It's always kind of an odd time in Florida, as we're staring at an oncoming hurricane.
It's too soon to know exactly what we might expect here in northeast Florida. It won't be "nothing," but the range is fairly wide.
Anyway, spent some time distracting myself this afternoon playing around with old computers.
As I mentioned before, the Apple IIe is working in nearly all respects. I think the McT SpeedDemon hung on its self-test. I'm guessing it may have something to do with the 8MB Aux Slot memory card. In all other respects, it works just fine.
I received some old DOS 3.3 disks today, the "original" disks Apple included with the IIe. They all booted and ran fine, even accelerated, so I'm pretty confident it's solid.
This morning, I cut and pasted the Applesoft code for the fidget spinner program into an Applesoft emulator web site. It ran in 25 seconds, which is about a minute faster than it ran on real hardware. I'd been wondering how fast it would run on real hardware using Beagle Compiler, a program that came out late in the II's life that would "compile" Applesoft programs and run them much faster, depending on how much floating point math they used. Floating Point didn't "accelerate" at all after compiling.
I had to do some hunting around for the "latest" (last) version of Beagle Compiler, because earlier ones had conflicts with later versions of ProDOS or BASIC.SYSTEM. In any event, I did find one from 1990 and it was able to compile the fidget spinner and it ran in about 30 seconds. I may be able to eek some additional speed out of it if I convert all the variables to integer variables. In normal Applesoft, there's usually no value in using integer variables, except to save space as array indices. Otherwise, Applesoft converts all numbers to their floating point representation, and handles them accordingly, which eats up some time. Beagle Compiler will treat declared integer variables (n%) as integers, and that saves a decent bit of time. Haven't done that yet, but probably will, just out of curiosity.
So then I wanted to see it run on the IIe with the SpeedDemon running. That's when the fun and games started.
I couldn't seem to get booted into ProDOS. It would get to the splash screen, then crash into the monitor. It would work if I turned off the SpeedDemon first, but that would kind of defeat the purpose of having the accelerator. That was with contemporary versions of ProDOS (2018). I thought I'd try with earlier versions, more contemporaneous with the SpeedDemon and see if I'd have more luck.
Well, after working successively back, I got the original ProDOS release, 1.0.1 from 1984, to boot. From there, I first ran the fidget spinner just using the accelerator. Ran in about 25 seconds. (I should have taken notes.) About as fast Javascript emulator.
I was able to get the compiler system into memory, I worried that it might have been incompatible with the early version of ProDOS. Crossing my fingers, I ran the fidget spinner from the compiler under the accelerator.
Ran in about 8 seconds! About 10x faster than regular hardware.
If I tweak the index variables, I might get another second out of it. We'll see.
It's a bit of a bear to pull the IIe out from under the monitor stand to open it up. I'll do that tomorrow sometime and pull the 8MB Aux Slot card and put the original 64K card and see if that makes any difference on the self-test for the SpeedDemon, and then see what happens with more recent versions of ProDOS.
I may have a conflict between cpus. The motherboard has the OG NMOS 6502A installed, while the SpeedDemon is running a 4MHz CMOS 65c02. I've ordered a 65c02 for the motherboard, which was part of Apple's original IIe "enhancement" kit. But I'll swap out the memory cards first and see what difference that makes.
Anyway, something to do rather than fret about hurricanes, war in the Middle East, half of America wanting a senile ex-gameshow host, wannabe dictator as their next president, and all the other stuff that makes life "interesting."
Originally posted at Nice Marmot 17:36 Monday, 7 October 2024