new ubuntu release - Ubuntu problem

jay
jay
Joined: 25 Jan 07
Posts: 99
Credit: 84,044,023
RAC: 0
Topic 228606

Problem (Linux)

EAH/Boinc not accepting link

/usr/lib/clc/verde-amdgcn-mesa-mesa3d.bc   has a soft-link to
     tahiti-amdgcn-mesa-mesa3d.bc ( in the same directory.)

https://einsteinathome.org/task/1383285041

says
OpenCL compiling FAILED! : -11 . Error message:
        fatal error:
        cannot open file '/usr/lib/clc/verde-amdgcn-mesa-mesa3d.bc':
        Opaque pointers are only supported in -opaque-pointers mode (Producer:
        'LLVM15.0.4' Reader: 'LLVM 15.0.4')

------

Well !!

The link was accepted prviuosly

I will re-install BOINC and try again.

 

Any other suggestions??

THANKS!!

Jay

jay
jay
Joined: 25 Jan 07
Posts: 99
Credit: 84,044,023
RAC: 0

A retry ahs the same

A retry has the same problem.  [edit typo]

Here is the full log

Stderr output

<core_client_version>7.18.1</core_client_version>
<![CDATA[
<message>
process exited with code 69 (0x45, -187)</message>
<stderr_txt>
16:44:49 (303660): [normal]: This Einstein@home App was built at: Jan 16 2017 08:09:16

16:44:49 (303660): [normal]: Start of BOINC application '../../projects/einstein.phys.uwm.edu/hsgamma_FGRPB1G_1.18_x86_64-pc-linux-gnu__FGRPopencl1K-ati'.
16:44:49 (303660): [debug]: 1e+16 fp, 3.6e+09 fp/s, 2895897 s, 804h24m57s18
command line: ../../projects/einstein.phys.uwm.edu/hsgamma_FGRPB1G_1.18_x86_64-pc-linux-gnu__FGRPopencl1K-ati --inputfile ../../projects/einstein.phys.uwm.edu/LATeah3012L12221101.dat --alpha 2.59819959601 --delta -0.694603692878 --skyRadius 1.890770e-06 --ldiBins 15 --f0start 892.0 --f0Band 8.0 --firstSkyPoint 0 --numSkyPoints 1 --f1dot -1e-13 --f1dotBand 1e-13 --df1dot 1.69860773e-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_LATeah3012L12221101_0900_30217032.dat --debug 0 --device 0 -o LATeah3012L12221101_900.0_0_0.0_30217032_0_0.out
output files: 'LATeah3012L12221101_900.0_0_0.0_30217032_0_0.out' '../../projects/einstein.phys.uwm.edu/LATeah3012L12221101_900.0_0_0.0_30217032_0_0' 'LATeah3012L12221101_900.0_0_0.0_30217032_0_0.out.cohfu' '../../projects/einstein.phys.uwm.edu/LATeah3012L12221101_900.0_0_0.0_30217032_0_1'
16:44:49 (303660): [debug]: Flags: X64 SSE SSE2 GNUC X86 GNUX86
16:44:49 (303660): [debug]: glibc version/release: 2.35/stable
16:44:49 (303660): [debug]: Set up communication with graphics process.
boinc_get_opencl_ids returned [0x18dc2e8 , 0x7f272a2f0d60]
Using OpenCL platform provided by: Mesa
Using OpenCL device "VERDE (, LLVM 15.0.4, DRM 2.50, 5.15.0-53-generic)" by: AMD
Max allocation limit: 536870912
Global mem size: 2147483648
OpenCL compiling FAILED! : -11 . Error message: fatal error: cannot open file '/usr/lib/clc/verde-amdgcn-mesa-mesa3d.bc': Opaque pointers are only supported in -opaque-pointers mode (Producer: 'LLVM15.0.4' Reader: 'LLVM 15.0.4')

OpenCL device has no FP64 support
OpenCL compiling FAILED! : -11 . Error message: fatal error: cannot open file '/usr/lib/clc/verde-amdgcn-mesa-mesa3d.bc': Opaque pointers are only supported in -opaque-pointers mode (Producer: 'LLVM15.0.4' Reader: 'LLVM 15.0.4')

------------------------------

I use the Kisak PPA.  I will try re-installing that, next.

 

Other suggestions ???

( Make a copy  of the tahiti into the verde - instead of a link.?)  will try that too

 

jay
jay
Joined: 25 Jan 07
Posts: 99
Credit: 84,044,023
RAC: 0

see

see https://einsteinathome.org/task/1383485524

had reloaded from kisak ppa:

sudo aptitude reinstall mesa-opencl-icd
The following packages will be REINSTALLED:
  mesa-opencl-icd
0 packages upgraded, 0 newly installed, 1 reinstalled, 0 to remove and 0 not upgraded.
Need to get 0 B/7,490 kB of archives. After unpacking 0 B will be used.
(Reading database ... 313449 files and directories currently installed.)
Preparing to unpack .../mesa-opencl-icd_22.2.4~kisak1~j_amd64.deb ...
Unpacking mesa-opencl-icd:amd64 (22.2.4~kisak1~j) over (22.2.4~kisak1~j) ...
Setting up mesa-opencl-icd:amd64 (22.2.4~kisak1~j) ...
Processing triggers for libc-bin (2.35-0ubuntu3.1) ...

will now replace the link with an actual copy.... and try

   still get error after 13 seconds...  :-(

 

mountkidd
mountkidd
Joined: 14 Jun 12
Posts: 176
Credit: 12,161,512,555
RAC: 8,068,312

Checkout the info in this

Checkout the info in this post and the two that follow it.  Might be a solution for you...

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3,899
Credit: 43,558,422,644
RAC: 63,046,193

best to stop using Mesa.

best to stop using Mesa. every time you sneeze at this system you need to futz around with it with some new kludge to get it limping again. just install proper drivers already lol.

_________________________________________________________________________

jay
jay
Joined: 25 Jan 07
Posts: 99
Credit: 84,044,023
RAC: 0

@mountkidd Thanks for the

@mountkidd

Thanks for the reference.

The ROCM won't apply to me. I have an older ATI : Fire HD 7750. ROC does not support it.

It runs along fine until someone applies an incorrect link flag  when creating code for "Verde" island processors.

I was using a LTS version of Ubuntu : 20.4. I reinstalled the opencl drivers .Still failed.

Tried Kisak PPA: still failed.

Upgraded my whole system to Ubuntu 22.10: still failed.

Upgraded to Kisak in 22.10. Still failed.

The error message is about a pointer. I replaced the pointer with the actual code referenced by the pointer. Still  failed. I suspect a bad or misleading error message.

Who are the developers - so that I can make a github error report?

Is there a better way to investigate or a better forum? For example libclc of llvm.org (https://libclc.llvm.org/)??  or https://discourse.llvm.org/t/how-do-i-obtain-the-original-data-type-of-a-pointer/66739..

I did them a post in the beginner's forum: https://discourse.llvm.org/t/where-to-start-with-opencl-error/66876

T H A N K S !!!

Jay

 

mountkidd
mountkidd
Joined: 14 Jun 12
Posts: 176
Credit: 12,161,512,555
RAC: 8,068,312

It's not about ROCm.  The

It's not about ROCm.  The problem is ProtectSystem=strict in boinc-client.service prevents opencl from starting up.

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3,899
Credit: 43,558,422,644
RAC: 63,046,193

mountkidd wrote: It's not

mountkidd wrote:

It's not about ROCm.  The problem is ProtectSystem=strict in boinc-client.service prevents opencl from starting up.

Mesa drivers is also likely a source of his problem.

_________________________________________________________________________

jay
jay
Joined: 25 Jan 07
Posts: 99
Credit: 84,044,023
RAC: 0

@

@ MountKidd

Greetings.

I did set ProtectSytem=no  a few days ago. I retested again today.

The startup when starting a WU still failed... Here is E@H log snippet:

Max allocation limit: 536870912
Global mem size: 2147483648
OpenCL compiling FAILED! : -11 . Error message: fatal error: cannot open file '/usr/lib/clc/verde-amdgcn-mesa-mesa3d.bc': Opaque pointers are only supported in -opaque-pointers mode (Producer: 'LLVM15.0.2' Reader: 'LLVM 15.0.2')

OpenCL device has no FP64 support
OpenCL compiling FAILED! : -11 . Error message: fatal error: cannot open file '/usr/lib/clc/verde-amdgcn-mesa-mesa3d.bc': Opaque pointers are only supported in -opaque-pointers mode (Producer: 'LLVM15.0.2' Reader: 'LLVM 15.0.2')

Couldn't compile OpenCL kernels (error: 1)
initialize_ocl returned error [2006]
OCL context null
OCL queue null
Error generating generic FFT context object [5]
15:34:13 (3622): [CRITICAL]: ERROR: MAIN() returned with error '5'

I am trying to find out to whom to report the error about opaque pointers.

Any suggestions??

Thanks, Jay

[[ Message of the day: If at first you don't succeed, you are human.]]

 

mountkidd
mountkidd
Joined: 14 Jun 12
Posts: 176
Credit: 12,161,512,555
RAC: 8,068,312

Not certain that

Not certain that ProtectSystem=no works.   Try ProtectSystem=off  (this works for me) or just #ProtectSystem.

Wedge009
Wedge009
Joined: 5 Mar 05
Posts: 117
Credit: 17,014,250,501
RAC: 7,412,378

Let's ignore the unhelpful

Let's ignore the unhelpful and pro-monopoly remarks against Mesa.

I seem to recall that at one point you did have Mesa working on your system - was it for this particular GPU? Do you remember what changed when it stopped working?

I understand Kisak's PPA, while relatively up-to-date, is mainly intended for users running games with Proton (I use it myself). I was interested in trying the rusticl implementation of OpenCL and was directed to try Oibaf's PPA https://launchpad.net/~oibaf/+archive/ubuntu/graphics-drivers, and it seems to be more of a daily build against current source trees - probably not something to use in a stable system unless as a last resort. On the other hand, I take it using the distribution-supplied Mesa is not working either?

Failing that, is there a particular reason you prefer Mesa's OpenCL implementation over amdgpu-pro? I don't remember what your reasons were, if you cited them before.

---

Okay, I actually read some of the other posts in more detail and if I understand correctly this is your situation:

  • Previously running successfully on Ubuntu 20.04 with Mesa OpenCL.
  • I'm assuming the GPU is a Radeon HD 7750 (Cape Verde-based) card, not FirePro, which is the professional workstation line.
  • Issue started when attempting to use Ubuntu 22.04 and reverting to Ubuntu 20.04 hasn't helped.

Is that correct?

Cape Verde is first-generation GCN, so it is indeed rather old. Old enough that AMD says it's only officially supported via fglrx which would require a now-ancient and unsupported distribution https://www.amd.com/en/support/graphics/amd-radeon-hd/amd-radeon-hd-7000-series/amd-radeon-hd-7750.

I used to run R9 280s, which were basically Tahiti GPUs refreshed from HD 7900 series, but that was years ago. I'm guessing trying to run the 'legacy' OpenCL implementation from amdgpu-pro wouldn't work - I use that to run my Fiji GPUs but they're third-generation GCN and still officially supported. It looks like official support for Linux only goes back to second-generation GCN: https://www.amd.com/en/support/kb/release-notes/rn-amdgpu-unified-linux-22-20

So if you have your heart set on contributing to Einstein, I think it may be necessary to invest in a newer GPU. Otherwise if you're just trying to make the most of what you have on hand, it does look like Mesa is the only option in this case.

Opaque pointers seem to be related to the code itself, possibly with how LLVM is compiling it. When you say 'it doesn't work' in Ubuntu 20.04, is that with a reformat and re-installation?

I suspect you may have more luck with the opaque pointers issue on Mesa project rather than here. It doesn't appear to be an issue specific to Einstein.

Soli Deo Gloria

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.