Idle DR-DOS | Hacker News Reader

Idle DR-DOS

os2museum.com

211 pointsingve29 comments

I continue to be amazed by Michal Necasek's ability to deeply care about things that literally nobody has even thought about for decades. Kudos!

Power management under DOS on early x86 laptops was downright primitive. There were no standards of any kind (I'd never even heard of the DR-DOS $IDLE$ device until now, but it doesn't seem like that gained any kind of traction at the time), not even BIOS-based abstractions. Just vendor-provided 'drivers' (in name only: it's not like DOS made any distinction between system and user code) that directly communicated with the hardware.

Unsurprisingly, that all broke down as soon as you did anything even remotely interesting. Run QEMM386? No power management for you! Have a certain PCMCIA card inserted? Ditto! And shipping any kind of updates took forever at that time (literally FedExing floppies...), so getting a working system was pretty much impossible.

What I find interesting is things like the Atari Portfolio and HP 200LX could run their versions of DOS for quite a while on a couple AA batteries.

IIRC they had some basic "suspend to RAM" where when they closed they just barely kept the RAM alive and didn't power anything else.

While Atari Portfolio used the CMOS version of 8088, I believe HP ones made the jump to 80186 (or compatibles) which AFAIK had, among other things, dynamic frequency setting

Correct - the x00LX (I think the 95 also) were one of the few mass produced computers to use the rare 186.

The HP 200LX and similar used a 186-compatible CPU called "Hornet". Using an actual 80186 in a PC compatible would have been damn near impossible; the 186 was a system-on-chip design that had built-in timers, an interrupt controller, and other gubbins. Some of these were mapped to devices that had other devices mapped to them in the PC design, so building a 186-based system that met the common threshold for PC compatibility (i.e. it ran the IBM versions of Lotus and Flight Simulator acceptably well) was a nearly insurmountable task.

Because it acted more like a real OS, with user programs calling through driver APIs rather than touching the hardware directly, with the right driver support Windows could run on a 186-based system; one (the Tandy 2000) was used to develop Windows 1.0's color modes because it had a high-resolution 8-color display before the EGA came out: https://www.1000bit.it/lista/t/tandy/tandy2000_ad.jpg

ISA-compatible "clones" of the 186, like the NEC V40 and the "Hornet", did not have the SoC problem and could be used in PC-compatible system designs.

In other words, it has a 186-class ISA but isn't the actual 186 SoC.

It was never really clear to me if the "Hornet" was a custom design Intel did for HP or if "Hornet" was just the codename for the 80C186EA, a 186 with added powersaving support. Either way those palmtops were quite cool machines.

For what it's worth, I've used a HP 200LX to confirm that some undocumented details (found by disassembling microcode) work as expected:

    REP BOUND             only compares against lower bound
    REP MUL               result in AH (byte) or SP (word)
    REP IMUL              result in CH (byte) or BP (word)
    REP IMUL reg,r/m,i16  first operand ignored, 32 bit result in DX:AX
    REP IMUL reg,r/m,i8   first op ign., AH=result bit 23..16, AL=bit 7..0
    AAM 00H               no "divide by zero" exception, AH=FFh, AL unchanged

So it's certainly based on the 80186 microcode, as none of these instructions do the same on any other x86 chip.

This is perhaps the comment that most aligns with a username I’ve ever seen ;)

Thanks for the best explanation of the '186 I've seen in 30 years. :-)

No problem! The more you know, right?

I mained a used Tandy 2000 for a while as a kid/teenager. Back when it came out, computers still came with full technical docs, including Intel's extensive 186 documentation. I've forgotten things about that beleaguered chip many enthusiasts never knew.

Research Machines made a sizable inroad into UK schools with their range of computers from the late 1970s to early 1990s.

One such example is the RM Nimbus: https://www.computinghistory.org.uk/det/41537/RM-Nimbus-PC-(...

It had a 186 from the very first models, and was later re-branded as PC-186 when they introduced their 286 model.

Indeed so. The BBC Micro Master 512 was the only other one I saw.

Ditto. Same feeling here.

Kudos, indeed!

This must be the first time in history when the technical information in a software patent was actually useful.

A requirement that a software patent compile and run would be a welcome one.

When a DOS VM does not have any power management, the laptop quickly lets me know by kicking up the fan speed.

An examination of the source code for the PC/AT BIOS reveals what it actually does when waiting for input on INT 16/AH=0 --- it polls the keyboard buffer in a loop that also appears to repeatedly update the keyboard LED indicator status, and one of the instructions that it executes (with interrupts enabled) is a NOP. Naturally, that makes one wonder whether simply replacing that NOP with a HLT would be sufficient to idle the CPU, and if so, why didn't IBM do that originally; and what newer BIOSes do.

The 6A69MV39 AWARD BIOS I inspected (from 1999) looks similar, but loops with a "delay" created by reading from I/O port 0EBh.

What would unHLT it? The next timer tick?

It stops execution until an interrupt occurs.

https://en.wikipedia.org/wiki/HLT_(x86_instruction)

This is an awesome article, but I was hoping for a copy of the final source and MAKE files on Github...

Spoon-fed kids nowadays, sheesh. In my days an .img file was a godsend. You'd took it as your precious and ran away with it. A bit of disassembler and you were good to go. I recommend Ghidra, you'll learn a valuable skill, namely reverse engineering, trust me, will give a very nice thrill of the hunt.

Believe me, I've done my time in the trenches. I've disassembled enough x86 machine code to get my sick kicks. And I did a lot of it using DEBUG.EXE, the monster that I am!

Memories of inputting small COM programs with debug so I could jump start serial communications and bring over bigger guns - good bad times!

Should I feel uncool now for having used gwbasic for that? ;)

Look at this fancy dude having gwbasic ha! We were lucky to find a DEBUG laying around, some poor slobs were stuck with EDLIN.

I remember back in the day magazines would provide small utility programs as statements to enter into DEBUG.

It was (and is!) amazing to realize how much power you could fit in an assembly language program small enough to be something you could type in by hand. Sure some heavy lifting was done by INT 21h but low-level programming is higher level than many think. Compare to godbolt where many C programs aren’t much longer in assembly.

This is pretty epic work, nicely done!

Reply on news.ycombinator.com