Topic: OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.0 - page 843. (Read 5805998 times)

New release version.

Source only at the moment:
http://ck.kolivas.org/apps/cgminer-1.2.5.tar.bz2

New features:
- I removed the -n option being no dynamic and left it how it was as number of devices
- Setting an intensity now of any value automatically disables the dynamic feature. It was basically completely overrided if dynamic was enabled, so these are now mutually exclusive features.
- I've added signal handling to try and clean up at the end and return the shell/console neatly if you press ctrl-c. I believe you need to use ctrl-break on windows still (when the binary comes out).
- There is a new option:
 --text-only|-T      Disable ncurses formatted screen output
- I've made it so that the watchdog thread only restarts threads if they're inactive in the setting of normal network conditions and will not try to restart them if the network connectivity is down.
- cgminer now tests on startup that it can connect at least once with the url and credentials to not try and reconnect indefinitely.
- There is no "fake work" at startup now with one hard lot of work handed to the gpu mining thread which should prevent the false positive hardware errors reported.
- CPU mining has been fixed, it was not resetting back properly on getting a new work item.
- The status window in curses mode will update more reliably and do less work now.
- Modifications to build on mingw32 for windows and now to build on osx courtesy of Ycros
- A status summary is now printed on exiting cgminer. It looks like this:

---
Summary of runtime statistics:

Started at [2011-07-15 01:05:28]
Runtime: 0 hrs : 12 mins : 13 secs
Average hashrate: 178.4 Megahash/s
Queued work requests: 33
Share submissions: 28
Accepted shares: 24
Rejected shares: 4
Reject ratio: 14.3
Hardware errors: 0
Efficiency (accepted / queued): 73%
Utility (accepted shares / min): 1.97/min

Discarded work due to new blocks: 4
Stale submissions discarded due to new blocks: 0
Unable to get work from server occasions: 0
Work items generated locally: 0
Submitting work remotely delay occasions: 0
New blocks detected on network: 2

Summary of per device statistics:

 GPU 0: [178.4 Mh/s] [Q:33  A:24  R:4  HW:0  E:73%  U:2.04/m]


... I think I need a little break now

No, this is what the curses UI looks like

Ah, it works if I leave off -n. But I can only run 1 instance in linux, 2 doesn't seem to work.

Without -n I'm about 7-8 mhash behind hashkill on my dedicated miner and 20-25 on my workstation

what exactly have you done? I had also issues on 64 bit natty, but now it works....

I don't know exactly why, but that made it. I re- ./configured and "make"ed again and now it works.

Thanks fot the help. Now i will do a bit of testing and optimizing to look what the miner can do in comparison to phoenix. But what is weird: the CPU usage of my triple Core CPU is about 30% on each core. While using 4 Instances of Phoenix, i used only 3 to 4 % of each core. So the XPU usage is quite higher with the cgminer. Any theories on that?


EDIT:
==> SOLVED:I forgot the -n flag. Now the CPU usage is around 2% for each core.

Trying to install on Natty 64bit.

I can build both tarball and git versions, but neither works.

The 1.2.4 tarball just says: Segmentation fault
The git repo says: 2 GPUs Detected. Then back to shell

trying against 2.4sdk, 11.6 catalyst. Dual 5830s.

I have no problems running hashkill, phoenix or poclbm on this box.

You can log to a file with the curses output. Just redirect stderr to a file.
./cgminer blah 2>mylogfile.txt

cgminer detects if stderr is going to a file or just the console and decides on whether to log to it or not.

Crashes are part of the game early on in code, and hopefully I can close in on all of them.

More code to come shortly (most in git tree already)...

I'm telling you, there's some kind of a weird display bug when a share gets solved in under 5 seconds, it shows hashrates way higher than what my card can do

Couple of issues (on windows)

1. --ndev | -n also uses -n, which is also used to turn off dynamic intensity.
2. ctrl-break doesn't restore the command prompt back to normal - no cursor, no scrolling, insert mode turned on, etc. and ctrl-c doesn't work.
3. default detected parameter for 6870 (whatever they are) are not optimal. -w 128 -v 2 is better. i think it detects -w 256.
4. LP doesn't work with bithopper's. don't know if this is because of bithopper or some issue w/ LP on cgminer side. it just loops LP requests.
5. no output at all if missing url, user, pass. For some pools pass is optional.
6. Even if it does not connect to host it is hashing...what is it hashing? this is with an invalid/incorrect url

I have a 64bit natty here, too, and it works. Could you try without the GPU mining stuff to see whether it compiles that way? If so, there might be a conflict between the specified LDFLAGS and the expected default.

Make install doesn't work, sorry. Just make it in its own directory and run it from there. Make sure to do ./configure again after you install any new libraries etc.

Every time i try to run it on my pc(under windows XP SP3 x32) i just get the following error:
http://img195.imageshack.us/img195/1565/screenshot0dq.png

Think so too, the only bottleneck performance wise will be how fast or how much ALU OPs a certain GPU can handle .

Dia

Hardly any data is actually moved between the gpu and the cpu. It would be extremely unlikely to be creating a bottleneck.

Question about the GPU computing source code:  What method is being used to move data between the CPU and the GPU?  If using a caching method, you've created a slight bottleneck.  Since the GPU memory is non-cachable, using a caching method will cause a major slowdown.  Granted, I'm not awfully familiar with GPU computing code, it is certainly something to keep in mind.

You cannot replace the kernels in cgminer with other kernels. They are incompatible. You will get mostly rejects and errors if you do that. The cgminer included kernel is the most advanced available phatk kernel at the moment anyway.

Hi, the phatk110711.cl is basicly my version (see: http://forum.bitcoin.org/index.php?topic=25860) with some mods to make it work with cgminer + added support for 4-component vectors, which was both added by ckolivas.

Dia

I really don't care so much about electricity. BTW is becuse I hve multiple machine set on different places and collect all for make some BTC.

If it's not possible, I would suggest UFASoft's miner on BTCGuild or Slush's pool.  Only if you're not paying for electricity, that is.

Also, what compels you to mine on a netbook?  You can't be getting more than 4-5 MH/s out of it....