With 2 days left for student application freeze, I am a bit concerned no one
has picked up the idea I posted on the wiki and wanted to start a discussion
for any interested candidates, about any concerns students may have.
The idea can be read at:
(http://elinux.org/BeagleBoard/GSoC/Ideas#MMC_and_DMA_Linux_performance)
and is pasted at the end of this email.
This is a good project for folks “familiar” with kernel programming and are
knowledgeable in the embedded Linux ways to as a starting point to contributing
code to upstream kernel and working on some interesting system level issues.
If students did read it on the wiki earlier, and the reaction was more of “Its
too tough”, instead of a reaction “I don’t think I’m that interested”, then
that certainly can be helped, and it’d be great to get your thoughts and
concerns out in this thread.
The deadline is fast approaching, so do respond soon!
Cheers,
-Joel
MMC and DMA Linux performance on Beaglebone
Improving performance of MMC driver by understanding issues, improving MMC, DMA
drivers and eliminating bottlenecks.
Goal: Both MMC and DMA are critical to high performance of I/O intensive
workloads on a Beagleboard/ARM platform, even fast system boot up depends on
it.
A good amount of performance improvement is possible just by identifying what’s
going on in hot paths and how things can be done more simply, without breaking
anything else. Also improvements are possible using innovative techniques such
as intelligent buffer allocation and reducing overhead where possible in
dependent components such as DMA. Cutting the fat in hot paths is definitely a
start.
Existing Project: Upstream Kernel Hardware Skills: Yes Software Skills: C,
Possible use of JTAG, ftrace, perf etc. Possible mentors: Joel Fernandes