[poppler] pdf rendering performance on the olpc test boards

[Krzysztof Kowalczyk]

On 11/7/06, Tomeu Vizoso <tomeu at tomeuvizoso.net> wrote:
> we have been testing the olpc pdf reader on an test board (based on
> evince and poppler) and have found the performance a little below our
> expectations. Simple text pages (for example [1]) take about 4 seconds
> to render.
>
> After some profiling of page rendering on the board with sysprof, this
> functions appeared using 45% of CPU time:
>
> _XcmsLookupColorName
> _XcmsSetGetColors
> _XcmsInitScrnInfo
> _XcmsResolveColorString
>
> We are currently using the 0.5.4 version with the cairo backend. The
> processor on the board is a Geode GX-500 [2], the floating point
> performance of which is quite poor. Composition is currently done by
> software.
>
> Does this info make some sense to anybody? Is poppler who is eating our
> CPU or should we look elsewhere? Can you recommend any compilation
> options that could improve performance?

Try splash backend and #define USE_FIXEDPOINT - this should use
non-floating point rendering.

In my profiling I never found the actual rendering performance to be
an issue - it was always parsing. You might try my poppler tree
(http://blog.kowalczyk.info/software/sumatrapdf/develop.html) to get
performance improvements in some cases (details of the changes are on
my blog and in bugzilla with patches, although my svn has some fixes
that are not in patches I originally submitted).

-- kjk