A Visual Approach to Investigating Shared and Global Memory Behavior of CUDA Kernels
Type
ArticleAuthors
Rosen, PaulKAUST Grant Number
KUS-C1-016-04Date
2013-07-01Online Publication Date
2013-07-01Print Publication Date
2013-06Permanent link to this record
http://hdl.handle.net/10754/597436
Metadata
Show full item recordAbstract
We present an approach to investigate the memory behavior of a parallel kernel executing on thousands of threads simultaneously within the CUDA architecture. Our top-down approach allows for quickly identifying any significant differences between the execution of the many blocks and warps. As interesting warps are identified, we allow further investigation of memory behavior by visualizing the shared memory bank conflicts and global memory coalescence, first with an overview of a single warp with many operations and, subsequently, with a detailed view of a single warp and a single operation. We demonstrate the strength of our approach in the context of a parallel matrix transpose kernel and a parallel 1D Haar Wavelet transform kernel. © 2013 The Author(s) Computer Graphics Forum © 2013 The Eurographics Association and Blackwell Publishing Ltd.Citation
Rosen P (2013) A Visual Approach to Investigating Shared and Global Memory Behavior of CUDA Kernels. Computer Graphics Forum 32: 161–170. Available: http://dx.doi.org/10.1111/cgf.12103.Sponsors
We thank Kristi Potter for her feedback. This work was supported by DOE NETL and KAUST award KUS-C1-016-04.Publisher
WileyJournal
Computer Graphics Forumae974a485f413a2113503eed53cd6c53
10.1111/cgf.12103