Denise bugfixes and improvements

[PiCiJi]

Originally Posted by amilo3438

I sorted the remaining problems according to similarity:

thank you for testing again. do another ticket and then I'll continue here.

[amilo3438]

Take as much time as you need, no need to rush!
(I sorted and reviewed it to make it a little clearer, but they are all already known from earlier, i.e. from another thread.)

[amilo3438]

Post #20 is updated according to latest nightly! (some gfx problem have been fixed)

[PiCiJi]

How many demos are there for the Amiga (OCS/ECS) ?

[amilo3438]

Originally Posted by PiCiJi

How many demos are there for the Amiga (OCS/ECS) ?

As I remember, of those mentioned in post #20 is ECS only: OLD_BULLS-Fugazi !
(but try to test it on winuae or a real machine if not sure)

[PiCiJi]

I meant the question regardless of emulators. How many demos (non AGA) are there for the Amiga ? (approximately)

[jbl007]

Originally Posted by PiCiJi

How many demos (non AGA) are there for the Amiga ? (approximately)

Good question. pouet.net lists 25 results on 690 pages (=17250 results). This number can easily be doubled I guess...

I was about to report gfx garbage in P.P Hammer: https://hol.abime.net/1129 but your latest update fixed that.

Still have one more: https://www.pouet.net/prod.php?which=3073
Demo refuses to start, some initial check fails. Msg from the demo: "no fast processors or aga chips allowed!"

[amilo3438]

Originally Posted by jbl007

Good question. pouet.net lists 25 results on 690 pages (=17250 results). This number can easily be doubled I guess...

Some 10 yrs. ago, while testing the demos for winuae, I stopped after some over 13500 demos.
(but I am not sure that it is the final number)

Originally Posted by jbl007

Still have one more: https://www.pouet.net/prod.php?which=3073
Demo refuses to start, some initial check fails. Msg from the demo: "no fast processors or aga chips allowed!"

Seems no prob. on winuae A500 1.3 512+512! (just tested)
Edit: Seems also no prob. at loading on the latest WinFellow! (except a lot of gfx prob.)

But there is also a final version that seems to work well in Denise: https://www.pouet.net/prod.php?which=61182

[amilo3438]

Glitches if performance and/or runAhead is enabled in the emu!
----------------------------------------------------------------------
TURNIPS-WorkForNothing_qdfix.adf ... at the end of the demo upper part glitches if performance and runAhead is enabled together!
(in "The Zone")(A500 ECS 1mb Chip ram)

HITTHEROAD_1.adf ... while in "Floffy2" part click on the Left Mouse Button and it crashes (resets) if runAhead is enabled!
(in "The Zone")(A500 ECS 512mb Chip + 512 Slow)


Glitches on a scrolling text!
------------------------------
Ozone-ShedTearsA.adf ... right side of the scrolling text glitches when a picture is shown!
(in "The Zone")(A500 ECS 512mb Chip + 512 Slow)

[PiCiJi]

Glitches if performance and/or runAhead is enabled in the emu!

in performance mode denise chip is emulated only partially. Programs that read back sprite/playfield collisions get problems. This cannot be remedied in this mode, because this is where the speed advantage comes about. It may also be another problem that can be fixed. we'll see.

edit: I missed the "or".

[amilo3438]

For TURNIPS, glitch happens only if performance & runAhead is enabled! If one of this is disabled = no glitches!

For HITTHEROAD problem is only if runAhead is enabled! (performance here doesn't matter)

[amilo3438]

Blitter problem!
-----------------
FLD-Demodisk021.adf ... E: Blitter - Storm = glitches!
(in "The Zone")(A500 ECS 512mb Chip + 512 Slow)

[amilo3438]

Stops after a while!
---------------------
FLD-Demodisk022.adf ... D: Party-Intro Defcon 1
(in "The Zone")(A500 1.3 512 +512)

[PiCiJi]

Stops after a while!
---------------------
FLD-Demodisk022.adf ... D: Party-Intro Defcon 1
(in "The Zone")(A500 1.3 512 +512)

Thanks for the support.

Scoopex Crash Landing fix
In V 79, the BPL PTR is written in H 0xda. This was ignored because the same DMA must not be used in the following cycle.
However, the change must go through, otherwise the display is broken. So in H 0xdb no BPL DMA is allowed to happen. The demo disables PBL DMA too early and this causes the last two BPL DMAs in 0xd9 and 0xdb no longer happen. To get more information, I wrote a test for this.
This switches off BPL DMA at various positions. The behavior in the next line is also interesting here.
In the lower part of the test, the blitter is still running and changes the display color to white when it is finished. It's a bit hard to see and happens in block 27.
Based on the test results, Denise was corrected. (latest nightly)

For comparison, the test was run on a real ECS Agnus Rev 6a.
My OCS Agnus Amiga unfortunately smoked out a few weeks ago.
I would be happy if someone could run this test on an OCS Agnus (Rev5 or smaller) and make a screen recording with the mobile phone. I had to modify H-Sync/H-Centre in the service settings of my CRT to make the right part of the image completely visible (overscan)
An A1000 comparison test would also be helpful, but more important is the A500 with OCS Agnus.
I am attaching the recording of the ECS Agnus to this posting. It is not the colors that are important, but the deviating line height on the right and left side.
In addition, I attach the test ADF.

In a quick test with WinUAE 5.0.0 (2023.05.14) (ECS Agnus, Cycle exact, kick 1.3, Overscan: Extreme) not everything was error-free. Maybe it's also because that I didn't set all the necessary options in WinUAE.

[Toni Wilen]

I guess I forgot to ask ross to do nasty bitplane DMA on/off test (EDIT: fixed, thanks )

OCS A500 photo uploaded (OSSC + LG OLED). A1000 is identical (as expected). If you need more test photos: only test executables will be accepted (must work on 1.3, 3.1 and possible real fast RAM must not affect test results)

[PiCiJi]

Originally Posted by Toni Wilen

I guess I forgot to ask ross to do nasty bitplane DMA on/off test

OCS A500 photo uploaded (OSSC + LG OLED). A1000 is identical (as expected).

Thank you. Fortunately, the OCS Agnus result is already consistent with the emulation.

[amilo3438]

Missing vertical green scroller!
-----------------------------------
FLD-Demodisk047.adf ... C: Connan / TPL!
(in "The Zone")(A500 1.3 512 +512)(compared to winuae)


Glitches on a vector objects!
--------------------------------
FLD-Demodisk050.adf ... A: Cementery / Possessed (wait after the 1st credit text and then later)
FLD-Demodisk092.adf ... C: Forgotten Realism / Mythos
(in "The Zone")(A500 1.3 512 +512)


Glitches on a scrolling text!
-------------------------------
FLD-Demodisk056.adf ... F: Molecules / IT (at 1st it seems ok, but then errors!)
(in "The Zone")(A500 1.3 512 +512)(compared to winuae)


Stops after a while!
----------------------
FLD-Demodisk083.adf ... G: Tropical Sunset / TSL (demo stops but music continue to play)
(in "The Zone")(A500 1.3 512 +512)(compared to winuae)


Blitter problem!
------------------
FLD-Demodisk089.adf ... F: Demo / Hammerhead = glitches in starscroller and the next scroller!
(in "The Zone")(A500 1.3 512 +512)(compared to winuae)


Missing diagonal white scroller!
-----------------------------------
FLD-Demodisk090.adf ... K: Intro / United Forces
(in "The Zone")(A500 1.3 512 +512)(compared to winuae)


Note: Post #20 is also updated with the latest findings!

[amilo3438]

Scroller jumping!
-------------------
FLD-Demodisk111.adf ... E: Relaxation / Reality
(in "The Zone")(A500 OCS/ECS 1.3 512 +512)(compared to winuae)

Note:
The 1st bigger scroller is ok in OCS and jumps in ECS = correct. (compared with winuae OCS & ECS)
The 2nd smaller scroller jumps in OCS and ECS = not correct. (does not jump in winuae OCS & ECS)



With this one I finished with FLD-Demodiskxxx collection testing!

One question regarding the emu: Is there existing a portable mode! (= config files are not saved in a system users folder than in emu folder)

[PiCiJi]

nightly:
OLD_BULLS-Fugazi-A.adf
2/3 bugs fixed

1. ECS Agnus uses 0.5 mb slow mem as chip mem. This fixes crash after 2nd green text. It already worked with 1 MB chip mem.
2. Blitter linedraw fix
3. todo: need to fix some black lines in animation directly after green texts

- checked all CPU\IPL related tests from vAmigaTS.
Now, these all work in Denise

@Toni
Some tests that didn't work in Denise, I tested in WinUAE and some failed like DIVS2, DIVU2, CHK1

I've worked with FX68k quite a bit. All tested opcodes with the typical IPL sampling during the prefetch cycle have the interrupt serviced after current instruction, if IPL line is changed no later than 1 CPU cycle before the prefetch cycle.

in CPU cycles:
-2 [sample limit] -1 [1 2 3 4] prefetch

[amilo3438]

Originally Posted by jbl007

Still have one more: https://www.pouet.net/prod.php?which=3073
Demo refuses to start, some initial check fails. Msg from the demo: "no fast processors or aga chips allowed!"

Seems fixed in the latest nightly!

Originally Posted by amilo3438

One question regarding the emu: Is there existing a portable mode! (= config files are not saved in a system users folder than in emu folder)

I was wondering if Denise supports portable mode? (some other emulators expect an empty file named portable to be created)