Ask the Experts - ARM Fellow Jem Davies Answers Your GPU Questionsby Anand Lal Shimpi on June 30, 2014 11:52 AM EST
- Posted in
- Ask the Experts
When we ran our Ask the Experts with ARM CPU guru Peter Greenhalgh some of you had GPU questions that went unanswered. A few weeks ago we set out to address the issue and ARM came back with Jem Davies to help. Jem is an ARM Fellow and VP of Technology in the Media Processing Division, and he's responsible for setting the GPU and video technology roadmaps for the company. Jem is also responsible for advanced product development as well as technical investigations of potential ARM acquisitions. Mr. Davies holds three patents in the fields of CPU and GPU design and got his bachelor's from the University of Cambridge.
If you've got any questions about ARM's Mali GPUs (anything on the roadmap at this point), the evolution of OpenGL, GPU compute applications, video/display processors, GPU power/performance, heterogeneous compute or more feel free to ask away in the comments. Jem himself will be answering in the comments section once we get a critical mass of questions.
Post Your CommentPlease log in or sign up to comment.
View All Comments
libv - Wednesday, July 2, 2014 - linkSo when will ARM try to upstream the Mali kernel driver and get it into the linus tree?
UrQuan3 - Friday, July 25, 2014 - linkJust wanted to thank you for addressing this question. I've done quite a bit of work on various Linux/Linaro project boards and the lack of video drivers has got to be the #1 problem. At one point I found the hardware vendor was having to modify the kernal to get the driver in (placing kernal support on the HW vendor instead of the community). Not good.
I don't personally care about source code for video drivers, but we certainly need available binaries.
libv - Wednesday, July 2, 2014 - linkAt the sunxi project, we have some mali-400 binaries whose origins are a bit grey, and we provided some framework, x driver integration and packaging around it. This is, for many people trying to use a gnu/linux (as opposed to android), the canonical source for useful binary drivers for mali-400 today.
libv - Wednesday, July 2, 2014 - linkJem, as a key member of the open source community around the Allwinner SoCs, i commented (on irc and our mailing list) on Allwinner joining linaro on some TV commitee thingie back in March. Allwinner had at that time started to more proactively work with the sunxi community, and we were hopeful for getting GPL violations solved and more documentation and code available. My cynical view was that this was all about to end with Allwinner joining (a part of) linaro. And it did. Linaro ticked the "Open Source" box for allwinner marketing, and no further attempts to appropriately participate with open source communities were undertaken.
I am glad to see that you too are hiding under the linaro umbrella here.
libv - Wednesday, July 2, 2014 - linkSan_dehesa, there seems to be little point to this, as our dear guest is squarely against open sourcing the mali. Please check out the end of my lima driver talk at FOSDEM this year for more details on this, there is even a whole section in a blog entry of mine (libv.livejournal.com/24735.html) :)
san_dehesa - Monday, July 7, 2014 - linkI've already read it. Pretty interesting stuff. By the way, do you have twitter or some other social network where I can follow your progress. You don't seem to update your blog pretty often :P
Tigran - Monday, June 30, 2014 - linkHello, Jem!
It seems we have Xbox 360/PS 3 like graphical performance in some mobile GPUs. What do you think:
1) What does prevent developers from making relative games on mobile platforms?
1) When shall we see these games on smartphones and tablets?
3) Is x86 with low energy consumption a serious competitor for ARM?
JemDavies - Monday, June 30, 2014 - linkGiven the levels of performance now available, I don’t see developers being prevented from making great games on mobile platforms: both smartphones and tablets. The games companies are making some fantastic games available on these platforms right now. Of course, the performance will always be less than something plugged into the mains, liquid cooled, consuming more than 300 Watts of power, but often the games developers are producing multiple versions of games targeted at platforms of different capabilities.
Delivering the best performance in the lowest power envelope possible continues to be a guiding design principle for ARM. While we take all competition seriously, ARM does not differentiate solely on power, but also a business model that enables an open ecosystem and fuels a rapid pace of innovation among our partners.
TETRONG - Monday, June 30, 2014 - linkDoes your company have any plans to support audio acceleration?
Obviously some workloads such as FFT, Reverb, EQ, etc.. are highly amenable to parallel computation. There were some efforts to run reverbs on GPU's but they never really went anywhere in terms of commercial products.
I suppose the dream scenario would be to have access to the GPU in such a way as to have the equivalent of a Nord G2 or Symbolic Sound Pacarana right there within the tablet and without having to program in OpenCL(painful).
I do experimental sound/graphic design in different programming languages such as Max/MSP, Touch Designer, etc.. and we are always looking for more firepower and hope not to be shut out of the fun this round. The computations we need to perform are very similar to what the Financial Engineering/Data Science community is looking for.. we make use of all the same mathematics in algorithmic/generative/procedural composition and visualization.
Not so much a question..just wanted to raise my hand and say - Yes, we are out here..in droves and want what you're selling!
JemDavies - Monday, June 30, 2014 - linkAt ARM, we’re always looking at ways of doing things in a more efficient (lower power or energy consumption) way. It is a digital world, and most problems can be expressed in terms of 1s, 0s and algorithms. Our CPUs are designed to be as efficient as possible at the vast range of applications that get thrown at them but some applications are sufficiently different that we think there is market there for a domain-specific processor. A GPU is one example of that (as graphics is specialised) but we also produce video processors to do decode and encode of digital video.
GPU Compute is useful for a number of applications that can be expressed as massively parallel algorithms, and there are already use cases in mobile devices where our GPUs are being used to accelerate use cases. I do believe some people are already doing research into using GPUs for audio processing, and you’re not being shut out at all – quite the contrary. The commonly-used languages for this are currently OpenCL and Google’s RenderScript.