WinUAE Voodoo 3 emulation testing

[White]

Surely it has nothing to do with it but for example the voodoo 2 do not need the bios then I had the doubt.
then i was trying various versions of the bios for voodoo 3 to see if somehow i could solve the problem with the 4.1
but nothing to do yet the 4.1 is fast and it also works well with voodoo and really accelerated to the difference of cvision64
I'm sorry again if I return to the subject :-)

with the latest beta version of winuae in this post
wipeout is almost perfect flicker and rare

[AC/DC HACKER!]

Originally Posted by Toni Wilen

Don't use JIT when testing anything strange?

I'll try to explain this..I've created some logs but I don't know if they'll be accepted. Mad-Matt is a lot better at explaining the graphics technical than I am.


JIT is always disabled by OS 4.x and 68K is shut off. What I noticed is..White is correct with WipEout IF you use Triple Buffer within the menu of the game. Double buffer produces 3D flickers. It's not the game, it's Voodoo or...new stuff in WinUAE. I don't know which. Payback WOS is almost playable but since there isn't any control over buffering, there are still Voodoo video glitches happening. I considered posting pictures. This doesn't have anything to do with debugging PPC games, this is WinUAE or Vooodoo or both.


68K: With JIT enabled or not, the games loading time are very slow. I don't understand why it's so slow from 68K but speedy and almost normal A4000 speed with PPC OS. Is that still something to do with the PCemu you've typed about in the Beta thread? You commented about the PPC running from Softpoint or something like that, but then...shouldn't the PPC OS 4.x stuff be really slow also?
Anyway, if the speeds were the same, I can see that the video glitches should be about the same with OS 3.x as it is with PPC OS 4.x.


Maybe as Matt eluded to...there's still something buggy about Zorro 2 and 3?? That also affected CV643D? I'm going to get some of my video stuff functioning again that I used to and see if I notice any differences with them. Right now, I agree, it's not about the "games" it's about getting it all to be smooth and functioning as should.


I rarely use JIT. If I'm just sitting to play a game, WHDLoad, Shogo with Picasso 4 (as I was some months ago) then I care about the games. I'm with Matt, I'd like to go a lot more with WinUAE...it's FASTER than my A4000, haha and more like how we all wanted the Amiga to be.


I AM Not Asking you to Debug WarpOS.. I'm not even discussing that at all. When I run Payback 68K...there are hesitations just before the main menu shows options. Moving through the menus, are some hesitations as well, which also have some graphic glitches...but on in all, other than being a little slower than it should be, the Voodoo 3D looks VERY good. The video for the game is almost smooth as an A4000.


So, I'm wonder if there's still a buffering bug in the mix still as Matt suggested?


I'll make a log of 68K video. I'll grab some screen shots soon, too.

[AC/DC HACKER!]

Originally Posted by Mad-Matt

The driver pulls in the gpu/memory clocks from the bios and wont allow you to overclock in the driver/p96 settings. You need to edit/reflash the bios if you want to change clock speeds. I've had to drop mem speeds on my card to avoid rainbow coloured screens.

Perhaps I'm missing something here. Are you refering to your Windows BIOS or reflashing the Voodoo? Is there a way to access the Voodoo BIOS from WinUAE??

[Mad-Matt]

I'm talking about a real card with windows editing/flashing tools.

[AC/DC HACKER!]

I figured, but wanted to be sure.

[Toni Wilen]

Originally Posted by AC/DC HACKER!

I'll try to explain this..I've created some logs but I don't know if they'll be accepted. Mad-Matt is a lot better at explaining the graphics technical than I am.

I meant the crash when voodoo bios is not loaded. It shouldn't happen in non-JIT modes.

[Toni Wilen]

Originally Posted by White

Surely it has nothing to do with it but for example the voodoo 2 do not need the bios then I had the doubt.

Voodoo 2 is totally different (no (S)VGA hardware) than Voodoo Banshee or 3+.

[White]

@Toni
Thanks for the reply, and the explanation

[AC/DC HACKER!]

WinUAE exits with or without JIT for 68K with voodoo.rom renamed.

68K version...

Pictures are from basic Workbench 3.1, JIT - ran very fast. Without bugs..it would be at correct normal A4000 speed.

The same Workbench without JIT and MMU active produced very few graphical bugs. It's just slow.. I do have Fast as Possible set for both logs.

Logs included of both sessions. Want some other options en/disabled?

[AC/DC HACKER!]

Here is the log when voodoo.rom isn't found and without JIT, but MMU is enabled.

[abu_the_monkey]

is anyone else seeing a pattern where programs that use ahi sound drivers crash with a requester but if you select 'suspend' rather than 'reboot' then it continues on and works as expected minus the sound?

under os3. i haven't tried os4

br

edit: this issue seems to be only when emulating real hardware. ie gfx cards ( zorro or pci). with generic uaegfx it is not a problem.
NB: the programs are not hardware 3D related.

edit: it seems only to be when emulating pci hardware. all cv64/3d configs seem to be ok. mediator/voodoo not so much.

[AC/DC HACKER!]

I know you're seeking Yes/No replies...so I'll reply this way..I haven't ventured into my normal Amiga areas in a while...so not yet. BTW, I adore that monkey, fun yet also mischievous. Ahh...to test the audio part of PCI..

[White]

error sorry

[Toni Wilen]

Originally Posted by abu_the_monkey

is anyone else seeing a pattern where programs that use ahi sound drivers crash with a requester but if you select 'suspend' rather than 'reboot' then it continues on and works as expected minus the sound?

Perhaps you have some AHI driver that does something bad when combined with PCI? Remove all unneeded ones temporarily.

I have noticed unexpected weirdness will happen sooner or later when testing different hardware (that is practically impossible in real world, like switching between A1200 and A4000 -only hardware using same HDF) and also slowly collecting more and more drivers without currently installed hardware.

EDIT: cleaning up old drivers is even more important when mixing PCI bridges. They have all their own unique drivers.

[Mad-Matt]

I,ve not had any major issues with add-on cards, but I think there is at least a bug with mediator sb128 in aos3. At least ahi v5 is required but also sound only plays correctly in a (PC) screenmode. On my 1200 Pc and non PC both work fine.

It wouldn't be much of an issue but at least some warp3d stuff I'm testing only looks correct in non PC modes.

@toni I can link an updated hdf if you want to test.

[AC/DC HACKER!]

Originally Posted by Toni Wilen

Perhaps you have some AHI driver that does something bad when combined with PCI? Remove all unneeded ones temporarily.

I have noticed unexpected weirdness will happen sooner or later when testing different hardware (that is practically impossible in real world, like switching between A1200 and A4000 -only hardware using same HDF) and also slowly collecting more and more drivers without currently installed hardware.

EDIT: cleaning up old drivers is even more important when mixing PCI bridges. They have all their own unique drivers.

This is why I rename libraries and such with versions at the end of the filename. Keeps them from being located when not used and locate versions easily without having to "versions" often.

[Toni Wilen]

Originally Posted by Mad-Matt

I,ve not had any major issues with add-on cards, but I think there is at least a bug with mediator sb128 in aos3. At least ahi v5 is required but also sound only plays correctly in a (PC) screenmode. On my 1200 Pc and non PC both work fine.

Non-PC screenmode + Mediator DMA hack DMA fixed.

There are still some mysteries to solve:
- Something weird is going on with byteswapped modes if voodoomem is less than 15. (Hang/crash due to invalid memory accesses if "wrong" screenmode)

EDIT: It seems if voodoomem is less than 15, "unused" voodoo vram is added as system ram:

This appears in system memory list:

52c00000 10 0 pcidma memory
Attributes 2005 First 52c00020 Lower 52c00020 Upper 52f00000 Free 3145696
52c00020: 52c00020-52f00000,00000000,002fffe0 (3145696)

But I can't understand how this can work because byteswapping mode affects VRAM reads and writes: if byte swapping mode changes, system crashes.. Is this really supposed to work this way? Or is this handled by MMU (VoodooGuide mentions NO_MMU and MMU variants).

- FM801 Mediator driver either is broken or something is still wrong with PCI emulation. For some unknown reason when AHI loads the driver, it sets FM801's PCI config structure's interrupt disabled bit -> AHI only plays first sample repeatedly.. SB128 (ES1370) works fine.

[Hedeon]

Originally Posted by Toni Wilen

But I can't understand how this can work because byteswapping mode affects VRAM reads and writes: if byte swapping mode changes, system crashes.. Is this really supposed to work this way?

The memory header is in VRAM and gets swapped (= corrupted list). CHIP ram is lower priority so is after voodoo system mem in the list. So both Alloc/Free of voodoo system mem (doesn't happen a lot as it is lower prio than FAST) and CHIP mem result in a guru (memory list corrupted I would think).

At least, that is one problem with byteswapping I encountered on the voodoo with the VRAM being added to system RAM. And then switching from PC to Non-PC or vice-versa.

Edit:

So the important stuff at 52c00000 - 52c00020 gets corrupted/swapped preventing the next node (CHIP RAM) to be read. I never encountered immediate crashes, however. Depended on if the memory list was traversed.

[Toni Wilen]

It should be almost immediate crash because WB calls AvailMem() periodically and it scans all memory lists..

[Hedeon]

Originally Posted by Toni Wilen

It should be almost immediate crash because WB calls AvailMem() periodically and it scans all memory lists..

Maybe I remember it wrong :-) But I am sure that byte-swapping corrupts the memory list as the node before CHIP is being swapped (and made nonsensical)

EDIT:

This was my post number 1337, so it must be true.