Parallella
Above: Picture of my Parallella on top of a Corsair Blue LED fan. Taken on 7/10/2019
Above: top command running on Parallella running Ubuntu 20.04.
Updated (12/12/2023)
Links
Building a Parallella Cluster by Mark Dewing
https://markdewing.github.io/blog/posts/building-a-parallella-cluster/
Parallella/Raspberry Pi Cluster Computing - scattershot
Parallellagram
OpenSHMEM for Epiphany
https://www.parallella.org/2017/04/26/openshmem-for-epiphany/
Using C++ to program the Parallella - CODUIN
http://blog.codu.in/parallella/epiphany/bulk/cpp/2016/05/06/parallella_cpp.html
RISC-V port to Parallella Board
https://github.com/eliaskousk/parallella-riscv
Benchmark
Below: Linpack (linpackc) Result on ARM cores on Parallella. (Code from: https://github.com/2000nickels/linpackc )
Result: Rolled Double Precision 98801 Kflops; 1000 Reps
Hello World on Parallella using gcc: (added on 2/15/2022)
Above: Classic C Hello World from: https://developer.arm.com/documentation/dui1093/a/Getting-Started/Compiling-a-Hello-World-example
You can compile by gcc helloworld.c
Then type ./a.out in the shell to run the compiled program. It should display "Hello World" on your shell.
Above: Geometric Decomposition on Epiphany.
Code from: https://github.com/mesham/epython/blob/master/docs/tutorial3.md
2020/2/23
Today I built the Parallella Lisp (plisp) It compiled fine with my Ubuntu 16.04 Parallella box.
Some test information:
> ( reverse ( quote ( 1 2 3 4 5 6 7 ) ) )
( 7 6 5 4 3 2 1 )
> ( testfun 101 )
202
> ( any numberp ( quote ( a b ) ) )
nil
processor id: 15
memory: 11700480
node size: 16
nnodes: 10967
nodemem: 175472
nnames: 1563
namemem: 62520
nstrings: 0
stringmem: 0
setflag message: Exited normally!
Updated: 5/11/2019
I got the Ubuntu 16.04 running on my Parallella, by first upgrading to Ubuntu 15.10 then to 16.04. It took several hours in total, making sure my ssh connection didn't die. I tried the eMesh Epiphany sample, and it compiled fine.
Some Parallella codes I wrote / modified.
Simple Code in Epiphany Basic (ebasic)
/*
Simple Printing and Core Identity and Loop Sample
By Tsubasa Kato 2015
*/
a=0
for i=1 to 10
a = a + 2
print a+" "+coreid
next
ePython script (Written on 2016-10-22)
import parallel
a=0
b=0
if coreid() > 0:
a=sendrecv(coreid(), coreid()-1)
if coreid() < numcores()-1:
b=sendrecv(coreid(), coreid()+1)
print "Values are "+a+" and "+b
c = sqrt(a)
d = sqrt(b)
e = c + d
print(e)
7/22/2018:
Me maintaining Parallella for use in 2018. vivid -> xenial . Working ok for now. Fingers crossed.
Parallella Ray Tracing Demo I run on the HDMI version.
Photo on left: Parallella Micro Server x 2 in Black box.
9/6/2018:
I've been experimenting with various ways to cool the Parallella since the summer was very hot this year. Below is a photo of one of the version. (The Parallella in a blue case)
4/30/2018:
I am planning to get the Parallellas up and running again in my lab. I want to be able to use it as a low power but many core experiment environment.
1/18/2018:
The other day I powered up my Parallella and connected it from my desktop computer. Didn't do much, just configured so it has static IP address. I need to make it so the other Parallella will have static IP address as well.
12/27/2017:
Searched about possible application of big data and Parallella and found one site. (refer above, CODUIN)
4/26/2017: Since I've been busy, I haven't been playing with the Parallella for a while, but I intend to; to make sure my skills don't get rusty.
4/8/2017: I measured the temperature using ztemp.sh and the temperature for both Parallella Micro Servers were 43-44.5 degrees Celsius recently. This is probably because it's getting warmer here in Japan. I am thinking of purchasing a fan that operates out of USB.
3/26/2017: I enclosed the other Parallella Micro Server in aluminum case. It took a while for it to be put in, but the temperature is stable at around 41.5-44.0 degrees Celsius.
3/21/2017: I enclosed one Parallella Micro Server inside the aluminum case available from Amazon.com . I am going to enclose another one soon.
Parallella with a fan. The fan is an old Intel fan connected to a USB port cable via an adapter. Cooled quite well.
Parallella epython Mandelbrot example running (video)
Below: Me testing a script (emesh_bandwidth_all2one).
The result of the all-to-one on-chip communication is 3889.00MB/s.
