Older Radeon/ATI GPU and Ubuntu drivers - revisited.

Greetings,

I wanted to share some things and ask for spirtual guidance in other areas.

Its about getting AMD drivers in Ubuntu.

Yes, AMDGPU-Pro works for the RXnnn GPUs but not others.

Thank goodness, the mesa video driver picked up OpenCL support.

I started this thread/topic in case there are others still trying to find the driver that fits.

A history, Pre Ubuntu 16.04, the OpenCL driver was fglrx. It became not-workablewhen new linkage to Xorg went into place. Drat.

So things, should be simple - but are not - at least to me.

Here is how something simple gets complicated. I hae two PC - both running Ubuntu Mate with Radeon/ATI 7750. I kept one with upgrading to a new release - so I could keep running Einstein GPU WorkUnits. The other - I just ran CPU WU.

Then along came the mesa driver with Open CL. After installing, BOINC could recognize the GPU, but the WU failed in the first four seconds.

I found a PPA that gave an updated driver that fixed theproblem. (Thanks Oibaf and Pauli Miguel Dias !!!)

Then, I was crunching merrily away on both machines.

Time passes.  After a while, there were bo updates to the Ubunt that ran FGLRX. Maybe I should upgarde it too? (Add Halloween buwahaha here.)

There are now at 4 PPAs for the Mesa drivers. Oh, I wish I had written the name of that PPA down. I would not have mattered (I think) because that PPA was now only supported for the Long Term Support (LTS) release and the current release.

The newest PPA did not work with the newest release. (ubuntu-x-swat/updates.) I couldn't get it to work for E@H  on the LTS either.

Paulo is running two versions of PPA - a weekly working vesion and a stable version.

I have two approaches to learning - the school of hard knocks - repeated trials of failure, slowly looking for cause and effect; and asking knowledgable people on the forums that undestand what is working. I think I waited too long.

I have tried several times with a freshly formatted partition for '/' and a new install. (I keep /home separated for this.) But I have not gotten a combination to work yet.

 

My attempt to understand stops when I try to understand how the PPA replaces packages and yet keeps all of the link dependencies to other code. I relize that I have to give an install command for an opencl driver related package for the PPA to kick in.

My worry is that I don't know the correct order. I have tried a shotgun approach, but that failed and ended up with  apt failures with locks.

Here is what I have done so far: installed Ubuntu 16.04;  Placed the PPA into apt; done the first update and upgrade (about 447 packages updated.) Installed BOINC; started BOINC and checked the event log.

BOINC doesn't recognize the GPU card (Found the OpenCL shared library file). This expected since I haven't installed a driver that BOINC would use for my Radeon/ATI card.

So I will stop the BOINC manager; stop the BOINC client and install the correct package.

aptitude versions mesa-opencl-icd shows 4 versions - regular and debug for both 64 and 32 bit.

aptitude install -s mesa-opencl-icd    doesn't show me which one aptitude will choose. I'll use the -dbg version because it only is supplied by the PPA.

Reboot. Hope a screen comes up after the VGA driver used when GRU starts goes blank for 15 seconds. Yay. Log in. Restart the BOINC Client with the xhost stuff. start BOIC and check the event log.

The PC with that works Has:

Sun 21 May 2017 11:59:46 AM EDT |  | OpenCL: AMD/ATI GPU 0: AMD CAPE VERDE (DRM 2.46.0 / 4.8.0-52-generic, LLVM 4.0.1) (driver version 17.2.0-devel, device version OpenCL 1.1 Mesa 17.2.0-devel, 2048MB, 2048MB available, 512 GFLOPS peak)

The PC I just installed has driver version 17.1.0-padokaPPA. Try a WU and see if it fails.......

Drat. It errors out.

See  https://einsteinathome.org/host/12533566

and then the task that was just returned:     https://einsteinathome.org/task/647985691

It says:

<h3>Stderr output</h3>

<pre><core_client_version>7.6.31</core_client_version>
<![CDATA[
<message>
process exited with code 69 (0x45, -187)
</message>
<stderr_txt>
02:59:20 (3920): [normal]: This Einstein@home App was built at: Jan 16 2017 08:09:16

02:59:20 (3920): [normal]: Start of BOINC application '../../projects/einstein.phys.uwm.edu/hsgamma_FGRPB1G_1.18_x86_64-pc-linux-gnu__FGRPopencl1K-ati'.
02:59:20 (3920): [debug]: 1e+16 fp, 2.3e+09 fp/s, 4619059 s, 1283h04m19s27
command line: ../../projects/einstein.phys.uwm.edu/hsgamma_FGRPB1G_1.18_x86_64-pc-linux-gnu__FGRPopencl1K-ati --inputfile ../../projects/einstein.phys.uwm.edu/LATeah0029L.dat --alpha 4.42281478648 --delta -0.0345027837249 --skyRadius 2.152570e-06 --ldiBins 15 --f0start 1092.0 --f0Band 8.0 --firstSkyPoint 0 --numSkyPoints 1 --f1dot -1e-13 --f1dotBand 1e-13 --df1dot 3.344368011e-15 --ephemdir ../../projects/einstein.phys.uwm.edu/JPLEPH --Tcoh 2097152.0 --toplist 10 --cohFollow 10 --numCells 1 --useWeights 1 --Srefinement 1 --CohSkyRef 1 --cohfullskybox 1 --mmfu 0.1 --reftime 56100 --model 0 --f0orbit 0.005 --mismatch 0.1 --demodbinary 1 --BinaryPointFile ../../projects/einstein.phys.uwm.edu/templates_LATeah0029L_1100_10097730.dat --debug 1 --device 0 -o LATeah0029L_1100.0_0_0.0_10097730_0_0.out
output files: 'LATeah0029L_1100.0_0_0.0_10097730_0_0.out' '../../projects/einstein.phys.uwm.edu/LATeah0029L_1100.0_0_0.0_10097730_0_0' 'LATeah0029L_1100.0_0_0.0_10097730_0_0.out.cohfu' '../../projects/einstein.phys.uwm.edu/LATeah0029L_1100.0_0_0.0_10097730_0_1'
02:59:20 (3920): [debug]: Flags: X64 SSE SSE2 GNUC X86 GNUX86
02:59:20 (3920): [debug]: glibc version/release: 2.23/stable
02:59:20 (3920): [debug]: Set up communication with graphics process.
boinc_get_opencl_ids returned [0x20fb258 , 0x7feef3f6aec0]
Using OpenCL platform provided by: Mesa
Using OpenCL device "AMD CAPE VERDE (DRM 2.43.0 / 4.4.0-78-generic, LLVM 4.0.0)" by: AMD
Max allocation limit: 1503238553
Global mem size: 2147483648
OpenCL compiling FAILED! : -11 . Error message: fatal error: cannot open file '/usr/lib/clc/verde-amdgcn-mesa-mesa3d.bc': No such file or directory

---

This is the first time I saw this error. Usually it is an 'Atomic' command or in invalid bitcast

I compared /usr/lib/clc.

The working pc has:

lrwxrwxrwx   1 0 0      15 Feb 25 00:30 sumo2-r600--.bc -> cedar-r600--.bc
lrwxrwxrwx   1 0 0      15 Feb 25 00:30 sumo-r600--.bc -> cedar-r600--.bc
-rw-r--r--   1 0 0 5382384 Feb 25 00:30 tahiti-amdgcn--.bc
-rw-r--r--   1 0 0 5406792 Feb 25 00:30 tahiti-amdgcn-mesa-mesa3d.bc
lrwxrwxrwx   1 0 0      17 Feb 25 00:30 tonga-amdgcn--amdhsa.bc -> amdgcn--amdhsa.bc
lrwxrwxrwx   1 0 0      18 Feb 25 00:30 tonga-amdgcn--.bc -> tahiti-amdgcn--.bc
lrwxrwxrwx   1 0 0      28 Feb 25 00:30 tonga-amdgcn-mesa-mesa3d.bc -> tahiti-amdgcn-mesa-mesa3d.bc
lrwxrwxrwx   1 0 0      15 Feb 25 00:30 turks-r600--.bc -> barts-r600--.bc
lrwxrwxrwx   1 0 0      18 Feb 25 00:30 verde-amdgcn--.bc -> tahiti-amdgcn--.bc
lrwxrwxrwx   1 0 0      28 Feb 25 00:30 verde-amdgcn-mesa-mesa3d.bc -> tahiti-amdgcn-mesa-mesa3d.bc

---

the PC wih the error has NO names that include "mesa"

This is my 4th or 5th dead end. I need a new approach.

Please offer positive suggestions.

Thanks,

Jay