python OpenCL bitcoin miner

[m0mchil]

Grinder, it would be really nice if you mention what your setup consists of. Thanks

[Grinder]

I updated my post almost immediately with some info. Do you need to know anything else? Parameters are -d 1 -w 128 -f 0 -r 2

Hm. No errors if I add -v. Same thing happens on a 5850 on the same system, although I did get one accepted before it exited when I tried.

[qed]

Working great on HD 6950s.

[vuce]

Works great. I get pretty much the same hashrate as with phoenix, but the number of submitted shares is ~7% higher (or maybe the last 3 hours were just extremely lucky).

[fasti]

http://support.amd.com/us/kbarticles/Pages/AMDCatalyst114ReleaseNotes.aspx

GPU Compute enhancements:

    The OpenCL runtime (included in AMD Catalyst 11.4) includes performance enhancements supporting zero-copy on APUs and increased performance for PCIe transfers between a CPU and discrete GPU.

Wonder if this can improve mining...

[kindle]

Hi m0mchil thanks for keeping your miner up to speed with the bit_int support. Out of curiosity does your latest version activate the bit_int support automatically? Is there a flag I have to add to activate it.

Also I noticed similar performance to that of phoenix. I guess essentially both miners are the same as it is running the same kernel, however is the hash checking the same for both?

[JWU42]

I am seeing same performance as Phoenix 1.3 as well - fwiw

Seems BFI_INT is on by default

[m0mchil]

@kindle
Yes, it is on by default if the hardware supports it.

Both miners are now the same in terms of hash checking. Phoenix has far better documented and structured code. It also cares about 'efficiency', something I'm tired of explaining that actually doesn't matter.

Also, Phoenix seems to have different way of load tuning (aggression) which on my particular setup results in more laggy behavior... and perhaps slightly better performance for dedicated mining rigs, not sure.

I hope I won't attract anger with blatantly copying their BFI_INT support. I tried something like this a month ago but didn't understand there is "elf within elf".

[TheShoura]

How do you use the --platform flag?

http://bitcointalk.org/index.php?topic=6745.0

[marcus_of_augustus]

New version is up. Changes:

- BFI_INT (~10% performance improvement)
- TCP keep-alive

thanks for the update.

Can confirm that poclbm is now up to speed with phoenix (maybe even slightly better but will have to watch shares found over time to confirm).

HD 5970, ubuntu 10.10, SDK 2.1, fglrx 11.2
seems that mem. under-clock speeds (775/300) with linux now works with m0mchill ... yay. Was this connected to the BFI_INT capability?

Faster, cooler, quieter.

[Cablesaurus]

I can confirm I'm getting greatly increased speeds.
On my XFX 5870's I went from ~330MHash to ~365MHash, each card, with no noticeable increase in temperature.

[error]

After updating to 20110428 I have the following error (Fedora 15 x86_64; Python 2.7.1). Previous version ran fine.

EDIT: Nevermind, I just git cloned the repo instead of downloading stuff manually as this outdated guide suggests (maybe it should be updated). Everything is working.

[CoinMan]

@m0mchil Thank you for the recent update to your miner!  It added 30 Mh/s to my miner.  How sweet that is!

[IlbiStarz]

How can you tell what version you are running?

Also, how come in cmd, I'm getting about 320Mhash/s (I have 2 cores), but in deeptbit, the average is only 400Mhash/s.

[LightRider]

Holy crap, I haven't updated in a whole month!

[wiktor]

I know this is a tired subject, but I'm another victim of the "100% CPU" problem. Though everyone else seemed to have it with AMD cards, and I have an nVidia 9500 running on amd64 Linux.

Code:

./poclbm.py --host=<...> --port=8332 --user=<...> --pass=<...> --frames=76 --vectors --device=0 --platform=0

It is actually using the GPU since it goes up to 5.5 Mhps (my CPUs only go up to 2 Mhps on their own, also I was running minerd at the same time on those CPUs), but still uses a large chunk of my CPU (1 whole CPU), so it's not that I have somehow selected the CPU driver.

I've tried playing with some of the settings, turning off vectors only made it worse (it ate 1 CPU and a bit of another), changing the value of frames to 0 or 30 had no difference.
I have tried running it without minerd hogging the CPUs but that made no difference…


If I get this to work I'll be able to pull off over 13 Mhps (4×2Mhps + 5.5 Mhps - rounding), but if the script itself takes up a whole CPU the minerd's performance halves, so I get hardly 9-10 Mhps.

Oh, and hi!, BTW ;-)

[LightRider]

I'm getting Long Poll Exceptions on BTCMine.com's pool.

Code:

30/04/2011 06:29:35, long poll exception:
Traceback (most recent call last):
  File "BitcoinMiner.pyo", line 259, in longPollThread
  File "BitcoinMiner.pyo", line 222, in request
  File "httplib.pyo", line 974, in getresponse
  File "httplib.pyo", line 391, in begin
  File "httplib.pyo", line 349, in _read_status
  File "socket.pyo", line 397, in readline
timeout: timed out

[JackSparrow]

Q: Does my video card / driver supports OpenCL?

AMD - 4xxx and up.

You mean HD-4xxx and up, do you?

https://secure.wikimedia.org/wikipedia/en/wiki/Radeon#Processor_generations

[RustyShackleford]

m0mchil, I've been using your miner for months now with no issues, but the latest version seems to be causing me an issue:

Exception in thread Thread-2:
Traceback (most recent call last):
  File "threading.pyo", line 525, in __bootstrap_inner
  File "threading.pyo", line 477, in run
  File "BitcoinMiner.pyo", line 272, in miningThread
  File "BitcoinMiner.pyo", line 373, in loadKernel
IOError: [Errno 13] Permission denied: '4cd591f539230b2adbff06c9d1a53ce6.elf'


Windows, 9800GTX+, using all the same flags I always have. Tried on two different pools just to make sure it wasn't a pool problem & both pools work fine with the version I was using before (3/11)

Any ideas?

[RustyShackleford]

m0mchil, I've been using your miner for months now with no issues, but the latest version seems to be causing me an issue:

Exception in thread Thread-2:
Traceback (most recent call last):
  File "threading.pyo", line 525, in __bootstrap_inner
  File "threading.pyo", line 477, in run
  File "BitcoinMiner.pyo", line 272, in miningThread
  File "BitcoinMiner.pyo", line 373, in loadKernel
IOError: [Errno 13] Permission denied: '4cd591f539230b2adbff06c9d1a53ce6.elf'


Windows, 9800GTX+, using all the same flags I always have. Tried on two different pools just to make sure it wasn't a pool problem & both pools work fine with the version I was using before (3/11)

Any ideas?

I solved this myself, It was a permissions/folder attributes issue. I don't know exactly what the correct permissions are, but an xcopy /O /K of the previous version's folder; subsequently updated with the contents of the new version; works fine.