On 1/29/21 1:07 AM, Ole Tange wrote:
Could you consider implementing a parallel merge, so I can retire
parsort?
Yes, improving that part of 'sort' performance has been on my long list
of things to do for quite some time. If someone else could take up the
task it'd be done quicker Anyway,
During the past 6 months I have been using GNU sort on a lot of files in
the GByte range on my 64 core machine. I have discovered, that GNU sort
is bad at using all the cores.
I have built parsort (now distributed as part of GNU parallel) as a
small wrapper around GNU sort and it performs 2-3x bet
