Samsung Improving Cairo's OpenGL ES 3.x Support, May Eye Vulkan In Future

[phoronix]

Phoronix: Samsung Improving Cairo's OpenGL ES 3.x Support, May Eye Vulkan In Future

Back in September there were developers from Samsung's Open-Source Group adding initial OpenGL ES 3.0 support to Cairo. The GLESv3 upbringing in Cairo is still ongoing and not yet fully vetted, but Bryce Harrington of Samsung OSG has now blogged about this effort...

http://www.phoronix.com/scan.php?page=news_item&px=Cairo-GLESv3-Work-OSG

[schmidtbag]

It's not that I don't appreciate Samsung's involvement, but I would much rather they focus on improving upon the GPU drivers they use in their Exynos series. They have the money and resources for it, and devices containing Exynos CPUs are are very appealing for Linux use.

[starshipeleven]

It's not that I don't appreciate Samsung's involvement, but I would much rather they focus on improving upon the GPU drivers they use in their Exynos series. They have the money and resources for it, and devices containing Exynos CPUs are are very appealing for Linux use.

They're using Mali or PowerVR, so they can't.

[schmidtbag]

They're using Mali or PowerVR, so they can't.

Can't, or won't? Seeing as Samsung is one of (if not the) largest distributors of these GPUs, I think ARM/SoftBank or Imagination would make a deal with them.

[starshipeleven]

Can't, or won't? Seeing as Samsung is one of (if not the) largest distributors of these GPUs, I think ARM/SoftBank or Imagination would make a deal with them.

That's not how it works.

ARM sells his GPU licenses and SDKs pretty well even without Samsung asking them a price to get access to the sources and license to use/improve them for their own products, and there isn't much competition (it's already the lower cost crap they can find) so Samsung can't say "yo I'll use a competitor's GPU if you don't let me".

They could flat-out refuse their offer and still sleep fine at night.

This means that the ones dictating price and conditions will be ARM, not Samsung.

Really, it would be MUCH easier and less expensive for Samsung to just reverse-engineer an opensource driver.

[schmidtbag]

ARM sells his GPU licenses and SDKs pretty well even without Samsung asking them a price to get access to the sources and license to use/improve them for their own products, and there isn't much competition (it's already the lower cost crap they can find) so Samsung can't say "yo I'll use a competitor's GPU if you don't let me".

That doesn't really relate to my point. I'm not talking about financing and competition. I'm thinking more in terms of a joint development where both companies get something out of it. Kind of like DX12 - it's MS's product but Intel, AMD, Nvidia, etc all help contributed toward its development. Does that make more sense?

[starshipeleven]

That doesn't really relate to my point. I'm not talking about financing and competition. I'm thinking more in terms of a joint development where both companies get something out of it.

This is an alien concept in the closed-source, intellectual-property-based corporate world.

And I repeat what I said, Samsung is not in a strong position in such a deal so they will get gouged or can even get a "no" answer. And for good reasons.

Imagination's anus is still bleeding after Apple pulled that trick with them, they had some kind of partnership, then Apple cut ties when they had "stolen" enough IP or trained enough internal engineers to be able to go on on their own.

There is no guarantee that Samsung will play a fair game, as Samsung is orders of magnitude bigger than either ARM holdings or Imagination, and just as Apple they do have their own internal engineering team designing silicon (that made the Exynos SoC) they could easily take the IP they need and run, just as Apple did.

So, for ARM holdings we have a "potential better driver" (which does not translate much in higher sales because they make cheap shit for embedded, not high-performance GPUs) and "relatively good source of $$$" (due to new licensing agreements with Samsung), vs "risk of losing your IP" and "risk of creating a much stronger potential competitor".

So it's not that likely they will allow Samsung to "help them" for cheap, if at all.

Seriously, an ARM spokeperson even said they don't have open drivers because they are afraid of people stealing their IP if they do so.

Kind of like DX12 - it's MS's product but Intel, AMD, Nvidia, etc all help contributed toward its development. Does that make more sense?

No, DX12 is MS's puppy and their puppy only. None really contributed to its development, and hardware manufacturers that don't conform to its diktats can stfu and git gud, as usual with Microsoft's standards.
If they can't get their hardware to run decently on Windows they can kiss goodbye all the PC consumer market, and for AMD and NVIDIA it still does matter.

Even if I correct "DX12" into "Vulkan" (which is technically more fitting, as it was made by an organization all major GPU manufacturers are a part of) it is an apples and bananas comparison. Since we are talking of drafting an open standard (= 0 risk of IP theft) and not writing a driver for a specific IP design.

[coder]

Kind of like DX12 - it's MS's product but Intel, AMD, Nvidia, etc all help contributed toward its development. Does that make more sense?

Seems to me it took AMD making Mantle and Khronos starting work on Vulkan before MS got off its big fat posterior to make DX12.

Imagination is probably in a place where someone with some real leverage could push them to open source their driver, but it'd have to be like a Samsung. And I don't see what's in it for Samsung.

[schmidtbag]

Seems to me it took AMD making Mantle and Khronos starting work on Vulkan before MS got off its big fat posterior to make DX12.

I'm aware of this, but that's besides my point.

Imagination is probably in a place where someone with some real leverage could push them to open source their driver, but it'd have to be like a Samsung. And I don't see what's in it for Samsung.

It's simple: more customers. Linux is popular on ARM despite all the shortcomings, but it loses a lot of popularity thanks to GPU driver issues. GPU drivers seem to be the #1 complaint that I've noticed. Exynos-based platforms would gain a lot of popularity if they could function as a complete desktop or laptop replacement. The main reason why the Raspberry Pi has been so popular was because of it's well-functioning hardware. Broadcom's hardware isn't that impressive, but it doesn't need to be since it has proper hardware acceleration.

And sure, Samsung would be helping out companies like Allwinner and Rockchip too, but they're usually only found in very low-end products; it wouldn't be much of a loss to Samsung.

[coder]

The main reason why the Raspberry Pi has been so popular was because of it's well-functioning hardware. Broadcom's hardware isn't that impressive, but it doesn't need to be since it has proper hardware acceleration.

Perhaps, but given that the low-cost (i.e. low-margin) Pi only shipped 10 M units as of Sept 2016, it's barely a drop in the bucket compared to their smartphone sales. And I don't see (but maybe I'm wrong) a lot of people using those Broadcom SoC's outside of the Pi.

Anyway, maybe the RISC-V folks will turn their attention to GPUs, next. I wouldn't hold my breath, but we can hope.