SpMV Profiling and Optimization Analysis
Journal Title: IOSR Journals (IOSR Journal of Computer Engineering) - Year 2015, Vol 17, Issue 4
Abstract
Abstract: Sparse matrix-vector multiplication is an important operation when it comes to sparse matrixcomputations. Very large and sparse matrices are used in many engineering and scientific operations. Hencethe matrix needs to be partitioned properly. Even though the matrix is partitioned and stored appropriatelythere still exists a possibility, the performance achieved is not significant. Thus, the need to address these issues.System proposes an integrated analytical and profile based performance modelling that accurately measures thekernel execution time of various SpMV CUDA kernels for a given target sparse-matrix. Based on this thedesigned optimal solution auto-selection algorithm automatically reports the SpMV optimal solution for a targetsparse-matrix. The system is evaluated on NVIDIA GeForce GTX 680 and NVIDIA Quadro 8000. The system isfurther extended to one more matrix storage format.
Authors and Affiliations
Aditi V. Kulkarni , Prof. C. R. Barde
Keywords
Related Articles
- EP ID EP127518
- DOI -
- Views 132
- Downloads 0
A Comprehensive Study On Handwritten Character RecognitionSystem
Abstract: Nowadays handwritten character recognition is still remain an open problem because of thevariability in writing style. Conversion of handwritten characters is important for making manuscripts intomachine...
Regression Test Suite Prioritization Using Hill Climbing Algorithm
Abstract: Regression testing is an expensive, but important action in software testing. Unfortunately, there may be bereft assets to acquiesce for the re-execution of all analysis cases during corruption testing. In this...
Effect of Angle Orientation of Flat Mirror Concentrator on Solar Panel System Output
Abstract: In this research two flat glass mirrors is used as concentrator of solar panel system. The mirrors increase's the concentration of sun light ray on the solar module. Anew model of solar panel system is designed...
Support Vector Machine Based Sentiment Analysis Process for Twitter Streams
Abstract: Sentiment Research in twitter is quite tough due to its short length. Attendance of emoticons, slang words and misspellings in tweets compelled to have a preprocessing pace beforehand feature extraction. There...
Various Techniques of Clustering: A Review
Abstract: Data mining is a method that is used to select the information from large datasets and it performs the principal task of data analysis. The Clustering is a technique that consist groups of data and elemen...