Eric Anholt Leaves Intel's Linux Graphics Team For Broadcom

[Drago]

Either way its a good sign for the future of Mesa.

I hope he will use Gallium this time.

[blackiwid]

1. the reason for the workplace change, he not even claimed but it could look like he searched a new interesting challenge.
I woul doubt that, the reason I would bet my money on is that, most of the time you cant get a higher salary without changing the employer, so he switched to broadcom.

I find it strange that intel really pays not as good as broadcom, but it seems thats the case.

2. I hate it:

2.1 because RPi is a very slow minimalistic hardware, its now soon 2 years old (version 2) it was very weak processor even for arm socks at the time it got to market. so yes for the old owners such a driver might be interesting but who will buy such bad hardware in now or in 1-2 years? When that driver is ready nobody will buy this pecase of hardware anymore.

2.2 Its one of the most unfree solutions out there, not becuase of the software-stack but because of the gpu-hardware, with basicly moving the half os into the firmware. look at debian statement they tell u not to buy rpi, and that it was just a bad hype.
If only a missing free driver would be the problem it would be great but that alone cant fix the many problems with that hardware.

2.3 rPi is basicly a own arm-standard: gpu-driver problems with other socs aside, there are already ARM-images of armhfp / arm7 images of many distros in the long run it will be availibly for most distros, like debian, like fedora... they are basicly the x86 images of the arm world, and will work with more or less every arm soc except rPi. So supporting this hardware with special images are extremly time consuming.

3. so if that rpi gets no BIG update with a way better soc but the same peace of crap as gpu, it would be better to not invest to much time of good people to invest in that hopefully in the mid/long term dieing plattform. And if it gets a big speed/feature update with the same unfree gpu, it would be even worse, it could become the new nvidia standard,

4. and even worse, devoloper have to invest even more time to make a good driver with this hardware, becuase they have to make workarounds around that bad gpu design, to make suff like gallium3d or even stupid normal x-drivers work with that hardware.

its just a big waste of sources, when there are way better socs out there which maybe as one soc has not so much hype but all are at least the same arm-standard and all are compatible except the gpu drivers, and even there many share the same ones.

5. I want ONE download for arm k differenciation between 32bit and 64bit ok, but who the hell wants a dell-inspiron-5877-x86-incompatible-special.iso to see on every os, and lets think that further, in the end u have for 1000 pcs out there 1000 different isos for every distro, what a waste of resources.

[caligula]

1. the reason for the workplace change, he not even claimed but it could look like he searched a new interesting challenge.
I woul doubt that, the reason I would bet my money on is that, most of the time you cant get a higher salary without changing the employer, so he switched to broadcom.

I find it strange that intel really pays not as good as broadcom, but it seems thats the case.

2. I hate it:

2.1 because RPi is a very slow minimalistic hardware, its now soon 2 years old (version 2) it was very weak processor even for arm socks at the time it got to market. so yes for the old owners such a driver might be interesting but who will buy such bad hardware in now or in 1-2 years? When that driver is ready nobody will buy this pecase of hardware anymore.

2.2 Its one of the most unfree solutions out there, not becuase of the software-stack but because of the gpu-hardware, with basicly moving the half os into the firmware. look at debian statement they tell u not to buy rpi, and that it was just a bad hype.
If only a missing free driver would be the problem it would be great but that alone cant fix the many problems with that hardware.

2.3 rPi is basicly a own arm-standard: gpu-driver problems with other socs aside, there are already ARM-images of armhfp / arm7 images of many distros in the long run it will be availibly for most distros, like debian, like fedora... they are basicly the x86 images of the arm world, and will work with more or less every arm soc except rPi. So supporting this hardware with special images are extremly time consuming.

3. so if that rpi gets no BIG update with a way better soc but the same peace of crap as gpu, it would be better to not invest to much time of good people to invest in that hopefully in the mid/long term dieing plattform. And if it gets a big speed/feature update with the same unfree gpu, it would be even worse, it could become the new nvidia standard,

4. and even worse, devoloper have to invest even more time to make a good driver with this hardware, becuase they have to make workarounds around that bad gpu design, to make suff like gallium3d or even stupid normal x-drivers work with that hardware.

its just a big waste of sources, when there are way better socs out there which maybe as one soc has not so much hype but all are at least the same arm-standard and all are compatible except the gpu drivers, and even there many share the same ones.

5. I want ONE download for arm k differenciation between 32bit and 64bit ok, but who the hell wants a dell-inspiron-5877-x86-incompatible-special.iso to see on every os, and lets think that further, in the end u have for 1000 pcs out there 1000 different isos for every distro, what a waste of resources.

There is still one good reason to use rpi. It's the cheapest. The better boards are $10 more expensive. Developer time is worth 0$ in open source. So it makes sense to spend years of work hours on rpi to make it work. It's also the most hyped de facto standard for these domains. So people want to support rpi not better boards. People even build computational clustets with 10x rpi when a single $40 board is faster than 10 x $25 rpi boards. :-)

[Veerappan]

I hope he will use Gallium this time.

One can hope...

I guess it's feasible that he'll go that route. Intel's big reason was always that they had an existing classic driver that was highly optimized and that they didn't want to spend time switching over to something that might not gain them any performance (or might end up losing them some). In the case of the Broadcom chip, the existing driver is there, but I don't believe that it's running off Mesa at all. Which means that Eric is going to have to build large parts of a new driver from scratch and gallium could end up being a time-saver here.

We know that Gallium *can* work for ARM-based GPUs like Freedreno (hi Rob!). The question now becomes whether it's going to save enough time/complexity and whether Eric's comfortable going that route (due to knowledge and theoretical performance).

This story is a good reminder that I've been planning to buy one of the Pis ever since the GPU code drop happened a few months ago. I need something small to tinker with.

[brent]

Well, the Raspberry Pi, despite somewhat sucky hardware and closed-source-ness, has *by far* the best support of any ARM development/hobbyist board. There is a team of developers working on a software stack and they actually fix bugs, try to make new things work, update to new kernels, and so on. No other ARM board has this level of support, not even close.

USB should also work quite well now, at this point they have basically rewritten the dwc_otg driver. All serious issues have been fixed and performance is better.

[curaga]

Either way its a good sign for the future of Mesa.

How so? If there indeed is dev poaching between mesa-contributing companies, that affects morale and creates an unhealthy environment. I don't see how it's a good sign for the future.

[ssvb]

How so? If there indeed is dev poaching between mesa-contributing companies, that affects morale and creates an unhealthy environment. I don't see how it's a good sign for the future.

Intel is already way too overrepresented in the OSS graphics space. Rebalancing the efforts and also taking care of the other hardware is good

[Kivada]

Intel is already way too overrepresented in the OSS graphics space. Rebalancing the efforts and also taking care of the other hardware is good

Yeah, and ARM GPUs are way under represented when it comes to OSS drivers, this is actually a great change. It was looking like the best hope for an ARM SoC with fully working OSS drivers was going to be AMD whenever they got around to releasing their Seattle server SoCs.

[blackiwid]

Well, the Raspberry Pi, despite somewhat sucky hardware and closed-source-ness, has *by far* the best support of any ARM development/hobbyist board. There is a team of developers working on a software stack and they actually fix bugs, try to make new things work, update to new kernels, and so on. No other ARM board has this level of support, not even close.

USB should also work quite well now, at this point they have basically rewritten the dwc_otg driver. All serious issues have been fixed and performance is better.

Thats the current state but in the future it can come fast to the oposite:

FEDORA:

http://fedoraproject.org/en/get-fedora#arm

what do we have here, 7 OFFICIAL spins for the general arm standard (armhf or arm7) and 1 UN-official Remix.

DEBIAN:

http://cdimage.debian.org/debian-cd/7.5.0/armhf/iso-cd/

same here, 4-5 spins only unofficial versions of raspbian are there.


The Problem is that, even if they would release also the boot blobs and so on, and all gets included to upstream to kernel and so on, they always will have to use this thirdparty things.

And this guys have to fork than basicly every single distro or at least the biggest ones. how much waste of time. And we see how fast we see drivers and stuff for linux hardware gpus and even how bad the linux support for tablets is.

You also will then only get later images of distros, because they will not get released by the distros itself, so some weeks maybe even longer later u get then finaly a image. And I even doubt that u just can easily apt-get dist-upgrade with this images.

We see on the tablet front how horrific the situation is, we see that much from technicaly people loved and hyped products like nexus 7 2013 is imposible to install a real linux onto it 1 year after release. Not even with proprietary blobs.

We need a standard and official support for arm from the distros. Everything else sucks, then u maybe can hope 10 years that maybe ubuntu makes a image for u, and then u have to use ubuntu because it only works with ubuntu etc.

I dont want a androidisierung of GNU/linux, seen android hate it, done with it.

And to the performance yes they can maybe workaround or fix some issues there, but they cant fix that this soc has only 1 Core with 700mhz and only 512mb max. And because its has no full arm7 instruction set its IPC is lower than other socs.



AND another point, we see on android that today no technical guy dont pays 50,- Euro more if he gets better software support, like with the nexus devices. Because else they are damned to to get their bad support of the vendor. And they can force users to buy new ones in not giving them the updates to newer android versions.

So no development time in opensource projekts are not for free they matter much. The biggest difference-point between a good soc arm device and a bad is today the software-support.


Is the rPi good enough for some tasks, maybe yes, will it get much better no. Yes of course they get maybe then 30fps in glxgears instead of 5, 2d performace is then maybe 10% faster in X or whatever shure. But u have to wait some months have to fiddle around with downloading fresh images from some bogus sites, invest some time, and in the end u are already near the top... while with a 2core 1.4ghz soc with 2gb ram u have 200-500% mehr speed in some cases. Without touching even one line of code. for 20,- Euro more... one less sd-card could have saved u the cost.

Again I understand people that bought that hardware at the beginning there were not much good alternatives. But I find it such a waste of resources to bring there 10% more speed on that crap-hardware instead of investing that in long term stuff.

As example lets take the adriano gpus (or how u spell it) there will be newer adreanos that will share much of the design, so your develop for a driver which comes years and years after it newer versions.

How much Broadcom-gpus will we see in future, hopfefully none... on tablets on smartphones as example? None? good. so let it slowly die and dont invest so much effort in such crap.

And thats another argument, yes for caseless socs rpi got sold often, but if you count in tablets and phones all the other socs / gpus are around much more people. and even if only a few % of them want to install a real linux on it. thats still more than rpi only sold devices.

Rpis sold so far 2,5 mio, adreno gpus are in every 3rd smartphone and tablet. Its hard to find good numbers but I found as example that: "http://www.androidauthority.com/next-gen-nexus-7-shipments-190788/" 7-8 mio expected and it got very good reviews I pretty shure it hit that mark easily.


So yes its only one person, but there arent hundret of thousends good GPU driver programmers out there so every single one vasted is one to much.

[robclark]

1. the reason for the workplace change, he not even claimed but it could look like he searched a new interesting challenge.
I woul doubt that, the reason I would bet my money on is that, most of the time you cant get a higher salary without changing the employer, so he switched to broadcom.

I find it strange that intel really pays not as good as broadcom, but it seems thats the case.

2. I hate it:

2.1 because RPi is a very slow minimalistic hardware, its now soon 2 years old (version 2) it was very weak processor even for arm socks at the time it got to market. so yes for the old owners such a driver might be interesting but who will buy such bad hardware in now or in 1-2 years? When that driver is ready nobody will buy this pecase of hardware anymore.

I guess it is safe to assume that whatever newer/faster next-generation SoC broadcom is cooking up will have a gpu that is based on what is in rpi.

2.2 Its one of the most unfree solutions out there, not becuase of the software-stack but because of the gpu-hardware, with basicly moving the half os into the firmware. look at debian statement they tell u not to buy rpi, and that it was just a bad hype.
If only a missing free driver would be the problem it would be great but that alone cant fix the many problems with that hardware.

Well, this is actually moving the gpu driver to the arm side, so it is a step in the right direction.

Plus various other goings-on.. like bcom joining linaro, and I know they've been hiring linux dev's.. so they seem to be trying to do the right thing. So I have some optimism that some future generation of RPi could be a very nice little board.

2.3 rPi is basicly a own arm-standard: gpu-driver problems with other socs aside, there are already ARM-images of armhfp / arm7 images of many distros in the long run it will be availibly for most distros, like debian, like fedora... they are basicly the x86 images of the arm world, and will work with more or less every arm soc except rPi. So supporting this hardware with special images are extremly time consuming.

3. so if that rpi gets no BIG update with a way better soc but the same peace of crap as gpu, it would be better to not invest to much time of good people to invest in that hopefully in the mid/long term dieing plattform. And if it gets a big speed/feature update with the same unfree gpu, it would be even worse, it could become the new nvidia standard,

4. and even worse, devoloper have to invest even more time to make a good driver with this hardware, becuase they have to make workarounds around that bad gpu design, to make suff like gallium3d or even stupid normal x-drivers work with that hardware.

AFAICT most of the issues w/ RPi GPU are really just software issues.. throwing away their existing driver, and writing a gallium + drm driver is actually the perfect solution.

its just a big waste of sources, when there are way better socs out there which maybe as one soc has not so much hype but all are at least the same arm-standard and all are compatible except the gpu drivers, and even there many share the same ones.

well, hopefully other mobile gpu vendors are paying attention, and reconsidering their position wrt. FOSS graphics drivers. But at the end of the day, one paid FOSS driver dev for a mobile gpu vendor is better than zero. Of course I hope that more gpu vendors jump in the game and start sponsoring FOSS driver development (or even just releasing docs!!), but there is really nothing to critisise here.

5. I want ONE download for arm k differenciation between 32bit and 64bit ok, but who the hell wants a dell-inspiron-5877-x86-incompatible-special.iso to see on every os, and lets think that further, in the end u have for 1000 pcs out there 1000 different isos for every distro, what a waste of resources.