Distributed Hardware Evolution Project

[MichaelDHEP]

Hello everyone,

We've (re)started the Distributed Hardware Evolution Project at http://www.dhep.ga . You can host an island of evolving digital circuit designs which will become the next generation super-reliable electronics for use where human lives are at stake (aviation, aerospace, medical equipment, power stations, etc.).

The project is now open for anyone who'd like to get a head start in the stats!

Cheers,
Michael

 

[Ken g6]

Interesting. Not a BOINC project, though?

 

[MichaelDHEP]

Not at the moment. Figuring out a way to get a cross-platform Java client onto BOINC. Will take some time I suppose.

However in the meantime if you'd like to get a head start it's all yours.

 

[crashtech]

Which advanced x86 extensions is this code capable of utilizing, e.g., SSE, AVX, AVX2, AVX-512?

 

[Ken g6]

Which advanced x86 extensions is this code capable of utilizing, e.g., SSE, AVX, AVX2, AVX-512?

I suspect the answer is, whatever Java is capable of utilizing. So, probably not much.

 

[crashtech]

So ideal for that old Core2 in the corner, perhaps.

 

[zzuupp]

My apologies for being paranoid, but I'm not sure I currently wish to join a new non-boinc java project.
I'll be a chicken and let someone else dive in first.

 

[MichaelDHEP]

Hello replying to the vector extensions this might answer your question: http://prestodb.rocks/code/simd/ . So Java will identify places in which they can be used and our code has several iterations similar to that so I suspect they'll be used a fair bit. However I'll look at it in more depth and see if we can make even more use of them.

And as you can see it has no access to do anything except connect to the Sussex University webserver and to dhep.ga . You may also be interested in our publications of our results so far: http://users.sussex.ac.uk/~mmg20/index.html In particular the doctoral thesis and the CSRP paper.

Kind regards,
Michael

 

[MichaelDHEP]

Fair enough. The updated link (they annoyingly keep shifting their linking system!) to the New Scientist article is https://www.newscientist.com/article/dn4012-circuit-design-evolving-in-distributed-digital-world/

I see the distributed computing scene has changed somewhat in the last 14 years when this project first ran for a couple of years. It appears to be somewhat saturated now. I suppose the whole internet has changed somewhat.

Feel free to check everything you need and don't hesitate to get in touch if you have any questions.

Cheers,
Michael @ DHEP

 

[Assimilator1]

Interesting project , & it's based not too far from me (I'm I Surrey), & featured in a magazine I'm sub'ed too! (I don't think I've read that 1 yet [edit] Ah, I see it was a while ago, I've probably forgotten then ).

What was the outcome of the original run?

 

[MichaelDHEP]

Hello Assimilator1, the outcome of the first run was very positive. You can see all the (open source) results here: http://www.dhep.ga/statstschof.php . (In summary evolved circuits require only 22% of duplication overhead, whereas the best conventional techniques require over 69% for the same benchmarks - this can be make or break when it comes to adding this kind of reliability in mass produced products) Our intention this time round is to tackle larger circuits which should be possible with faster CPUs and a bigger crowd. These sized circuits would make a huge impact in areas needing high-reliability electronics - which are increasingly ubiquitous around us. We will also be tackling a wider range of circuits including Sequential and a new class of "Absolutely Fault-Secure" circuits for which there are no conventional design techniques. We envisage the project now been here to stay for the long-term.

 

[StefanR5R]

Somebody suggested DHEP to be added to Formula Boinc in 2019:
http://formula-boinc.org/forum/viewtopic.php?f=15&t=131

Not a good idea IMO, until DHEP is updating and exporting team stats (1.) at all, (2.) on a regular basis, (3.) frequently enough such that not too much of work done in year n is credited in year n+1.

(I don't have a forum account at the FB site, so I am lazily posting it here.)

Edit:
Hmm, on second look, free-dc and boinctstats both have team stats, actually. Our free-dc based weekly stats left me with the mistaken impression that there were 0 updates to the team credits.

Edit 2:
On third look, maybe getting 0 credits would still be preferable to what many teams actually get: less than 0 credits.

Maybe bcavnaugh was just trying to be funny with his suggestion, and I didn't get it.

 

[StefanR5R]

At the time of my previous post I could not reach the DHEP project server. But now I found the forum thread which explains the recent negative credits updates: There was a brief period during which some hosts received mistakenly inflated credits. These mistaken credits were manually reverted (not sure with what precision). So, this was hopefully a one-off glitch.

 

[Assimilator1]

I see this project has been available on BOINC since July!

I've just joined.

 

[Assimilator1]

Damn these WUs have a long run time! After a night of running it, it's currently showing 12d 8hrs!!

Is this normal??
If so I may have to stop running it if it blocks other projects running, or will BOINC pause them for other projects?

 

[StefanR5R]

Damn these WUs have a long run time!

These are virtually infinite* WUs which give continuous results and continuous credit:
https://dhep.ga/boinc/forum_thread.php?id=31

• Credit is awarded hourly for processing done within workunits.
• Workunits are placeholder dummies of non-defined duration with no required validation.
• Percentage progress can be ignored as is meaningless for tasks of non-defined duration.

More explanation is in that forum post.

In other words, loading one placeholder WU is like loading one normal WU, completing it, loading the next WU, completing it, etc..

While you run the WU, you should get credit every hour. Once you have enough, you can abort the WU without harm. Or suspend the DHEP WU, run something else for a while, and later resume that same DHEP WU.

*) "infinite" = terminates when the science goal is modified at the server, which happens every few weeks according to the forum post

If so I may have to stop running it if it blocks other projects running, or will BOINC pause them for other projects?

I haven't run DHEP myself yet. BOINC has an inbuilt algorithm which lets it switch itself between projects, determined by several factors. However, I suspect that the nonstandard behavior of the DHEP WUs can easily confuse the BOINC client, and it may be necessary to switch between projects manually if you want to run several at a certain proportion.

 

[Assimilator1]

RE DHEP, ah good to know, thanks Stefan .
Eventually found TA stats here, & I've got 54k credits so far .

Re BOINC, guess I'll have to keep an eye on it, which I do anyway really.

 

[Assimilator1]

It does eventually share resources with other projects, although I then choose no new tasks for SETI so I can give most of the CPU to DHEP .

Current RAC is 109k with 10 threads running on my i7, clocked up 1.6 mill credits .
I see that Kaebu is close to me in RAC, & not that far behind on total credits, with the firepower he has listed he could easily blow past me!

 

[ao_ika_red]

I didn't follow conversation about running DHEP from last year. So I need to ask, how do I run DHEP? I imagine it's not as simple as registering myself on DHEP and let BOINC Manager sort things out. I remember I need to install something like LHC with their VirtualBox.

 

[Howdy]

I didn't follow conversation about running DHEP from last year. So I need to ask, how do I run DHEP? I imagine it's not as simple as registering myself on DHEP and let BOINC Manager sort things out. I remember I need to install something like LHC with their VirtualBox.

I just registered on the DHEP site, added the program in BOINC and hit "run" and it worked. Nothing special was done on my part. If it required VBOX, guaranteed you would see me asking more questions on how to run it....................

 

[ao_ika_red]

I just registered on the DHEP site, added the program in BOINC and hit "run" and it worked. Nothing special was done on my part. If it required VBOX, guaranteed you would see me asking more questions on how to run it....................

Okay, I solved my problem. As usual, the culprit is my landline ISP. I bet it has something to do with unfamiliar domain like DHEP's .ga or other eastern european domain.

And I was right. By looking at BOINC's download list, it's downloading Java Runtime Environment, which I believe it has to be separately installed before DHEP merged with BOINC last year.

 

[ao_ika_red]

Okay, I solved my problem. As usual, the culprit is my landline ISP. I bet it has something to do with unfamiliar domain like DHEP's .ga or other eastern european domain.

Somehow, tasks upload / download process are working again with landline ISP. I think the problem was limited to downloading intial files.

Spoiler: DHEP initial files download list

22/05/2019 19:15:28 | Distributed Hardware Evolution Project | Started download of wrapper_dhe-custom-2R_windows-x86_64.exe
22/05/2019 19:15:28 | Distributed Hardware Evolution Project | Started download of jreitcwin_job_3.xml
22/05/2019 19:15:31 | Distributed Hardware Evolution Project | Started download of ITClientv1.10.jar
22/05/2019 19:15:32 | Distributed Hardware Evolution Project | Started download of java.policy.evm
22/05/2019 19:15:33 | Distributed Hardware Evolution Project | Started download of jre-1.8-win64.zip
22/05/2019 19:15:33 | Distributed Hardware Evolution Project | Started download of input

 

[Assimilator1]

Lol, weird.

And yea I did nothing special to get DHEP running.
Btw I later stopped running it as it stopped other projects from running!