Skip to Main Content
 

Global Search Box

 
 
 
 

ETD Abstract Container

Abstract Header

COMPARISON OF THE PERFORMANCE OF NVIDIA ACCELERATORS WITH SIMD AND ASSOCIATIVE PROCESSORS ON REAL-TIME APPLICATIONS

Abstract Details

, MS, Kent State University, College of Arts and Sciences / Department of Computer Science.
Basic tasks for Air Traffic Control will be implemented using NVIDIA’s CUDA language on a NVIDIA device and compared to the performance of an Associative SIMD processor doing the same tasks. To do this, we create a simulation of an airfield with constantly moving aircrafts. The tasks that will be used in the evaluation are: tracking and correlation, collision detection, and collision resolution. These are the most compute intensive of the Air Traffic Control tasks, so they will give us a good measure of the capabilities of the NVIDIA device. The first task is tracking and correlation of the aircrafts in a 256 nautical mile by 256 nautical mile bounding area on a 2D plane with varying altitudes. This task is executed once each half second during each 8 second major cycle period and uses radar to correlate the exact location of the aircraft and its flight records. During every 8 second cycle, Batcher’s algorithm is used to check if any aircraft’s projected path has a possibility for collision. If a potential collision is possible within the next 20 minutes, we first locate a collision free path for one of them and then have it switch to this path. In previous research, the ability of a multicore system to perform basic ATC tasks was investigated. The graph showing its performance increased rapidly as the number of aircraft increased, which is consistent with the general belief that all large real-time systems require exponential time. In contrast, in our earlier research, an associative SIMD system was shown to be able to execute these basic tasks in linear time with a graph that had a very small slope. Additionally, the multicore regularly missed a large number of deadlines while the SIMD system did not miss a single deadline. Our goal here was to determine whether we could get SIMD-like results using a CUDA implementation of the same real-time system involving basic ATC tasks on a NVIDIA accelerator. Our research shows that our NVIDIA accelerators can provide a SIMD-like implementation of this real-time system. Moreover, using curve-fitting with MATLAB, the graph showing the NVIDIA accelerators performance increases only slightly faster than a linear graph.
Johnnie Baker, Dr. (Advisor)
Gokarna Sharma, Dr. (Committee Member)
Ye Zhao, Dr. (Committee Member)
79 p.

Recommended Citations

Citations

  • Shaker, A. M. (2017). COMPARISON OF THE PERFORMANCE OF NVIDIA ACCELERATORS WITH SIMD AND ASSOCIATIVE PROCESSORS ON REAL-TIME APPLICATIONS [Master's thesis, Kent State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=kent1501084051233453

    APA Style (7th edition)

  • Shaker, Alfred. COMPARISON OF THE PERFORMANCE OF NVIDIA ACCELERATORS WITH SIMD AND ASSOCIATIVE PROCESSORS ON REAL-TIME APPLICATIONS. 2017. Kent State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=kent1501084051233453.

    MLA Style (8th edition)

  • Shaker, Alfred. "COMPARISON OF THE PERFORMANCE OF NVIDIA ACCELERATORS WITH SIMD AND ASSOCIATIVE PROCESSORS ON REAL-TIME APPLICATIONS." Master's thesis, Kent State University, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=kent1501084051233453

    Chicago Manual of Style (17th edition)