Request for increased awareness: Open Source FPGA acceleration
Hi all,
Those who participated in our past Sprint in Lyngby have met Ruben. He
is the one who kindly packaged all that needed to be packaged for the
first and yet only completely Open Source toolchain for FPGA computing.
Details are on http://wiki.debian.org/FPGA/Lattice .
Ruben started that page which takes you by the hand to get some LEDs
blinking on a 20something $/€ device that attaches to your USB port. It
features a series of I/O ports to do just about anything or just compute
(i.e. accelerate) on the device itself. I got a "thank you for that
page"-email yesterday (Ruben started it, I just added some references).
It is about the first time this ever happened to me :o) Anyway, so it
seems like there is some accumulating momentum and people are kind of
uncertain about what to do. I tell you: There is no risk in spending
those 25 $/€, which actually should be some 80 $/€ for a better
performing 8K device. While I have little doubt that both for Xilinx and
Altera we will also see something evolving in the near future (anybody
tried to get a free webpack license for the only Spartan-6 compatible
older ISE tools? something needs to happen on that front). The
yosys/arachne-pnr/icetools packages of the free pipeline instantaneously
sky rocketed to almost 140 users.
Anyway - I think we are on the forefront of a significant movement here.
This combination of free programming tools and cheap devices for
programmable hardware is completely new. Please keep your eyes and ears
open. There are closed-source implementations of Smith-Waterman on FPGA
out there, not so much on BLAST. Those first devices do not have the
memory to help much with accelerating sequence comparison - the usb port
is too lame and there is not enough memory on the device to keep the
data. But this is all just a matter of time. Hack along! And maybe make
some noise with your engineers on site for whom a PCI/USB3 card with
such an FPGA and a couple of gigs of memory on board is likely little
more than a master thesis away.
Cheers,
Steffen
Reply to: