-
Notifications
You must be signed in to change notification settings - Fork 38
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
v0.7.0 intermittantly hangs on Windows, segfaults on Linux at startup #32
Comments
I have mor einfo about the problem. In v0.7.0 my GPUs are reported as CL devices as well as Cuda, v0.6.2 only detects it coin@sys27:~/miners/verthash/VerthashMiner-0.7.0/build$ ./VerthashMiner -l Device list:OpenCL devices: CUDA devices: coin@sys27:~/miners/verthash/VerthashMiner-0.6.1/build$ ./VerthashMiner -l Device list(raw):OpenCL devices: None CUDA devices: It looks like --no-restrict-cuda is default. How to I force the miner to use Cuda? |
I have a workaround that solved the problem for my case with Cuda 10 but I don't think Cuda 10 is the issue because the I disabled the "Nvidia GPU to CUDA restrictions" check in main.cpp by commenting out the entire #ifdef HAVE_CUDA block. Update: The workaround may not fix the issue. In an ironic twist one system that had not seen the crash (started ok the Now I'm just confused. The only thing that is conststent is that it never crashes on v0.6.2. |
Hello. You can always select devices manually. |
I apologize for the confusion above, intermittant problems tend to be difficult to define. I'm not sure anything above is of any value except that v0.7.0 crashes intermittantly at startup on both Linux & Windows I've tried --cl-devices n and --all-cu-devices, it makes no difference. I see this on all my rigs. Is it just me that has this problem? |
jay, i too have encountered this. if you launch your windows "reliability monitor" you should see an entry for vertcoin crashing. it probably matches mine from three different machines https://i.imgur.com/PaN4aU7.png |
I was unable to reproduce this problem. |
Unfortunately it doesn't seem to have improved. It took 2 attempts to get 0.7.1 to start on Windows10 and after 4 tries Maybe this is a problem only with Maxwell cards. My rig with Pascal also had a Maxwell so it maybe it was the Maxwell that There seems to be some inconsistency in reporting available devices. Sometimes it also reports Nvidia as available for OpenCL Update: The reporting of Nvidia as OpenCL device doesn't seem to make a difference. In the latest test on one particular |
It's looking like it's a Maxwell only problem. v0.7.1 hasn't crashed with no Maxwell cards selected. |
I was able to test the latest version on Maxwell GPUs and there were no issues so far. |
GTX960 on ubuntu-20.04, compiled from source fails because it can't find a file but it exists in src/kernels/, console session `$ gdb --args ./VerthashMiner -o stratum+tcp://mine.zergpool.com:4534 -u x -p x --verthash-data verthash.dat --all-cu-devices --verbose For help, type "help". Thread 5 "VerthashMiner" received signal SIGSEGV, Segmentation fault.
|
Apologies for the garbled output for the CL test, but I don't think it was useful. It looks to me like the crash is due to data misalignment of job_id when calling the AVX2 version of strdup. Upon further thought it may be job_id is uninitialized considering it's the first job. Yup confirmed it. I added a log just before the crash. When it crashes the sctx job_id pointer is null. ./VerthashMiner -o stratum+tcp://mine.zergpool.com:4534 -u x -p x --verthash-data verthash.dat --all-cu-devices |
I added the following to the miner thread just before enterring the loop. It seems to prevent the crash.
[2021-03-15 17:39:47] INFO Found 0 OpenCL devices. Edit: you might want to tweak this a little. A multi-gpu rig will output a log for every thread. I also didn't check if |
So far no crashes with the job_id check at startup. While investigating this problem I noticed a couple of things you may be interested in following up on. You don't subscribe to stratum extranonce. All of the supporting code is present but never used. Other verthash miners support The other is calling stratum_gen_work from the miner thread. That's only necessary when the thread runs out of nonces It's just a couple of suggestions in case you're interested. As for the issue of the startup crash, I consider it technically resolved and await a new release with a fix. |
I have same issues with v0.7.2 release on Linux with Tesla M40 (Maxwell 2.0) GPU |
Windows 10, GTX-970.
Ubuntu 20.04, various Maxwell and Pascal cards.
Very often the miner hangs (Windows) or segfaults (Linux) on startup after the stratum connection but before the worksize set
message is displayed. Sometimes it starts ok. Once it's running it seems stable.
v0.6.2 doesn't have this problem, it always starts reliably, tested on Windows only.
Edit: Ubuntu 20.04 was tested with v0.6.1 with no segfaults.
Windows was tested with the precompiled binary. Ubuntu was tested using the precompiled binary and compiled from source.
Update: the problem is more persistent on Windows. I can't get v0.7.0 to work at all on my 2 Windows PCs, v0.6.2 has no
problems. The failure on Windows includes Turing cards.
The text was updated successfully, but these errors were encountered: