[EMAIL PROTECTED] writes:
>>Now, another important number would be the run-time memory use. I would
>>expect the run-time footprint to increase somewhat when a bidi document
>>is encountered and processed, but I would hope that it doesn't increase
>>much when a non-bidi document is processed.
>The messages I read on the porkjockeys newsgroup confirm the
>impression I was getting from my own tests: footprint is very variable
>from one test to another. I would need to run lots of tests to get a
>statistically significant sample, but my impression so far is that the
>increase in footprint usage when loading a non-bidi document with a
>bidi build is not larger than the variations between one load and
>another with the same build.
footprint is best measured by vss and rss on an idle system, and
is significantly dependant on exact order of allocations and
fragmentation. Be very careful of what tool you use to measure this.
See the footprint project data on allocators by Waterson on mozilla.org.
--
Randell Jesup, Worldgate Communications, ex-Scala, ex-Amiga OS team ('88-94)
[EMAIL PROTECTED]
