Main Menu
About us
Project Description
Quantitative Results
Research Lines
Research Results
Impact on Society
Press room
Contact us
Secure Login
Events Calendar
« < April 2019 > »
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 1 2 3 4 5

subTask 1.6 (UM): Application development on CMPs

Leader: José Manuel García

1. Brief Description of the Goals

Graphics Processing Units (GPUs) have been at the leading edge of increasing chip-level parallelism over the last decade, evolving into sophisticated massively parallel co-processors, fully programmable for throughput-oriented, data-parallel workloads. GPUs typically provide about one order of magnitude speedup over CPU counterpart versions, and in special cases, yielding speedups of several orders of magnitude.

The advent of programming models such as CUDA™ or OpenCL has facilitated the application of GPUs to many real-world computational domains. However, developers have to deal with a new parallel programming paradigm that is quite different than the traditional ones, and therefore, applications need to be redefined and even redesigned to leverage all GPU capabilities.

The use of GPU devices is becoming increasingly widespread as an effective way of tackling computational problems with high performance requirements, with applications covering many different subject areas. However, the software used on GPU devices requires significant effort and knowledge to produce fully optimised implementations that are capable of fulfilling the computational potential. Developing an understanding of how to make best use of these devices is an on-going research question.

2. Scientific and Technical Developed Activities

The work developed in this task covers the use of GPU devices in the field of general purpose, high performance computing. The work carried out in this task offers three examples of biological-inspired applications in which the use of GPU computing has been successfully implemented: Membrane computing, Swarm intelligence, and Virtual Screening.

The two first applications, linked to the emerging field of Novel Computing, benefit from the innate parallelism of biologically inspired models and are, therefore, interesting candidates for computation on GPU devices. In each case, we have identified several fruitful paths for optimisation of each algorithm and have successfully benchmarked the resulting performance. In breaking down the algorithm into component stages, we have shown how the GPU device can be used most effectively to give the best possible performance. This has resulted in several novel approaches to implementing the GPU based solutions, opening a new window for high-performance applications in the Novel Computing field. 

The third application, Virtual Screening, comes from Bioinformatics field. VS methods can considerably aid clinical research, predicting how ligands interact with drug targets. Most VS methods suppose a unique binding site for the target, usually derived from the interpretation of the protein crystal structure. However, it has been demonstrated that in many cases, diverse ligands interact with unrelated parts of the target and many VS methods do not take into account this relevant fact. We have developed a novel VS methodology called BINDSURF that scans the whole protein surface in order to find new hotspots, where ligands might potentially interact with, and which is implemented on the Nvidia’s Tesla C2050, codename “Fermi”, GPU architecture using CUDA. Additional features, such as implicit solvation models in the scoring function, must be included to improve the accuracy prediction of our high-throughput blind docking methodology BINDSURF, thus increasing exponentially the computational requirements. We have developed a novel method for those calculations called MURCIA, which is at the moment one of the fastest methods in the range 10 – 17000 atoms and which also provides information for the visualization of molecular surfaces in standard molecular graphics program like VMD, Pymol and Chimera.

Publications: [Cecilia09a]  

Projects funded by Public Calls: 

External collaborations Academia: Manuel UjaldonMartyn Amos.

External collaborations Industry: --

Company Agreements: --

PhD dissertations: --

Patents: --