INTL 3:770/1 3:770/3
REPLYADDR not@telling.you.invalid
REPLYTO 3:770/3.0 UUCP
MSGID: 2e2a0e57
REPLY: 3139:770/1.0 2303dac3
PID: SoupGate-Win32 v1.05
Daniel wrote:
> Before saying anything, I want to point out that there is no pretense
> of expertise in this subject. I'm just a curious bean. As the growth
> of retro computing matures, projects to resurrect the platforms by
> building vice boxes gets more common. The C64-mini, the zx spectrum,
> sega.. Otherwise, the 8-bit guy is taking off-the-shelf components to
> build himself a modern juiced up Vic20 to sell at some point beyond
> vaporware. They're creating the basic interpreter and kernal for their
> system. All's well and good. This brought me to an interesting thought
> with a similar notion. What stops anyone from doing the same thing
> with a modern cpu and memory/bus system? Is it the complexity of the
> modern cpu? In retro systems, the developer controlled memory
> allocation such. I'd assume the difficult part would be to micromanage
> every bit of memory management on a complex system. Am I on the right
> track?
There are a few things in play that prevent similar start-up times
to old 8bit computers being easily acheived.
First remember that technically nothing's instant, indeed there's a
decidedly perceptible delay between turning on a C64 and seeing a
command prompt, it's just not long enough to be inconvenient.
Also a C64 only has to read from its own ROM, which is equivalent in
performance to RAM and is ready and waiting to be accessed as soon as
turned on. PCs almost always have to load their OS from a storage
device into RAM, and then begin execution. Storage devices are
slower, need initialisation (BIOS code), and limited by an interface
bus that does not connect directly to the CPU like RAM/ROM.
A more fair comparison is when the C64 is asked to load GEOS from a
floppy drive, where of course the results are less in its favour.
It is possible for the OS kernel to be included in a custom BIOS
(eg. CoreBoot) that is written over the default one. Some of the
old PCs actually made by IBM had their own BASIC interpreter in
BIOS.
But next you have to do all of the detection and initialisation
for modern devices, which all have their own little bits of code
that need to run on a per-device basis. Here you can compare with
MSDOS, which on eg. a fast early 90s PC with a quick BIOS (some were
very quick, eg. on laptops) could approach something like the
start-up time of a C64. The advantage that DOS had is that it relied
on the BIOS to handle the HDD, display, and keyboard initialisation,
and for what else the user wanted to use they configured a specific
driver that didn't have to do automatic detection because the poor
confused user was usually forced to figure out the details and
pre-configure it manually. There often wasn't all that much else
that needed to be initialised on a typical DOS system either.
A modern OS is designed to detect everything automatically, and
generally ignores any device interface provided by the BIOS in
preference to its own set of device drivers which are probably
faster and/or more reliable once they've been loaded.
Still there are roads to take if you want to pursue a PC that boots
in a non-inconvenient amount of time, albeit maybe a little longer
than an old 8bit. I've already mentioned DOS, and while drivers for
modern things like networking and USB will slow it down, you can find
out for yourself what the limit is with FreeDOS. TinyCore Linux is a
very fast minimal Linux distro that loads itself entirely into RAM on
start up, and the start-up time is thereby proportional to the
packages that you configure to be loaded, after which they run very
quickly. KolibriOS is an x86 OS written entirely in assembly - I've
only booted it from a Floppy or a CD so I haven't really seen the
start-up time from HDD, but it probably has potential. I've already
mentioned CoreBoot, which can optimise the first step - loading the
OS kernel. I'll ignore stand-by type systems on the basis that
they're kind-of cheating.
All of those options have potential, but it's not the direction where
mainstream OS development has gone. Fact is that it's easier to write
an OS that boots slowly, while doing everything that users demand
from one today. Said users generally put up with it, so that's what
we get.
Oh and there are programs for PC like GRUBinvaders that run without
an OS at all, using just the BIOS interface to devices. That's akin
to a cartridge ROM running on a C64, except that the program has to
be loaded from HDD into RAM first. Though you could probably put it
in a CoreBoot BIOS image and have a PC that boots really quick but
just plays text-mode Space Invaders and nothing else. :)
- Idea: Connect the BIOS EPROM socket up to a female edge connector
and make cartridges with CoreBoot images for different x86
OS-less games?
--
__ __
#_ < |\| |< _#
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)
SEEN-BY: 1/123 14/5 15/0 2 19/36 90/1 104/115 106/201 114/224 702
SEEN-BY: 114/705 706 116/18 120/340 601 123/131 140 128/2 73 187 253
SEEN-BY: 153/7715 218/700 222/2 226/16 30 227/114 229/101 275 424
SEEN-BY: 229/426 664 1014 230/150 152 240/1120 5832 249/1 206 307
SEEN-BY: 249/317 400 250/1 261/38 100 266/512 267/155 275/100 282/1031
SEEN-BY: 282/1056 291/100 111 298/25 305/1 3 310/2 312/2 317/3 320/119
SEEN-BY: 320/219 322/757 340/400 341/66 342/200 396/45 640/1321 712/848
SEEN-BY: 801/161 189 2320/105 3005/1 3634/12 5020/1042
PATH: 770/3 1 712/848 261/38 15/0 317/3 229/426
|
