Re: [tablix-list] Scope for small (5%) speed improvement in Tablix

From: Tomaz Solc <>
Date: Mon Jul 26 2004 - 20:40:00 CEST


Thank you for pointing this out. I was thinking about changing data
structures a while ago, but I thought it would not be worth the effort.
If you can get 5% improvement only from changing tuplemap, then I guess
I was underestimating the overhead of the double indirection.

The same change can be done with *map structures in data.c Also the
integers tid, cid and sid in tuplemap could be changed to pointers that
point directly to the relevant data structures.

> 1) Memory use. Instead of using between 4k and 24k of RAM the new
> version uses a fixed 20k. Both I feel are small enough not to worry
> about.

The amount of unused allocated memory could be reduced by reducing
TUPLEMAX and then allocating additional memory when the number of used
tuples exceeds TUPLEMAX (similar to existing code in add_tuple.
> 2) All modules which reference tuplemap need to be upgraded. This
> took me less than 5 minutes. "vi `grep -l tuplemap *.c`

Perhaps it could be automated with a sed script... Anyway, this change
will have to wait. The next release will probably be 0.0.8 with some
minor changes and bug fixes. After that I'm planning to make some more
radical changes which will most likely include this fix of the data

Best regards
Tomaz Solc

Received on Mon Jul 26 20:54:20 2004

This archive was generated by hypermail 2.1.8 : Tue Aug 16 2005 - 20:42:17 CEST