4 Jun
2013
4 Jun
'13
2:04 p.m.
On 06/04/2013 01:52 AM, Paul Birkel wrote:
> Does anyone read/study Don Knuth's "Art ..." (TAOCP) any more?
Sorting
> using tape drives ... what fun! Memory, what memory ... don't need no
> stinkin' memory!
I wish that more were at least willing to look through it.
7 years ago, when they discontinued all of the "advanced" programming
classes, I was re-assigned to be a librarian. I insisted on the library
getting Knuth and K&R. The head librarian tried to veto them, "They are
WAY TOO OLD! We should only get current stuff." I pushed it. Hard.
We got them over her objections.
I have always had a copy at home, a copy at my business, a copy at my
college office. Now that I am RETIRED, maybe I'll eventually get rid of a
few copies. NOT YET.
On Tue, 4 Jun 2013, Chuck Guzis wrote:
There, for a lot of things, Knuth went out the window.
Selection sorts
for about n up to 1000 were fastest using vector instructions.
Quicksort was horrible for most practical values of n.
That's a great illustration of an extremely important concept, of adapting
the sort to the data and hardware configuration, and not being pushed to
look at a sort as some sort of sealed black box, based on assumptions of
"normal" hardware and "random" data.
I saw that particular fail most often with people who could not accept
that the "best" sort was NOT best when dealing with other situations,
particularly the situation where there already exists significant amount
of partial existing order.
--
Grumpy Ol' Fred cisin at xenosoft.com