Bug 1129160 - openblas ppc64 segfault invalid memory ref when ./dblat3 < ./dblat3.dat
openblas ppc64 segfault invalid memory ref when ./dblat3 < ./dblat3.dat
Status: RESOLVED FIXED
Classification: openSUSE
Product: openSUSE Tumbleweed
Classification: openSUSE
Component: Basesystem
Current
PowerPC Other
: P5 - None : Normal (vote)
: ---
Assigned To: HPC Bugzilla
E-mail List
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2019-03-14 07:10 UTC by Michel Normand
Modified: 2019-09-06 08:27 UTC (History)
0 users

See Also:
Found By: ---
Services Priority:
Business Priority:
Blocker: ---
Marketing QA Status: ---
IT Deployment: ---


Attachments
openblas_gnu_hpc_standard__ppc64_201903140629.log (3.23 MB, text/x-log)
2019-03-14 07:10 UTC, Michel Normand
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Michel Normand 2019-03-14 07:10:52 UTC
Created attachment 800012 [details]
openblas_gnu_hpc_standard__ppc64_201903140629.log

openblas ppc64 segfault invalid memory ref when ./dblat3 < ./dblat3.dat

since a while with following backtrace
===
[  769s] OMP_NUM_THREADS=2 ./sblat2 < ./sblat2.dat                                             
[  769s]                                                                                       
[  769s] Program received signal SIGSEGV: Segmentation fault - invalid memory reference.       
[  769s]                                                                                       
[  769s] Backtrace for this error:
[  769s] #0  0x7fff82c53ddf in ???                                                             
[  769s] #1  0x7fff82c52877 in ???                                                             
[  769s] #2  0x7fff82e304c7 in ???                                                             
[  769s] #3  0x118b5c5d0 in dtrmm_ounucopy                                                     
[  769s]        at generic/trmm_uncopy_4.c:93                                                  
[  769s] #4  0x118af8e0f in dtrmm_RNUU
[  769s]        at /home/abuild/rpmbuild/BUILD/OpenBLAS-0.3.5/driver/level3/trmm_R.c:254       
[  769s] #5  0x118af327b in dtrmm_
[  769s]        at /home/abuild/rpmbuild/BUILD/OpenBLAS-0.3.5/interface/trsm.c:381
[  769s] #6  0x118ae91c7 in dchk3_                                                             
[  769s]        at /home/abuild/rpmbuild/BUILD/OpenBLAS-0.3.5/test/dblat3.f:1059
[  769s] #7  0x118af1f43 in dblat3                                                             
[  769s]        at /home/abuild/rpmbuild/BUILD/OpenBLAS-0.3.5/test/dblat3.f:292
[  769s] #8  0x118ae4e8b in main
[  769s]        at /home/abuild/rpmbuild/BUILD/OpenBLAS-0.3.5/test/dblat3.f:355                
[  769s] /bin/sh: line 1: 21622 Segmentation fault      OPENBLAS_NUM_THREADS=1 OMP_NUM_THREADS=1 ./dblat3 < ./dblat3.dat
[  769s] make[1]: *** [Makefile:67: level3] Error 139
===

This is the same signature for TW or science projects (same openblas 0.3.5)
https://build.opensuse.org/package/live_build_log/openSUSE:Factory:PowerPC/openblas:gnu-hpc/standard/ppc64
https://build.opensuse.org/package/live_build_log/science/openblas:gnu-hpc/openSUSE_Factory_PowerPC/ppc64
Comment 1 Michel Normand 2019-03-14 08:28:08 UTC
I do not know if related but there is an upstream issue with Aix 
that reports a similar segfault backstrace at
https://github.com/xianyi/OpenBLAS/issues/1997#issuecomment-467381931
Comment 2 Egbert Eich 2019-03-14 12:31:02 UTC
PPC64 BE is a bit out of scope - so build errors will usually just get ignored.
There is no resolution to the upstream issue, yet, people are still discussing.

Let's keep this open for a while at least and see what happens.
Comment 3 Michel Normand 2019-09-06 08:27:49 UTC
as today with versions updates 0.3.6/0.3.7 this problem is solved.