I’ve never seen anything like this. Maybe run it in GDB so we can at least get a location of the crash?
Help please, I received the following error below while attempting to run the program on Fedora 18:
OpenCL error at platform_ids: -1001
Am I missing a library for Nvidia’s OpenCL runtime?
I’m ready to release open source GPU miner for beta testing, but can’t thread… restriction for newbies=) What I must do for thread start access rights getting ?
I have not tried on Fedora. I am using the Nvidia binary drivers from the NV website instead of debian packages. On debian if you try to use the packaged drivers there is a whole myriad of stuff that must be installed. In particular the opencl-icd’s. If fedora has some version of clinfo you can check to make sure your GPU is being detected as an available processing device. It may also freak out if there is a non GPU opencl device, but I can fix that.
[quote=“jonpry, post:44, topic:2147”]@SuperComputer
I have not tried on Fedora. I am using the Nvidia binary drivers from the NV website instead of debian packages. On debian if you try to use the packaged drivers there is a whole myriad of stuff that must be installed. In particular the opencl-icd’s. If fedora has some version of clinfo you can check to make sure your GPU is being detected as an available processing device. It may also freak out if there is a non GPU opencl device, but I can fix that.[/quote]
Awesome, thank you. It was a bad library path in the following file:
/etc/OpenCL/vendors/nvidia.icd
I found your code at https://github.com/eXtremal-ik7/xpmminer/tree/master/src/OpenCL some of it is pretty good. What are you doing in FermatTest384 with the precomputed values? This reduces number of times through the loop somehow?
GTX 780 Ti Classified @ 1.084 GHz (Kepler)
Val/h: 3.34795260 - PPS: 51 - SPS: 287.59017944 - ACC: 0 - Primorial: 41 Chain/Hr: 6: 133.00 7: 12.84 8: 1.83
New Block: 492397 - Diff: 10.68791342 / 7.00000000
Valid/Total shares: [ 15 / 16 ] - Max diff: 8.16436386
6ch/h: 129.40433351 - 145 [ 46 / 45 / 54 ]
7ch/h: 12.49421151 - 14 [ 5 / 6 / 3 ]
8ch/h: 1.78488736 - 2 [ 0 / 2 / 0 ]
Share Value submitted - Last Block/Total: 0.00000000 / 3.64999938
Current Primorial Value: 41
Not looking too good for the fastest Kepler card.
Only using 5 SMX though.
GTX 750 Ti @ 1.254 GHz (Maxwell)
Val/h: 2.70819187 - PPS: 89 - SPS: 495.82644653 - ACC: 0 - Primorial: 41 Chain/Hr: 6: 238.32 7: 21.67 8: 1.08
New Block: 492494 - Diff: 10.68797654 / 7.00000000
Valid/Total shares: [ 21 / 21 ] - Max diff: 8.42951488
6ch/h: 240.45310160 - 226 [ 79 / 75 / 72 ]
7ch/h: 21.27903554 - 20 [ 9 / 8 / 3 ]
8ch/h: 1.06395178 - 1 [ 1 / 0 / 0 ]
Share Value submitted - Last Block/Total: 0.00000000 / 2.49999952
Current Primorial Value: 41
Looking good for little Maxwell
Yes, it reduces total number of times (16 + 384/4 = 112) and removes threads divergence, speedup about 50% compare to /standard/ version.
Yes, if you optimize your code for big Kelper, the Primecoin mining throughput will be at least 4x that of little Maxwell’s.
Yes, it reduces total number of times (16 + 384/4 = 112) and removes threads divergence, speedup about 50% compare to /standard/ version.[/quote]
Thanks! eXtremal.ik7!!
any updates to nvidia miner?
./jhprimeminer1.02 -o http://ypool.net:10034 …
SM#:2 Thread: 163 Origin: ba267183cf2224726e8f8bdaebbd1cdf1a58e7f11b0013511993f95e41cad052822645bb5e052a600
Error: -9999
Type: 2
Name: fermat
OpenCL error at clQueue_wait: -9999
SM#:1 Thread: 407 Origin: 6f07ab39614b78f0e7c7c0f0a705132da5558ff8c186cbe5cd8550fa0a567a07445c7a420bef09e000
Error: -9999
Type: 2
Name: fermat
OpenCL error at clQueue_wait: -9999
Yes, it resolves the issue after restarting the miner, but it happens frequently and usually within a few minutes into mining. I have not yet exceeded 90 minutes of running the miner without it crashing.
I run the miner from within a script that restarts it. So mostly I just try to catch errors and exit. This is a bigger problem if the miner is not exiting. I figure if it restarts 1 time per hour you losing about 1 second of mining capacity which just doesn’t seem like a problem to me. I can try increasing some things. Fermat goes nuts if too many primes are found in the sieve.
I will use your provided script and give it a try.
if this is freq enuff, can make use of cronjob to kill and restart the script every 1 hour.
However, any plans to fix this bug or any new updates ?
any ways to overclock 750ti?