Re: [RFC] 2.5.8 sort kernel tables

[Matthias Andree]

On Thu, 18 Apr 2002, Keith Owens wrote:

> + * Do not assume that the table from the linker is correct, sort it at boot
> + * time.  Since 90%+ of the entries will be sorted, a bubble sort is good
> + * enough, it only runs once per table per boot.  The sort only does binary
> + * keys and only sorts in ascending order.

Any real-world figures on how long this sort process would take on big
tables on some sparc or i586 class box? (Just trying to figure if bubble
is really adequate. It is if the table is indeed essentially sorted with
only like 10 reversed neighbours or if it's short.)