Matteo Frigo. Portable High-Performance Pro- grams. PhD thesis, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, June 1999.  Home/Search   Document Details and Download   Summary   Related Articles

....specified processors GC speed up full GC speed up adapt avg. used processors 8 16 32 48 3.8 4.9 6.1 6.3 3.8 4.5 5.8 6.0 8.0 12.4 19.9 21.8 Table 4. The result of automatic regulation of GC parallelism, in BH O2K. costs than on irregular programs. The Cilk performance model [1, 8] estimates the par allel running time of both regular and irregular programs that are executed in LTC strategy. We utilize this model to estimate T , the parallel running time without cache misses cost. The recent model [8] by Frigo analyzes the costs of cache misses. However, current Frigo s ....

....BH O2K. costs than on irregular programs. The Cilk performance model [1, 8] estimates the par allel running time of both regular and irregular programs that are executed in LTC strategy. We utilize this model to estimate T , the parallel running time without cache misses cost. The recent model [8] by Frigo analyzes the costs of cache misses. However, current Frigo s model ignores contention costs, which heavily degrade the performance of parallel programs, especially on DSM. Frigo s model accounts for the increase of cache misses by parallelization. However, it tends to overestimate the ....

Matteo Frigo. Portable High-Performance Pro- grams. PhD thesis, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, June 1999.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC