No, the source for the new phatk kernel came from phateus himself and is as up to date as it gets.

Oh yes there is no way to clean up after a dead GPU restarted thread. It's a dramatic manoeuvre where I disable the thread and create an entirely new context. See the problem is that when code hits the GPU and NEVER RETURNS there is no reliable way of resuscitating the code that called it. So I disable the thread (in case it ever comes back) and create an entirely new thread. If I free any of the ram from the original thread, it will often crash if the thread returns and the memory has been de-referenced from under it. This is all because it's done in c low level.

Executive summary: If you get lots of GPU sick scenarios that restart the GPU, decrease the intensity level with -I till it stops happening.

Now that one is a driver crash. Nothing to do with cgminer directly. It even caused a linux kernel warning about a prolonged soft lockup. Perhaps try a new kernel ati driver and ati sdk.

They are most definitely there for a purpose. 2 per GPU, Workio thread, staging thread, longpoll thread, watchdog thread, gpu restart thread, cpu restart thread, and then there's a thread spawned (and then killed) for every getwork request and every submit work.

The new kernel works things -even harder- so if you were getting sick GPUs before, you still will now. The difference is it should recover by itself. If you get sick GPUs all the time, I suggest dropping the intensity down till you find where they stop being sick.

Too good to pass up a new release even with only 2 real changes.

Version 1.5.6 (you all know where by now - links in first post).

Executive summary: New faster phatk kernel (2.2) courtesy of phateus porting it (thanks!) and with my own custom modifications on top. Fixed multiple pool longpoll management.

New phatk and poclbm kernels in git tree for those who've been waiting for them.

Hopefully new release just around the corner with that

If you want to restore your tree without losing your changes, create a new branch and reset the master to the last one before your commits.

git checkout master
git branch newphatk
git reset --hard 58eb4d58599521933a3fef599e1dcba4f996dadc
git pull

that will pull my changes into the master branch and your personal changes will be in newphatk. Unfortunately your github account has a messed up master now so

git push -f

will force the changes to propagate. Do not use this command normally as it makes it impossible for people pulling from your branch to keep in sync.

Unfortunately there is not a way to do that at the moment. In the readme you'll see it recommends running 2 instances of cgminer selecting which GPUs to use. Use one instance in dynamic mode and the other with a set intensity.

Very good work. Nice of you to figure out how to do git and all as well. Don't worry about the merge, I've taken care of everything and cherry picked your changes as I needed to. I've modified a few things too to be consistent with cgminer's code and there is definitely a significant speed advantage thanks to your changes. Note that if you're ever working on git doing your own changes, do them to a branch that's not called master as you may end up making it impossible to pull back my changes since I won't necessarily take all your code. Thanks again, and I'm sure the cgminer users will be most grateful.

19 new blocks in 17 minutes? That doesn't even sound like bitcoin...

Seems to me like you've got it all under control, so I'll leave you to finish up. Thanks for your involvement. However I don't want multiple phatk kernels so just replace the current one in-situ and don't bother enumming a different kernel. As for the output code, I prefer to use 4k so feel free to do it your way, but be aware I plan to change it back.

Yes indeed, that's what I've spent weeks working on to get right. To restart it if it can only. The 5 minutes was a nominal choice based on the theory that if it gets sick, giving it a few minutes to cool down will make it more likely to respond. Note that if it gets sick a LOT, you may benefit from decreasing the intensity and sacrificing a few MH just to get a more stable throughput (eg -I 7 instead of .

Good idea. Though I might just put the pool number rather than the whole URL because the URL can be quite long leading to a very long log message. ?

You could just tell me what to do to interface it with cgminer (i.e. what new variables you want) and I'd copy most of your kernel across. Only the return code and define macros are actually different in cgminer in the kernel itself.

New version. NOTE the windows binary should CPU mine again (-O3 was unstable, but -O2 works) and includes a cpumine only executable as well.

Source:
http://ck.kolivas.org/apps/cgminer/cgminer-1.5.5.tar.bz2
http://ck.kolivas.org/apps/cgminer/cgminer-1.5.5.tar.lrz

Ubuntu:
http://ck.kolivas.org/apps/cgminer/cgminer-1.5.5-x86_64-built.tar.bz2
http://ck.kolivas.org/apps/cgminer/cgminer-1.5.5-x86_64-built.tar.lrz

Windows 32 bit binary:
http://ck.kolivas.org/apps/cgminer/cgminer-1.5.5-win32.zip


Version 1.5.5 - August 16, 2011

- Rework entirely the GPU restart code. Strike a balance between code that
re-initialises the GPU entirely so that soft hangs in the code are properly
managed, but if a GPU is completely hung, the thread restart code fails
gracefully, so that it does not take out any other code or devices. This will
allow cgminer to keep restarting GPUs that can be restarted, but continue
mining even if one or more GPUs hangs which would normally require a reboot.
- Add --submit-stale option which submits all shares, regardless of whether they
would normally be considered stale.
- Keep options in alphabetical order.
- Probe for slightly longer for when network conditions are lagging.
- Only display the CPU algo when we're CPU mining.
- As we have keepalives now, blaming network flakiness on timeouts appears to
have been wrong.     Set a timeout for longpoll to 1 hour, and most other
network connectivity to 1 minute.
- Simplify output code and remove HW errors from CPU stats.
- Simplify code and tidy output.
- Only show cpu algo in summary if cpu mining.
- Log summary at the end as per any other output.
- Flush output.
- Add a linux-usb-cgminer guide courtesy of Kano.

Courtesy of znort:
    Add automated benchmark of the CPU hashers
   
        The --algo switch now accepts the "auto" argument.
   
        When "auto" is passed to --algo, cgminer starts by benchmarking
        all the CPU algorithms it nows about and picks the fastest.
   
        This is useful for benchmarking, but also for folks who run
        cgminer on a large number of heterogeneous computers because
        it saves them from having to configure each instance optimally.
   
        Caveat emptor: depending on the platform, some algorithms will
        fail with "illegal instruction" (e.g. via padlock code on non via
        platforms, or SSE4 code on non SSE4 platforms).
   
        To protect against this, cgminer runs the benchmarks in a child
        process. The crash, if any occurs in the child, and the parent
        marks this algo as "fails" and continues benchmarking the next
        algorithm.

You need pkg-config installed.

Well it's the wrong thing for the pool to be accepting stale shares since that work is likely useless, so it's in the pool's best interest to not do so. On the other hand, if you're mining for such a pool, I guess you get to "cheat" with this useless work, so I can see how it would be an advantage to you. I could easily add another switch to the command line such as --submit-stale .

edit: This option has been committed to the tree.

Okay so I had some generous people offer me BTC online to get the windows binary version working, and I DID see someone offer a BTC bounty to get the windows binary working, so I've fluffed around for the last 2 days on a windows virtual machine to build some win32 binaries.

http://ck.kolivas.org/apps/cgminer/cgminer-1.5.4-win32.zip

Hopefully people appreciate the effort.

I see. Well I can't even get it compiling on mingw32. Can't figure out how to install the remaining dependencies of libcurl though I've installed it 10 different ways and even ended up building it myself which took hours. Gotta remember I haven't used windows in over a decade in any meaningful way.