David Balažic wrote: > More details: top reported 99% CPU usage of the cmp process. > iostat reported about 88 MB/s throughput, when simple reading would > give 130 MB/s.
Can you profile to find out which part of cmp is gobbling up the CPU? It could be that it's system time, in the kernel, in which case multithreading might not help. Is 'cmp -s' significantly faster than plain 'cmp'?
