Bookmark and Share


Nowadays application processors for mobile devices pack a lot of various hardware, including general-purpose processing cores, graphics cores, image processors, controllers and so on. End-users demand improvements in all aspects of usage, hence mobile chips are getting incredibly complex. With the rise of system-on-chips powered by modern graphics architectures that support GPGPU technology, the need for many special-purpose components will vanish, but the doors to new features will open.

Thanks to the fact that contemporary graphics processing units (GPUs) feature a relatively high number of comparatively simplistic processing units, it is possible to use them for highly-parallel general-purpose (GP) tasks. Such usage model is called GPGPU and it is on the rise on personal computer these days. Thanks to the fact that the latest ARM Mali T600, Imagination PowerVR 6 and Qualcomm Adreno 300 support GPGPU through OpenCL application programming interface, they can be used to make better photos of videos. Provided that the GPUs are not too power hungry, their usage eliminates necessity to integrate special-purpose hardware, such as image processors or video stabilizers, thus, reducing cost of the chip.

“You might be able to save some cost somewhere in the SoC by cutting out a bit of hardware that you had and run it on the GPU instead. So cost reduction and an improved capability. So maybe they will cut out some of the ISP and they will do it on the GPU because the silicon is already there, it is power efficient, it's a quicker way of doing [it], you get a cost reduction, performance goes up,” said Ian Smythe, director of marketing at ARM, in an interview with The Inquirer web-site.

Alternatively, GPGPU technology can be utilized to process demanding desktop-class applications on mobile devices, such as smartphones or tablets, a clearly interesting opportunity.

The main question is when the software developers actually start to take advantage of OpenCL and GPGPU technologies. At present there are not a lot of devices based on the latest graphics architectures, so there are no immediate financial gains because of the implementation. However, a year from now hundreds of millions of devices will use ARM Mali T600, Imagination PowerVR 6 or Qualcomm Adreno 300 graphics cores and all of them will be ready for a new generation of apps. It looks like software developers have to start the work now.

Tags: ARM, Imagination, Qualcomm, Mali, PowerVR, Adreno, GPGPU


Comments currently: 5
Discussion started: 03/28/13 10:27:10 PM
Latest comment: 03/29/13 10:35:06 AM
Expand all threads | Collapse all threads


The main question is when the software developers actually start to take advantage of OpenCL and GPGPU technologies.

That's the wrong question. The right question is how can the industry evolve HSA so that all/most of the technical intricacies are transparent to software developers. As a simple example, today a program may use integers and floats. The developer just uses what program requires, and doesn't even think about the fact that floating point calculations are routed to the floating point unit while integer calculations are handled by the integer units. The compiler takes care of those details.

We need the same thing for HSA. The developer should be able to focus on what algorithms are needed to address a problem. The tools (compiler or maybe some higher level intermediate tool between the source code and the compiler) should determine whether that algorithm is best handled by what today we consider the CPU or the GPU. In the future, that distinction will dissolve just like it did once the floating point unit was merged into the CPU.
0 0 [Posted by: guyr  | Date: 03/28/13 10:27:10 PM]
- collapse thread

wouldn't a better method be that we make the languish higher again. so if you use the function (in java) array.sort() on a large (>=10000) int array than it would start using a multitreaded sort. than if you add or multiply a very large float/ double array then it would use opencl. like array.add() (or matrix.add()).

this would also allow better vector processor optimisations. so instead of going down and durty to get 90% scaling let the standard library do the work.
0 0 [Posted by: massau  | Date: 03/29/13 02:23:29 AM]
*HSA is an open standard
*I also understand that in the AMD Excavator core due early 2015, the APU will decide itself whether to offload calculations onto the CPU or GPU without special coding hooks.

Further reference:
1 0 [Posted by: linuxlowdown  | Date: 03/29/13 02:28:10 AM]

I like the way the whole industry is moving towards GPGPU tech. GPU is simply a powerful magic silicon that can do wonders. The most funny fact is that ego driven Intel is getting cornered day by day by its x86 tech and unwilling to admit GPU's processing power..........
1 0 [Posted by: tks  | Date: 03/29/13 04:44:36 AM]
- collapse thread

isn't the main problem whit the cpu that the cache is becoming so large that there is more power and die used for cache than logic.

while the gpu has many small and slower cores but this enables high yields. because you could have a gpu whit 100 "bad cores" and still sell it as a lower card.

intel tried the vliw (EPIC itanium) ach but failed because of the compiler.
0 0 [Posted by: massau  | Date: 03/29/13 10:35:06 AM]


Add your Comment

Related news

Latest News

Wednesday, November 5, 2014

10:48 pm | LG’s Unique Ultra-Wide Curved 34” Display Finally Hits the Market. LG 34UC97 Available in the U.S. and the U.K.

Wednesday, October 8, 2014

12:52 pm | Lisa Su Appointed as New CEO of Advanced Micro Devices. Rory Read Steps Down, Lisa Su Becomes New CEO of AMD

Thursday, August 28, 2014

4:22 am | AMD Has No Plans to Reconsider Recommended Prices of Radeon R9 Graphics Cards. AMD Will Not Lower Recommended Prices of Radeon R9 Graphics Solutions

Wednesday, August 27, 2014

1:09 pm | Samsung Begins to Produce 2.13GHz 64GB DDR4 Memory Modules. Samsung Uses TSV DRAMs for 64GB DDR4 RDIMMs

Tuesday, August 26, 2014

10:41 am | AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU. AMD Unleashes Radeon R9 285 Graphics Cards, Tonga GPU, GCN 1.2 Architecture