Amiga emulators: is it worth it?

[amigaksi]

>by DiskDoctor on 2009/1/18 9:57:56


>Hello,

...
>Also, some time ago I encountered some post on some other forum stating that "no one has ever managed to re-create the original chipset as a Virtual Machine because it was SO PERFECT it is hardly possible if ever."

If you go by PC standard hardware, it is impossible to do the cycle-exact emulation of the Amiga.  Now if you have some specialized PC hardware like a multi-channel audio card, sprite-based video card, digital joystick interface, PC w/HPET timers, etc. that are all superset of the hardware of your Amiga in every respect and have software that uses these directly (not through an API or buffered scheme), then it is possible.  

>My problem is EVERY time I managed to set up the proper configurations, e.g. games (it's all about games) were playable but the music was getting stalled every time on rapid pseudo processor's usage.

>Classic Amiga music is crucial to me, something one might not accept unfidel.

From what I have read, emulators rely on buffering and do not do cycle-exact one-to-one mapping of the hardware; i.e, if I change the volume register of the Amiga dynamically at some set frequency based on Copper or IRQ, you can bet there will be latency or distortions in the emulation (without even trying it).

>Or maybe it is my fault because of not optimizing configuration set-up properly?

For some things, it does not matter how much you play with the configuration, it just won't work.  For example, if you were relying heavily on joystick port for input or output via CIA chip ($BFE001/$BFD000) or $DFF00A/$DFF00C/$DFF036 to cause "fill-ins" for your music manually or controlled by another device, you can bet that using the PC Gameport or even USB version of joystick will give you inferior performance.

[amigaksi]

>by persia on 2009/1/18 10:38:43

>The various versions of UAE are almost technically perfect. That's what some complain about, 80's computer was clunky and a bit unstable, running an Amiga has always been a bit like building your own ultra lite aircraft, you were exposed to the environment and never knew when you would experience a crash...

Believe it or not, Amiga computers are stable-- it's the software on any machine that makes the computer look unstable.  Early Windows were also crashing all the time; but now with memory protection, the crash is localized to the application running.  I have plenty of software on Amiga that never crashes.

>Real Amigas are always on the edge of disaster, with cobbled together parts that were never meant to work together. They click, they pop, they wheeze. Experiencing an Amiga through the clean safety of a modern OS is somehow almost cheating.

If the emulator is cycle-exact, it should crash the emulated software just like the real amiga.

[amigaksi]

>by AmiKit on 2009/1/19 5:13:35

>Quote:
--------------------------------------------------------------------------------

>>The problem with emulating a computer is that there will always be latency. Usually this is most visible in sound.

--------------------------------------------------------------------------------

>I don't think so. Even if there was any latency I doubt any human can notice it.

There IS latency since they purposely introduced it in their buffering scheme so they get more time to translate from one machine to another.  You will notice it -- it depends on the application and which hardware resources it uses.

Quote:
--------------------------------------------------------------------------------

>>You'll never get the complete Amiga look and feel via emulation, no matter how good UAE is.

--------------------------------------------------------------------------------

>Well, a simple Turing test denies your statement. http://www.amigaforever.com/tutorials/goodorbad/

That article is subjective (understandably) and flawed.  You can also put a person who mostly does BASIC language programming in front of an Atari ST set up like an Amiga but that does not make an Atari ST an Amiga.  There's no way to try all the possible applications that can be written for a machine.  CPU power increasing over many years does not make all the other hardware automatically better.  And if the emulation was 100% accurate on PC, there's no reason to use the emulation-- just get the PC version of the software which avoids the translation layer completely and get a more efficient result.  It's easy enough to write a program that works on all OCS amigas but not on an emulator given the standard PC hardware.

[amigaksi]

by AmigaHeretic on 2009/1/19 11:44:46

...
>I use WinUAE and it's emulation is great. The only thing that I never could get to work before was one of the NewTek demos where it shows "4096 color Hi-Res" (on OCS) and it has the picture of the guys face closed up, well, the latest version of WinUAE show it perfectly now. So I guess it's WinUAE IS perfect now.

I guess the earth is flat now-- I just looked out the window and saw in every direction I could.

[amigaksi]

by AmiKit on 2009/1/20 2:16:08

Quote:
--------------------------------------------------------------------------------

That article is subjective...

--------------------------------------------------------------------------------

>lol, a human being IS subjective! And the Turing test, in this case, is about human perceiving. You did not get it.

No, the URL you gave is stating much more than just someone being subjective.  

[amigaksi]

by AmiKit on 2009/1/20 3:14:58

Quote:
--------------------------------------------------------------------------------

If you go by PC standard hardware, it is impossible to do the cycle-exact emulation of the Amiga.

--------------------------------------------------------------------------------

>Hmmm, strange because I saw Cycle-Exact option in WinUAE/Chipset settings. In addition there's "100% accurate" sound emulation option in sound settings.

I can also put many options in software that work partially or not at all.  I don't follow blindly.  You are not even giving experimental results what to speak of deductive logic of whether it's do-able on a particular hardware.

[amigaksi]

>Experimental results? Lol, you still don't get it, do you? It is the subjective human perceiving that counts here. Not the small difference in exact numbers, voltage or whatever.

Okay, if you can live with the inaccuracies of those softwares you are using the emulator for, great.  But it does not take away from the FACTs that some software won't work given those inaccuracies and other humans can perceive differences where you may not.  Some people I know think the lossy MP3 is same as 16-bit lossless linear audio CDs.  They may live with it, but objectively studied-- you can compare the original digitized wave file to the audio CD and MP3 file and show the distortions caused by MP3.

>Look, I understand that from objective point of view the emulated Amiga on PC hardware is not Amiga. But my favourite games behave just like they behaved on my real Amiga. At least my impression is the same. So from MY point of view it IS Amiga. I really don't care about the PC HW that is under the desk, I don't care about cycle-exact issue or whatever. I don't need to do experimental HW research in order to increase my subjective impression of the game, etc. That's it.

You are studying a small sample only.  I can write software where even YOU will notice a difference.

[amigaksi]

>by AmiKit on 2009/1/20 5:29:32

>@spirantho

>Are we still talking about A500 games?
>Even WHDLoad on a real machine (A1200 and better) can be considered as emulation when running A500 games - in strict point of view.

But there's a big difference.  AGA is a superset of the OCS hardware (and backward compatible).  You can write software for OCS that works fine on AGA w/o modifications.  It's just that some software used the CPU for timing or called OS routines at absolute locations or something like that which you need adjust to make them compatible.

>If those things are imperceptible, what makes the difference in your feelings? Just the knowledge that it is the emulation? If so, that is exactly what Turing test eliminated.

>--Jan

You can tell certain differences if you are really addicted to a game and know its nuances rather than if you just played it once or twice or just saw it.  Audio and Video is not the only thing of our experience, subtle timings also is in our experience which you seem to be throwing under the "subjective" catagory.  And even for a buffered 60Hz display, if you don't properly sync up the frames to the same VBlanking used by Amiga, you can notice the 1/60 difference.  And for audio, it's much higher frequency that one can notice.

[amigaksi]

>by AmiKit on 2009/1/21 5:56:17

>>@amigaksi
>>...humans can perceive differences where you may not.

>Exactly... wait a sec, does it mean I am not a human?  Just kidding.

There was the word "other" before humans.  People addicted to a game (or application) get used to the audio/video/timing so what they did not perceive initially they perceive later.  And with MP3, it's similar-- if you show them the distortions by selecting the appropriate parts of the wave and original wave and playing them one after another, they can notice the difference.  So it's not a constant.

>>Some people I know think the lossy MP3 is same as 16-bit lossless linear audio CDs.

>You mix the knowledge and the experience. If they cannot differentiate between lossy MP3 and WAV, in their world lossy MP3 is OK. No need for "objective" studies.

Yes, there is.  When people create a product they are targetting the general populace with various interests/levels of perception.  You have to take an objective approach although different people may have different levels of requirements.

>>You are studying a small sample only.

>Yep, but it doesn't mean it's bad. Seems we're getting to the core finally. While you're talking about nomothetic approach I am talking about ipsative or individual approach. Now the question is what approach is more suitable when talking about emulated games. I think that with regards to the impression of the game the INDIVIDUAL EXPERIENCE is more important than subtle differences in video, sound or timing (if there are any, of course).

There is a difference so let's not put that into doubt as we only need to show one example to prove they are different.  And we already know the latency is purposely there.  What you need to do is change the "the" in front of "the INDIVIDUAL EXPERIENCE" to "MY".  "The" would imply an objective approach.

>If there really is any 1/60 difference then the fact whether the human can notice or not depends on the intensity of the original value. This is called Weber-Fechner law. Needles to say that such a 1/60 difference, even though noticed, won't affect your game experience at all. It objectively affects your measurement device only.

Sorry, it does not depend on any "intensity".  It's not so vague.  Just move a screen large object a few pixels every 1/60 of a second.  So if you are in 320*240 mode, say you can move object 320 pixels in 2 seconds.  Well, if your emulator does not know when that position changes, 1 frame will show part of the object moving and another frame will show the rest of the object.

>>I can write software where even YOU will notice a difference.

>That would be interesting.
 
Not "would be".  It is already existing.

[amigaksi]

>by shoggoth on 2009/1/20 11:33:18

>>amigaksi wrote:
>>If you go by PC standard hardware, it is impossible to >>do the cycle-exact emulation of the Amiga.

>I'd like to see you explain that statement. Emulation granularity is *not* dictated by hardware capabilities such as timers, sprites, multi-channel audio chips. Generally that's not how you solve these things, at least not in any of the emulators I've worked on.

It's good to have conversations and learn things but when you start insulting and poking fun at the other party, that "learning" experience goes away and it's more to do with biased emotional sentiments.

Accurate emulation depends on hardware capabilities.  I'll explain it further if I feel like it.

[amigaksi]

>by shoggoth on 2009/1/22 8:11:28

>No, you're making false statements, and you don't appreciate when people correct them.

That's a false statement.  Claiming that cycle-accurate should mean 1/7.16Mhz timing accuracy on OCS Amiga is false right?  Perhaps, your term is misleading others-- care to think of it that way?

>If I do have a bias, it favours amiga and other machines of that era. I appreciate the complexity and capabilities of these machines, but I don't let that blind me or prevent me from accepting technology today.

I never said technology today should not be accepted.

>I claim that you have decided that these machines cannot be emulated accurately, based on wrongful assumptions about how emulators work internally. For that reason, I've explained how they work. If you don't want to accept that, you're simply rejecting reality in favour of your wrongful assumptions.

Based on facts about hardware of target platform and source platform (Amiga), I made my statements-- I can verify all my statements.  I have nothing to gain by speculating something imaginary nor are you stating what those "wrongful assumptions" are.

>>Accurate emulation depends on hardware capabilities. I'll explain it further if I feel like it.

>No, it does not.

Yes, it does.  See below.

>You need enough CPU power, that's it. You don't need high resolution timers or any of the other stuff you've stated.

Regardless, of how fast your CPU is, it won't make your beep speaker (1-bit resolution) do 4-channel 16-bit 44Khz audio.  Regardless if your CPU is Pentium IV at 4Ghz, it won't improve your timers.  And you can't time things equally or better with 1.19Mhz timer vs. a 3.57Mhz timer-- just seems to be violating some laws.

>Marat Fayzullin has written a simple yet very informative tutorial about writing computer emulators. You can find it here: http://fms.komkon.org/EMUL8/HOWTO.html

He is claiming any computer can be emulated because he defines "emulate" as an ATTEMPT to imitate the target machine.  That means, you should also accept: Atari ST can emulate Amiga, Atari 800 can emulate Pentium IV dual core, Apple I can emulate the MAC II, etc.  

[amigaksi]

>by AmiKit on 2009/1/24 4:44:21

>@amigaksi

>>There is a difference so let's not put that into doubt as we only need to show one example to prove they are different. And we already know the latency is purposely there.

>What is this discussion about actually? While you are still pointing at the subtle and objective difference between real and emulated game, I am saying it doesn't matter.

I was pointing out that differences exist even in cases where YOU did not perceive them.  I gave you the example of MP3 vs. linear uncompressed audio.  But the small differences is not the entire story.  Major differences also exist depending on target hardware used by emulator and it's spec.

>Let's say I admit that the difference exists. But from the subjective point of view the result is the same, because:
>1) the difference is subtle
>2a) the difference is hardly noticeable by human perceiving
>2b) even though the difference is noticeable, it won't affect your game experience or a whole enjoyment of the game because of what is mentioned in 1).

No, the difference is "subtle" for YOU that's why you don't notice it unless YOU are trained.  As I stated, some people like some musicians can easily notice difference between MP3/uncompressed 16-bit audio.  Objectively, you can measure the difference and that approach has to be taken since YOU are not the only user.  

>Yes, you can train people to identify those subtle differences but why would you do that? Just to be a winner who can say: "I told you there was a difference!". And I would say: "I don't care, let me play."

No, I already win even if I accept you don't notice any differences since it's better to have real Amiga that works 100% for everyone than an emulator which is good for some games for some people.

>"...the operational principle of the brain is holistic, parallel, and analog, with self-organizing tendencies; or, that the whole is different from the sum of its parts. The Gestalt effect refers to the form-forming capability of our senses, particularly with respect to the visual recognition of figures and whole forms instead of just a collection of simple lines and curves."

That's true but it actually supports me.  Emulator is not an Amiga as a whole although some parts (like the static visual output) look the same.

>>I can write software where even YOU will notice a difference.

>>Not "would be". It is already existing.

>Link please?

I can post source code or an ADF to a sample code I wrote.  I don't have a record of everyone else's work.

[amigaksi]

>by shoggoth on 2009/1/23 3:51:30

>>by amigaksi on 2009/1/22 12:25:08
>>That's a false statement. Claiming that cycle-accurate should mean 1/7.16Mhz timing accuracy on OCS Amiga is false right?

>No, cycle accuracy at 1/7.Mhz granularity is perfectly possible without using timers. Get those timers out of your head.

This is where we differ and you also differ from others who also claim you need higher precision timers than just a 60Hz WM_TIMER message.  I won't be getting timers out of my head since my work is mostly timing stuff.  I time stuff through the I/O ports in interfacing between Ataris/Amigas/PCs.  Here your WM_TIMER bullcrap won't cut it nor your updating frames at 60Hz be enough.  But I am trying to give you cases where this I/O is not involved since everyone knows that's not emulated.

>... based on facts about hardware - but not based on how emulators actually work. You've *assumed* that they work in a certain way using timers - which they generally don't (because it would be completely retarded to do it that way).

Again that's your flawed idea.  You can make the emulators cycle-exact if you had an interrupt of 1/7.16Mhz accuracy and did everything cycle at a time instead of frame at a time which is only makes it "visually" appear the same except for the VBI not being in sync with WM_TIMER difference.

>>Regardless, of how fast your CPU is, it won't make your beep speaker (1-bit resolution) do 4-channel 16-bit 44Khz audio.

>Now you're just being silly. Of course you need a soundcard to get proper sound. That wasn't the point - we were discussing emulation accuracy.

Emulation-- to equal or excel in my dictionary.  That means you have to equal or excel in every catagory including audio (DAC output/music/dynamic effects/etc.), video (color depth/sprites/blitter/etc.), timing, reading joystick ports (for games), etc.
 
>>Regardless if your CPU is Pentium IV at 4Ghz, it won't improve your timers. And you can't time things equally or better with 1.19Mhz timer vs. a 3.57Mhz timer-- just seems to be violating some laws.

>Drop timers. You don't use timers to achieve accurate emulation. That's *in* your *head*. You've chosen to compare this aspect of computers because it suits your ideas, obviously.

If your defintion of emulation means "an ATTEMPT" to mimic the target machine, yeah you can drop timers.  Otherwise, you should drop your ideas that you only need CPU speed to emulate any machine.

>LOL! So by this you claim that Marat Fayzullin has no idea what he's talking about?  Do you know who this guy is?

>"... as an ATTEMPT to imitate" - well that's *exactly* what emulation is. And emulation accuracy is defined by compatibility and perceived user experience. Both which can be realized on any PC today. If you ignore the latter, even a ZX81 can accurately emulate a modern quad core x86 class machine, given enough time and memory.

Okay, as long as you stick to this definition.

>I'll try to refrain from further comments now, since there is no point in discussing a topic such as this one with you. You're either incredibly ignorant or trolling, or both.

Your the one who keeps insulting and making straw man arguments like I am claiming "not accepting technology" or "Amigas can never be emulated" or "their divinity is preventing them from being emulated".  Your taking them out of context.

[amigaksi]

>by DiskDoctor on 2009/1/23 14:18:58


>@ shoggoth
>&@ amigaksi
...
>Old Amigas were excellences, though nothing divine. Sure it is theoretically possible to re-create this. But I do not think it is doable on the current PC stuff especially with all WinLinMac stars as middleware... Nowadays, like software (OS), like hardware.
...

I mostly agree but it's hard to assume it's "nothing divine" without some further evidence.

About using stand-alone boot-able Amiga that uses the hardware resources directly-- that would be a better approach.

[amigaksi]

>by Hammer on 2009/1/23 15:30:49

amigaksi wrote:
>>by DiskDoctor on 2009/1/18 9:57:56


>>Hello,

...
>>Also, some time ago I encountered some post on some other forum stating that "no one has ever managed to re-create the original chipset as a Virtual Machine because it was SO PERFECT it is hardly possible if ever."

>>>If you go by PC standard hardware, it is impossible to do the cycle-exact emulation of the Amiga. Now if you have some specialized PC hardware like a multi-channel audio card, sprite-based video card, digital joystick interface, PC w/HPET timers, etc. that are all superset of the hardware of your Amiga in every respect and have software that uses these directly (not through an API or buffered scheme), then it is possible.
(SNIP)

>To have design for Windows Vista logo on PCs, the PC must have HPET timers.

>Refer to http://msdn.microsoft.com/en-us/library/bb931844.aspx

Someone declaring a standard is different from being a standard (in reality).  I can declare the new timing standard for Amigas is 28.6363Mhz not 3.579545Mhz but that does not mean everyone has it.  By standard PC hardware, I mean you go to almost anyone's home and they have that hardware.  The 1.19Mhz timer is there in every PC.

I can offer some upgrade h/w that uses the 28.6363Mhz crystal on the Amiga MB and uses it to increment some register but it's specialized hardware.  Perhaps, I can get Microsoft to declare it an Amiga standard.