ReactOS and older computers...

Here you can discuss ReactOS related topics.

Moderator: Moderator Team

Z98
Release Engineer
Posts: 3379
Joined: Tue May 02, 2006 8:16 pm
Contact:

Re: ReactOS and older computers...

Post by Z98 »

nute wrote:First off, lower color density is faster. Second off, having a monochrome monitor even if you are running on
an AMD Core 2 duo is a good way to keep people from playing games.
An AMD Core 2 Duo. Right..........................

You have an odd set of priorities. Keep people from playing games? Why the blazes would we care about that?

Let's see, base system requirements for ROS these days is 100-200 MHz with 32MB of RAM. Whether you'd be able to do anything useful on such a system is another matter entirely, since there'd be no RAM left over for applications to use.

I'm seriously questioning your intelligence here. You're making claims here that any liable to fail you in any introductory computer engineering or introductory programming class if you tried to actually carry them through, at least the ones I took.

Programming on an older system does NOT mean you are forced to program more efficiently. If you are that obsessed with this so-called optimization, then go learn embedded programming. Then maybe you'll realize just how much functionality you lose when trying to program things under such resource constraints. People programming on those systems take a different approach, one that is often not directly applicable to the programming of desktop or server systems. They essentially do less with less, wherein on a desktop system the idea is to do more with less. On embedded systems, the more part is usually not possible.

The military has a radiation hardened version of the Pentium designed to survive EMP blasts due to detonating nukes.

You "encourage designing ROS so that Pentium specific instructions are isolated." And what exactly do you think this project has been doing? Usage of assembly code is kept to an absolute minimum so that porting to different architectures is easier. But we explicitly chose to ignore the pre-Pentium Intel processors because there's no reason to bother supporting them. Any modern Windows NT application that you would want to run is almost guaranteed to not work on those systems and an OS is useless without applications. Hell, I wouldn't even have the patience to test applications on something that old and weak, much less use it for daily activities.

Let's see. Prove that using a subset of the Pentium's instructions slows things down. Well, not using the cmpxchg instruction would definitely slow things down. Then there's the lack of MMX and SSE as you already pointed out, though that's more an application side loss. But hell, you've already taken a major performance hit with the loss of the atomic compare and exchange instruction. The rest are just more nails in the coffin.

So far I have yet to see any actual reason for wanting to support pre-Pentium Intel processors. They're too slow to actually run even Windows 95 era software and they're actually incapable of running applications that make use of vector processing extensions.

An additional note from Timo. 32bpp is actually faster than 16bpp since on (relatively) modern system 32bits is the natural machine word size.

nute
Posts: 251
Joined: Tue Aug 29, 2006 1:30 am

Re: ReactOS and older computers...

Post by nute »

Z98, why are you being such a jerk? Do you have nothing better to do
than be a nasty European who calls an American stupid? Figures.

What is wrong with pressing the question of why doesn't this run on older
computers? Okay, so maybe there are instructions that a 486 would have
to emulate and that would be slow. Thing is, those instructions aren't going
to exist potentially when you move to an: Alpha, Sun Sparc, or a PowerPC.
As far as using those instructions, I don't see why it's so hard to optimize
without making the code totally dependent on one computer architecture.

Windows 95, no it wasn't NT but yes it was an OS, ran in 16 megs of ram
on 486s. Windows 98SE needs more ram and the specs say 486 DX2-66,
but that's a bit tight running it on one. Both of these versions of Windows
running on top of dos are 32 bit systems. Question, if a 32 bit version of
Windows exists for 486s, why not ReactOS on the 486? These systems
could run software. Windows NT 3.51 existed when Windows 3.1 was
popular. I assume that it ran on 486s.

I want to know why ReactOS has to use so much processing power and
memory to even run. Sure people go out and buy ever faster computers
with more and more memory, but will ReactOS need the latest computer?

Sure a Pentium II or Pentium III will probably do, but why does ReactOS
need even this much computer?

The way that people are complaining about instructions specific to the Pentium
makes me think that ReactOS is being written in assembly which isn't portable.
Last I checked, ReactOS is being written in C and C++. I would think smart
linking at run time would be enough to choose the optimizations that fit the system
at hand. If you are on a Sparc and a certain instruction doesn't exist, you still have
to support it which suggests that you have to emulate it.

People complain that an older computer will never run firefox, which may not be
important depending on one's needs. One of the major complaints against Windows
is that you needlessly have to go out and buy the fastest computer with the most
memory. I hope ReactOS doesn't follow the model, we are going to write to the
latest processor instructions implemented only in the latest CPUs. People are
going to wonder why they have to buy a new computer every 3 years where
hopefully ReactOS will alleviate that somewhat.

I hear GCC isn't the greatest optimizing compiler in the world, but an optimizing
compiler like GCC should take care of optimizing the code to a certain degree for
you. Granted, there are some things an optimizing compiler can't do considering
that it's decisions are made at compile time.

Lone_Rifle
Test Team
Posts: 802
Joined: Thu Apr 03, 2008 2:17 pm
Contact:

Re: ReactOS and older computers...

Post by Lone_Rifle »

nute wrote:Z98, why are you being such a jerk? Do you have nothing better to do
than be a nasty European who calls an American stupid? Figures.
He's American. You're quick to jumping to conclusions. Perhaps this is why you're so ignorant. You don't listen, you make your own assumptions, you get upset when people try to correct you, you're stubborn. All this makes you ignorant. May I ask the highest level of education you've attended? I may be able to explain why your academic records are so woeful.

I didn't bother reading the rest of your mindless drivel because I have better things to do, one of which is to tell you of this story: My friend's friend had a chihuahua which never knew its place in life. Day in, day out, it would bark at anything and everything, even things that do not threaten it, and after a while it got quite smug about itself. One day, the friend's friend's St Bernard had just about enough of the barking. It went up to the chihuahua, gawked at it yapping away full of itself, lifted its paw, and swiped the chihuahua. The little dog skidded across the marble floor for about 100m. In my version of the story it then collided with a cardboard which then fell on it, crushing it to death.

You remind me of the chihuahua.

nute
Posts: 251
Joined: Tue Aug 29, 2006 1:30 am

Re: ReactOS and older computers...

Post by nute »

There is an attitude of superiority and frankly disdain on the part of developers.
This doesn't impress me. It makes me think that a lot of you are deeply prejudiced.

I know that registers are faster than: the L1 cache, the L2 cache, and main
memory. The registers or L1 cache takes about a processor cycle for an access to
happen, the L2 cache is about 10 cycles away, and main memory is roughly 100
cycles away. This is why there is a translation lookahead buffer in the CPU, it
sometimes saves the CPU from having to go to main memory to read a page table
entry. Every Intel processor from the 386 up to at least the Pentium 4 is a 32 bit
processor. Instructions have been added and newer processors do branch
prediction to try and speed things up, but beyond that there is a core architecture
that has remained essentially the same.

Companies are starting to write software for Linux. Rick Rocket and Dirk Dashing
are 2 examples of games. I have yet to see a port of AutoCAD that is reasonably
priced for Linux, but who is to say that that won't happen? Playing the game
reverse engineer proprietary software so that you can run the programs written
for it is bound to be a frustrating. Maybe this whole project needs to be
reevaluated. It started in 1996 and to this day there are no stable releases. In
that time, we have gone from Windows 95 and NT4 to Vista and soon
Windows 7. Windows Vista and Windows 7 are arguably abusive OSes
in that they have all this DRM crap in them, but what is going to change that?
There are no laws to allow ReactOS developers to look at Windows source
code, even NT4 code. Never mind that Microsoft doesn't sell NT4 anymore
and never mind that it doesn't support it either. Never mind that Windows has
almost 99% market share. When a company is so successful that it overtakes
all of it's competitors, it's product should be placed in the public domain.

I don't know if a project like this could be finished sooner if it wasn't trying
to be completely compatible with Windows, I guess the Syllable and Haiku
projects are the ones to watch for the answer to that question. If even the
oldest version of NT became public domain, that would probably help this
project along considerably.

Linux, at least the older releases, runs on old computers. It's one of the
arguments for using Linux. Linux has become very capable over time
and frankly gnome the way it is set up by default feels an awful lot like
Windows. There seems to be some momentum to program to non
Microsoft API's. The importance of Windows is questionable these
days. ReactOS doesn't run much Windows software yet, and
it may be too early to tell if that will change significantly. Linux has
big companies like IBM behind it, what companies are behind ReactOS?

nute
Posts: 251
Joined: Tue Aug 29, 2006 1:30 am

Re: ReactOS and older computers...

Post by nute »

I am doing well in college. As far as the dog story,
dogs are notoriously dumb. If it's mindless drivel I
am writing, how did I write it? It's easy to attack
a faceless person on a message board like this,
but that doesn't make it right to do so.

This is great publicity for ReactOS. ReactOS, the
system backed by developers who are meaner than
Ballmer. The developers who will convince you to
run to Microsoft because at least Microsoft doesn't
call you stupid.

Public relations, who is responsible for PR in this
community? Anyone? Does anyone on here who
is being so nasty to me want to take responsibility
for their behavior and try to make peace, or is that
too much to ask?

Great idea, act out because of coding frustrations.
Alienate people. Insult someone's intelligence.
And while you're all at it, go ahead and announce
that the ReactOS project is shutting down indefinitely.

hto
Developer
Posts: 2193
Joined: Sun Oct 01, 2006 3:43 pm

Post by hto »

What is amazing, that these troll attempts produce much heat. :)

nute
Posts: 251
Joined: Tue Aug 29, 2006 1:30 am

Re: ReactOS and older computers...

Post by nute »

"What is amazing, that these troll attempts produce much heat. :)"

What is so great about there being a heated and frankly quite nasty
discussion?

Who are you calling the troll? How is that helping the situation?

I'm surprised the PR people haven't locked this thread before it gets worse.
Clearly, it doesn't take much provocation to get certain people in the
ReactOS community to flame. That is too bad.

hto
Developer
Posts: 2193
Joined: Sun Oct 01, 2006 3:43 pm

Post by hto »

Clearly, it doesn't take much provocation to get certain people in the ReactOS community to flame. That is too bad.
+1

Lone_Rifle
Test Team
Posts: 802
Joined: Thu Apr 03, 2008 2:17 pm
Contact:

Re:

Post by Lone_Rifle »

hto wrote:What is amazing, that these troll attempts produce much heat. :)
What's even more amazing, it seems, is that some people seem to become all defensive when I form an opinion about what they say.

Haos
Test Team
Posts: 2954
Joined: Thu Mar 22, 2007 5:42 am
Contact:

Re: ReactOS and older computers...

Post by Haos »

Why wouldnt anyone listen to me more than year ago when i asked so nice just to ban him? We wouldn`t have be putting up with the crap he produces...

Or at least PLEASE, close this thread down, before it turns really nasty...

Ged
Developer
Posts: 925
Joined: Thu Sep 29, 2005 3:00 pm
Location: UK

Re: ReactOS and older computers...

Post by Ged »

I love a heated debate as much as the next man, but this one is getting a little too personal now.

Z98
Release Engineer
Posts: 3379
Joined: Tue May 02, 2006 8:16 pm
Contact:

Re: ReactOS and older computers...

Post by Z98 »

One final thing, since you're right that I never did explicitly address the whole "why ROS takes up so much resources" argument you've been making, even I've implicitly addressed it multiple times. And yes, I know the topic is locked.

Your assertion only works if one regards 100 MHz and 32MB of RAM as a significant amount of resources. Considering the current generation of desktops, how you can make that assertion with a straight face is beyond me, except for one possibility which I'll cover later.

You are trying to compare operating systems that provided significantly less functionality than the current generation of NT operating systems and claiming that the resources needed for both should not change. Any new functionality is going to increase the code size, which then increases memory usage. Considering how much has been added to NT compared to the original NT and even the DOS systems, there is no possible way to have not increased the resources required to provide that functionality. In spite of this, we've gotten memory usage down below 32MB of RAM. How one can possibly go any lower is beyond me, since then you would have to start eliminating the base functionality that the 5.x NT family provides. Since the entire point was to create an OS compatible with the NT5 family, that defeats any purpose whatsoever.

Now the only way the assertion that 100 MHz and 32MB can be considered a significant amount of resources is if you tried to use ROS in a situation where we never designed it to be used for. We explicitly state the minimum system requirements needed and if you choose to ignore those, then it's your problem to try and get it to work. We provided that list of requirements so that people understand if they are working outside those bounds, we are unlikely to provide any assistance. The same goes with the requirement for a Pentium-class processor. We intended for ROS to be used on a Pentium-class or later. We chose not to support the pre-Pentiums because it offers no advantages whatsoever. ROS is intended to be a desktop or server OS, not an embedded OS. And even embedded systems these days offer more processing power and memory than a pre-Pentium processor.

You can continue pressing the question of why ROS won't run on completely obsolete systems, and we will continue responding that we never intended for ROS to run on such systems. We see no advantages to supporting them and you have yet to provide any convincing ones. The argument that other OSes ran on them is a moot point, since we're not those other OSes and they're as equally obsolete as the processors you keep bringing up and we are not trying to achieve function parity with them. We're in fact trying to surpass the ones you mentioned.

I'll stop here since this topic is locked and the rest of your assertions get more and more ridiculous. Your arguments are demonstrating a serious lack of understanding about this project and its goals under the guise of worrying about its future. If you insist on not informing yourself about the project's present, then there's nothing we can do to correct your misconceptions.

Locked

Who is online

Users browsing this forum: Bing [Bot], Semrush [Bot] and 1 guest