26-Apr-2024 16:52 GMT.		 UNDER CONSTRUCTION
Anonymous, there are 19 items in your selection
[News] Elbox release new Mediator driversANN.lu
Posted on 30-Mar-2001 00:46 GMT by Christian Kemp19 comments
View flat
View list
Darrin wrote: "I just lifted this from Amiga-News.de... It seems that Elbox have been working VERY hard on the Mediator software and are proud to present a new PCI library, Voodoo card driver and Ethernet driver."

ELBOX Computer: Neue Software für Mediator PCI-Busboard

ELBOX COMPUTER - Krakow, 29 March 2001

REVOLUTION BEGINS:

COMPLETELY NEW SOFTWARE

FOR MEDIATOR PCI BUSBOARD

pci.library ver.3.0

Voodoo.card ver.2.0

MediatorNET.device ver.2.0

ELBOX Computer programmers have been committed to preparing this new revolutionary software for the last couple of months.

Enhancements made in the programming stage offer the user performance that has never been available for Amiga users before.

The new pci.library ver.3.0 enables access to the Config Space, I/O Space and Memory Space in PCI cards directly in the address space of the turbo card processor. Access to the PCI Memory Space is executed through the continuous 1.7 GB area.

The new PCI library supports all the turbo cards with processors: 68030, 68040 (& PPC603e), 68060 (& PPC603e).

The new Voodoo.card driver ver.2.0 enables full control over the memory of the graphic card used as a frame buffer by the P96 system.

This allows using the maximum resolutions offered by the Voodoo3/4/5 cards.

And now something special: the new Voodoo driver offers the possibility of overclocking memory in Voodoo cards for the first time in Amiga.

We are sure this new piece of software will make your work and enjoyment much more effective and playful.

Mariusz Wloczysiak ELBOX COMPUTER, Press Department

Elbox release new Mediator drivers : Comment 1 of 19ANN.lu
Posted by Amifan on 29-Mar-2001 22:00 GMT
Continue memory space uh? Maybe they mean: the bankswitching is completely handled by the PCI.library, so the whole 1.7 Gb is useable. Just call the PCI.library with your memory requirements and you've got it to your avail......
Anyway...no memory barriers anymore :o)
Elbox release new Mediator drivers : Comment 2 of 19ANN.lu
Posted by Teemu Kärkkäinen on 29-Mar-2001 22:00 GMT
Yeah! This is good news!
Now, when is that MediatorUp archive going to arrive? ;)
[-W]
Elbox release new Mediator drivers : Comment 3 of 19ANN.lu
Posted by Amifan on 29-Mar-2001 22:00 GMT
In reply to Comment 2 (Teemu Kärkkäinen):
I'm checking mail every hour :o( Nothing yet...Anyway..I'm curious how they will handle the PowerPC side, if at all.....
Elbox release new Mediator drivers : Comment 4 of 19ANN.lu
Posted by phill wooller on 29-Mar-2001 22:00 GMT
In reply to Comment 1 (Amifan):
I assume they are using the mmu to automatically select each page.
This would still have a speed overhead, but programs / device
drivers won't have to worry about it anymore. It's probably
taken them this long to get it working ( as I can't believe
they hadn't thought about doing it till now .... ).
Phill
Elbox release new Mediator drivers : Comment 5 of 19ANN.lu
Posted by Thomas Frieden on 29-Mar-2001 22:00 GMT
In reply to Comment 1 (Amifan):
The bankswitching is part of the hardware, it will *never* go away with software. The probably use a MMU to switch the window automatically.
If that's true, you can kiss PPC-native drivers goodbye.
Regards, Thomas
Elbox release new Mediator drivers : Comment 6 of 19ANN.lu
Posted by Christian Michael on 29-Mar-2001 22:00 GMT
In reply to Comment 5 (Thomas Frieden):
Ok Thomas. Bankswitching is inevitable you say.
Is bankswitching the equivalent expression of using the mediator offset registers for transfers?
Elbox has previously stated that using the offset registers has no practical performance hit.
Eyetech has claimed that this approach would reduce transfer speed significantly (2.7 MB/s in worst case scenario)
Maybe you, as a developer with knowledge of the mediator HW could elaborate?
best regards,
Christian Michael
Elbox release new Mediator drivers : Comment 7 of 19ANN.lu
Posted by Victor on 29-Mar-2001 22:00 GMT
In reply to Comment 5 (Thomas Frieden):
> The bankswitching is part of the hardware, it will *never* go away with software. The probably use a MMU to switch the
window automatically.
> If that's true, you can kiss PPC-native drivers goodbye.
You're contradicting yourself here all right.
In case of the really native PPC drivers using MMU with access
on the side of 68k is not important at all.
Elbox release new Mediator drivers : Comment 8 of 19ANN.lu
Posted by Christian Michael on 30-Mar-2001 22:00 GMT
In reply to Comment 6 (Christian Michael):
Ok - no need to explain anyway. I do understand that bankswitching refers to setting zorroII window(=bank) offset, so that it points to the appropiate PCI address.
However, I can not see the big problem in relation to Warp3D, unless the screen buffers are not continous. Assuming the buffers are continuous, it would just require the "window/bank" offset to be changed once in the case of a "bank/window" overlap?
Regards,
Christian Michael
Elbox release new Mediator drivers : Comment 9 of 19ANN.lu
Posted by Hans-Joerg Frieden on 30-Mar-2001 22:00 GMT
In reply to Comment 7 (Victor):
> You're contradicting yourself here all right.
Think again:
LockBitmap (or P96 equivalent) will return a virtual address that the PPC does not see, because there is a hole in the address space.
So the MMU setup on the 68K side *is* very important unless the complete system (i.e. RTG system, OS, everything) is on the PPC side, in which case a 68k pci.library becomes a moot point anyway...
Elbox release new Mediator drivers : Comment 10 of 19ANN.lu
Posted by Hans-Joerg Frieden on 30-Mar-2001 22:00 GMT
In reply to Comment 8 (Christian Michael):
Try running QuakePPC (or any direct-access WarpUp or PowerUp program) on the new drivers (I think E2240 was quoted on the Mediator list) and ask yourself why it crashes...
Best regards, Hans-Joerg.
Elbox release new Mediator drivers : Comment 11 of 19ANN.lu
Posted by Electric Head on 30-Mar-2001 22:00 GMT
Well on the plus side, and for the record.. I just got my updated driver from Elbox this morning, and its awesome. After overclocking my Voodoo 3 and cranking workbench up to 1280x1024, I've been on a high all day ;) Very very fast, much more stable, and not one single bit of graphics trashing. The only complaint I have now is that my icons are a bit too small and I have to move my mouse a bit further to get to amidock :o)
Elbox release new Mediator drivers : Comment 12 of 19ANN.lu
Posted by Christian Michael on 30-Mar-2001 22:00 GMT
In reply to Comment 10 (Hans-Joerg Frieden):
Have Blizzard 1260 clocked at 69 MHz so there is no mmu-problems here.
However the mmu modification done by the voodoo.card makes it impossible to set the cachemode to imprecise, so I am using the 1.8 version again.
Imprecise makes writes 20% faster with bustest. WritePixel routines benefits the most (+60% in with P96speed).
Has anyone got an idea how to modify mmu tables when the new driver is active?
(0x82000000 - 0x823fffff is remapped to 0x00200000 and marked as cacheinhibit according to MuScan)
regards,
Christian Michael
Elbox release new Mediator drivers : Comment 13 of 19ANN.lu
Posted by victor # on 30-Mar-2001 22:00 GMT
In reply to Comment 12 (Christian Michael):
Could you run a bustest within the mentioned region (0x82000000 - 0x823fffff)?
Couldn't the PPC problem somewhat solved if _its_ MMU is setted up similarly?
Elbox release new Mediator drivers : Comment 14 of 19ANN.lu
Posted by Hans-Joerg Frieden on 30-Mar-2001 22:00 GMT
In reply to Comment 13 (victor #):
Sure, if the PPC MMU would have a similar Setup, this would solve the problems. However, this is not trivial. Both MMU's must be kept in sync.
The danger is that this will produce a lot of context switching, and a lot of overhead, since the pci.library is 68k only...
Regards, Hans-Joerg.
Elbox release new Mediator drivers : Comment 15 of 19ANN.lu
Posted by Darrin on 30-Mar-2001 22:00 GMT
In reply to Comment 14 (Hans-Joerg Frieden):
Dear Hans-Joerg,
Do you have any ideas as to what Elbox should do next to get around this? Have Hyperion been contacted by Elbox with regards to making the Mediator more "software developer friendly". After all, PPC WarpOS support is essential to them in the long run and usefull to you as an aid to selling more quality software.
All the best,
Darrin
Elbox release new Mediator drivers : Comment 16 of 19ANN.lu
Posted by christian.michael@groenjord.dk on 30-Mar-2001 22:00 GMT
In reply to Comment 13 (victor #):
No problems with bustest. I tried addr=82000000 or 82100000 and the results were exactly the same as addr=200000 or 300000 with the 1.8 software version (incl gfx trashing in top of the screen).
Performance: max 9.4 MB/s. With the old update AND imprecise cacheinhibited mode it is 11.3 MB/s)
regards,
Christian Michael (blizz1260/69MHz)
Elbox release new Mediator drivers : Comment 17 of 19ANN.lu
Posted by Christian Michael on 30-Mar-2001 22:00 GMT
In reply to Comment 14 (Hans-Joerg Frieden):
A PPC native PCI library could solve the problem on BlizzPPC/CybPPC efficiently then? - Hope elbox listens..
Regards,
Christian Michael
Elbox release new Mediator drivers : Comment 18 of 19ANN.lu
Posted by Hans-Joerg Frieden on 30-Mar-2001 22:00 GMT
In reply to Comment 15 (Darrin):
The same mechanism must be implemented on the PPC side, otherwise things will not work from the PPC side. Pointers are incoherent, i.e. a bitmap is mapped to high memory on the 68k side where there is no memory in the PPC version.
Elbox release new Mediator drivers : Comment 19 of 19ANN.lu
Posted by phill wooller on 02-Apr-2001 22:00 GMT
In reply to Comment 14 (Hans-Joerg Frieden):
It shouldn't require any context switching if done properly, but whether this it is possible to do anything properly with the current ppc situation is another matter.
The memory window "problem" is really a software issue & if everyone could play together it would be easily solved. Unfortunately Elbox doesn't seem to be in with the popular kids who want us to use their hardware that has no software issues, but many hardware ones ( released hardware not just announced ).
If the Shark/PPC gets released with warpos/powerup compatibility then Elbox will have won.
Phill
Anonymous, there are 19 items in your selection
Back to Top