29-Mar-2024 10:21 GMT.
UNDER CONSTRUCTION
Anonymous, there are 105 items in your selection [1 - 50] [51 - 100] [101 - 105]
[Rant] AOS4 on AOne slow? Perhaps, perhaps not.ANN.lu
Posted on 23-Sep-2003 09:53 GMT by Fabio Alemagna105 comments
View flat
View list
Ok, I haven't had a chance to look at those videos until a few minutes ago, and I thought I'd give some reasons for which it appears slow, basign my judgments on what I see and what I know about the AmigaOS internals.

First of all, I must say that, although there was perhaps nothing more to see, those videos only show one little part of the GUI system, which, from what we know, could perhaps be the only one a bit flawed.

In essence, what we are shown are 2 things: the speed with which Reaction GUIs are drawn (in Solid-Resize and GUI-Reactivity), and the way (a faulty one - but read on) opaque window moving is implemented (in Opaque-Move).

Indeed, Reaction GUIs are drawn quite slowly, but I can't help but notice that Reaction GUIs have always been inherently slow, at least on my UAE setup, much slower than any other GUI, even MUI. The reason for this can be researched in the fact that Reaction uses a completely different approach than MUI, which is also the reason for which I greatly dislike Reaction in favour of MUI. However, it's obvious that the redraw is quite slow also considering Reaction's faults, and this is possibly due to the fact that the gfx library was emulated, along with a basic chipset support (some blitter thingies, like the BLTDONE flag in DMACONR register), which surely slowed everything down a lot.

Using this argument to say that AOS4 is at the same stage MOS was 2 years ago is pure flamebait for mainly 2 reasons: 2 years ago MOS already had a native gfx subsystem (which renders the comparision useless, although it might seem that it makes the situation even worse for AOS4), and also because this gfx subsystem is a temporary one, and we don't know at which state of development the new one is (but I reckon it's close to completion, or on the way to it). In any case, such comparisions are meaningless because AOS4 has now things which MOS didn't have 2 years ago, and vice versa.

About the opaque moving, instead, it can be noticed that there's a flaw in its implementation, which basically makes it *very hard* to *impossible* for the damaged windows to refresh themselves until the movement has stopped. Read Georg Steger's explanation of this phenomenon here

.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 1 of 105ANN.lu
Posted by Peter Gordon on 23-Sep-2003 08:15 GMT
Basically, its not important and the videos don't show what OS4 will be like with the final PPC native graphics system.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 2 of 105ANN.lu
Posted by Rassilon on 23-Sep-2003 08:16 GMT
@Fabio

..... Congratulations!! I think that has to be the first unbiased review made of the AOS4 movie clips!

Rassilon
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 3 of 105ANN.lu
Posted by Anonymous on 23-Sep-2003 08:19 GMT
@Fabio

Very nice! I have to admit I'm pleasantly surprised, but you excelled yourself here - the only unbiased report I've seen.

Ian
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 4 of 105ANN.lu
Posted by Jürgen Schober on 23-Sep-2003 08:27 GMT
Please also check this:

<A HREF="http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=1399&forum=4&start=20&viewmode=flat&order=0">
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=1399&forum=4&start=20&viewmode=flat&order=0</A>

Some things are already explained herein.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 5 of 105ANN.lu
Posted by Crumb // AAT on 23-Sep-2003 08:33 GMT
"I can't help but notice that Reaction GUIs have always been inherently slow, at least on my UAE setup, much slower than any other GUI, even MUI."

really? I've always felt Reaction to be quite faster than MUI... at least in my 4060/60 and in my laptop with UAE. And no, I don't use textures in the MUI windows... And I've always felt that ClassAct was faster too
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 6 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 08:43 GMT
In reply to Comment 1 (Peter Gordon):
> Basically, its not important and the videos don't show what OS4 will be like
> with the final PPC native graphics system.

I didn't say that, nor that's a possible conclusion you can draw from my words. It is important that Reaction is inherently slower than MUI, for instance, and it's important that opaque window moving is implemented the wrong way. It's not SO important, though, and it doesn't NECESSARILY reflect the way things will be when the product is sold.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 7 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 08:49 GMT
In reply to Comment 4 (Jürgen Schober):
From a Rogue's post in that thread:

"Furthermore, there is a certain overhead when changing between native code and
emulation, making these calls more expensive since layers is PPC."

This means that there's still some sort of "context switch" penalty when going from/to emulated code. As a comparision, MOS doesn't have this problem.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 8 of 105ANN.lu
Posted by Jürgen Schober on 23-Sep-2003 09:03 GMT
In reply to Comment 7 (Fabio Alemagna):
>This means that there's still some sort of "context switch" penalty when going
>from/to emulated code. As a comparision, MOS doesn't have this problem.

How comes. Can you explain that ? A simple trap is overhead compared to a simple jmp. Even, the system code (graphics.library) runs interpreted without L2 caches on, yet. So, this gives you an overhead, indeed.
Can you tell me hwo you define "context switch" here ?
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 9 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 09:06 GMT
In reply to Comment 8 (Jürgen Schober):
> How comes. Can you explain that ? A simple trap is overhead compared to a simple
> jmp.

Is that trap handled by the MMU? That's the overhead then. If not, I don't see what overhead there could be. MOS doesn't have any such special overhead, it's just a bit longer than a jmp, but not as long as it would be if the MMU were involved.

If you don't trust me, just ask any MOS Dev.

In other words, giving the emulation context switch as a reason for the thing to be slow is meaningless unless that is really a bottleneck. I was taking Rogue's word for granted and assuming that if he said it's a bottleneck then it is.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 10 of 105ANN.lu
Posted by Peter Gordon on 23-Sep-2003 09:07 GMT
In reply to Comment 6 (Fabio Alemagna):
OK... I should have preceded my statement with "This is not a summary of Fabios points, it is my opinion on the whole thing".
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 11 of 105ANN.lu
Posted by Kjetil on 23-Sep-2003 09:20 GMT
In reply to Comment 9 (Fabio Alemagna):
And how slow is PPC MMU interrupt compared with a jmp and some 68k register argument conversions?
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 12 of 105ANN.lu
Posted by Rassilon on 23-Sep-2003 09:26 GMT
In reply to Comment 9 (Fabio Alemagna):
@Fabio

Unfortunately I have to take part what I said earlier about this being the first unbiased review. From your later posts it seems you are hovering dangerously close to the edge of being MOS biased. I am saying this because your statements in your later posts see you making asumptions as to why AOS4 may be worse than MOS. If you are right then fine. But you don't know for definate and wrog assumptions canlead to major mistakes!!

Rassilom :)
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 13 of 105ANN.lu
Posted by Eva on 23-Sep-2003 09:26 GMT
I will no comment the slowness of Aos4 compared also to MOS 0.8 on cyberstorm PPC, but simply add a point.
What are the andvantages behind the wait of another year to see something comparable to Mos in the actual state?
Actually we can only see market advantage for Hyperion and AmigaInc ... or?

---
In any case, such comparisions are meaningless because AOS4 has now things which MOS didn't have 2 years ago, and vice versa.
---
Absolutely no. Actually Mos is a commercial, mature and usable product and can't be compared to what ALL we saw in EMpoli. There are no things that Mos don't have compared to Aos4 simply because Aos4 rest a project, not a product. Really nevermind to the Cyberstorm version of A0s4! Because must be clear that Aos4 version on Cyberstorm PPC is also really far to be complete.

And then, who are you Fabio that can speak about Aos4 or Mos? What are your references? (PS Excuse me Fabio, LOL but I just readed a Bouma post on one of the Pianeta Amiga report about references LOL That guy need a psichiatrik, or a cure to his magaloman idea to be the official Hyperion&AmigaInc news cersor.)
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 14 of 105ANN.lu
Posted by Rassilon on 23-Sep-2003 09:31 GMT
In reply to Comment 13 (Eva):
Hey i'm on a roll today!!!

@Eva

If they whole world was to take your view on everything then no-one would have bothered producing a different car after the Model-T Ford!

Rassilon
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 15 of 105ANN.lu
Posted by samface on 23-Sep-2003 10:00 GMT
In reply to Comment 6 (Fabio Alemagna):
>It is important that Reaction is inherently slower than MUI, for instance

Is that a fact? I'm sorry but I simply have to disagree. I've never experienced any slower GUI than MUI in my entire life. Compare a fully configured MUI and Reaction on the same 68k computer and you will see a major difference, and it's NOT benefitial on the behalf of MUI. For example, I used MUI for file requesters on my Amiga (can't remember what it was called) and when a file requester opened, I could literally see the window beeing drawn on the screen. My Amiga was equipped with a 68040@32MHz processor and a BVisionPPC. I can tell you that it was not a fault of the hardware considering that MUI only uses 16 colors and it didn't matter if I removed whatever eyecandy that I had applied.

Reaction is flying in comparison to MUI.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 16 of 105ANN.lu
Posted by Anonymous on 23-Sep-2003 10:03 GMT
In reply to Comment 13 (Eva):
You're asking Fabio about references? And what are yours? You don't even post with a real name. I don't thing anyone has ever seen anything other than trolling and flamebaits from you.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 17 of 105ANN.lu
Posted by brotheris on 23-Sep-2003 10:08 GMT
In reply to Comment 15 (samface):
on 30/50 reaction freezes pointer when drawing more complex gui and I can see how it draws, takes around 0.4s (freezing because of input.device). MUI doesn't freeze o redraw. mouse pointer moves, redraw took a bit shorter time. MUI guis tend to be more complex.
asl.library uses gadtools, it doesn't use Reaction. So yuo are comparing different things.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 18 of 105ANN.lu
Posted by itix on 23-Sep-2003 10:13 GMT
In reply to Comment 7 (Fabio Alemagna):
Well, in the final product overhead should be inexistant because when both Intuition and Graphics are native PPC code. But then 68k apps have this context switch overhead of course.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 19 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 10:19 GMT
In reply to Comment 15 (samface):
> Is that a fact? I'm sorry but I simply have to disagree.

It's a fact for what I can see on my setup. Also, it's a fact if you consider how Reaction and MUI work: Reaction not only is slower, due to the overhead of the locking mechanism that BOOPSI gadgets need, it also slows down the whole system, since reaction gadgets run in the input.device's context.

AOS3.5 prefs programs are just slow, even if compared to the MUI prefs program.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 20 of 105ANN.lu
Posted by samface on 23-Sep-2003 10:20 GMT
In reply to Comment 17 (brotheris):
The comparison was in the fact that MUI has the mere ability to end up in a situation where drawing is noticable, and I used the file requester example since it is must be one of the most basic tasks for a GUI rather than something complex. I've never experienced such thing with Reaction on my AmigaOS3.9 setup, ever.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 21 of 105ANN.lu
Posted by Eva on 23-Sep-2003 10:21 GMT
In reply to Comment 16 (Anonymous):
Lol I was ironic, read the PS :D
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 22 of 105ANN.lu
Posted by Graham_nli on 23-Sep-2003 10:23 GMT
Thanks for your thread, but the reasons it is slow have been covered by Hyperion over on AmigaWorld.net already.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 23 of 105ANN.lu
Posted by samface on 23-Sep-2003 10:24 GMT
In reply to Comment 19 (Fabio Alemagna):
You must have a really weird setup. Anyone ever calling MUI fast is completely unreal to me.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 24 of 105ANN.lu
Posted by brotheris on 23-Sep-2003 10:27 GMT
In reply to Comment 20 (samface):
You are lying. Freezing mouse pointer is userunfriendly as it can be.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 25 of 105ANN.lu
Posted by Hans-Joerg Frieden on 23-Sep-2003 10:28 GMT
To be honest, I don't know what the whole discussion is about. AmigaOS 4 is work in progress, and as such, it still has its flaws that are being worked on (hence the "work in progress"). If you where looking at a final version with these flaws, then granted it would be a reason for ranting. However, it isn't.

Before anyone says that we will never get rid of these problems, I want to remind you that similar "experts" have said that "OS4 will never work at all" and "OS 4 will never work on the AmigaOne". It does both (in fact the version you have seen is the result of approximately two weeks of work; I had predicted 3 months of the migration).

As I have stated over at AW.net, we're working on ways to improve the speed of the update. Parts of the problem is that the Workbench used on the A1 demo was still the 68k version, and the prefs programs are, too. Another part of the problem is that graphics is still 68k. The impact on performance is not only that emulation is inherently slower than native code, it is also that switching between emulation and native code incurs an overhead that especially for short functions adds up considerably, more so since Layers is already PPC-native (there is also no L2-Cache support yet; we didn't have time to add it to the demo). As Georg also pointed out, the region code in graphics isn't very hot, and also emulated. Also, as Fabio correctly pointed out, the graphics library used is the original one (I only did three basic patches to prevent lockups in WaitBlit and WaitTOF and WaitBOVP), the rest goes via an exception handler that throttles most accesses and emulates a few (the bfe001 mouse button issue, for example). And graphics library does a lot of custom chip accesses, and they all trigger an exception.

Like I said, OS 4 is in Beta (Alpha even for the A1 version), and things are added/ported/fixed on a daily basis. The biggest showstopper is the still-68k graphics subsystem, which is what we are currently working on. So I would like to ask people to keep an unbiased look at things and accept that we're still underway and not there yet - no one claimed we are.

Thanks for reading.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 26 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 10:30 GMT
In reply to Comment 23 (samface):
> You must have a really weird setup. Anyone ever calling MUI fast is completely
> unreal to me.

If there's anyone unreal here, sammy, it's you. I've explained you why things are like I say they are. Now, do some stopwatch tests and come back with the results. You'll be surprised.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 27 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 10:31 GMT
In reply to Comment 22 (Graham_nli):
> Thanks for your thread, but the reasons it is slow have been covered by Hyperion
> over on AmigaWorld.net already.

I prefer to judge things by myself, from what I can see, rather than blindly believing the parties involved. I have enough in-dept knowledge to be able to do that.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 28 of 105ANN.lu
Posted by Hans-Joerg Frieden on 23-Sep-2003 10:32 GMT
In reply to Comment 7 (Fabio Alemagna):
The overhead on a switch to the emulator is that the emulator requires a certain setup to run, and a certain register layout. This means that you need to save some registers and reload some others. The overhead isn't big, but it tends to add up when a lot of small functions are emulated.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 29 of 105ANN.lu
Posted by Hans-Joerg Frieden on 23-Sep-2003 10:34 GMT
In reply to Comment 9 (Fabio Alemagna):
There is no MMU involved. But even a simple function call means a certain overhead.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 30 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 10:35 GMT
In reply to Comment 25 (Hans-Joerg Frieden):
> To be honest, I don't know what the whole discussion is about. AmigaOS 4 is work
> in progress, and as such, it still has its flaws that are being worked on (hence
> the "work in progress"). If you where looking at a final version with these
> flaws, then granted it would be a reason for ranting. However, it isn't.

Indeed mine wasn't intented to be a rant :-O, I expected it to be moderated in the forum section, but well... I mainly tried to stress the point that AOS4 is WIP and that there's nothing to complain about as of now.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 31 of 105ANN.lu
Posted by samface on 23-Sep-2003 10:35 GMT
In reply to Comment 22 (Graham_nli):
What the... How dare you enter an ANN.lu thread and post a comment with intelligent reasoning?

;-)
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 32 of 105ANN.lu
Posted by itix on 23-Sep-2003 10:36 GMT
In reply to Comment 23 (samface):
You cant use IBrowse, YAM, AmIRC then? How about FTP clients, games, mp3 players/streamers?
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 33 of 105ANN.lu
Posted by David Scheibler on 23-Sep-2003 10:37 GMT
In reply to Comment 23 (samface):
Holger Kruse once did some speed tests with his GUI modules for MIAMI (MUI,
Reaction) which also showed that MUI is faster than Reaction.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 34 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 10:38 GMT
In reply to Comment 29 (Hans-Joerg Frieden):
> There is no MMU involved. But even a simple function call means a certain
> overhead.

That's ok, I was just drawing an, admittedly close-to-wild, conclusion from those words. I just don't think that that kind of overhead is the real culprit here, I don't think there can be that _many_ function calls which add up for that so BIG slow down when drawing the gui. I'm more inclined to believe that it's the whole thing of the emulation and the fact that there's no L2 cache which causes troubles.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 35 of 105ANN.lu
Posted by Johan Rönnblom on 23-Sep-2003 10:57 GMT
On 060 and CVision, I find MUI to be fast enough, and I think I have
rather high demands in this department. :-)

In MUI3.8, there are some quirks though that cause bad responsiveness
in some cases, this is a bit annoying (never stopped me from using MUI
apps, but these things should be solid imo). In the version that comes
with MOS, these issues are gone, in my experience. I suspect the
version for OS4 has these problems solved as well, as it seems to be
mostly the same version.

As for Reaction, my problem with it is not that it's slow, but that
it's quirky and well.. doesn't represent what a good UI should be,
IMHO. But this is a matter of taste. I like MUI, especially with the
aforementioned quirks being fixed (and on hardware that is certainly
fast enough to handle it). If others like Reaction, that's no problem
for me. :-)
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 36 of 105ANN.lu
Posted by catohagen on 23-Sep-2003 11:02 GMT
wonder how this place would look like if everyone started their own thread about
how they feel about this or that operating system....
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 37 of 105ANN.lu
Posted by Fabio Alemagna on 23-Sep-2003 11:05 GMT
In reply to Comment 36 (catohagen):
> wonder how this place would look like if everyone started their own thread about
> how they feel about this or that operating system....

Surely it would be a better place than if everyone started making random complaints about this or that people's opinion. More constructive, at least.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 38 of 105ANN.lu
Posted by Anonymous on 23-Sep-2003 11:08 GMT
In reply to Comment 37 (Fabio Alemagna):
If you learned to differentiate between opinions and facts in how you present your arguments people might feel like being more constructive towards you Fabio.

You have a long history of being irrational and ill tempered. This will dog you for the rest of your Amiga community contribution.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 39 of 105ANN.lu
Posted by catohagen on 23-Sep-2003 11:21 GMT
In reply to Comment 37 (Fabio Alemagna):
>Surely it would be a better place than if everyone started making random complaints about this or that people's opinion. More constructive, at least.

no, i doubt filling/flooding this site with random 'thoughts' by random normal
people about operating system would make this a better place, there already was
several threads about the OS4 demostration, and you could easy share your views there...
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 40 of 105ANN.lu
Posted by samface on 23-Sep-2003 11:28 GMT
In reply to Comment 24 (brotheris):
I agree, freezing mouse pointer is not very user-friendly. However, I've never experienced mouse freezing on my AmigaOS3.9 setup when Reaction loads or refresh and that is not a lie. It's obvious that we simply have different experiences with Reaction and I'd say it would be better if we simply agreed to disagree rather than accusing each other of lying.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 41 of 105ANN.lu
Posted by Eva on 23-Sep-2003 11:28 GMT
In reply to Comment 25 (Hans-Joerg Frieden):
Hans, 2 years passed since Hyperion declred "2 months to see Aos4..." and it is neighter a usable product under LEGACY MACHINE.
And if we want to call about "experts" you have so many in your "camp"

>> "The first besta of AmigaOs4 will be releasen in 2 months" (Ben Hermann, November 2001)

>> Posted by Mike Bouma (Trusted user) on 22-Sep-2003 06:14:24
You can judge the "SLOW" fractals for yourself here:
http://www.lightelements.com/gallery/03-os4/jt-os4e.avi
They are not slow ...

If this are not FUD or wise experts words ...
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 42 of 105ANN.lu
Posted by krize on 23-Sep-2003 11:35 GMT
I have used MUI on 030/50,04/25 and now 060/50 and I have never understood the people complaining about it, I think its one of the best GUIs out there (ever!!!)

Samface: you dont use yam,ibrowse,amirc,amftp,voayger +++ ???? what do you use ??
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 43 of 105ANN.lu
Posted by brotheris on 23-Sep-2003 11:35 GMT
In reply to Comment 40 (samface):
Well, it works same way on all Amigas. Agreeing to disagree.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 44 of 105ANN.lu
Posted by Kjetil on 23-Sep-2003 11:39 GMT
In reply to Comment 41 (Eva):
http://www.lightelements.com/gallery/gallery/03-os4/mandelbrot.avi

The web pages are updated!!!
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 45 of 105ANN.lu
Posted by catohagen on 23-Sep-2003 11:54 GMT
In reply to Comment 41 (Eva):
>The first besta of AmigaOs4 will be releasen in 2 months" (Ben Hermann, November 2001)

i don't remember any quotes or dates promised so i'm not arguing with you there,
but OS4 handles betatesting in a closed list with NDA's so the public doesn't know
whats starts where or when, no beta's of OS4 where ever promisted delivered to the
public.
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 46 of 105ANN.lu
Posted by catohagen on 23-Sep-2003 11:56 GMT
In reply to Comment 45 (catohagen):
>The first besta of AmigaOs4

assuming besta means beta
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 47 of 105ANN.lu
Posted by Eva on 23-Sep-2003 12:02 GMT
In reply to Comment 42 (krize):
Lol he uses IE, FLashFxp, Mirc, Outlook :D
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 48 of 105ANN.lu
Posted by Anonymous on 23-Sep-2003 12:26 GMT
1.
MUI is not a standard. It does not follow the AmigaOS Classes system (official way of extending the Amiga's GUI), and will not benefit from any future OS enhancements. Case in point: OS 3.5.
2.
It's SLOW. Gadtools, in comparison, is about 2-3 times faster in screen redraws alone. There's little point in a feature-laden GUI if it's a pain to use.
3.
It installs a whole new set of library (and more) files on your hard drive--do Windows .DLL files come to mind? ...same philosophy, similar results. Different types of classes for the same purpose (text editors, lamp/led image classes) give you more choice, but then begin to cause problems when they're not compatible with one MUI app, but are required by another MUI app to run. The result of NOT using a lot of third party tweeks & add-ons is a clean and stable system. Unfortunately, AmigaOS 3.5 falls victim to the 'system GUI stored on hard drive' problem; the new 'getfont.gadget' caused problems for old applications when communication failed between the AmigaOS developers and the author of New York-II. On a much larger scale, Windows 2000 is a perfect example of this problem; it stores about 200-300MB of hidden system files that it copies back over illegal third-party installation attempts to overwrite it's system .DLL files. Microsoft released a patch for Internet Explorer 5, which failed to work because Windows 2000 saw the patch as a third-party attempt to overwrite it's files...makes the Amiga communication problem seem small, but this method is susceptible to all manner of conflicts all too often.
4.
MUI uses it's own system of screen modes, which means needless hassle, more wasted time.
5.
Consistency in a GUI is important; any professional software designer will tell you this. MUI doesn't help the Amiga's GUI standards by creating it's own little world. "But!" I hear you say, "it makes things so much easier to program." Maybe, but a well designed Gadtools program is worth the effort. Gadtools responds/redraws faster, doesn't require extra support files, and still looks good. This is what helps make the Amiga unique among personal computers--tight code, minimal files.

http://www.gregdonner.org/soapbox_amiga/soapbox_amiga.html
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 49 of 105ANN.lu
Posted by Kjetil on 23-Sep-2003 12:28 GMT
Well the speed depends on many things, if the conversion form RGB big ended to RGB little ended is written inn C and not assembler, if bitter burst is handled by pixel write and not handled by gfx card, if there is some 68k assembler part need to be emulated, and the emulation is heavily debugged,

One of this scenarios will slow down the system, the most used function call are often the source to slow speed, not there are lot of things in the system that eats CPU like icon.library that is 100% 68k emulation,

There is easy way to test low level functions, and that is to record number of pixel writes and reads with in a type frame of 5min, dividing number of pixel by 5 min will give you a average pixel write speed, if that checks out, then there most be problem with the programs using that function classact or some thing else.

The same procedure should be repeated on every function commonly used, Hypersion should add counters inn the function to measure the number of time function calls are used (only for debugging), the most calls should be the most optimized functions!!!

It might be a bit early to sagest any speed optimization until the product is full PPC / C / ASM
AOS4 on AOne slow? Perhaps, perhaps not. : Comment 50 of 105ANN.lu
Posted by Crumb // AAT on 23-Sep-2003 12:29 GMT
In reply to Comment 17 (brotheris):
sorry, but OS4 ASL USES Reaction.
Anonymous, there are 105 items in your selection [1 - 50] [51 - 100] [101 - 105]
Back to Top