Problematic AGA demos.

[amilo3438]

Originally Posted by ED-209

Have been supplied a very rare trackmo from exolon called "Dreams 3" by TRSI.
It always freeze half way through demo when I assume it is trying to access disk 2.
Have tried many different config including quickstart A1200 basic and itterations.
The demo does not even start without CE enabled.

It is not very rare!

Cubic Dream (Demo) by TRSI -> http://janeway.exotica.org.uk/release.php?id=3406

Note: It says "CUBIC DREAM (DREAM³) *AGA* tRACKMO [1/2]"

Tested on WinUAE 3.2.2 (Quickstart A1200 Basic) to work fine!


Also found ADF versions here ...

Cubic Dream (19xx)(TRSI)(AGA)(Disk 1 of 2) -> problematic! (EDIT: Tested it again (2x) and now it works fine! ... Hmm, what happened at 1st time!?)
Cubic Dream (19xx)(TRSI)(AGA)(Disk 2 of 2)

Dream 3 (19xx)(TRSI)(Disk 1 of 2).zip -> works fine!
Dream 3 (19xx)(TRSI)(Disk 2 of 2)


EDIT2:
It appears that the demo was already mentioned on this forum but a long time ago -> http://eab.abime.net/showthread.php?t=26928
and even in this topic -> http://eab.abime.net/showpost.php?p=937537&postcount=27 (BLITTER: register modification during linedraw!).

Originally Posted by ED-209

Oh no.. that is a shame

Suppose it was just a lack of knowledge of topic! (just kidding)

[vagrant]

well.. that is welcome news! thanks amilo
It was amongst some other rarities and I'd never seen Cubic Dream before.
In fact, apparently I already have the working version on my HD lol!
Excessive festive drinking is a likely fault here... happy new year to you both

[amilo3438]

Originally Posted by ED-209

Excessive festive drinking is a likely fault here... happy new year to you both

so, that was the reason ... happy new year to you 2

[amilo3438]

Massive by skarla -> http://www.pouet.net/prod.php?which=56885

Used config. = QS A4000 040, 3.1, 2MB+8MBFast + 64MBZ3

Tested on WinUAE versions:

winuae_2710b2 -> http://eab.abime.net/showpost.php?p=932881&postcount=20
winuae_2710b3 -> http://eab.abime.net/showpost.php?p=933236&postcount=31
winuae_2710b4 -> http://eab.abime.net/showpost.php?p=933561&postcount=33

The demo seems still fine on WinUAE2710b2 (pic.1)!

Small problem appeared from WinUAE2710b3 (pic.2)!

And the problem started from WinUAE2710b4 (pic.3)! (and its still same in current WinUAE version!!!)


Note: The problem happens also if JIT=OFF!!! (so, regardless of the JIT)

[Toni Wilen]

Originally Posted by amilo3438

Massive by skarla

Fixed. Finally AGA problem that had absolutely nothing to do with timing or CPU

[amilo3438]

That was pretty fast! (Nice)

p.s. Just tested with latest winuae.zip (21:13) and it works fine now! (thx)

[vagrant]

Nebula by Dual Crew Shining
http://janeway.exotica.org.uk/release.php?id=11797

Tested with A4000 / 030 / 8mb fast quickstart (also tested under 020/030 cycle exact)

Main problem is that after the intro, the music stops/skips during every effect transition.

Also it does not like JIT or fastest possible, because lots of 'juddering' in some scenes, (but I guess this is just some pointless timing related issue)

[Toni Wilen]

Originally Posted by ED-209

Nebula by Dual Crew Shining
http://janeway.exotica.org.uk/release.php?id=11797

Tested with A4000 / 030 / 8mb fast quickstart (also tested under 020/030 cycle exact)

Main problem is that after the intro, the music stops/skips during every effect transition.

Does it also happen if you run the demo from ram disk? (I can't duplicate it)

Also it does not like JIT or fastest possible, because lots of 'juddering' in some scenes, (but I guess this is just some pointless timing related issue)

I guess it is simple caused by inaccurate calculations (demo most likely uses fixed point calculation, FPU is not used) which is more visible when movement is smooth enough (=fast CPU).

[vagrant]

Originally Posted by Toni Wilen

Does it also happen if you run the demo from ram disk? (I can't duplicate it)

Yep, same problem.
The first time it happens is transition after the 3D rose. A more accurate description would be that some samples are not triggered at the beginning of the pattern. Cycle exact / slower cpu actually make the problem worse.

There are 3 captures on youtube. 2 of those are winuae captures and it happens on both. The other capture [ Show youtube player ] is from real amiga and there is no problem.

[amilo3438]

New testing of CPU timing problematic demos after the post #97 --> http://eab.abime.net/showpost.php?p=...2&postcount=97

Now tested in WinUAE 3.4.0 !


Some observations regarding the demo Olympe (Trackmo) by Gods -> http://janeway.exotica.org.uk/release.php?id=10612

I tested it again in the quickstart A1200 (Basic non-expanded config) by playing with the CPU frequency and the CPU cache:

At CPU Frequency 4x (A1200) --> It stops after loading the track 77 !
But if now enter into WinUAE debugger and disable the cache with "r CACR 0" and exit with "x", the demo will continue to work !

At CPU Frequency 2x (A500) --> It does not stop after loading the track 77 !
But if enter into WinUAE debugger and disable the cache before reaching the track 77 with "r CACR 0" and exit with "x", the demo stops at track 77 !


Summary:

At CPU Frequency 4x (A1200) the demo stops at the track 77 when the CPU cache is enabled and pass the track 77 when the CPU cache is disabled !
At CPU Frequency 2x (A500) the demo pass the track 77 when the CPU cache is enabled and stops at track 77 when the CPU cache is disabled !



More CPU timing problematic demos that seems to work fine if the CPU cache is disabled before reaching the problematic part:

Gods-Olympe --> it does not stops at track 77 if CPU cache is disabled ! (see for more findings above)
CHRYSEIS&POSEIDON-botm --> it does not glitch at PHOTO$ part if CPU cache is disabled !
DAMAGE-Uquala --> the "Kolorki" part seems fine if CPU cache is disabled !
US-YulBrynnerwSkinh --> not quite better but it less glitches on the "pink" rotation part ! (Note: It seems fixed from the winuae3410 Beta0 2017-01-10 test version)
SHD-Revelation --> it doesnt glitch on plasma part if CPU cache disabled !
X_DREAM-Borntro --> not quite better but it less glitches if CPU cache disabled !


And in these CPU problematic demos disabling the CPU cache did not helped:

C_LOUS-Cosmicglide --> it still glitches on ring rotation part !
INFECT-Harmagedon --> after a while only black screen with music !
Insane-Despair --> stops after a while at text part !
ZOMO-Speed --> it still glitches !


Maybe it might be helpfull one day !


EDIT:

Problematic demos that work even if the CPU cache is enabled but with specific Custom Frequency:
(Reminder from the post #92)

------------------------------
CPU Freq = Custom 12.737829
------------------------------
Gods-Olympe --> the highest freq for demo to work! (otherwise it stops after loading at track 77)
CHRYSEIS&POSEIDON-botm --> the highest freq for demo to not glitch! (otherwise it glitches on PHOTO$ part)
DAMAGE-Uquala --> the highest freq for demo to not glitch! (otherwise it glitches on KOLORKI plasma part)
US-YulBrynnerwSkinh --> the highest freq for demo to not glitch! (otherwise it glitches on "pink" rotation part)
(Note: It seems fixed from the winuae3410 Beta0 2017-01-10 test version)

------------------------------
CPU Freq = Custom 12.171703
------------------------------
C_LOUS-Cosmicglide --> the highest freq for demo to not glitch! (otherwise it glitches on on ring rotation part)

[amilo3438]

Just for info...

SysInfo v3.24 speed comparisons between a real A1200 and WinUAE3400 QS A1200 with used "special" Custom Frequency ! (see attached picture below)

With Custom Frequency = 12.737829 MHz the SysInfo values is some 3% lower than on a real A1200 !?
With Custom Frequency = 12.737830 MHz the SysInfo values is some 2% higher than on a real A1200 !?

Notice the difference of Custom Frequency = 0.000001 MHz !

Seems impossible to adjust the WinUAE Custom Frequency to show the values close as on a real A1200 !?

Notice that the Custom Frequency = 12.737830 gives the same values in SysInfo v3.24 as CPU Frequency = 4x (A1200).
( see in attachment on http://eab.abime.net/showpost.php?p=...6&postcount=27 )

So the "best" values for WinUAE QS A1200 seems placed somewhere in between two Custom Frequencies that is currently not possible to adjust !

Wonder is it possible to adjust the current CPU Frequency of 4x (A1200) to be closer (or equal) to a real A1200 values ? (as current values is some 2% higher and that makes a problem to some CPU problematic demos)


EDIT:

Seems its possible to have the WinUAE A1200 config SysInfo v3.24 values more closer to a real A1200 if the CE (Full) = OFF and CE (DMA / Mem) = ON !!! (see on the 2nd pic)

With CPU Frequency = 4x (A1200) + CE (Full) = OFF and CE (DMA / Mem) = ON the SysInfo values is now only 1.4% lower than on a real A1200 !?

This fixed the most CPU timing problematic demos (6/10) mentioned in the previous post #110, except in these 4 still problematic:

SHD-Revelation --> it still glitches on plasma part !
X_DREAM-Borntro --> it still glitches !
Insane-Despair --> it still stops after a while at text part ! (but the glitch looks now much better!)
ZOMO-Speed --> it still glitches !


EDIT 2:

I manage to improve the previous config even more by playing with the cpu_throttle setting in the config file ! (see on pic.3)
I found that SysInfo v3.24 shows the correct CPU Frequency of a real A1200 @15.20 MHz when the cpu_throttle range from 283.3 to 307.7 (mid=295.5).
I choose to take the middle value of cpu_throttle = 295.5 and on new testing found that 2 more problematic demos work fine now:

SHD-Revelation --> it does not glitches on plasma part !
Insane-Despair --> it does not stop after a while at text part !

So now with this new config 8/10 most CPU timing problematic demos work fine, except 2 still problematic:

X_DREAM-Borntro --> it still glitches !
ZOMO-Speed --> it still glitches !

With CPU Frequency = 4x (A1200) + CE (Full) = OFF and CE (DMA / Mem) = ON and cpu_throttle = 295.5 the SysInfo values is now only 0.94% lower than on a real A1200 !?

And I think (IMHO) that this is currently the most compatible A1200 configuration possible on WinUAE ! (Note:tested on latest WinUAE3400)
(Note: Such config may introduce some new glitches here and there (because it's not Full CE), but most time will work pretty good!)


EDIT 3:
If change the cpu_throttle from 295.5 to 298.0 one more CPU problematic demo will start to work fine:

ZOMO-Speed --> does not glitches anymore !

So we have 9/10 CPU timing problematic demos that now work fine and only 1 still problematic:

X_DREAM-Borntro --> it still glitches !

Ps.
If cpu_throttle = 498.0 the problem in X_DREAM-Borntro will be fixed, except then might some other demos start to glitch like f.e. SHD-Revelation, but it seems its the only one that will glitch. So it is still not possible to have all 10 mentioned problematic demos fixed at once.
But the SysInfo speed values looks now almost perfect to those from a real A1200. (see on pic. 4)

[amilo3438]

Summary of post #111

CPU timing problematic demos tested in WinUAE 3400 A1200 (non-expanded) configs:

Case 1: CPU Frequency = 4x (A1200) (standard QS A1200 setting)
Case 2: CPU Frequency = 4x (A1200) + CE (Full) = OFF, CE (DMA / Mem) = ON, cpu_throttle = 298.0
Case 3: CPU Frequency = 4x (A1200) + CE (Full) = OFF, CE (DMA / Mem) = ON, cpu_throttle = 498.0

Note: The cpu_throttle does no work when CE(Full) is enabled ! (only when its disabled)


Problematic demo fixed ................ Case 1 .......... Case 2 .......... Case 3
----------------------------------------------------------------------------
CHRYSEIS&POSEIDON-botm ............ NO ............... YES .............. YES
C_LOUS-Cosmicglide ..................... NO ............... YES .............. YES
DAMAGE-Uquala ........................... NO ............... YES .............. YES
Gods-Olympe ............................... NO ............... YES ............... NO ........ (problematic if cpu_throttle > 331.5)
INFECT-Harmagedon ..................... NO ............... YES .............. YES
Insane-Despair ............................. NO ............... YES .............. YES
SHD-Revelation ............................ NO ............... YES ............... NO ........ (problematic if cpu_throttle > 419.2)
US-YulBrynnerwSkinhIt .................. NO* ............. YES+ ............ YES+
ZOMO-Speed ............................... NO ............... YES .............. YES ........ (problematic if cpu_throttle < 297.9)
X_DREAM-Borntro ......................... NO ................ NO .............. YES! ........ (problematic if cpu_throttle < 497.1)
----------------------------------------------------------------------------
Fixed ......................................... 0/10 .............. 9/10 ............. 8/10


* Possibly fixed in WinUAE 3.4.1 version !
+ At the end of the demo it still glitches !
! With cpu_throttle = 497.0 the Sysinfo values is below and with cpu_throttle = 497.1 is above then on a real A1200 !


Note: Sysinfo values with Case 2 is slightly lower and with the Case 3 is slightly higher than on a real A1200 !
(the Case 1 gives much higher Sysinfo values than on a real A1200)

Note: Currently the Sysinfo shows the CPU Frequency same as on a real A1200 (i.e. 15.20 MHz) if cpu_throttle is >=283.3 and <=307.7 !

Conclusion: The best setting is the Case 2 as the most of cpu timing problematic demos work fine with it ! (9/10) (See EDIT below!)
(The Case 2 seems to valid if cpu_throttle range from 297.9 to 331.5 !)


EDIT: +Other examples (just for info):
----------------------------------------------------------------------------
PB-SWING .................................. YES ............... NO ............... YES! ...... (problematic if cpu_throttle < 497.1)
Freezers-Lech ............................. YES ............... NO ............... YES! ...... (problematic if cpu_throttle < 497.1)
STONE_ARTS-Vidar ....................... NO? .............. NO ................ NO! ........(problematic if cpu_throttle < 992.3) ?
Oops-Anniversary ......................... NO ............... YES ............... YES
MirageUK-SubliminalXtc .................. NO ............... YES ............... YES
QUARTEX-40kPartyWatEenGoreKop .. NO ................YES ............... YES
C_LOUS-Mustard .......................... NO? ............... NO ................ NO
Void-Dusted ................................ YES* ............. YES ............... YES
----------------------------------------------------------------------------
Fixed ......................................... 3/18 ............ 13/18 ............. 14/18


? Does STONE_ARTS-Vidar and C_LOUS-Mustard work on a real A1200 non-expanded ? (because on Case 1 does not glitch only with +4MB Fast)
* Void-Dusted needs WB to start! (Note: Its the first demo that requires CE (Full) to work fine, otherwise crashes!)


Conclusion: Now the best setting is Case 3 but with cpu_throttle = 497.1 as is the most closest to a real A1200 Sysinfo values ! (and the most compatible)
(still above a real A1200 but not too much as it happens with the Case 1)

I guess with all 3 configs most possible issues in current A1200 emulation is covered ! (except in a few cases mentioned)
Note: The A1200 (Case 1) config is the same as quickstart A1200 (non-expanded) ! (so not attached)

[Seiya]

demos require exactly cpu frequencies to run properly.
If a demo require 14 Mhz 68020 or 68EC020, if you cannot emulate exactly 14 Mhz timing the demos always will have strange and problematic issue.
Most demos are sincronised to music and audio with cpu cock.
WinUAE emulate very well basic A500, but from A1200 there always some difference: too fasto or too slow. I never found a config that be like real A1200 at 14 Mhz.
i have create a config to emulate real A3000 with 68030 at 25 Mhz. In same case it is very similar, in other more slower or slighter faster..
Demos are more sensitive to these values than games

[amilo3438]

Originally Posted by Seiya

I never found a config that be like real A1200 at 14 Mhz.

Check configs attached in post #112; A1200 (Case 2) and A1200 (Case 3) !
(IMHO both configs are very close to speed of a real A1200 and pretty compatible.)

[Seiya]

however for demos i use winuae v2.3.1. for now i have no found many problems.
from tonight i want to put in streaming these demos to watch all these issue.

[amilo3438]

Note that mentioned configs are tested only in latest official WinUAE 3.4.0 ! (so should be used only with this version)

[Seiya]

i tried 3.4.0, but sysinfo crash emulation.
I tried Aibb and i see a good emulation: slighter faster than real 1200. and 18 Mhz emualted cpu.
but i don't know if it recognise it properly.

i hope in the next winuae releae a slider to choose frequencies like 8, 10, 12, 16, 20, 25, 28, 33, 40, 42, 50, 66 Mhz with right cpu timings to a great new emulation experience

[Toni Wilen]

This is totally pointless.

[amilo3438]

Originally Posted by Toni Wilen

This is totally pointless.

Anyway, it helped to find Cases 2 and 3 that cover (i.e. fix) most (if not all) before cpu timing problematic demos ! (so how it can be pointless)

Just check how many cpu problematic demos run on current quickstart A1200 config ? -> none !

Obviously the current quickstart A1200 (non-expanded) runs too fast in comparison to a real A1200!
I wonder if it could be slow down to Case 3 values in order to increase the compatibility ?

p.s. I am not going to use the quickstart A1200 (non-expanded) until it becomes better ! (as Cases 2 or 3 works already fine for me)

[Toni Wilen]

It immediately gets totally pointless when trying to match some reported cpu speeds and trying to match those. It means nothing. It is impossible to get correct speeds. It may be possible to get some demo working by tweaking CPU speeds but it does not improve anything.

Nothing has changed since last time I replied with similar reply