when you make changes to the prefs and save them. But it also condenses lots of weird GUI hacks into one - gives you finer control over pen mapping, lets you mess with button borders, lets you tweak prop gadget appearances, and weirdest of all, lets you play with the window border gadgetry. Its "Future" style is a shocker: the close gadget and others sit WITHIN the title bar, instead of being part of it, for an appearance just like Windows 95 or NeXT. It lets you "square up" the border gadgets, too, so if you're running in Topaz 8, your window gadgets don't look squished. It lets you tweak the sizes of things - border scrollbars, for instance. And it gives you some creative freedom with your buttons - it's not hard to give them a Xen look, a Windows 95ey look, or even a slightly Macintoshey look. (Macintosh? Sure - rounded corners and the Xen style have a slightly Copland/System 8 look.) It's way cool if you can't stand your Amiga looking like everyone else's. :-)
Tonight I installed Ace 2.4 on it. This is a brilliant move for two main reasons: 1) writing BASIC programs is more productive than Freecell, and 2) 5MHz 68000-like performance ought to prevent me from being a lazy coder. It does take two minutes to do a sizable compile, i. e. anything that uses the includes, but hey, that gives me time to think of all the bugs I need to fix before the next compile. My first project is going to be a Toolmanager-like dock program, with a somewhat more intuitive prefs editor that lets you simply drag the buttons around. I have other ideas that can wait - but this at least should give me plenty of firsthand experience with Gadtools.
But amidst all this tinkering with the emulation, two things came to mind. First, running at such casual speeds gave me a new appreciation for the 50MHz firebreathing monster at home. And second, the virtual filesystem is proof that the emulation can "think outside the box". That opens a LOT of possibilities.
Already people are working on RTG drivers for the Win32 edition of UAE - this will allow Picasso96-compatible software (which includes Workbench) to open screens directly on PC video hardware and use an Intel-native version of graphics.library to draw everything. There is also talk of a way to open Intuition windows directly on the desktop of whatever platform you're running on - Amiga Intution windows alongside Mac or PC apps, for instance - and the eerie rumor that Stefan "MUI" Stuntz is writing an Intel-native muimaster.library for use with UAE. If you thought running 68K Mac apps on a PowerPC was a crazy idea, just wait until people start buying Pentiums to use as Amigas. That day's probably not far away - Pentium 100s are today selling for about what A500s sold for in 1989, and a P100 will deliver slightly-above-A500 performance. Add the RTG drivers and virtual filesystems, and you're talking about an Amiga ready for the next century.
I don't advocate leaving behind the concept of a hardware Amiga. But I do think this presents a unique opportunity to think of new roles for the Amiga in 1997 and beyond. First off, here's a short list of applications where UAE is highly useful:
What is Java? It was designed for embedded applications, it now is supposed to be a cross-platform CPU-independant bytecode system, that can deliver the same compiled executable to any machine regardless of OS or CPU type. It fails more often than not; there are a number of slightly incompatible 'flavors' of Java out there, not least of which are two competing toolkit APIs from Microsoft and Sun. It has marginal security. It has no windowing toolkit to speak of. It's slow as hell and it crashes the host machine way too often. It has a small installed base of applications.
Enter UAE. Sound nutty? It is. But think about it. You don't need to add much to it - just an abstracted network layer, and RTG and sound drivers. It's platform-independant - it's a "virtual machine" with crashproofing, it runs on ANYTHING, any CPU, is workably fast on modern machines, it has 12 years of applications (including arcade games and 3D modelers) to draw from, and not only has a decent windowing toolkit, it has a complete OS inside it - and still draws very little RAM.
Consider for a moment a way to encapsulate UAE into a Web browser plugin. You have on your machine a Kickstart ROM, or Kickstart-compatible ROM, and an OS install. A Web page delivers an Amiga "applet" to you - it opens up in the emulator, it can display itself in the Web page in 320x200 or whatever size it wants, by default the Amiga's emulated video hardware "pops up" in the browser, like Java applets do now. Or it can make RTG calls to a native Intuition-emulator - and bust out of the box, opening windows and menus right there on your machine like it was born there. Full-screen 24-bit stuff? Suuure. Native Pentium- or PPC- or Alpha-based math libraries? But of course. Shapeshifter? Why not.
That's just one possibility - and one that makes a frightening amount of sense, because suddenly the "network computer" concept busts wide open. An A1200-like machine with an 030 and an Ethernet port, hooked up to a VGA monitor, will be one SCREAMING "Amiga virtual machine" - nothing emulates an Amiga better than an Amiga. Sell those as NC's, and sell or give away the OS and Amiga emulator "plugin" for everyone else. Market share? You got it. Amigas on every desk? Just watch. Amiga achieves immortality? Done. And there'd be plenty more wishes-come-true where those come from - Sun and Oracle have already done all our advertising for us.
But that's just a possibility.
Now let's get back to something more conventional for awhile. Specifically, "real" Amiga hardware: I think it's time we reevaluated the PowerPC as the CPU for future Amiga models.
Why? After all, Phase 5 and the others have invested a considerable amount of time and money into making PowerPC Amiga products - right?
Phase 5 has shoved a PowerPC chip on the end of a Cyberstorm card, written a couple of DSP-like libraries and a blitter library, and invested considerable time and money in developing this "A\box" thing that no one has seen yet. A\box is not an Amiga. Neither is PIOS One. 'Course, PIOS One isn't a Macintosh either - it's not likely Apple will EVER certify a Mac clone design that outrageously brilliant.
But the PowerPC is faster than a Pentium, right?
A 200Mhz PowerPC 603e is just about neck and neck with a 200MHz Intel Pentium - the PPC has the edge in floating point, the Pentium has the edge in integer. The PPC604e is similarly neck-and-neck with Pentium Pros at the same clock speed, with a slightly wider edge in floating point and a substantially larger gap in price.
But the PowerPC has CHRP - the Common Hardware Reference Platform - behind it, right?
No. Apple's dragging its feet on CHRP certification - it has yet to ship a PowerPC Platform compliant OS. IBM similarly never shipped OS/2 for PowerPC. Microsoft has already dropped Windows NT for PowerPC. What's left? Be? You can run Be just fine on a Power Mac or Mac clone - and it'll be available on Intel before the end of the year anyway. Compatibility is not an issue - and once you talk PowerPC, you have to talk Apple. Apple, as I discussed last time, is now a puppet of Microsoft - "Vichy Apple."
So what's left on the CPU menu?
Intel chips are decently fast and decently cheap, you can get wholesale board/CPU combos for $500 or less and roll your own system. But unfortunately, Intel chips mean you must cope with the bass-ackwards PC clone architecture: the bios, the 1970's interrupt scheme, "Plug-N-Play" that only works if ALL components in a system are PNP, COM port weirdness (IRQ 3 = COM2 = COM4, IRQ 4 = COM1 = COM3?), and a zillion other problems, some of which are inherent in the chips. Besides, you don't really want Intel to get any bigger, do you? :-)
Hewlett-Packard PA-RISC? Uh... no. They had an edge in 1994, but not no more. Besides, you can expect a future (1998) generation of PA-RISC to be one and the same chip with the Intel Merced - the "80786". (The Pentium II is just an outgrowth of the P6-series.) HP and Intel have been codeveloping the Merced project for some time, and part of the agreement is that it must run both Intel and HP PA-RISC software. See above about Intel.
MiPS? Sure, if you don't mind spending that much for Pentium-like performance. They were hot shit in 1993, but are behind the times now.
SPARC? If you expect Sun to last that long, and don't mind paying MiPS-like prices for one of the weirdest RISC architectures around, go ahead.
My choice? Digital Alpha.
Nuts, I suppose - but consider. Digital (they don't like being called DEC anymore) wants market share. And they just happen to keep cranking out the fastest CPUs on the planet. If you asked them nicely to produce a consumer-grade 32-bit Alpha in the 60MHz range, for $40 per chip, they'd probably do it. Besides, there's an awful lot of "slowpoke" Alphas floating around - the 21064 and such - in the 100MHz range, stuff that's "unsellable" but still faster than a 68060 and probably cheaper too. Find some that aren't already bolted to an expensive and unsellable workstation, and away you go.
The business plan would look like this. Gateway/AI has the source code to the Amiga OS. Bring some programmers on board to do a rebuild. Add memory protection, resource tracking, SMP, and full RTG and AHI support. Leave in support for existing "weird" Amiga hardware - but make it removable. Recompile it all so it doesn't run existing Amiga binaries - and strap UAE and some RTG drivers onto it, so there's at least software to run on it.
Then design two lines of computer. There's a consumer end, selling for $600, it would have a 32-bit Alpha onboard, 16MB RAM, a hard drive, maybe one expansion slot, either AGA or an SVGA chipset, and connect to either a VGA monitor or a TV. It would have a breakout panel on the back, behind which are two "communications slots" designed to hold any two of the following: an analog modem, an Ethernet adapter, an ISDN modem, a cable modem, a satellite transceiver, an IR transceiver, or whatever else would fit nicely in there. The machine typically ships with one slot filled - usually a phone-line modem - and the user can swap or add another. Ship it in distinctive-looking cases, Boing balls, Lego, whatever - if it doesn't look like a PC, people won't expect it to act like one, and will be pleasantly surprised when it outperforms their $1600 Wintel box.
Then there's the high end. Where high-end PCs leave off, there you will find Amigas. Start with the 433Mhz 21164a, 64MB RAM, some old 24-bit card, and a SCSI-3 drive; out the door at $5000. Go on up, to the high-end eight-CPU 600Mhz monsters with gigs of RAM and a nonlinear video solution for "only" five figures. With a nice fresh OS running on them, such boxes will be delightfully zippy - no more wasting resources trying to give Windows NT room to stretch out. The OS is there to support the computer, not the other way around - and the high-end rendering people have been wishing for an OS besides Windows NT for a long time now. Build it and they will come - the developers who're pissed at the Microsoft Alpha compiler will love it, and the users will similarly dig it.
A recipe for success.
Now then. Who's got that source code again?
If you click on the "Message Board" link at the bottom of this homepage, you will find my handy-dandy sorta-threaded Web message board. I wrote it myself, don't expect any miracles. Anyway, in the next few days I will be placing on that board a petition for Gateway 2000 to support the Digital Alpha in future Amiga models and operating systems. Sign it. I don't yet know where to send such a petition when we're satisfied with it, whether I should try to coordinate with the IC/OA Steering Committee (they all seem to be PowerPC people over there), or what - but at least I can get the petition in circulation.
This is our future. It's time we started thinking outside the box about it.
[John's Homepage]
[Sarah McLachlan Stuff]
[Donna Lewis]
[Cabinet of Curiosities]
[Squidprojects]
[About John]
[John's Art]
[Email John]
[Guestbook]
[Message Board]