Scientific Programming / 2017 / Article / Tab 5 / Research Article
A Heterogeneous System Based on Latent Semantic Analysis Using GPU and Multi-CPU Table 5 Results of execution time in a truncated SVD procedure applied in the hLSA system using GPU-accelerated procedures and the LSA system using CPU sequential-only procedures for 2000 documents.
Number of hLSA LSA Matrix S (ms) Matrix D (ms) Matrix T (ms) Total (ms) Matrix S (ms) Matrix D (ms) Matrix T (ms) Total (ms) 25 0.158 7.241 59.635 67.033 0.094 11.151 254.472 265.717 50 0.134 6.540 59.228 65.902 0.113 12.756 251.627 264.496 75 0.144 6.870 59.576 66.590 0.127 10.842 250.042 261.011 100 0.149 4.216 60.443 64.808 0.118 10.594 251.530 262.242 125 0.118 5.064 60.644 65.826 0.115 11.048 262.973 274.136 150 0.176 4.657 61.183 66.016 0.127 10.387 261.292 271.806 175 0.113 6.631 61.259 68.002 0.120 10.626 257.150 267.896 200 0.162 4.195 62.297 66.653 0.135 10.148 265.645 275.928 225 0.174 4.204 63.519 67.897 0.140 10.298 266.845 277.283 250 0.177 4.941 63.100 68.219 0.142 10.090 256.517 266.749 275 0.177 4.599 65.915 70.691 0.140 10.917 257.991 269.048 300 0.174 4.829 64.671 69.675 0.147 10.869 269.890 280.906 325 0.174 6.413 64.655 71.242 0.150 10.774 254.650 265.574 350 0.178 4.269 67.372 71.818 0.149 10.570 253.184 263.903 375 0.157 4.843 68.000 72.999 0.194 10.630 252.340 263.164 400 0.209 5.808 66.520 72.537 0.181 10.776 252.570 263.527 425 0.209 6.443 66.587 73.239 0.174 12.384 265.330 277.888 450 0.243 5.847 69.124 75.213 0.162 10.868 258.973 270.003 475 0.198 4.866 70.410 75.474 0.169 10.796 265.007 275.972 500 0.201 5.765 70.373 76.338 0.172 10.656 256.289 267.117