[llvm-bugs] [Bug 35910] New: Einstein at home FGRPB1G OpenCL app crashes with ::SelectionDAG::~SelectionDAG(): Assertion `!UpdateListeners && "Dangling registered DAGUpdateListeners"' failed.

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=35910

            Bug ID: 35910
           Summary: Einstein at home FGRPB1G OpenCL app crashes with
                    ::SelectionDAG::~SelectionDAG(): Assertion
                    `!UpdateListeners && "Dangling registered
                    DAGUpdateListeners"' failed.
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: AMDGPU
          Assignee: unassignedbugs at nondot.org
          Reporter: pavel.ondracka at email.cz
                CC: llvm-bugs at lists.llvm.org

Created attachment 19657
  --> https://bugs.llvm.org/attachment.cgi?id=19657&action=edit
compressed full backtrace

The Einstein at home FGRPB1G OpenCL app crahes with:
hsgamma_FGRPB1G_1.18_x86_64-pc-linux-gnu__FGRPopencl1K-ati:
/home/ondracka/graphics/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp:917:
llvm::SelectionDAG::~SelectionDAG(): Assertion `!UpdateListeners && "Dangling
registered DAGUpdateListeners"' failed.

GPU: AMD CAICOS
This is with the latest git version of all components:
llvm: 9901953429f6e4519760d62761d5555cbafbd24f
mesa: a1901d092c
libdrm: de807d1a740effe60da62d07dcfeae7c01012d15
libclc: d6384415ab854c68777dd77451aa2bc0d959da99
kernel: 4.14.11-300.fc27.x86_64

I'm actually unsure whom to blame here, the llvm seems like the best candidate,
however I can repost this bug at mesa if needed.
Full backtrace attached. Please note that the part before the assertion failure
with:

OpenCL compiling FAILED! : -11 . Error message: input.cl:7:26: error:
unsupported OpenCL extension 'cl_khr_fp64' - ignoring
input.cl:10:30: error: unknown type name 'double2'; did you mean 'double'?
input.cl:10:30: error: use of type 'double' requires cl_khr_fp64 extension to
be enabled
OpenCL device has no FP64 support

should be be harmless. According to the developers, the app tries if the device
can do double precision math, however if it can not, it should fall back to CPU
for the double precision part.

How to reproduce:
This is actually little complicated, since the app and input data (the app
searches for binary pulsars in real radio observatory data) is not available
standalone, but only through the BOINC platform. I'm happy to provide any
needed info and/or test patches, however in case anyone wants to try himself,
the following steps describe how to reproduce with Fedora.
1) install boinc-client and boinc-manager
2) start boinc in a new directory (check in the output that the AMD GPU is
detected, for some reasons it works here only if the binary is executed with
the full path, eg. /usr/bin/boinc )
3) start boincmgr (GUI), it needs to be started in the same directory as the
client to connect, create account and attach the boinc to the Einstein at home
4) pause the project immediately (or kill the client) otherwise the project
will send a lot of (unrelated to this bug) work
5) go to the the https://einsteinathome.org/ and log in
6) go to account -> preferences -> project and in the list of applications
leave  only the Gamma-ray pulsar binary search #1 (GPU). Save changes.
7) start boinc again / unpause the project
8) Wait for the needed files to download
9) When you get the GPU tasks they should be finishing immediately due to error
10) Kill the client and cd to newly created projects/einstein.phys.uwm.edu/
11) The stderr.txt file should contain the error output including the full
command line needed to run the tasks, alternatively this can be obtained at the
 https://einsteinathome.org/account/tasks/ when clicking on the task id of the
failed tasks (it may take some time for the website to update)
12) use the obtained command to run the app downloaded in
projects/einstein.phys.uwm.edu/ in standalone mode and debug

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20180111/8c2652a3/attachment.html>