CGminer v3.8.5 on Broadcom based DD-WRT / OpenWRT

[Zillions]

Excuse me while I subscribe to this thread ... I have a feeling once I get my hardware in I'm going to want quick acess to this thread.

[GoldZuGeld]

otrw2 should work just fine.
I haven't done it, so I can't give instructions. Let me know how it goes.

yeah otrw2 works. I've used this instructions http://www.dd-wrt.com/phpBB2/viewtopic.php?t=156013

[Mobius]

P_shep,

Will you be releasing an updated version with cgminer 2.7.5?

[P_Shep]

Here you go:

https://github.com/downloads/pshep/cgminer/cgminer-2.7.5-mipsel.tar.gz

[Mobius]

Here you go:

https://github.com/downloads/pshep/cgminer/cgminer-2.7.5-mipsel.tar.gz

Thank you.

[rampone]

So I had it running today on a 3200. darn, it does no openwrt.
But what I immediately noticed, that the idle time between work was easy noticable (bflsingles led going out for about 100ms to 300 ms. That is a lot actually, as with my athlon x2 250 cpu on ubuntu it is barely the idle time is barely noticable.

I am not into cgminer, but maybe work delivery could be tuned, as it seems like sth is calculated, just the delivery of data should not take much longer on that router. (tuning in meaning: send new already prepared work before checking the result of the work before).

Maybe i have to fire up the 500 mhz celeron laptop, and check the "idle times" there.

I did not let it run for a long time, to see if hashrate is still the same, as i guess it wont be, what idles, (even though not long), does not hash. My guesstimate of hash loss is sth like 5 to 10 %?

Damn, i wanted this to be my solution, love router hacking/openwrt/mesh nets/those little computers

Maybe ask the cgminer guys, what they think of that effect, and if it could be tuned. (asking in irc right now)

[P_Shep]

(bflsingles led going out for about 100ms to 300 ms.)

I certainly don't see that. just a flicker for me. 300ms would be VERY noticeable. In my version of cgminer I run (home_brew in my git) I have timers around all the comms routines. Avg time is around 0.015s, max time 0.024s over the past 2 days with many, MANY throttles (been HOT here).

The best way to check is to compare hash rates between systems over a day or so. Should be consistent.

[rampone]

so I  got told, that you can check the timings of the bfl via the api, the minimum time is only 0.00003 @e3200 vs 0.00001 @x2 cpu.

that value still does not explain that idle time.

I noticed, that the serial ports are hooked in differently under dd-wrt then under a normal distro (not /dev/ttyUSB -> /dev/usb/tty/0), is that the cause?

[rampone]

I really am not able to time it, 300 ms could be too much, but it is definitely more noticable than with the "pc".
The api readings also give me faith, that it is caused by sth else.
As the e3200 had to actually have another use, which it can't handle (pptpinternet connection fails), i will probably send it back monday, and try with a tp-link and openwrt again, oh that endian...

But to not be so negative: p-sheps work is really slick with integrated service etc.
http://youtu.be/xxxNnIS2cuQ

http://youtu.be/PaH2CWPVVZ8

maybe my guesstimate with more than 5% idle is a bit high, but you can see the difference.

P_shep is your "blinking" less?

[P_Shep]

Definitely an observable difference there. Mine blink slower then your unbuntu there, probably as slow as your e3200. I can say that's less then 100ms though, 100 would be quite clear. Guess this is simply the processing time of the router... not much you can do about that.

[rampone]

the funny thing is the min time for idling reported by cgminer via the api is way less. thats why i think, it could be sth else. maybe i test it with my celeron 500 usb 1.1 notebook setup sometime, see, if its blinking the same like with a fast cpu, or does the blinking like with the router. but not today.


For anyone with just a few bfls and already owning a linksys/dd-wrt compatible mipsel router and no server running 24/7 this is definitely still the way to go.

[Mobius]

(bflsingles led going out for about 100ms to 300 ms.)

I certainly don't see that. just a flicker for me. 300ms would be VERY noticeable. In my version of cgminer I run (home_brew in my git) I have timers around all the comms routines. Avg time is around 0.015s, max time 0.024s over the past 2 days with many, MANY throttles (been HOT here).

The best way to check is to compare hash rates between systems over a day or so. Should be consistent.

Can you point me to a link to be able to setup my workstation to compile the cgminer git code for dd-wrt

[rampone]

I also got told, that bfgminer has more fpga optimized code than cgminer right now? maybe run a compile of that one, should "behave" the same with api, config etc.? At least worth a try, im going to sleep now though.

[P_Shep]

Can you point me to a link to be able to setup my workstation to compile the cgminer git code for dd-wrt

At the bottom of the readme is a link to the compiler, and the option I built with. There's no guide on how to do it... there's not enough time in the world to write one! It took me a few weeks to get it all working and this is part what I do for a living.

[P_Shep]

I also got told, that bfgminer has more fpga optimized code than cgminer right now? maybe run a compile of that one, should "behave" the same with api, config etc.? At least worth a try, im going to sleep now though.

I've compiled bfgminer and will run that for a couple of days. see what the results are. so far after 15hrs, my U is about 0.5 lower than cgminer, but still early to say.

[efiniti]

can someone explain what the setup is?  i don't understand why you'd want to run cgminer on a router.  you're not trying to mine with the router's proc right?..

[irritant]

can someone explain what the setup is?  i don't understand why you'd want to run cgminer on a router.  you're not trying to mine with the router's proc right?..

no, you mine on the FPGA attached to the router via USB, attached to the router instead of a pc because a router uses very little electricity

[P_Shep]

Over 2 days:

CGminer:

Code:

 [2012-09-08 20:47:05] Started at [2012-09-06 13:11:23]
 [2012-09-08 20:47:05] Runtime: 55 hrs : 35 mins : 41 secs
 [2012-09-08 20:47:05] Average hashrate: 4961.5 Megahash/s
 [2012-09-08 20:47:05] Solved blocks: 0
 [2012-09-08 20:47:05] Queued work requests: 25805
 [2012-09-08 20:47:05] Share submissions: 231002
 [2012-09-08 20:47:05] Accepted shares: 230735
 [2012-09-08 20:47:05] Rejected shares: 267
 [2012-09-08 20:47:05] Reject ratio: 0.1%
 [2012-09-08 20:47:05] Hardware errors: 0
 [2012-09-08 20:47:05] Efficiency (accepted / queued): 894%
 [2012-09-08 20:47:05] Utility (accepted shares / min): 69.17/min
 [2012-09-08 20:47:05] Work Utility (diff1 shares solved / min): 69.25/min
 [2012-09-08 20:47:05] Discarded work due to new blocks: 1187
 [2012-09-08 20:47:05] Stale submissions discarded due to new blocks: 7
 [2012-09-08 20:47:05] Unable to get work from server occasions: 379
 [2012-09-08 20:47:05] Work items generated locally: 231185
 [2012-09-08 20:47:05] Submitting work remotely delay occasions: 13
 [2012-09-08 20:47:05] New blocks detected on network: 521

BFGminer:

Code:

 [2012-09-13 08:31:11] Started at [2012-09-10 20:36:08]
 [2012-09-13 08:31:11] Runtime: 59 hrs : 55 mins : 3 secs
 [2012-09-13 08:31:11] Average hashrate: 4917.5 Megahash/s
 [2012-09-13 08:31:11] Solved blocks: 0
 [2012-09-13 08:31:11] Queued work requests: 54000
 [2012-09-13 08:31:11] Share submissions: 247619
 [2012-09-13 08:31:11] Accepted shares: 247177
 [2012-09-13 08:31:11] Rejected shares: 442
 [2012-09-13 08:31:11] Reject ratio: 0.2%
 [2012-09-13 08:31:11] Hardware errors: 0
 [2012-09-13 08:31:11] Efficiency (accepted / queued): 458%
 [2012-09-13 08:31:11] Utility (accepted shares / min): 68.76/min
 [2012-09-13 08:31:11] Discarded work due to new blocks: 29545
 [2012-09-13 08:31:11] Stale submissions discarded due to new blocks: 16
 [2012-09-13 08:31:11] Unable to get work from server occasions: 359
 [2012-09-13 08:31:11] Work items generated locally: 252177
 [2012-09-13 08:31:11] Submitting work remotely delay occasions: 25
 [2012-09-13 08:31:11] New blocks detected on network: 523

CGminer has it by a whisker.

[Luke-Jr]

CGminer has it by a whisker.

2 days is nowhere near long enough for that amount of precision.

[-ck]

Looks good to me. Clearly any claim that jesusminer is better is pure fud.