python OpenCL bitcoin miner

[ciuciu]

Hi,

I try to install the GPU miner under Ubuntu and I get this error:

Traceback (most recent call last):
  File "poclbm.py", line 23, in <module>
    platform = cl.get_platforms()[0]
pyopencl.LogicError: clGetPlatformIDs failed: invalid/unknown error code

I would appreciate some help.

I'm using an 5850 and Ubuntu 10.4.

Thank you.

[davout]

Hey momchil, i'm getting loads of these while running your latest git code, weird thing is that I'm getting these messages on only one of the two cores of my 5970, I don't really know what it means, should I downclock it a bit ?

Code:

16/12/2010 23:57, 1a8eed81, invalid or stale
16/12/2010 23:57, 68c13670, invalid or stale
16/12/2010 23:59, 910435fc, invalid or stale
290592 khash/s

[btchris]

... on only one of the two cores of my 5970 ... should I downclock it a bit ?

Code:

290592 khash/s

So you're the reason I can't mine any coin... 

*jealously sulks off*

[m0mchil]

Moved actual search to separate thread to avoid IO overhead. It wasn't a problem when used with local bitcoin client, but become a problem with slush's mining pool.

Davout, check if crossfire is switched off. Or try with lower clock. Xelister reported on #bitcoin-dev similar issues with 5970 and Diablo's miner.

[Elanthius]

D:\Bitcoin\poclbm_20101218>poclbm.exe --help
Traceback (most recent call last):
  File "poclbm.py", line 3, in <module>
  File "pyopencl\__init__.pyc", line 3, in <module>
  File "pyopencl\_cl.pyc", line 12, in <module>
  File "pyopencl\_cl.pyc", line 10, in __load
ImportError: DLL load failed: The specified procedure could not be found.

Maybe missing a dll?

[slush]

Hey momchil, i'm getting loads of these while running your latest git code, weird thing is that I'm getting these messages on only one of the two cores of my 5970, I don't really know what it means, should I downclock it a bit ?

Code:

16/12/2010 23:57, 1a8eed81, invalid or stale
16/12/2010 23:57, 68c13670, invalid or stale
16/12/2010 23:59, 910435fc, invalid or stale
290592 khash/s

Oh! It is exactly what I reported you! I find in server logs that one of your miner is sending corrupted results. Only one, second one is OK. Please try to downclock it a bit.

[m0mchil]

Elanthius, in previous version I forgot to remove OpenCL.dll from the py2exe distribution. It is not there in current version and you should check that you have Stream SDK 2.2 installed and proper version of OpenCL.dll loadable by miner. Be sure there aren't multiple versions - for example if an nvidia one from previous drivers load first it could give you this error.

[Elanthius]

Elanthius, in previous version I forgot to remove OpenCL.dll from the py2exe distribution. It is not there in current version and you should check that you have Stream SDK 2.2 installed and proper version of OpenCL.dll loadable by miner. Be sure there aren't multiple versions - for example if an nvidia one from previous drivers load first it could give you this error.

Hmm, I have an nvidia card that is OpenCL capable. I tried copying over the OpenCL.dll from system32 but that still crashes. In the end I just copied the dll out of the previous version and that seems fine. I'm getting 20Mh/s on a GeForce 8800 so its definitely using OpenCL appropriately.

[BitLex]

Hey momchil, i'm getting loads of these while running your latest git code, weird thing is that I'm getting these messages on only one of the two cores of my 5970, I don't really know what it means, should I downclock it a bit ?

Code:

16/12/2010 23:57, 1a8eed81, invalid or stale
16/12/2010 23:57, 68c13670, invalid or stale
16/12/2010 23:59, 910435fc, invalid or stale
290592 khash/s

just got a HD5970 and of course had to try all kinds of settings, drivers and sdk versions to see what works best for me and i came across some similar problems.


running poclbm_py2exe_20101218, or poclbm_py2exe_20101214 pre-compiled binaries (on XP64),
device=1 works fine as usual (<300M),
device=2 though, as soon as it finds the first winning hash, gets this

Code:

19/12/2010 14:56, 0cfc0e17, accepted
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
19/12/2010 14:56, fbe37df7, invalid or stale
...
...

this only happens on device=2, no matter if device=1 is used or not, no matter what clock-speed, even below stock-settings device=2 floods its "invalid or stale"-msg


running device=2 (or both devices) on poclbm_py2exe_20101126 works perfect, both GPUs get <300M, find hashes, which are added to different "pool.Worker-shares" as expected, no troubles at all.

it seems something happened in between 20101126 - 20101214, which effects HD5970s, or multi-gpu-cards,
had no problems before running 2 single-gpu cards.

think i'll keep 20101126 running for now, or am i missing some new required features here?
don't need to know if any single hash has been accepted, as long as >95% are, but i need to run both GPUs, not just one. 

[davout]

I'm getting these too, didn't get them at all before git pulling latest miner version...

[slush]

Are you still using separate workers for each chip/gpu?

To m0mchil - how many getworks() do you prefetch now? I probably have to tune "max_jobs" for those cards.

I see _this_ problem is on my side, because I register every getwork for later checks. Due to optimizations, I have 10 jobs per worker, which worked very well for all miners and 5970. But when m0mchil made some changes, maybe he is prefetching more jobs and I throw away oldest one; it leads to 'stale' errors.

I will increase max_jobs to 20 for your card. Please report if it goes better.

this only happens on device=2, no matter if device=1 is used or not, no matter what clock-speed, even below stock-settings device=2 floods its "invalid or stale"-msg


running device=2 (or both devices) on poclbm_py2exe_20101126 works perfect, both GPUs get <300M, find hashes, which are added to different "pool.Worker-shares" as expected, no troubles at all.

it seems something happened in between 20101126 - 20101214, which effects HD5970s, or multi-gpu-cards,
had no problems before running 2 single-gpu cards.

think i'll keep 20101126 running for now, or am i missing some new required features here?
don't need to know if any single hash has been accepted, as long as >95% are, but i need to run both GPUs, not just one. 

[slush]

I'm getting these too, didn't get them at all before git pulling latest miner version...

Your limit is already over 20 parallel jobs (increased before days). I hope m0mchil's miner is not so hungry .

[davout]

I was getting the very same messages even before trying pooled mining

[BitLex]

its not only the pool, its the same on local getwork-servers, aka mainline-clients.

[slush]

its not only the pool, its the same on local getwork-servers, aka mainline-clients.

Oh! Good to know! I didn't realized that. So it should not be related to 'max_jobs' solution in pool.

[BitLex]

Oh! Good to know!

i guess it is, should have mentioned that earlier.

i tried -testnet, cuz i didnt want to wait a day or 2 to let it happen and it's the exact same invalid-floods as soon as the first hash is found (the first one even makes it to the wallet sometimes).

[ColdHardMetal]

Hello,

I downloaded the miner and followed the instructions from the OP. I feel as though I have it set up correctly. When I try to run the miner a new CMD window flashes on my screen too rapidly to read anything, or even see if it says anything. I don't get any errors in the CMD window I'm using to launch the miner but I really don't feel like it's working. I'm also not sure how I'm able to tell if it is working.

Any assistance would be appreciated.

Thanks.

[dbc]

Looking at the opencl kernel, can the belowOrEquals function not avoid the endian related comparison of seperate bytes, instead moving the switch into the python code when creating targetH and targetG. Then less branches in kernel and perhaps better stream usage?

[gohan]

Finally got my hands on a GTX 275. Reporting stats:

CPU: Core2 Extreme QX6700, 2.67 GHz
OS: Gentoo Linux 2.6.34
Client: Classic 0.3.19 (local build)
khash/sec: ~5000

GPU: Nvidia GeForce GTX 275
Miner: m0mchil's OpenCL (2010-12-20)
Options: -r 3 -f 10 -w 256
khash/sec: ~49500

I have an old motherboard (PCIe 1.0a, etc.) which has a lot of strain on it, so your results with the same CPU or GPU could be better. It seems I should've gone for an ATI radeon though.

[distant]

Does anyone know what the optimal settings are for a GTX 580?

Thanks,
DiSTANT