Am 2014-08-09 02:17, schrieb valdis.kletni...@vt.edu:
On Fri, 08 Aug 2014 13:28:45 -0700, Arlie Stephens said:
On Aug 08 2014, valdis.kletni...@vt.edu wrote:
There's a big difference between knowing how to change the spark plugs on
a VW Beetle, and being able to walk into a Formula One pit and make tuning
suggestions that actually help the performance.

And yes, there's *that* big a gap between the usual beginner programmer
and some parts of the kernel.  In fact, I'll go out on a limb and say that
there are more people in this world that really understand Formula One
engines than people who really understand the Linux scheduler. :)
Now that's depressing.
There are 11 teams competing in the 2014 Formula One series.  Say 10 engine
jocks on each team - that leaves us 120 or so people who *really* know
the engines. (And that's probably an under-estimate - McLaren's total
engineering staff is around 240 people, so they probably

Looking at next-20140807:

for i in kernel/sched/*.[ch]; do git blame $i; done | cut -f2- -d'(' | awk '{print 
$1" "$2}' | sort | uniq -c | sort -nr

There's only 235 entries total.

The top 20:

    8922 Peter Zijlstra
    1629 Ingo Molnar
    1483 Paul Turner
    1317 Dario Faggioli
    1282 Linus Torvalds
    1202 Juri Lelli
     999 Frederic Weisbecker
     703 Gregory Haskins
     643 Rik van
     621 Mel Gorman
     612 Paul Gortmaker
     517 Mike Galbraith
     492 Li Zefan
     472 Steven Rostedt
     346 Nicolas Pitre
     338 Thomas Gleixner
     309 Kirill Tkhai
     282 Tejun Heo
     272 Rusty Russell
     210 Suresh Siddha

The cutoff for "less than 40 lines" is at spot #54, and "less than 10 lines" is
at spot #98, after which point the next 137 people have contributed
single-digit amounts of code.  (Lots of well-known names down in that
single-digit club, too - but those numbers smell more like people who have
changed a kernel API and just fixed up the scheduler uses of the API rather
than doing deep understanding of the kernel).

So less than 100 kernel scheduler contributors, to 120 F1 engine designers.

(And yes, I'm glossing over people who have written big chunks of scheduler
code that have since been replaced.  Feel free to dig through git history
and do your own numbers if you want something more accurate :)

So yeah.  Go ahead and be depressed. :)




_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
But this is just a result of demand & supply: there are more F1 engineers because you can generate money by hiring them. but the scheduler is already done - no one needs any more scheduler developers than there are right now. so the conclusion is not that the Linux scheduler is more complicated than F1 engines. hell, there are even more F1 engineers than people who know the internals of the software i'm currently developing on. does it mean it's more complicated? nope, it just means the company i'm working for only needs a handful of developers. So don't be depressed, just do what you're interested in and success will follow.
_______________________________________________
Kernelnewbies mailing list
Kernelnewbies@kernelnewbies.org
http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies

Reply via email to