Pages:
Author

Topic: My improvements to poclbm (Read 7544 times)

hero member
Activity: 698
Merit: 500
June 24, 2011, 09:13:50 AM
#21
any chance that we get compiled binaries  Embarrassed

fgpaminer compiled them in his version (which also uses the phatk-kernel).
You can find it here: http://forum.bitcoin.org/index.php?topic=19169.0

read http://forum.bitcoin.org/index.php?topic=19169.msg272518#msg272518 on that thread

and for ati 5770, poclbm > phatk
Ali
member
Activity: 84
Merit: 10
June 24, 2011, 06:12:32 AM
#20
any chance that we get compiled binaries  Embarrassed

fgpaminer compiled them in his version (which also uses the phatk-kernel).
You can find it here: http://forum.bitcoin.org/index.php?topic=19169.0
hero member
Activity: 698
Merit: 500
June 24, 2011, 05:37:40 AM
#19
any chance that we get compiled binaries  Embarrassed
member
Activity: 295
Merit: 98
June 23, 2011, 07:51:01 AM
#18
Great work luke-jr

I made some minor tweaks to the status display line to show three digits of precision on the display of reject and efficiency percentages.

A lot of times rejects is between %0 - %1 and being able to see the fraction of a percent is handy (well to me anyway..  Wink  )

 here's and examples of the modified display...

Code:
[401.772 MH/s (~444 MH/s)] [Rej: 13/1635 (0.795%)] [GW: 2704 (Eff: 59.985%)]


and here's the diff of BitcoinMiner.pl that implements it...
Code:
bigfoot@brontosaurus:~/bitcoin-luke-jrs-poclbm$ diff ../luke-jr-poclbm/bitcoin-luke-jrs-poclbm/BitcoinMiner.py BitcoinMiner.py
149,150c149,150
<                       eff = self.shareCount[1] * 100 / self.getworkCount
<               self.say('[%.03f MH/s (~%d MH/s)] [Rej: %d/%d (%d%%)] [GW: %d (Eff: %d%%)]', (rate, round(estRate), self.shareCount[0], totShares, self.shareCount[0] * 100 / totSharesE, self.getworkCount, eff))
---
>                       eff = float(self.shareCount[1]) * 100 / float(self.getworkCount)
>               self.say('[%.03f MH/s (~%d MH/s)] [Rej: %d/%d (%0.03f%%)] [GW: %d (Eff: %0.03f%%)]', (rate, round(estRate), self.shareCount[0], totShares, float(self.shareCount[0]) * 100 / float(totSharesE), self.getworkCount, eff))

full member
Activity: 129
Merit: 100
June 22, 2011, 10:31:43 PM
#17
What does the Efficiency value mean?
Same thing it's always meant with regard to mining. Percentage of getworks that result in a share. With roll-ntime, fast miners can easily milk lots of shares out of each getwork.
ok so 100% is 1 share for 1 getwork?

edit: ok i get it now. eff = accepted shares/getwork*100 - simple really Wink
legendary
Activity: 2576
Merit: 1186
June 22, 2011, 10:29:50 PM
#16
What does the Efficiency value mean?
Same thing it's always meant with regard to mining. Percentage of getworks that result in a share. With roll-ntime, fast miners can easily milk lots of shares out of each getwork.
full member
Activity: 129
Merit: 100
June 22, 2011, 10:27:32 PM
#15
What does the Efficiency value mean?
yeah, i'm wondering the same thing Wink

is 359%,380%: good, great, ok, awesome Tongue
Ali
member
Activity: 84
Merit: 10
June 22, 2011, 03:42:43 AM
#14
I'm using your branch now fpgaminer. The phatk-kernel gives me a boost of about 12 MH/s.

What does the Efficiency value mean?
hero member
Activity: 560
Merit: 517
June 22, 2011, 03:32:16 AM
#13
Quote
That's what I do too but it raises the error.
Yikes, yeah. Well, to be honest, pyopencl looked like a bit of a pain to compile when I took a look at it last, and I couldn't find anything pre-compiled. Hopefully someone else can chime in here.

Other than that, you may want to take a look at my branch, for which I did put together a ready to run Windows exe:
http://forum.bitcoin.org/index.php?topic=19169.msg254264#msg254264

For that one I copied over the pyopencl binaries from the original poclbm packaging, so I didn't have to compile pyopencl.  Grin

My branch includes luke jr's modifications, but it also uses the phatk kernel (which may give you a speed boost).
Ali
member
Activity: 84
Merit: 10
June 21, 2011, 02:39:43 AM
#12
You use it by doing "python2.7.exe bitcoin-luke-jrs-poclbm\poclbm.py"

That's what I do too but it raises the error.

Code:
Traceback (most recent call last):
  File "poclbm.py", line 3, in
    import pyopencl as cl
ImportError: No module named pyopencl

How to install pyopencl? I never used python.
hero member
Activity: 560
Merit: 517
June 20, 2011, 11:08:53 PM
#11
Quote
poclbm looks like python too, but has binaries. and how do i use the source?
The Windows release of poclbm uses py2exe to create an EXE for us lazy Windows users Smiley

You use it by doing "python2.7.exe bitcoin-luke-jrs-poclbm\poclbm.py"


Luke-Jr: Great work! I merged this with my branch and it seems to be working wonderfully so far. I haven't measured connection performance, but it's very nice to have it tally up the rejected statistics and such. Thank you for this great improvement to poclbm  Cheesy
newbie
Activity: 46
Merit: 0
June 20, 2011, 10:33:09 PM
#10
Please, post also instructions Smiley
yep too blind to find the download button

I were assuming a set of patches, but apparently not. No reason to be an jackass for a valid point.
no, i can't find the binaries either, just the source files.
It's Python. There are no binaries.
poclbm looks like python too, but has binaries. and how do i use the source?

Code:
c:\Python27>python2.7.exe bitcoin-luke-jrs-poclbm\BitcoinMiner.cl
  File "bitcoin-luke-jrs-poclbm\BitcoinMiner.cl", line 2
    typedef uint2 u;
                ^
SyntaxError: invalid syntax

c:\Python27>python2.7.exe bitcoin-luke-jrs-poclbm\BitcoinMiner.cl  -d 0 -o mining.eligius.st --port=8337 --user=asdfghjkl --pass=x
  File "bitcoin-luke-jrs-poclbm\BitcoinMiner.cl", line 2
    typedef uint2 u;
                ^
SyntaxError: invalid syntax

c:\Python27>python2.7.exe bitcoin-luke-jrs-poclbm\poclbm.py  -d 0 -o mining.eligius.st --port=8337 --user=asdsfgghhjkl --pass=x
Traceback (most recent call last):
  File "bitcoin-luke-jrs-poclbm\poclbm.py", line 3, in
    import pyopencl as cl
ImportError: No module named pyopencl
legendary
Activity: 2576
Merit: 1186
June 20, 2011, 08:47:12 PM
#9
Please, post also instructions Smiley
yep too blind to find the download button

I were assuming a set of patches, but apparently not. No reason to be an jackass for a valid point.
no, i can't find the binaries either, just the source files.
It's Python. There are no binaries.
newbie
Activity: 46
Merit: 0
June 20, 2011, 08:04:22 PM
#8
Please, post also instructions Smiley
yep too blind to find the download button

I were assuming a set of patches, but apparently not. No reason to be an jackass for a valid point.
no, i can't find the binaries either, just the source files.
sr. member
Activity: 402
Merit: 250
June 20, 2011, 07:43:42 PM
#7
Please, post also instructions Smiley
yep too blind to find the download button

I were assuming a set of patches, but apparently not. No reason to be an jackass for a valid point.
newbie
Activity: 46
Merit: 0
June 20, 2011, 06:24:15 PM
#6
Please, post also instructions Smiley
yep too blind to find the download button
sr. member
Activity: 402
Merit: 250
June 19, 2011, 04:31:54 PM
#5
Please, post also instructions Smiley
hero member
Activity: 767
Merit: 500
June 18, 2011, 02:14:39 PM
#4
Have you ever found a pool that has accepted a share that your modified client has attempted to resend?

Normally a rejected share won't be fixed by resending it.
Rejected shares aren't resent. That's only for network issues.

Okay understood. Sounds cool.

Will
legendary
Activity: 2576
Merit: 1186
June 18, 2011, 02:11:08 PM
#3
Have you ever found a pool that has accepted a share that your modified client has attempted to resend?

Normally a rejected share won't be fixed by resending it.
Rejected shares aren't resent. That's only for network issues.
hero member
Activity: 767
Merit: 500
June 18, 2011, 02:08:26 PM
#2
Have you ever found a pool that has accepted a share that your modified client has attempted to resend?

Normally a rejected share won't be fixed by resending it.

Will
Pages:
Jump to: