One of the core competences at Charonite is GPU computing and high performance computing solutions. Our highly skilled team of HPC developers have extensive experience in developing and applying GPU based solutions in a variety of industries, with a special focus on the transportation, data mining, search engine and image processing industries.
The use of GPU computing to speed up algorithms provides a low-cost way of achieving performance levels to that approaching entry-level supercomputing. The main appeal of GPU computing is the fact that the cost barrier to entry is very low: most graphic cards used by gamers already can run GPU programs and compared to the cost of specialised computing hardware, the cost is minimal.
The speedup that can be achieved by implementing a GPU version of a particular algorithm can be extremely significant - we have often achieved speedups well in excess of 100%, sometimes even in the 1000% to 5000% range! Naturally, it takes experience in both algorithm analysis, highly parallel programming and an almost obsessive attention to detail in order to achieve the best performance out of the computing hardware.
We prefer the use of the nVidia CUDA programming language as it has the best support so far in the industry (we can also provide solutions using OpenCL but this is less in demand and has less support for developers in the type of solutions we have implemented so far).
Charonite's expertise and services that are offered in the area covers various aspects including:
- Algorithm analysis and conversion to parallel mode
- Profiling algorithm performance to identify the best areas for improvement and speedup
- Extensive CUDA Programming experience
- New algorithm design and R&D
- Integration with OpenCV and OpenGL
- Use of OpenCL if necessary to use non-nVidia hardware
- Theoretical and practical aspects of formal algorithm design and development
- Application of GPU and HPC concepts in a variety of industries
- Image Processing, Data Mining and Search Engine specialisations
- Expertise in applying GPU and HPC solutions in the traffic and transport industries
 |
| Typical GPU Core used for our CUDA solutions |
Charonite has delivered a variety of data mining applications, including customer segmentation for e-Commerce applications and user segmentation in a variety of B2C sites. New modules that natively support GPUs are being added to the Obulus Platform, making it possible for Obulus applications to benefit from the GPU speedups out of the box without any GPU expertise needed on the part of the developers who develop Obulus applications.
An example of recent work that we have carried out in this area, we have recently completed some outsourcing expert work for the
ODICIS project in collaboration with
Dr. Ing. David Zammit Mangion the
University of Malta, which is an EU funded
FP7 project that aims to develop a single display cockpit by employing state of the art technologies.
Another example is the use of CUDA in our traffic analysis products, where we have received EU
ERDF funding to develop advanced image processing and computer vision algorithms that automatically track a variety of vehicles, while using low-power GPU hardware that can lead to improvements in traffic flow, monitoring and control.
Kindly contact us for more information about our GPU programming and development expertise at info@charonite.com.