In article <199901062113.WAA06180@kaa.cs.few.eur.nl> pk@cs.few.eur.nl (Paul Kranenburg) writes:
>> So, the reason for that change was quite well documented: the old
>> version ran about 10x slower.  I even proposed a more general
>> solution, which none of the people who bitched about it at the time
>> bothered to implement: move the mknods into shell functions, and
>> provide two implementations of those functions, one using shell
>> arithmetic and one using expr.
>
>I did just that last weekend but was appalled to find that the result
>ran 5x slower again (though still 2x faster than the wholy expr based
>script).  Anyone with more shell script skills care to look for
>something more efficient?
>
>This is what I did (I'll gladly send the whole MAKEDEV file
>(for sparc) on request):
>

I think the problem is the forking for the evaluation of the backquotes.

christos