VA2000 Support Thread

[mark_k]

For starters, use ShowConfig to see which memory address range(s) the board uses.

How do other Amiga RTG cards handle this? Do they appear as two logical AutoConfig boards, one in Zorro II RAM space ($200000-$9FFFFF), and hardware registers in (uncacheable) I/O space $Exxxxx? That could be a good approach for a future hardware/firmware revision maybe.

[trixster]

In the sonnet thread we're using mmu.library to cache inhibit the space the sonnet uses.

So pciinfo returns the following:

Pci Device 7:
VendorID: $1057
VendorName: Motorola
DeviceID: $4
DeviceName: MPC107 PCI Bridge/Memory Controller for PPC
ClassCode: $e0001
RevisionID: $12
InterruptPin: 4
SubsystemVendorID: $0
SubsystemID: $0
MemSpace0: $70000000 - $7bffffff
MemSpace1: $60400000 - $60400fff

And in "sysrefs/mmu-configuration" we're using something like:

SetCacheMode from 0x07000000 size 0x08F80000 Copyback
SetCacheMode from 0x70000000 size 0x0BB00000 CacheInhibit
SetCacheMode from 0x00000000 size 0x00200000 CacheInhibit NonSerial Imprecise
SetCacheMode from 0x64D00000 size 0x00200000 CacheInhibit

The 2nd line inhibits the sonnet's space, the fourth line inhibits the voodoo memory the mediator uses.

Dunno if this info is helpful or not but it might be useful to someone.

[klx300r]

can someone please tell me where exactly the va200CX board goes on the A4000 daughter board

thanks!

[mark_k]

OK, I checked the va2000cx driver source code. From there:

if (cd = (struct ConfigDev*)FindConfigDev(cd,0x6d6e,0x1)) {
  memory = (uint8_t*)(cd->cd_BoardAddr)+0x10000;
  regs = (MNTVARegs*)(cd->cd_BoardAddr);

To preserve the best performance, instead of marking the entire board address space as uncacheable, it could/should be enough to mark only the first 64KB uncacheable.

So from trixter's post above you'd need a line like this in the mmuconfiguration file:
SetCacheMode from 0x00200000 size 0x00010000 CacheInhibit

$200000 is just an example. Your board may be configured to a different address, e.g. $400000, $600000 depending on which other Zorro II boards are in your machine. Use ShowConfig to check your board's base address.

[amigoun]

Thanks mark_k, will definitely try this tonight.

[klx300r]

Att: A4000 users...want to get this installed this weekend:

Originally Posted by klx300r

can someone please tell me where exactly the va200CX board goes on the A4000 daughter board

thanks!

[trixster]

If you install mmulibs I think it will create mmu-configuration in sysrefs/envarc. If it doesn't then I think there's an example mmu-configuration file in the install archive. I'm not at my Amiga to check at the moment though.

[amigoun]

Installed MMULib adjsted mmu-config and it works. Freeze is gone
My only concern is quite poor performance of the graphics itself. I am quite sure that with the stock firmware graphics were significantly more responsive. But this is not the fault of the MMULib.
For reference my current P96Speed bench is this (MMULib installed):

P96Speed 1.2 - ©`97-99 by Jens Langner
.-----------------------------------------.
| Computer......: Amiga 3000/ECS |
| CPU...........: 68030/25 MHz |
| OS / WB.......: V40.68/V40.42 |
| SetPatch......: V43.7 |
| Chip/Fast.....: ~2.0MB/15.5MB |
| Graphics card.: NO GRAPHICS CARD |
| GFX system....: Picasso96 v2.1e |
| Resolution....: 640 x 480 x 8 |
| Depth/Colors..: 256 colors |
| Testlength....: 13 |
+-----------------------------------------+
.============= SPEEDRESULTS ==============.
| RectFill()................ 307 op/s |
| RectFill() Pattern........ 27 op/s |
| WritePixel().............. 10966 op/s |
| WriteChunkyPixels()....... 271 op/s |
| WritePixelArray8()........ 270 op/s |
| WritePixelLine8()......... 1994 op/s |
| DrawEllipse()............. 1187 op/s |
| DrawCircle().............. 1436 op/s |
| Draw().................... 428 op/s |
| Draw() Hor/Ver............ 1123 op/s |
| ScrollRaster() X.......... 6 op/s |
| ScrollRaster() Y.......... 9 op/s |
| PutText()................. 449 op/s |
| BlitBitMap().............. 290 op/s |
| BlitBitMapRastPort()...... 269 op/s |
| BitMapScale()............. 31 op/s |
|--------------- Intuition ---------------|
| OpenWindow().............. 17 op/s |
| MoveWindow().............. 37 op/s |
| SizeWindow().............. 22 op/s |
| CON-Output................ 17 op/s |
| ScreenToFront()........... 43 op/s |
`========================================='

[klx300r]

Originally Posted by klx300r

Att: A4000 users...want to get this installed this weekend:

solved thanks to wlemonds over at amiga.org

[mark_k]

amigoun: Could you run MuScan and paste the output here?

[amigoun]

Here the results:

MuScan 46.1 (02.07.2016) © THOR

68030 MMU detected.
MMU page size is 0x400 bytes.

Memory map:
0x00000000 - 0x00007FFF CopyBack Single Remapped to 0x07100000
0x00008000 - 0x001FFFFF CacheInhibit Imprecise NonSerial
0x00200000 - 0x00BBFFFF Blank
0x00BC0000 - 0x00BFFFFF CacheInhibit I/O space
0x00C00000 - 0x00D7FFFF Blank
0x00D80000 - 0x00DFFFFF CacheInhibit I/O space
0x00E00000 - 0x00EFFFFF Blank
0x00F00000 - 0x00F7FFFF CacheInhibit
0x00F80000 - 0x00FFFFFF ROM CopyBack Remapped to 0x07F78000
0x01000000 - 0x06FFFFFF Blank
0x07000000 - 0x07F77FFF CopyBack
0x07F78000 - 0x07FF7FFF ROM CopyBack
0x07FF8000 - 0x07FFFFFF CopyBack
0x08000000 - 0x3FFFFFFF Blank
0x40000000 - 0x41FFFFFF CacheInhibit I/O space
0x42000000 - 0xFFFFFFFF Blank

[mark_k]

What is the base address of your VA2000 board? It could be that setting the VA2000 memory region to CopyBack would improve performance on 68040/060 (no effect on 68030).

[amigoun]

ShowConfig says it is 0x40000000 (in my setup).

[mark_k]

Ah right. So what you want to do, is mark $40000000 to $4000FFFF as uncacheable, and $40010000 to $41FFFFFF as cacheable, even CopyBack.

Try running these commands, then run P96Speed again. (And maybe MuScan to confirm the changes you made.)
MuSetCacheMode ADDRESS=$40000000 SIZE=$10000 CACHEINHIBIT IOSPACE
MuSetCacheMode ADDRESS=$40010000 SIZE=$1FF0000 COPYBACK

[amigoun]

Did so. But no noticeable difference in speed as you wrote (I have stock A3000 030/25).
The VA2000 driver needs some optimization for sure.

[mark_k]

Is your data cache enabled? Do CPU CACHE to enable it. (Just double-checking, could you perhaps have disabled it globally before???)

[amigoun]

Yes, data cache is enabled.

[amigoun]

I wonder, is there any way to force zorro II/III on VA2000? Because the card itself went slow for me, when I out of curiosity tried to insert Commodore A3660 card in my A3000. It remained slow when cpu card was removed. Maybe it somehow switched for Zorro2/3 operation?

[JimDrew]

WHY do you have to disable caching for the VA2000CX? That really makes no sense (coming from someone who designed Zorro II hardware).

[IvanEBC]

I’ve had this issue since day one. Tonight i FOUND the issue but no solution

After my Amiga 4000/va2000 got mentioned in an Amiga blog with images, it again highlighted the issue and I’m stuck wondering, how do I fix this? - http://amigax1000.blogspot.com.au/

Issue: My workbench screen (1024x768) will perfectly go edge to edge on my monitor but anything using native screen (generally meaning ANY game or demo) will sit in the top left hand corner. I can tell it will not be correct because even during the Workbench loading phase, you can see a slight greenish box in the same top left corner position.

Lukas had claimed it was the monitor itself, but i didn't believe that. But with this in mind, I have started to test a bunch on monitors.

I use a BenQ 912 for my A4000 (Also my A1200 with an indivision AGA and my screens DO get centered). I also run a small 10cm cable from the VA2000 (HDMI) to the rear Zorro slot of the 4000 that gives me a standard VGA out, mostly because the old BenQ912 4:3 doesn’t have HDMI in.

List of screen I’ve tested with tonight with VGA, all producing the same issue over and over.
BenQ FP537s (small 14” LCD)
LG Flatron 1919SP
IBM Thinkvision
LG Flatron T170BH (This is a CRT monitor, and it STILL had this issue)
BenQ E2200HD Widescreen

AT THIS POINT I TRIED SOMETHING DIFFERENT – Opening my 4000, I connected the next monitor directly to the HDMI port … no VGA Adapter.
ASUS VW266H using HDMI (Widescreen)
The green tinted box during loading FILLED the screen – oookay!
Tried a number of games and demos – and FULL screen (mostly)

The issue I now face and wonder how everyone has got around this is this.
Obviously the Amiga wasn’t really made for widescreen so everything looks stretched out in my only successful test. Finding a 4:3 screen that’s more native to the Amiga that has HDMI, I believe could be near impossible.

I wonder how others have got around this situation? I'd like to hear some thoughts, and if this is something that can be corrected via a firmware update or some other setting?

Thanks guys...

SOLUTION: I switched to a HDMI to DVI adaptor instead, and now everything is full screen. Maybe this helps someone else.