3 Sep
2013
3 Sep
'13
8:05 a.m.
As an aside, cat should write to a disk device node in a minimum size of one page for the CPU in question so it's performance should be almost the same as using dd with the bs=1m option. I say should because I submitted a patch to make it do so about 10 years ago, I recall that the patch was accepted and I hope it hasn't been backed out.
Presumably when not using the oflag=direct option, the writes are merged anyway. James