I have not yet quantified it, but if anyone is sufficiently interested I
will do a benchmark. The app is rather specialised but it goes something
like this:-
I have a directory tree with directories (which can go to a depth of up
to 8) containing small files that themselves contain a filename of the form
/ers/Atoz/A/9999999.sof (any resemblence between this and a win95
explorer URL shortcut is, unfortunately, entirely intentional); this file
is then opened and read and closed, then the shortcut file is closed.
Subjectively the overall speed with two people accessing this 'database'
via win95 and samba (containing about 10,000 references to approx 4,000
.sof files) has about halved between 1.3.xx and 1.2.13 (it is sometimes
better than this, but it varies a lot).
Is there any benefit in caching say the last 256 inode lookups at the
namei level (or thereabouts) rather than going thru all the FS dependant
code? (this is a trick which has speeded up a lot of op. sys. in the past
on various mainframes).
-- Dirk-Jan Koopman Tel/Fax: +44 1362 696076 Mobile: +44 973 333806 Computer Consultant Email: djk@dirku.demon.co.uk or G1TLH@GB7TLH.#35.GBR.EU "The typewriting machine, when played with expression, is no more annoying than the piano when played by a sister or near relation." --Oscar Wilde