I got SGL to work in Windows, and wrote some simple code with it: https://www.youtube.com/watch?v=azfQHgUKOrk
Do you have the SDK? I believe it is posted here on Vogons somewhere.
I can give you my code, so you can play around with it (Visual C++ 6.0): https://www.dropbox.com/s/2ir10zpnbp27lj5/SGL.zip?dl=0

Scali wrote:

I got SGL to work in Windows, and wrote some simple code with it: https://www.youtube.com/watch?v=azfQHgUKOrk
Do you have the SDK? I believe it is posted here on Vogons somewhere.
I can give you my code, so you can play around with it (Visual C++ 6.0): https://www.dropbox.com/s/2ir10zpnbp27lj5/SGL.zip?dl=0

Ah, you misunderstood my previous post. I've written some code for SGL Direct including MiniGL (OpenGL DLL reimplementation which wraps using SGL Direct) which I mentioned here in my previous post by using SDK posted on VOGONS under both VC++6.0 and MinGW.

The problem is that I can't make SGL to run (execute) under Windows on my emulated PCX2 card in PCem. 😀 It works on real card, but on emulated even if I respond correctly to all registers, initialize correctly IO memory mapped region they still don't work - executing any SGL application crashes with message that SGL.dll wasn't loaded. It's the problem in DllMain in SGL.dll that loads dynamically PVROS which gives SGL interface with use of SGLMID5, it seems that I'm missing some early register and it makes driver go to different code path that makes SGL not working. SGL for DOS (Tomb Raider) is more simple and is working good, because writes to good registers and populates VRAM with texture data, waits for "is render completed" command and then continues etc. Haven't tried Direct3D on emulated PCX2 under PCem though.

Thanks for the status update, always nice to hear from you Tuxality 😀

I couldn't get my emulation to get the driver to work either (not even with debug drivers), and plus pressing the alt-key sends me into a bsod hell for some odd reason

It behaves like that because you did not emulated this card, only faked that it exist... anyway:

I managed to get SGL and Direct3D to work on my emulated NEC PowerVR PCX2 card! 😀 😀 😀

I mean, holy shit - it works! Driver no longer rejects my card, null rendering for now (shows only garbage from host card VRAM), but at least SGL is working and now I can proceed with reverse engineering more serious PCX registers. 😉

Tuxality wrote:

I managed to get SGL and Direct3D to work on my emulated NEC PowerVR PCX2 card! 😀 😀 😀

Ossom!
So what did you have to do to make it detect the card? Was it trying to read some special kind of identification string or such?

Nah, I just made a byte order typo in only one of the PCI configuration space header field, so lame of me. 😁

Actually, apart from PCI configuration space it detects which type of card is it by reading two R/O registers which are used by driver before initializing card. After detecting which PowerVR card is it and which revision/programming interface it has, initializes basic registers and soft resets the chip. I've written in the past some code for direct PowerVR CLX2 programming on Sega Dreamcast, basic registers are quite similar but the rest is quite different sadly.

And how do you do the emulation? Are you building a software renderer that works exactly like the PowerVR? Or are you using VRAM and the registers to sniff out drawing commands and geometry/texture data, and feed that to OpenGL/D3D/Vulkan or such?

It's not done yet, but I'll just implement software rasterizer from scratch that works like PCX2 after reverse engineering meaningful registers.

There is no point in using OpenGL or another API combined with host GPU as a renderer (although it's of course possible), for this I have my wrappers that allows to make games working on modern systems and look a little better.

Tuxality wrote:

It's not done yet, but I'll just implement software rasterizer from scratch that works like PCX2 after reverse engineering meaningful registers.

Yea, would be great if it actually renders with the same quality as a real PCX2.

I'm more into wrappers than real emulation. I'm not a purist. I love wrappers like dgvoodoo or nglide, you can obtain amazing results with some old games, using some fancy effects of modern graphic cards.

New registers discovered, this time including target framebuffer phys addr and other important things. A few, simple tiles blitting test shown on image, click for full size:

Dimensions might not be correct, because it's blitted in wrong color space, but it's not that important for now. 😀

Myloch wrote:

I'm more into wrappers than real emulation.

PCem's the kind of project that can't have wrappers. Everything is an internal cpu crunch.

also powervr has some crazy depth and alpha sorting which wouldn't be easy to be accurate unless it's a wrapper done in Vulkan. The DX11 renderer in Demul is very interesting but that became closed source unfortunately, and that's the closest PowerVR-looking anything i've seen in hardware emulation to date.

My brain seeks answers...
One elderly gentleman told me that it's possible to run two Matrox m3D cards in SLI through the PCI bus and that it increases fps in games. Has anybody actually tried it? And what is the difference between Matrox m3D revisions 001A and 002?

Dreamer_of_the_past wrote:

My brain seeks answers...
One elderly gentleman told me that it's possible to run two Matrox m3D cards in SLI through the PCI bus and that it increases fps in games. Has anybody actually tried it? And what is the difference between Matrox m3D revisions 001A and 002?

I don't believe PowerVR ever developed technology like that.

gdjacobs wrote:

Dreamer_of_the_past wrote:

I don't believe PowerVR ever developed technology like that.

Fun fact: It was actually planned for the PowerVR based arcade board (which never came to marked).

But no, you can't link two PCX2 cards together.

@Tuxality Would you consider open-sourcing your code on Github or something? It could be very useful to others!

vetz wrote:

But no, you can't link two PCX2 cards together.

You wouldn't need to.
They just render one tile at a time, and then send the tile to the VRAM of another video card via a DMA bus transfer.
You could easily do this with as many PCX2 cards as you like in parallel, but of course the PCI bus would become a bottleneck at some point.

Perhaps their idea for the arcade machine was to use a different type of bus.

They eventually did have some redundant scaleability in the NAOMI though but that's crossing into Series2 territory....

elianda wrote:

I put my driver installation approach there: ftp://78.46.141.148/driver/Matrox/m3D/ read howtoinstall.txt I did this because th […]
Show full quote

I put my driver installation approach there:
ftp://78.46.141.148/driver/Matrox/m3D/
read howtoinstall.txt
I did this because the default install was rather troublesome with incompatibilities to certain games.
The dll replacing and registry settings fixed this. I don't know how other cards behave though.

The original drivers worked for me. Can you go into detail what your "installation process" fixed.

Thx!