Hello everybody,
I just got myself an Iliad v2, and I immediately noticed the sluggishness...
even though it's a nice gadget overall.
Anycase, I though I ought to try to do something about it, so did some research - and recompiling.
As the Iliad's PXA255 CPU doesn't have an FPU, all floating point operations
are hideously slow, and should be avoided at all cost.. at least use
-ffast-math gcc option to minimize the impact.
Additionally, I noticed that poppler has an embedded fixed-point library,
so I compiled both ipdf (with
jharker's alpha fullscreen patch) and
poppler again, and at least to my eyes the system seems a lot faster, especially when going back and forth.
(This means I/O is still quite a bottleneck - maybe a readahead cache?..;)
OK, this is what I did:
Makefile.am: (both) added
Code:
AM_CFLAGS='-O4 -ffast-math'
AM_CXXFLAGS='-O4 -ffast-math'
Popper configure flags:
Code:
./configure --host=arm-linux --build=i386 --prefix=/usr --enable-fixedpoint --disable-cairo-output --disable-gtk-te
st --disable-poppler-qt
Any case, this seemed worthawhile at least to my eye. I can attach
the ipdf and the lib if anyone wants to benchmark without compiling,
just let me know..;)
In any case, I'd love to hear you comments!
//HS