You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am reporting a bug others will be able to reproduce and not asking a question or requesting a new feature.
System information including:
OpenCoarrays Version: caf version 2.9.2-13-g235167d
Fortran Compiler: GNU Fortran (GCC) 11.2.0
C compiler used for building lib: gcc (GCC) 11.2.0
Installation method: make; make install
All flags & options passed to the installer: cmake -DCMAKE_BUILD_TYPE=Release
Output of uname -a: Linux thelio.indiana 5.14.18-100.fc33.x86_64 #1 SMP Fri Nov 12 17:38:44 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
MPI library being used: MPICH 3.3.2
Machine architecture and number of physical cores: 12 core AMD Threadripper 2920X
Version of CMake: 3.19.7
To help us debug your issue please explain:
What you were trying to do (and why)
I am developing code to perform a halo exchange associated with domain decomposition methods for PDEs and have encountered a problem. To debug the problem I've extracted the problem loop into a subroutine and wrapped it with a main program that initializes some structure arrays with data from a debugging data set for 4 images. The subroutine loop is performing the equivalent of an MPI_Alltoallv or MPI_Neighbor_alltoall.After calling the subroutine, the program compares the result of the subroutine with reference data to check for a successful completion. Here is the problem reproducer:
What happened (include command output, screenshots, logs, etc.)
When compiled with -fcheck=bounds the runtime reports a spurious out-of-bounds error:
At line 69 of file gfortran-20220127.f90
Fortran runtime error: Index '15' of dimension 1 of array 'buffer%array' above upper bound of 14
Error termination. Backtrace:
#0 0x402b81 in ???
#1 0x4050b5 in ???
#2 0x40545e in ???
#3 0x7f325e2981e1 in ???
#4 0x40253d in ???
#5 0xffffffffffffffff in ???
Error: Command:
`/opt/lib/gcc-11_2/mpich/3.3.2/bin/mpiexec -n 4 ./a.out`
failed to run.
What you expected to happen
The program should run to completion without error. That there is no actual out-of-bounds reference can be confirmed by uncommenting the write statement and verifying the index in each iteration of the loop lies within the array bounds -- until it aborts on the following line. Also the NAG and Intel compilers report no out-of-bounds access.
Step-by-step reproduction instructions to reproduce the error/bug
System information including:
OpenCoarrays Version: caf version 2.9.2-13-g235167d
Fortran Compiler: GNU Fortran (GCC) 11.2.0
C compiler used for building lib: gcc (GCC) 11.2.0
Installation method:
make; make install
All flags & options passed to the installer:
cmake -DCMAKE_BUILD_TYPE=Release
Output of
uname -a
:Linux thelio.indiana 5.14.18-100.fc33.x86_64 #1 SMP Fri Nov 12 17:38:44 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
MPI library being used: MPICH 3.3.2
Machine architecture and number of physical cores: 12 core AMD Threadripper 2920X
Version of CMake: 3.19.7
To help us debug your issue please explain:
What you were trying to do (and why)
I am developing code to perform a halo exchange associated with domain decomposition methods for PDEs and have encountered a problem. To debug the problem I've extracted the problem loop into a subroutine and wrapped it with a main program that initializes some structure arrays with data from a debugging data set for 4 images. The subroutine loop is performing the equivalent of an
MPI_Alltoallv
orMPI_Neighbor_alltoall
.After calling the subroutine, the program compares the result of the subroutine with reference data to check for a successful completion. Here is the problem reproducer:What happened (include command output, screenshots, logs, etc.)
When compiled with
-fcheck=bounds
the runtime reports a spurious out-of-bounds error:What you expected to happen
The program should run to completion without error. That there is no actual out-of-bounds reference can be confirmed by uncommenting the write statement and verifying the index in each iteration of the loop lies within the array bounds -- until it aborts on the following line. Also the NAG and Intel compilers report no out-of-bounds access.
Step-by-step reproduction instructions to reproduce the error/bug
The text was updated successfully, but these errors were encountered: