From: msevior@physics.unimelb.edu.au
Date: Sun Sep 21 2003 - 09:00:57 EDT
> I could speed up loading the sample (RTF spec) in 5291 again
> by 20% ( now 1min18s, before 1min39 (cvs HEAD with configure
> --enable-extra-optimization)) on my laptop.
>
> The trick is to first read a memory location and only write
> to it if it is really necessary.
>
> Robert
>
> diff -u -r1.14 fp_ContainerObject.cpp
> --- fp_ContainerObject.cpp 19 Sep 2003 15:42:19 -0000 1.14 +++
> fp_ContainerObject.cpp 21 Sep 2003 05:33:10 -0000
> @@ -127,7 +127,9 @@
> */
> void fp_Container::clearBrokenContainers(void)
> {
> - m_pMyBrokenContainer = NULL;
> + if(m_pMyBrokenContainer) // avoid unnecessarily dirtying of
> memory pages
> + m_pMyBrokenContainer = NULL;
> +
Ummm, I'ma bit scared by this patch. After you load that doc, can you edit
it, especially tables broken over many pages, without weird effects?
Martin
> UT_uint32 i =0;
> for(i=0;i<countCons();i++)
> {
This archive was generated by hypermail 2.1.4 : Sun Sep 21 2003 - 09:20:20 EDT