Open-Source VIA Linux Driver Still Wants To Be Merged, But Pursuing Acceleration First

[phoronix]

Phoronix: Open-Source VIA Linux Driver Still Wants To Be Merged, But Pursuing Acceleration First

The OpenChrome DRM driver has been in development for over a decade for providing open-source display driver support for VIA's aging x86 chipsets. For years now OpenChrome development has been down to one developer left on the project, Kevin Brace, and months ago he hoped to get the driver finally merged into the Linux kernel. He still holds out on those ambitions but will first aim to get basic acceleration working with a stable user-space API/ABI before mainlining...

https://www.phoronix.com/news/OpenChrome-DRM-XDC-2022

[sravn]

The effort that Kevin put into OpenChrome development is impressive as well as his persistence - keep up the motivation!
It is obvious from the history, that Kevin nicely summarizes in the presentation, that downstream development on OpenChrome is not easy.

Accepting a split where the base part of the driver is upstream - and improved upon in upstream - would likely bring in a lot of benefits. Everything from new contributors that can help do trivial updates to refactoring that do not break the driver, to performance improvements.

With some of the optimizations done upstream we may see improvements in the performance so the base driver may be useful for most users in the end.
And with the driver upstream the user base may increase as well, so the driver see more actual use and more testing.

So is was bad news when Kevin announced that development will continue downstream due to the conflicting interest around acceleration.

[roviq]

I really appreciate the efforts put into this driver. As my first PC was a Cyrix CPU, I always followed their products and the VIA's when they acquired/merged/whatever-happened with Cyrix/IDT/Centaur.

I had a few VIA C3, C7, Nano based machines, all Mini-ITX (which was designed by VIA as well!), but the last one I had, a top-of-their-line quad core Nano, was so unbelievably slow even compared with its contemporary low end alternatives from AMD and Intel, that I decided to sell my VIA collection. Even with their own GPU drivers on Windows 10, some applications didn't work (even hardware tools like WD Dashboard for checking drive's health and some Java visualization tools) because they required an OpenGL profile (embedded? ES?) that didn't exist back then. Probably the hardware can support it and be enabled, but the driver didn't publish the capability, and that OpenGL profile seems to be used by many modern apps that require just basic 2D/3D acceleration.

Checking eBay I see very few and scattered models and Thin Clients using those CPUs, but there are plenty of VIA chipset with integrated-graphics motherboards that could benefit from this effort, but, all of it a decade or more old.

Is someone doing something similar with SiS chipsets/GPUs?

[libv]

Hrm, i wrote in 2005, after openchrome forked away because i did not want to accept VBE code back in (while i was fleshing out this useless set of concepts i erroneously called modesetting) :
"Sure, this driver currently doesn't have a few glittery bits, but ask yourself: what use are they if you are unable to display them in the first place? Will you still care about those a year or two from now? If an X developer, 10 years from now, passes over this driver to adjust it to X-side changes, what will he think about those glittery bits and will they survive?"

Whoever wants to game on VIA hw will use windows XP for the full retro experience. No-one really wants wayland or anything. If there is anyone left who still wants "openchrome" today, he needs a terminal.

[SilverBird775]

This stuff is... real old. The people of my country say the best way of preserving of your childhood home is to cherish it in your memory, not patching a rusty walls. One way or another it will keep prone to breakage because of usability obsolescence. The aged BIOS bastion barely holds on, IA-32 almost died. I'm indeed guilty for keeping the aged hardware for as long as it is possible but at least i have the latest kernel and all the browser bells and whistles on 1080p monitor. There was time i wrote a custom BIOS for my S3 trio card, patched the windows binaries to fix bugs on this card, did amazing hacks and truly enjoyed this card at fullest. But you had to let it go. When the time comes, just let it go.

[Eirikr1848]

I have two mITX boards with VIA CPUs/GPUs and have long been waiting (13+ years) to find something to do with them under Linux.

As far as repurposing old hardware is concerned, this is a win. Even if it’s just a retro emulation station: it’s running with hardware acceleration under a modern, secure kernel.

Any less waste is a good thing. Most hardware can find a use (within reason) so long as support is built in to Linux

[Svyatko]

I have two mITX boards with VIA CPUs/GPUs and have long been waiting (13+ years) to find something to do with them under Linux.

As far as repurposing old hardware is concerned, this is a win. Even if it’s just a retro emulation station: it’s running with hardware acceleration under a modern, secure kernel.

Any less waste is a good thing. Most hardware can find a use (within reason) so long as support is built in to Linux

You can buy ATI Radeon HD 2000 or 3000 series for $5-$10, and use it with open OpenGL 3.3 driver.

Meanwhile, openSUSE developers need to patch 32-bit Tumbleweed to get rid of SSE2 instructions. SSE2 instructions are very common, because they are a part of AMD64 (x86-64) standard.

[Adarion]

Kudos to Kevin. He's really persisten in saving these (old) puppies.
Yes, these machines weren't very power hungry so if they're still good they might make a fairly good purpose-box. I still do have several C7 boards around, and some thing client stuff with successors. Maybe even some old C3/CLE266 combo.

[libv]

Yes, these machines weren't very power hungry so if they're still good they might make a fairly good purpose-box.

I would not back up that power hungry statement. It might have been true in the early 2000s, just after intel and amd were having their GHz race. But compared to intel atom cpus from the mid 2010s they were unbelievably wasteful. And that's before you start looking into arm devices. Those run rings around old via hw as well.

I also have several via fanless boards which always felt too warm and which just went belly up after a while.

I think the low power thing was an attempt by VIA to find some relevance, relevance which they quickly lost with their terrible execution (both on hw and software).

[Eirikr1848]

You can buy ATI Radeon HD 2000 or 3000 series for $5-$10, and use it with open OpenGL 3.3 driver.

Meanwhile, openSUSE developers need to patch 32-bit Tumbleweed to get rid of SSE2 instructions. SSE2 instructions are very common, because they are a part of AMD64 (x86-64) standard.

The problem is additional power consumption. The CPUs I have are 15W dual-core 32-bit VIA CPUs. With a tiny industrial fanless case and tiny power supply they still have a use case either for emulation or for small industrial purposes -- especially with all the legacy ports. The issue comes with using any software that requires a GUI for legacy serial-port or parallel-port connected hardware due to the lack of acceleration.

It's still kinda useable as-is, but will be more useable once this is done.

Those legacy PCI cards cost money - even if cheap - and the goal is to repurpose existing hardware without spending additional funds, or taking up additional room with an even bigger case to accommodate a dGPU.