A memory and blocks of threads management in parallel computation using CUDA architecture

Jacek Widuch

Abstract


With the propagation of a multi-core processors a parallel data processing becomes more accessible to a wide range of users. An example is CUDA architecture developed by NVIDIA, which is a multi-core GPU architecture. The GPU can be treated as a SIMD processor with shared memory. The influence of memory management and blocks of threads management on time of computation using CUDA architecture was researched on the basis of matrix multiplication.

Keywords


CUDA architecture; graphics processor; multicore processor; shared memory

Full Text:

PDF (Polski)

References


Cormen T. H., Leiserson C. E., Rivest R. L.: Wprowadzenie do algorytmów. WNT, Warszawa 2000.

Kirk D. B., Hwu W. W.: Programming Massively Parallel Processors: A Hands-on Approach. Morgan Kaufmann, 1st edition (February 5,2010).

Mattson T. G., Sanders B. A., Massingill B. L.: Patterns for Parallel Programming. Addison-Wesley Professional, 1st edition (September 25,2004).

Pacheco P.: Parallel Programming with MPI. Morgan Laufmann, 1st edition (October 15, 1996).

Parberry I.: Parallel Complexity Theory. John Wiley & Sons Inc (September 1987).

Parhami B.: Introduction to Parallel Processing: Algorithms and Architectures. Springer,1 st Edition (January 31,1999).

Rauber T., Runger G.: Parallel Programming for Multicore and Cluster Systems. Springer, 1st edition (March 10,2010).

Sanders J., Kandrot E.: CUDA by Example: An Introduction to General-Purpose GPU Programming. Addison-Wesley Professional, 1st edition (July 30,2010).

Wittwer T.: An Introduction to Parallel Programming. VSSD, 1 st edition, 2006.

NVIDIA CUDA Best Practices Guide Version 3.0. NVIDIA Corporation (February 4, 2010), http://www.nvidia.pl/object/cuda_getjpl.html.

NVIDIA CUDA Programming Guide Version 3.0. NVIDIA Corporation (February 20, 2010), http://www.nvidia.pl/object/cuda_getjpI.html.

NVIDIA CUDA Reference Manual Version 3.0. NVIDIA Corporation (February, 2010), http://www.nvidia.pl/object/cuda_get_pl.html.

Timeout Detection and Recovery of GPUs through WDDM. Windows Haulwjn Developer Central, Microsoft Corporation, (Updated: April 27, 2009), http:.microsoft.com/whdc/device/display/wddm_timeout.mspx.




DOI: http://dx.doi.org/10.21936/si2010_v31.n4A.341