Par4all: Auto-Parallelizing C and Fortran for the CUDA Architecture
Par4all: Auto-Parallelizing C and Fortran for the CUDA Architecture
Par4all: Auto-Parallelizing C and Fortran for the CUDA Architecture
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
•Results ◮<br />
(II)<br />
Speed up<br />
100<br />
10<br />
�Par4All in <strong>CUDA</strong> — GPU conference 10/1/2009<br />
1<br />
Tesla 1060 240 streams<br />
GTX 200 192 streams<br />
8c Intel X5472 3 GHz (OpenMP)<br />
2c Intel Core2 6600 2,4 GHz (OpenMP)<br />
1c Intel X5472 3 GHz<br />
DOUBLE PRECISION<br />
200 300<br />
Matrix size (Kbytes)<br />
Reference 1c Intel 6600 2,4 GHz<br />
HPC Project, Mines ParisTech, TÉLÉCOM Bretagne, RPI Ronan KERYELL et al. 35 / 46
Change language