1 / 5

GPU VSIPL: Core and Beyond

GPU VSIPL: Core and Beyond. Andrew Kerr 1 , Dan Campbell 2 , and Mark Richards 1 1 Georgia Institute of Technology 2 Georgia Tech Research Institute. Goal. An application development environment for embedded high performance computing that achieves

nedaa
Download Presentation

GPU VSIPL: Core and Beyond

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. GPU VSIPL:Core and Beyond Andrew Kerr1, Dan Campbell2, and Mark Richards1 1Georgia Institute of Technology2Georgia Tech Research Institute

  2. Goal • An application development environment for embedded high performance computing that achieves • Portability: same code usable with different processors, processor generations, and vendors • Productivity: disciplined programming model, leverage highly optimized libraries for signal processing+linear algebra • Performance: employ highly advanced processors,~1 TFLOPS Approach • Adopt the VSIPL API for open standard portability and productivity • Develop a state-of-the-art GPU-VSIPL library to leverage CUDA-enabled GPU performance

  3. GPU-VSIPL Functional Coverage VSIPL API • What’s covered from VSIPL Core • Data Types • real, complex, integer, boolean, index • View Types • Matrix, vector • Element-wise Operators • arithmetic, trigonometric, transcendental, scatter/gather, logical, and comparison • Signal Processing • FFT (in-place, out-of-place, batched) • Fast FIR filter, window creation,1D correlation • Random number generation, histogram • Linear Algebra • generalized matrix product • QR decomposition, least-squares solver • What’s Not (yet) • Linear Algebra • LU, Toeplitz, least-squares solvers • What’s Added Beyond VSIPL Core • Scalar and matrix versions of element-wise vector operators • Matrix utility functions Core Profile GPU VSIPL Core Lite Profile

  4. Performance Examples:Signal Processing 1D FFT, In-Place 1D Correlation

  5. Performance Examples:Linear Algebra Matrix-Vector Product QR Decomposition

More Related