Phoenix - Efficient, fast, modular miner

[Internet151]

I'm still having occasional issues even with the latest version of phoenix not automatically reconnecting to bitcoinpool.com when it goes down for brief periods of time. Either the mining window stops doing anything or it says work queue is empty and long poll messages every ~10 minutes endlessly.

[JorgePasada]

Just got the following error . . .

Code:

bitcoin@bitcoin:~/poclbm$ ./poclbm.py
Traceback (most recent call last):
  File "./poclbm.py", line 27, in <module>
    platforms = cl.get_platforms()
pyopencl.LogicError: clGetPlatformIDs failed: invalid/unknown error code

Any Help?

Just followed the 11.04 thread guide http://bitcointalk.org/index.php?topic=7514.0 at that link.

Got the error above. I went with SDK 2.4

Should I be using 2.4 or 2.1? I'm mainly running on 5xxx series hardware.

Thanks to everyone who's contributing to this thread, sorry for the chatter, trying to learn is slow and painful . . .

[evilpumpkin]

Hi,

I get "Warning: work queue empty, miner is idle" both on mining.bitcoin.cz (RPC) and bitcoinpool.com (RPC+LP) with different queue sizes.
The following just occured with queue size 5:


[17/05/2011 22:17:05] Currently on block: 124681
[17/05/2011 22:17:05] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:05] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:05] New block (WorkQueue)
[17/05/2011 22:17:05] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:05] New block (WorkQueue)
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:06] New block (WorkQueue)
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:06] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:16] Server gave new work; passing to WorkQueue
[17/05/2011 22:17:16] New block (WorkQueue)
[17/05/2011 22:17:32] Result ...0000000019288c5d accepted
[17/05/2011 22:17:53] Result ...0000000045b2e911 accepted
[17/05/2011 22:18:17] Warning: work queue empty, miner is idle
[0 Khash/sec] [15 Accepted] [0 Rejected] [RPC]


It seems that the miner neglected to get new jobs, the last time work on a new block started.

Can anyone tell me where this problem might come from or maybe even how to fix it?


Thanks in advance

[benjamindees]

Just to post my experience, I have switched from poclbm-mod to phoenix (poclbm), and my reported hashrate is 10% higher with the GPU running 5 degrees C cooler.  I'll have to see whether that translates into a higher payout or not.

[Shiver]

I have the same experience, but I have to use the mod version of poclbm because otherwise I get the "miner is idle" reports every several hours.  I was about to install the Phoenix app as an alternative until I read this thread and seeing that it has the same issue.  Perhaps I'll have to get a wrapper application that will baby sit the miner and restart it when necessary.

Just to post my experience, I have switched from poclbm-mod to phoenix (poclbm), and my reported hashrate is 10% higher with the GPU running 5 degrees C cooler.  I'll have to see whether that translates into a higher payout or not.

[shivansps]

Jedi95 there something can be done about this?


Coding my C program to parse it and get the mhash/s is easy, (last line, seach for [ ), but the file size increase in about 1KB per second!

[yrral]

Jedi95 there something can be done about this?


Coding my C program to parse it and get the mhash/s is easy, (last line, seach for [ ), but the file size increase in about 1KB per second!

I think the problem is that your language interprets the \b control character as some character. Change your system locale or just pipe the output through sed.

[jedi95]

Jedi95 there something can be done about this?


Coding my C program to parse it and get the mhash/s is easy, (last line, seach for [ ), but the file size increase in about 1KB per second!

Your best option would be to modify ConsoleLogger to log only the information you want. The hashrate/accepted/rejected are reported separately from log messages so it should be easy to make it output exactly what you want and nothing else.

[d3athrow]

I love this miner, this + the phatk kernel i get 10 - 25mh/s more with my 5830 than i do with poclbm or diablominer

[jedi95]

Version 1.48 has been released.

Changes:
1. Fixed RPC connection lost callback failing. This *SHOULD* fix the remaining issues with the miner idling.
2. Added phatk kernel to release packages. The included version has the same FASTLOOP changes as the poclbm kernel.
3. Added a warning if WORKSIZE is set incorrectly and the default is used instead.
4. Adds https:// support for connecting to SSL enabled bitcoin clients.

[jondecker76]

Going to try 1.48 right now,  the idling problem has been absolutely killing me, having to restart things 10+ times per day!  
I did an svn update yesterday and it still went idle on me over the night again (it was r86)... Just did an SVN update again just now, but it only pulled in one change file (RPCProtocol.py).  Now at r94, will report back if the ugly idle problem reappears again!



EDIT:
WOW!  Worse than ever!  Goes idle on all 3 cards within a couple of minutes..  Tried several times, going to reboot and try again

EDIT 2:
Still nothing good,  All 3 sessions go idle quite often, but sometimes within a minute or so it will start hashing again, but eventually it ends up just sitting there idle

[jedi95]

My problem is that I have been unable to reproduce any sort of idle problems with 1.46 and later.

Right now the only thing I can do is look at the code and try to figure out WTF is going on. What pool are you connecting to? I can't fix it unless I can see it fail for myself or someone posts enough details for me to figure out what exactly is going wrong.

[benjamindees]

try bitcoinpool.com

[jondecker76]

connecting to deepbit with all my idle problems

[jedi95]

Thanks, I'll run 1 5870 on each pool overnight with lots of extra debug messages enabled and see what happens. Hopefully I will find at least one of them idling so I can get to the bottom of this.

Interestingly I have tested with both those pools before. I blocked packets to the IPs to simulate the server being down/unresponsive and in each case the miner recovered once I unblocked the IP again.

[jondecker76]

let me try one more thing....  I've switched in the last week to a wireless connection (usb wifi adapter)  and the connection seems super slow (so slow that I have problems even sshing into the box) and it just occurred to me that there might be a link there..  Just went back to wired and restarted the miner, I'll report back if that solves my problems so you don't go wasting your time


thanks

[jedi95]

let me try one more thing....  I've switched in the last week to a wireless connection (usb wifi adapter)  and the connection seems super slow (so slow that I have problems even sshing into the box) and it just occurred to me that there might be a link there..  Just went back to wired and restarted the miner, I'll report back if that solves my problems so you don't go wasting your time


thanks

That could be it, since if you can barely SSH into the box it's possible that it's not able to get work fast enough to keep the GPU fed with work. Even so, I already setup the miners for testing so might as well let them run.

[jondecker76]

I switched back to wired and everything has been doing great so far - so it looks like it was definitely related to a flaky wireless adapter!  Good news for both of us

[kwukduck]

I'm just a big noob here, running ubuntu natty but can't seem to install pyopencl

whenever i try to install package python-pyopencl it starts bitching about the package nvidia-current, which is unable to install, without this package installed it can't configure pyopencl somehow...

This is what installer returns if it tries to install nvidia-current

Code:

kwukduck@home2:~$ sudo apt-get install python-pyopencl
Reading package lists... Done
Building dependency tree       
Reading state information... Done
The following extra packages will be installed:
  nvidia-current
Suggested packages:
  python-imaging-tk
Recommended packages:
  nvidia-opencl-icd
The following NEW packages will be installed:
  nvidia-current python-pyopencl
0 upgraded, 2 newly installed, 0 to remove and 0 not upgraded.
Need to get 0 B/50.3 MB of archives.
After this operation, 154 MB of additional disk space will be used.
Do you want to continue [Y/n]? y
(Reading database ... 143428 files and directories currently installed.)
Unpacking nvidia-current (from .../nvidia-current_270.41.06-0ubuntu1_amd64.deb) ...
dpkg: error processing /var/cache/apt/archives/nvidia-current_270.41.06-0ubuntu1_amd64.deb (--unpack):
 trying to overwrite '/usr/lib/libOpenCL.so', which is also in package amd-app 2.4
Selecting previously deselected package python-pyopencl.
Unpacking python-pyopencl (from .../python-pyopencl_0.92-1ubuntu1_amd64.deb) ...
Processing triggers for man-db ...
Errors were encountered while processing:
 /var/cache/apt/archives/nvidia-current_270.41.06-0ubuntu1_amd64.deb
E: Sub-process /usr/bin/dpkg returned an error code (1)

[jedi95]

Test results are in:

No idle miners, but I did find a completely unrelated bug in long polling. If the server sends a message over long polling, then long polling will stop working until you either restart the miner or the connection to the server is lost/regained. This has been fixed in SVN already, and it will be included in the next release.

I'm just a big noob here, running ubuntu natty but can't seem to install pyopencl

whenever i try to install package python-pyopencl it starts bitching about the package nvidia-current, which is unable to install, without this package installed it can't configure pyopencl somehow...

This is what installer returns if it tries to install nvidia-current

You have to recompile pyopencl without the nvidia-current dependency.