mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   GpuOwl (https://www.mersenneforum.org/forumdisplay.php?f=171)
-   -   How to set up for running gpuOwl under Ubuntu (and other Linux) with OpenCL (https://www.mersenneforum.org/showthread.php?t=25601)

M344587487 2020-11-07 22:00

You may have more luck with the 3200G if you still have it, the A8-9600 is pre-Ryzen but more importantly pre-Vega so I'd be surprised if you can get iGPU compute working. Even the Vega iGPU's are not particularly well supported so going before them is a longshot.

DrobinsonPE 2020-11-07 23:05

I finally realized that GPUOWL was not compiling and that was why it was not running. Apparently openCL does not quite work.

I went back and ran /opt/rocm/opencl/bin/clinfo
It finds the integrated graphic but stops running in the SVM capabilities.
[CODE] Number of devices: 1
Device Type: CL_DEVICE_TYPE_GPU
Vendor ID: 1002h
Board name: Wani [Radeon R5/R6/R7 Graphics]
Device Topology: PCI[ B#0, D#1, F#0 ]
..............
SVM capabilities:
Coarse grain buffer: Yes
Fine grain buffer: Yes
Fine grain system: No
Atomics: No
Preferred platform atomic alignment: 0
Preferred global atomic alignment: 0
Preferred local atomic alignment: 0
^C[/CODE]
I still have the 3200G but I moved it to a different motherboard and it is currently running Windows 10. I am testing it with Prime 95, mfakto, and GPUOWL on Windows 10 right now but will switch the SSD to Linux soon and follow the same installation process that almost worked on the A8-9600.

MOD -- please feel free to delete my previous post because GPUOWL still did not work. That is unless you want to continue memorializing my failures.

Xyzzy 2020-11-08 00:01

[QUOTE=DrobinsonPE;562571]That is unless you want to continue memorializing my failures.[/QUOTE]Your post might help someone in the future, even if all it shows is what you tried to do. Sometimes any hit for information via a search engine is appreciated!

:tu:

DrobinsonPE 2020-11-08 05:02

[QUOTE=M344587487;562565]You may have more luck with the 3200G if you still have it, the A8-9600 is pre-Ryzen but more importantly pre-Vega so I'd be surprised if you can get iGPU compute working. Even the Vega iGPU's are not particularly well supported so going before them is a longshot.[/QUOTE]

I thought I would give it a try because ROCm said it might work. The A8-9600 is a “Bristol Ridge” APU.

[url]https://github.com/RadeonOpenCompute/ROCm#Hardware-and-Software-Support[/url]

The integrated GPUs in AMD APUs are not officially supported targets for ROCm. As described below, "Carrizo", "Bristol Ridge", and "Raven Ridge" APUs are enabled in our upstream drivers and the ROCm OpenCL runtime. However, they are not enabled in the HIP runtime, and may not work due to motherboard or OEM hardware limitations. As such, they are not yet officially supported targets for ROCm.

paulunderwood 2020-11-13 09:50

Ubuntu vs Debian
 
With the lastest ROCm Debian could not cope with the need for python3.8. After many hours of fiddling I ended up using Ubuntu instead. With the help of Ernst's guide -- see OP (needs to add user to render group) -- I got something clean working running the latest gpuOwl.

I lost 2 days GPU computing to this upgrade.

PhilF 2020-11-13 14:01

[QUOTE=paulunderwood;563085]With the lastest ROCm Debian could not cope with the need for python3.8. After many hours of fiddling I ended up using Ubuntu instead. With the help of Ernst's guide -- see OP (needs to add user to render group) -- I got something clean working running the latest gpuOwl.

I lost 2 days GPU computing to this upgrade.[/QUOTE]

I've been fighting that too. It appears that AMD is no longer supporting Debian -- on purpose.

Anyway, which version of Ubuntu did you use?

paulunderwood 2020-11-13 17:22

I opted for 20.04 LTS..

M344587487 2020-11-13 18:55

I've had trouble with 20.10 on the latest ROCm, the iGPU crashes under compute load and amusingly hangs the entire system which I haven't seen in years.

paulunderwood 2020-12-08 06:11

Debian/Ubuntu/Centos/Fedora
 
Debian has a problem with ROCm's latest drivers -- requires and uninstallable version of python.

Ubuntu was a disaster for me. OS slowed down after a few days then ROCm upgrade screwed the system,

Centos -- libraries too old

Fedora 33 -- bang on OS.

ZFR 2021-02-12 20:53

Thanks for the thread.

[QUOTE=ewmayer;547643]

[B]Use -maxAlloc to avoid out-of-memory with multi-jobs per card:[/B]

If you run multiple gpuowl instances per card as suggested in general for both performance and should-one-job-crash reasons, you need to take care to add '-maxAlloc [(0.9)*(Card memory in MB)/(#instances)]' to your program-invocation command line.[/QUOTE]

Just a question. Where exactly is the multiple instance suggestion taken from? I couldn't find anything about it in the readme. Are 2 instances enough, or should I run more?

Each instance needs its own folder, right? And -pool can be used so they share worktodo?

M344587487 2021-02-12 21:33

Consumer gear is limited to two instances per card, at least AMD cards are, trying to run a third will compile the kernel but never run it. Depending on the card model there may be a small throughput benefit to running two instances of gpuowl, test your card to find out. AFAIK you're correct about how to do it.


All times are UTC. The time now is 11:19.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.