Denise bugfixes and improvements

[amilo3438]

Originally Posted by PiCiJi

You think the loaded save game is causing the error? Maybe I didn't store the last changes properly in the savegame.
I'll take a look.

I don't think that's the reason, because during testing I didn't need to reload the recorded location.

The way I tested it was to record on the spot before the next race track is read, so if I don't pass, I can repeat. But during the game there was no need for it, the game simply froze after loading one of the racetracks and pressing fire, as shown by the savestate file.

Since after playing the game several times I was unable to reproduce that problem, I don't know what to say, just mysteriously everything suddenly works without problems.


Other:
There are some games that do not work in Denise because of dongle protection!
----------------------------------------------------------------------------------------
B.A.T.II, Cricket Captain, Leader Board, RoboCop 3, Rugby Coach
----------------------------------------------------------------------------------------
(Note: WinUAE supports dongle protection in Host/IO ports)

[PiCiJi]

Originally Posted by amilo3438

I've also noticed that with the Threaded Renderer enabled, keyboard input on Archer Maclean's Pool works better for some reason than when it's disabled. (even with static)

good find.
The threaded renderer performs the rendering in an extra thread. VSync doesn't block the emulation, but it's still fully active.
This difference should not make the keyboard query in Archer Pool run better or worse.
Finally I was able to figure out what the problem is here.

but it seems 20 usec for KCLK (lo -20- hi -40- lo ) is wrong. game protection code input works with 10 usec (lo -10- hi -20- lo ) or something hasn't been understood here yet.
20 usec is ok.

The game handshakes the keyboard each frame, doesn't matter if a key was pressed or not.
When the handshake happens while transfering a key stroke, the current transmission is interrupted and the the whole byte is resend.
fixed in next nightly.

Crazy Cars:
Ok, that's hard to debug. At the time of the loaded savegame, it is clear whether it will crash later or not.
Tonight I'm going to try 2 - 3 times up to the Arizona track on my real Amiga.
Did the problem ever occur when using 512 + 512 ?

Originally Posted by amilo3438

There are some games that do not work in Denise because of dongle protection!

ok maybe it's not that much to emulate.

Edit: Do you use Turbo option for loading Crazy Cars ?

[amilo3438]

Originally Posted by PiCiJi

Crazy Cars:
Ok, that's hard to debug. At the time of the loaded savegame, it is clear whether it will crash later or not.
Tonight I'm going to try 2 - 3 times up to the Arizona track on my real Amiga.
Did the problem ever occur when using 512 + 512 ?

According to my post #199:
Crazy Cars (19xx)(Titus)[compilation Action d'Enfer][2960] ... stops at some point (Note:CrazyCars in save state and A500 KS13 ECS 512+512 config)

So, it happened on the other game version but I somehow managed to delete that savestate file.

I'm not sure if it's necessary to spend more time on that game! But from the same of the savestate file, you can see the reason for the collapse at that time, I suppose.

Later, I tested many times original version of the game exactly on ECS KS13 512+512 and every time without problems and never later tested the compilation version.

EDIT:
But one more thing, in the meantime, since the game was freezing, I had changed some settings in the emulator itself (unfortunately, in that frustration I don't remember what I changed all that), so I don't know if that could be the reason why everything works fine since then!? (And it really works well, I had some completely different problems that I postponed reporting, but now they seem to have disappeared as well.)

I'm also not sure if the current antivirus program could be causing some of the previous problems!? (in the meantime, it was also updated)

[amilo3438]

Originally Posted by PiCiJi

Edit: Do you use Turbo option for loading Crazy Cars ?

That's right, I totally forgot about that. Otherwise, in order to save time while testing games, I had a disk at 8x, but now it is at 1x. I think we have maybe found it, just need to confirm with another test.

EDIT: Confirmed! Tested again on A500 ECS KS13 512+512 but with disk at 8x and it froze on Arizona track!

PS.
Now I'm a bit embarrassed that I didn't think of it myself... but as soon as you mentioned it, I immediately knew it could be that. (I wasn't aware that 8x could have an impact)

[amilo3438]

Cybernoid II - The Revenge (1989)(Hewson)[0003] ... something is happening here! (power LED go crazy)
Damocles - Mercenary II (1990)(Novagen)[0643] ... In ECS during loading time the screen is different than on WinUAE! (otherwise works fine in OCS)

[jbl007]

Sorry to interrupt your discussion about compatibility.
But shaders do not work on my system (linux/mesa/radeonsi): Error message:

GLSL Vertex Shader #0 compilation error: 0:3(16): error: interface blocks with an instance name are not allowed in GLSL 1.40 (GLSL 1.50 or GLSL ES 3.00 required)

Shaders did actually work fine up to commit 9cf2aab. After that it fails. Seems be something wrong with d854b70. I hope this will help to find the problem.

[PiCiJi]

Originally Posted by amilo3438

(unfortunately, in that frustration I don't remember what I changed all that), so I don't know if that could be the reason why everything works fine since then!?

unlikely if it is no setting in the "system management" view.
I couldn't recreate this with ipf 2959. (2960 untested)
I tried it three times in emulation and two times with real machine.
forgot that Disk Turbo is remembered in savestate. and your crash states weren't in turbo mode.

Originally Posted by amilo3438

Cybernoid II - The Revenge (1989)(Hewson)[0003] ... something is happening here! (power LED go crazy)

same on real machine and disturbing sound too.

Originally Posted by jbl007

shaders did actually work fine up to commit 9cf2aab. After that it fails. Seems be something wrong with d854b70. I hope this will help to find the problem.

The shaders from the last Denise versions are rewritten as slang shaders to be compatible with the RA shader system.
The slang shaders are then translated into openGL shaders by spirvcross. For reasons of effort, I don't want to maintain 2 shader systems.
Do you mean the internal or external shaders ?

[amilo3438]

About shaders... post #206 is not my post!

Other:
I managed today (post #204) to freeze the game with disk turbo at 8x immediately!
Yesterday and this morning after several tests it worked fine all the time with the disk at 1x. I don't know what else to say.
Maybe I should test again with 8x to see if it happens randomly... tomorrow.


EDIT:
I don't know if I will be able to do that tomorrow, so I checked today...

1st time ... start the emulator, set the disk to 8x, wait for the hand to appear, load the game via Amiga/Load Software = no problem!

2nd time ... starting the emulator, setting the disk to 8x, waiting for the hand to appear, drag&drop the game to the screen = no problem!

3rd time ... Amiga/Hard reset + eject disk (from previous 2nd time), drag&drop the game to the screen = no problem!

I think it might be possible to make it crash if you test several different games before that, and after each one do a hard reset + eject disk and then drag & drop!
Because that's how it was today when the game froze... I didn't turn off the emulator before that, I just changed it to 8x and continued with hard reset + eject disk and drag & drop.

Also, I don't remember the game ever freezing when the emulator was freshly started (after testing it many times). I don't know what else to say.

[jbl007]

Originally Posted by PiCiJi

I don't want to maintain 2 shader systems.

I totally get that.
Again, it worked fine with the "new_shader" branch merge. One commit later it starts to fail.

Do you mean the internal or external shaders ?

Both fail and work the same way (same error message).

[PiCiJi]

nightly
- fix Archer Maclean's Pool Keyboard detection
- fix GLX version detection

Originally Posted by amilo3438

About shaders... post #206 is not my post!

... fixed.

Originally Posted by jbl007

Again, it worked fine with the "new_shader" branch merge. One commit later it starts to fail.

Right, I thought "spirvcross" doesn't generate GLSL code for older drivers/cards.
good find. I did the version detection incorrectly.
EDIT: maybe clear the shader cache before

Originally Posted by amilo3438

Damocles - Mercenary II

lost sync, black screen, floppy is stepping very fast

[ross]

Damocles - Mercenary II

Originally Posted by PiCiJ

lost sync, black screen, floppy is stepping very fast

Game that only works by chance in OCS (the "copper setup" is totally wrong for several reasons and in several places).

No emulator has anything wrong, it's the game that needs fixing

[jbl007]

Yay, I can use shaders now! Thank you PiCiJi.

[PiCiJi]

Originally Posted by amilo3438

I think it might be possible to make it crash if you test several different games before that, and after each one do a hard reset + eject disk and then drag & drop!
Because that's how it was today when the game froze... I didn't turn off the emulator before that, I just changed it to 8x and continued with hard reset + eject disk and drag & drop.

was RA always enabled ? and have you checked option: "disable RA on power" ?

it's not a problem with the internal emulation state. changing chipmem contents is enough to fix the savestate.
so the real problem "seems" to happen before this savestate.

it sounds like a softreset (doesn't initialize memory) problem, but you are using hardreset.
... inserting disk from beginning or WB screen could make a difference too.
by the way, amiga keyboard reset (pressing the 3 keys) is a soft reset too.

edit:

Originally Posted by amilo3438

I wasn't aware that 8x could have an impact

I don't think so in this case, but in general. It's not an acceleration of the entire emulation, like Warp.
The data is pushed into memory faster than normal. However, it will rarely lead to a problem with ADF.
With IPF, the turbo is temporarily disabled for all tracks that have timing information. The probability increases here that the bitcell width is measured.

[amilo3438]

Originally Posted by PiCiJi

was RA always enabled ? and have you checked option: "disable RA on power" ?

Run Ahead was disable all the time. (I am not using it) and "disable RA on power" was checked. (performance mode and prevent dynamic input was also checked)

Originally Posted by PiCiJi

it sounds like a softreset (doesn't initialize memory) problem, but you are using hardreset.

Hard reset + eject (using all the time) ... I don't know if it's possible for something to be disturbed after using it several times or maybe some program somehow disturbs it!?

Edit: I know that I couldn't make a mistake here by choosing soft reset instead of hard reset + eject disc because after such a reset the WB screen always appears and after the soft reset the disc would stay in the drive and start up.

Originally Posted by PiCiJi

... inserting disk from beginning or WB screen could make a difference too.

I know about that... there are some examples in winuae demo testing thread.

Edit: Maybe it could affect that game, but it's really hard to provoke a freeze even with a WB screen. I'm giving up on further testing that game.

And yes, 8x makes no difference with .ipf because it slows down when needed. (as I noticed during testing)

EDIT:
One question: Can using Pause have any effect!? Probably not as it just stops internal loop, I guess, but it's worth asking: Does the same internal state return after returning from pause!? Probably not!

But can moving mouse at some point, i.e. during loading at WB phase, can have any effect!? (maybe I moved the mouse unconsciously, let's check in one more time)
Edit: Checked and it works fine! (then it must be something with a hard reset... I don't know what else to check)

--------------------------------------------------------------------------------------
EDIT:
Well, I experienced something similar now, testing DesertStrike after a hard reset and eject, and with several tested games without shutting down the emulator. The game simply stopped/frozen after inserting and loading disc 3. Then I did a hard reset without ejecting the disc and it happened that the game loaded disc 3 but very slowly and then froze on track 83. Hmm, I said let's try on WinUAE and everything ok there.
Then I started Denise, waited for the WB screen and inserted disc 1, then disc 2 and disc 3, and the game works without problems. Doesn't this confirm that after a longer operation of the emulator and after several tests of games with a hard reset that something might not be right!?

PS. I had already experienced a similar thing in the days when I was testing demos on WinUAE... the only solution was to exit and restart the emulator. (I don't know is a such behavior of the emulator is expected.)

I suppose then that the only thing that would be correct would be to exit the emulator after testing each game before starting another, compared to the way I do it now, which is to use a hard reset and not exit the emulator.
--------------------------------------------------------------------------------------

[amilo3438]

One question:
I notice while Denise loading the Detroit game (and on some previously tested ones) that the disk drive led color changes from orange to red and back again. Maybe the game wants to record something or is it some kind of protection. In case the game wants to record something, how is it solved with the .ipf file in denise!? (does it create a new file)

[PiCiJi]

Originally Posted by amilo3438

Well, I experienced something similar now, testing DesertStrike after a hard reset and eject, and with several tested games without shutting down the emulator. The game simply stopped/frozen after inserting and loading disc 3.

happens for me too, when insertig disk while warping. disable warp and the disc will be recognized.

Originally Posted by amilo3438

I suppose then that the only thing that would be correct would be to exit the emulator after testing each game before starting another, compared to the way I do it now, which is to use a hard reset and not exit the emulator.

I have tested Crazy Cars at least ten times with hard reset + eject and loading other games befores and so on. I couldn't never crash the game at track 4. (only from savestate )

I suspect you use the emulator somehow differently and thus provoke the error situation, e.g. warp when changing disks.
One option would be to change the operation and see if the problems go away. In this way, the problem may be identified. It's always hard to pinpoint a problem that can't be reliably recreated.
I drag'n'drop the image onto the bottom card. In this case, a hard reset will be performed with the disc inserted (no WB screen). If you operate it like this, will it get better?
EDIT: I don't mean trying crazy cars again, but rather general use. A lot of patience and luck is necessary for this kind of bug.

Originally Posted by amilo3438

I notice while Denise loading the Detroit game (and on some previously tested ones) that the disk drive led color changes from orange to red and back again. Maybe the game wants to record something or is it some kind of protection. In case the game wants to record something, how is it solved with the .ipf file in denise!? (does it create a new file)

At the moment there is no way to permanently save data in IPF.
However, the writing process is completely emulated, but not permanently saved when the emulator is closed. However, savestates can make these changes permanently available, but not the IPF itself.

Edit: Another known problem are hotkeys which shares with amiga keyboard keys. The key press then goes to hotkey and Amiga keyboard. Some games got problems this way.
e.g.: I am using ALT + W for Warp. Pinball doesn't load table (black screen) (unconfimed, need to check this on real one)
e.g.: Fate - Gates of Dawn: pressing any key during title screen, prevent loading the game (confirmed)

[amilo3438]

Originally Posted by PiCiJi

I suspect you use the emulator somehow differently and thus provoke the error situation, e.g. warp when changing disks.

The only other way I do it is maybe to drag&drop on the screen and not on one of the icons that are shown. In addition, I can sometimes stay in the explorer after drag&drop to delete the previous file that I tested.

No, I don't use warp (it's off all the time), but I noticed that there was another option "Input check finish Warp" that was turned off.
Since I turned it on, and the emulator has been running for at least 2 hours without exiting and with several dozen already tested games, I haven't experienced any problem, although I still use my method.
I wonder why that option "Input check finish Warp" is not permanently enabled? (is there any reason)

Btw. Continue using my method and with that option turned on to see if there are any differences compared to before when it was turned off.
I mean, will there be any problem that used to happen with the option turned off. (so far it seems good, or I was lucky)


---------------------------------------------------------
Double Dragon II - The Revenge v4.16 (1989)(Virgin)(ECS)(NTSC)(US)[2635] ... Try to run it on A500 ECS KS13 512+512 PAL and see what happens!
---------------------------------------------------------


EDIT:
Something else:
From Amiga mode go to the C64 tab and do hard reset from there (that's how to enter to C64 mode from the amiga mode). Then on the C64 mode press the combination of hotkeys for hardreset or softreset that is defined in amiga configuration hotkeys (nothing is defined in C64 configurations hotkeys) and it switches back to the amiga mode from the C64 mode!? Is this how it should work or is there something wrong?
I feel that C64 mode should not be aware of Amiga hotkey combinations and vice versa.

[PiCiJi]

nightly:
- improve disk change behaviour (DesertStrike)
- improve lost sync detection (Damocles)
- fix a bug calculating disk head position when stepping to a track with different size (e.g. IPF, EXT ADF)

Originally Posted by ross

Damocles
Game that only works by chance in OCS (the "copper setup" is totally wrong for several reasons and in several places).

And BEAMCON is written accidently, which set frame lines to never initialized VTOTAL.
----------------
Desert Strike:
That's right, an undetected disc change happens without warp too. (stepping while waiting for change disc is correct)
I have increased minimum wait time to 1,9 seconds.
This improves detection. Even in WinUAE, it is not always recognized. It doesn't make sense to cross-check this with a Gotek now. I don't have KryoFlux to write it to real floppy disks. Presumably, the ADF version is sufficient as a test candidate.
If it is not recognized, simply drag and drop the same disc over it again.
-------------

I also found a potential bug when switching disks, which can lead to an incorrect memory area being accessed, e.g. when switching from IPF to ADF when the last IPF track is longer. The probability of occurrence is rather low.

Originally Posted by amilo3438

but I noticed that there was another option "Input check finish Warp" that was turned off.

However, this only works if the automatic warp is activated. Normally, the automatic warp is controlled by the floppy motor.
Stopping the motor late can cause you to lose a life after warping because the game is running a few seconds too fast.
"Input check finish Warp" also checks whether the game is asking for input and then turns off the warp in time.

Originally Posted by amilo3438

I wonder why that option "Input check finish Warp" is not permanently enabled?

However, this option also has a disadvantage. For example, if the game queries the inputs during loading, the warp runs too short or not at all.

Originally Posted by amilo3438

The only other way I do it is maybe to drag&drop on the screen and not on one of the icons that are shown.

In this case the option "always autostart by drag'n'drop" decides between hardreset + insert or insert only.

Originally Posted by amilo3438

]Double Dragon II - The Revenge v4.16 (1989)(Virgin)(ECS)(NTSC)(US)[2635] ... Try to run it on A500 ECS KS13 512+512 PAL and see what happens!

It's NTSC only.

Originally Posted by amilo3438

From Amiga mode go to the C64 tab and do hard reset from there (that's how to enter to C64 mode from the amiga mode). Then on the C64 mode press the combination of hotkeys for hardreset or softreset that is defined in amiga configuration hotkeys (nothing is defined in C64 configurations hotkeys) and it switches back to the amiga mode from the C64 mode!? Is this how it should work or is there something wrong?
I feel that C64 mode should not be aware of Amiga hotkey combinations and vice versa.

There is basically no Amiga/C64 mode. At the time, we had thought about switching quickly from one system to another.
Only if both hotkeys are assigned to the same keys, you cannot switch to the other system.

[amilo3438]

Eye of the Beholder (1991)(SSI)[0116] ... some glitch occurs in one part of the animation!

Like this: Wait for the part where a group of people enter the palace and until a sheathed knife appears and immediately after that glitch appears on the left side! (also checked on WinUAE where it does not appear)
Edit: The glitch was in the form of a small line that quickly appeared and then disappeared.

PS.
I tried a few more times with the disk inserted and with it not inserted and I can't provoke it to that glitch. I don't know, it's another one of those difficult to diagnose problems!? (after a long time of running the emulator)
Otherwise, apart from this, I haven't noticed any problems during testing so far.

[PiCiJi]

Originally Posted by amilo3438

Eye of the Beholder (1991)(SSI)[0116] ... some glitch occurs in one part of the animation!

confirmed on real hardware.
result differs on soft reset.

Edit: Pinball Dreams: pressing a key after table selection stops loading (crash) (confirmed on real hardware)
Edit: Eye of the Beholder: toggle option "PowerUp memory pattern" in WinUAE