Helping Valve to Power Up Steam Devices
80 points by bkardell
80 points by bkardell
Excellent to see that Valve is using Turnip (/ Freedreno) instead of Qualcomm's proprietary drivers (which last I checked require signing an NDA to donwload). I'm sure this gives them more design flexibility but it's also great for users.
Does the open source driver have Qualcomm's official blessing, or is it just "allowed to exist" by them? If the proprietary driver requires a NDA to download, I am kind of surprised Qualcomm would just let Valve use the open source one. Shows how much power Valve holds, I guess?
It's reverse engineered. The only leverage Qualcomm has here is in letting Valve actually buy the chips, and presumably they got that in writing before they chose this chip. Valve has competent lawyers.
It has been historically reverse engineered, but these days Qualcomm has paid employees/contractors working on Mesa Turnip/Freedreno (and the kernel drm/msm side) full time. Most recently and famously, Rob Clark (who was previously on the Chromium OS team).
Wild to see the Steam ecosystem turn into an open source funding engine that lifts all boats.
Valve has, for a while now, been funding some open source projects that are relevant to videogames. If I remember correctly, SDL2 was largely funded by Valve.
I didn't realize that Igalia was working with Valve on SteamOS stuff. It's really sick seeing some of my favorite people working on some of my favorite software!
Cool company, had not heard of them before
Igalia does impressive work on projects, and the organisation structure is also interesting. It's a worker cooperative!
Valve is meant to be a flat company, but you've still got Gabe Newell gathering enough to own a fleet of yachts and submarines. Valve brings in over $3.5 million per employee, way above other big tech companies.
It's great that they're working together. Really interesting work and a mildly interesting contrast in ownership models.
Everyone knows
People who are really serious about software should make their own hardware.
but in my opinion Valve shows
People who are really serious about software should make their own open source drivers (for their hardware).
It just give you so much leverage to develop, test & optimize everything high up to deep down in the stack.
Under normal circumstances, this would mean that only games compiled to run on ARM chips could be played on the Frame. In order to get around this barrier, a translation layer called FEX is used to run applications compiled for x86 chips (which are used in nearly all gaming PCs) on ARM chips by translating the x86 machine code into ARM64 machine code.
I'm surprised AMD nor Intel don't seem to try enforcing the patents that supposedly guard most of X86_64 implementation work.
Has there been any recent analysis of relevant patents? AMD64 is now old enough that any patens for the original should have expired, but I recognize modern processors have a lot more than what AMD64 offered (AVX2, etc.).
I also wonder if some related patents wouldn't apply to a software implementation, since presumably the patents would have been focused on hardware implementations.
I also wonder if some related patents wouldn't apply to a software implementation, since presumably the patents would have been focused on hardware implementations.
That's my first thought too, though courts have been known to make blatantly wrong decisions about applying hardware patents to software implementation. The one that comes to mind is the patent on implementing unaligned loads&stores in hardware on MIPS-I being abused to extract money from Lexra who built a CPU that did not implement those instructions at all on the basis that they could be emulated in software. (Caveat: I'm not sure that was actually a bad court decision so much as Lexra being forced into an unfavourable settlement because it didn't have a big enough cash reserve to fight a court case that it would probably have won.)