Exploiting Massive Parallelism for Scientific Datasets on the GPUs

Now GPU is being successfully and widely used as high performance accelerators for many general-purpose computations in various fields

However, there are many challenges such as irregular memory access patterns, recursive back-tracking function calls, limited memory capacity, transfer overhead between CPU and GPU, etc. These make hard to maximize the utilization of GPUs

We resolve these problem by proposing novel indexing schemes and platforms. Please refer "projects" or "publictions" sections for more information.

Poster presented in SC2015

Posted on by Jinwoong Kim

"GLOVE: An Interactive Visualization Service Framework with Multi-Dimensional Indexing on the GPU"(Collaborated with KISTI) has been accepted by the top international conference, Supercomputing

  • [More Info] [PDF]

  • Paper published in IEEE TPDS 2015

    Posted on by Jinwoong Kim

    "Exploiting Massive Parallelism for Indexing Scientific Datasets on the GPU" has been accepted by IEEE Transactions on Parallel and Distributed Systems and also selected as the featured paper of Aug. 2015 issue.

  • [More Info] [PDF] [Source Code(Github)]

  • Paper published in JPDC 2013

    Posted on by Jinwoong Kim

    "Parallel Multi-dimensional Range Query Processing with R-Trees on GPU" has been accepted by Journal of Parallel and Distributed Computing.

  • [More Info] [PDF] [Source Code(Github)]