[libdvbpsi-devel] libdvbcsa - NEON acceleration

[Nikolay Nikolaev]

Hello,

I am sending a patch to libdvbcsa whcih implements ARM NEON acceleration.
All development and verification is done with chroot/qemu-arm-static on a
x86.
All test are passing fine.

I don't have currently ARM board where I can test the speedup. My benchmarks

show slower speed with NEON, which probably is due to the complex emulation
of
the vector instructions.

The current implementation uses 64 bit wide batch size. I am also working on
a
128 bit version. Since I am doing my first steps in NEON - any comments and
recommendations are more than welcome.

If someone can run the benchmarks on a real board and post some numbers,
I'll be
very thankful.

Lastly, I send this mail to these 2 mailing lists as I was recommended to do
in
#videolan channel. If this is not the proper place - please point me where I
can
post/discuss these matters.

best regards
Nikolay Nikolaev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/libdvbpsi-devel/attachments/20110719/b126f8f8/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: neon.patch
Type: text/x-patch
Size: 4941 bytes
Desc: not available
URL: <http://mailman.videolan.org/pipermail/libdvbpsi-devel/attachments/20110719/b126f8f8/attachment.bin>