KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Settings

Subject: search.filters.subject.Openmp

Search Results

Now showing 1 - 2 of 2
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Analysing and optimizing snphap using radix-2 computation and openmp
    (2013-07-12)
    Ranokphanuwat, Ratthaslip
    ;
    Rattanatranurak, Apisit
    ;
    Kittitornkun, Surin
    ;
    Tongsima, Sissades
    In this paper, the run time complexity of SNPHAP, which is a haplotype inference tool, is extensively examined. The analysis is based on our previous work in terms of profiling and run-time complexity function. To reduce the run time complexity and enhance its performance, a Radix-2 computation and OpenMP multithreading are applied. The optimized results are compared with both original and compiler optimized versions on an AMD A6-3650 Linux machine. Due to the Radix-2 technique, the complexity is drastically reduced. In addition, the theoretical Speedup is consistent with the experimental one. Furthermore, up to 1,303% Speedup is achievable as a result of OpenMP multithreading.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Multithreading bioinformatics software with OpenMP: SNPHAP case study
    (2010-12-01)
    Ranok, Udom
    ;
    Kittitornkun, Surin
    ;
    Tongsima, Sissades
    This paper presents a parallelization framework for infer- ring haplotypes using an expectation maximization (EM) algorithm. Our framework utilizes GProf profiling tool, OpenMP library, and ompP profiling tool to parallelize the algorithm by determining the hotspot functions, mul- tithreading, and executing them on the Multi-core CPUs. In our experiments, we choose the SNPHAP program for this case study and run it on an 8-core Xeon Linux ma- chine. The results show that our framework can signifi- cantly speedup up to 214% on a large data set with 151 loci of a 10,000 data samples. In addition, deep profiles of multithreaded SNPHAP support our discovery that maxi- mum speedup can be achieved when the number of parallel threads equals to the number of physical cores.