Generating GPU Code from a High-level Representation for Image Processing Kernels

Richard Membarth, Anton Lokhmotov, Jurgen Teich
Hardware/Software Co-Design, Department of Computer Science, University of Erlangen-Nuremberg, Germany
Proceedings of the 5th Workshop on Highly Parallel Processing on a Chip (HPPC), 2011


   title={Generating GPU Code from a High-level Representation for Image Processing Kernels},

   author={Membarth, R. and Lokhmotov, A. and Teich, J.},

   journal={HPPC 2011},




Download Download (PDF)   View View   Source Source   



We present a framework for representing image processing kernels based on decoupled access/execute metadata, which allow the programmer to specify both execution constraints and memory access pattern of a kernel. The framework performs source-to-source translation of kernels expressed in highlevel framework-specific C++ classes into low-level CUDA or OpenCL code with effective device-dependent optimizations such as global memory padding for memory coalescing and optimal memory bandwidth utilization. We evaluate the framework on several image filters, comparing generated code against highlyoptimized CPU and GPU versions in the popular OpenCV library.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: