“This utility program is a Cilk++ program providing the Unix/Linux “wc” word count utility functionality. wc-cilk processes individual files (specified on the command line) in parallel, whereas wc processes the files sequentially.
Unlike many Cilk++ examples, this example is not CPU intensive; the word counting involves a simple file scan, treating the file content as a text string. Nonetheless, wc-cilk performance scales very well with the number of CPUs, since application speed is limited by file access and memory access.”
1 response so far ↓
1 Greg Pfister // Oct 16, 2008 at 3:58 pm
Don’t miss the fact that his performance results are, as he states, for cases where all the files are already in memory.