Tracy Reed
2017-03-07 09:08:57 UTC
We are having a co-meeting with the San Diego High Performance Computing
meetup!
This session will include high speed, low latency communication and come
kernel driver hacking - so it should be fun for both groups. As
heterogeneous computing becomes more prevalent, the biggest challenge is
getting data in and out of devices like GPUs, FPGAs and DSPs. We are
going to touch on all three. Gary Scantlen and Shawn McKenney will walk
you though a kernel driver we've developed to enable PeerDirect
communication between DSPs on multiple hosts. We will also show some
GPUDirect communication and we all will get down into the hardware and
program a PCIe Block Device onto a Xilinx Spartan-6 FPGA. Then we can
all play around with the toys for a bit. We'll be ready to modify some
drivers (and lock up some machines), tweak some VHDL and monitor some
PCIe packets with external LEDs. This will be very informal and
hands-on. We have some pretty cool projects we are working and would
love for some extra involvement. Come join us!
meetup!
This session will include high speed, low latency communication and come
kernel driver hacking - so it should be fun for both groups. As
heterogeneous computing becomes more prevalent, the biggest challenge is
getting data in and out of devices like GPUs, FPGAs and DSPs. We are
going to touch on all three. Gary Scantlen and Shawn McKenney will walk
you though a kernel driver we've developed to enable PeerDirect
communication between DSPs on multiple hosts. We will also show some
GPUDirect communication and we all will get down into the hardware and
program a PCIe Block Device onto a Xilinx Spartan-6 FPGA. Then we can
all play around with the toys for a bit. We'll be ready to modify some
drivers (and lock up some machines), tweak some VHDL and monitor some
PCIe packets with external LEDs. This will be very informal and
hands-on. We have some pretty cool projects we are working and would
love for some extra involvement. Come join us!
--
Tracy Reed
Tracy Reed