Current Issue : January - March Volume : 2013 Issue Number : 1 Articles : 4 Articles
We propose a fast data relay (FDR) mechanism to enhance existing CGRA (coarse-grained reconfigurable architecture). FDR\r\ncan not only provide multicycle data transmission in concurrent with computations but also convert resource-demanding\r\ninter-processing-element global data accesses into local data accesses to avoid communication congestion. We also propose\r\nthe supporting compiler techniques that can efficiently utilize the FDR feature to achieve higher performance for a variety of\r\napplications. Our results on FDR-based CGRA are compared with two other works in this field: ADRES and RCP. Experimental\r\nresults for various multimedia applications show that FDR combined with the new compiler deliver up to 29% and 21% higher\r\nperformance than ADRES and RCP, respectively....
The Viterbi algorithm is one of the most used dynamic programming algorithms for protein comparison and identification, based\r\non hidden markov Models (HMMs). Most of the works in the literature focus on the implementation of hardware accelerators\r\nthat act as a prefilter stage in the comparison process. This stage discards poorly aligned sequences with a low similarity score and\r\nforwards sequences with good similarity scores to software, where they are reprocessed to generate the sequence alignment. In\r\norder to reduce the software reprocessing time, this work proposes a hardware accelerator for the Viterbi algorithm which includes\r\nthe concept of divergence, in which the region of interest of the dynamic programming matrices is delimited. We obtained gains\r\nof up to 182x when compared to unaccelerated software. The performance measurement methodology adopted in this work takes\r\ninto account not only the acceleration achieved by the hardware but also the reprocessing software stage required to generate the\r\nalignment....
Performing runtime evaluation together with design time exploration enables a system to be more efficient in terms of various\r\ndesign constraints, such as performance, chip area, and power consumption. rSesame is a generic modeling and simulation\r\nframework, which can explore and evaluate reconfigurable systems at both design time and runtime. In this paper, we use the\r\nrSesame framework to perform a thorough evaluation (at design time and at runtime) of various task mapping heuristics from\r\nthe state of the art. An extended Motion-JPEG (MJPEG) application is mapped, using the different heuristics, on a reconfigurable\r\narchitecture, where different Field Programmable Gate Array (FPGA) resources and various nonfunctional design parameters,\r\nsuch as the execution time, the number of reconfigurations, the area usage, reusability efficiency, and other parameters, are taken\r\ninto consideration. The experimental results suggest that such an extensive evaluation can provide a useful insight both into the\r\ncharacteristics of the reconfigurable architecture and on the efficiency of the task mapping....
The nature of modern astronomy means that a number of interesting problems exhibit a substantial computational bound and\r\nthis situation is gradually worsening. Scientists, increasingly fighting for valuable resources on conventional high-performance\r\ncomputing (HPC) facilitiesââ?¬â?often with a limited customizable user environmentââ?¬â?are increasingly looking to hardware acceleration\r\nsolutions. We describe here a heterogeneous CPU/GPGPU/FPGA desktop computing system (the ââ?¬Å?Chimeraââ?¬Â), built with\r\ncommercial-off-the-shelf components. We show that this platform may be a viable alternative solution to many common computationally\r\nbound problems found in astronomy, however, not without significant challenges. The most significant bottleneck\r\nin pipelines involving real data is most likely to be the interconnect (in this case the PCI Express bus residing on the CPU\r\nmotherboard). Finally, we speculate on the merits of our Chimera system on the entire landscape of parallel computing, through\r\nthe analysis of representative problems from UC Berkeleyââ?¬â?¢s ââ?¬Å?Thirteen Dwarves.ââ?¬Â...
Loading....