Re: moving back to 32-bit swap_filens ?

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Tue, 14 Aug 2001 12:37:33 +0200

The intention from start was to make the file and dir-number field sizes
adjustable at build time.

Regarding structure sizes: The rules are mostly the same on all
platforms. It is not as platform dependent as it may seem. It is caused
by a combination of alignment requirements and malloc overhead.
Currently StoreEntry is at the edge (or at least was last time I looked
at it). Increasing it by one byte is almost the same as increasing it by
16 bytes, which unfortunately is quite a lot in the context.

I have a branch attempting to shrink the overhead of StoreEntry quite a
bit. See compactsentry. Unfortunately there currently is some bug left
there, and due to my other workloads the branch been idle for quite a
while now.

Changing the sizes while working with COSS is perfectly fine. However,
I don't think these adjustments should be in HEAD until we commit a good
working COSS there.

--
Henrik
Received on Tue Aug 14 2001 - 04:36:42 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:14:11 MST