[Openmp-dev] Debug assert trigger in OpenMP + MPI

Johannes Doerfert via Openmp-dev openmp-dev at lists.llvm.org
Tue May 26 09:02:27 PDT 2020


Thanks! Can you also commit the reproducer as a test?


On 5/23/20 4:04 PM, Joachim Protze via Openmp-dev wrote:
> I think, I found the issue and posted a fix at:
>
> https://reviews.llvm.org/D80480
>
> - Joachim
>
>
> Am 21.05.20 um 18:12 schrieb Raúl Peñacoba Veigas via Openmp-dev:
>> Hello again,
>>
>> I've managed to remove MPI from the equation. It seems a race 
>> condition in the runtime.
>>
>> int main(int argc, char **argv)
>> {
>>          int TIMESTEPS = 10;
>>          int BLOCKS = 100;
>>
>>          int nranks = 4;
>>
>>          int DATA;
>>
>>          #pragma omp parallel
>>          #pragma omp single
>>          {
>>                  for (int t = 0; t < TIMESTEPS; ++t) {
>>                          for (int r = 0; r < nranks; ++r) {
>>                                  for (int b = 0; b < BLOCKS; ++b) {
>>                                          #pragma omp task depend(in: 
>> DATA)
>>                                          { }
>>                                  }
>>                          }
>>
>>                          #pragma omp task depend(inout: DATA)
>>                          {}
>>                  }
>>                  #pragma omp taskwait
>>          }
>> }
>>
>> To run it execute:
>>
>> clang -fopenmp t1.c -o t1
>>
>> for i in {1..5000}; do echo $i; OMP_NUM_THREADS=3 ./t1; done
>>
>> Regards,
>>
>> Raúl
>>
>> El 21/5/20 a las 9:57, Raúl Peñacoba Veigas escribió:
>>> Hello everyone,
>>>
>>> Writing an OpenMP + MPI code I've triggered a debug assert in 
>>> __kmp_task_start:
>>>
>>> KMP_DEBUG_ASSERT(taskdata->td_flags.tasktype == TASK_EXPLICIT);
>>>
>>> I attach a simpler code that does not do anything special with 
>>> additional info.
>>>
>>> #include <mpi.h>
>>>
>>> #include <assert.h>
>>> #include <stdio.h>
>>> #include <stdlib.h>
>>> #include <string.h>
>>> #include <unistd.h>
>>>
>>> int main(int argc, char **argv)
>>> {
>>>         int TIMESTEPS = 10;
>>>         int BLOCKS = 100;
>>>
>>>         MPI_Init(&argc, &argv);
>>>
>>>         int rank, nranks;
>>>         MPI_Comm_rank(MPI_COMM_WORLD, &rank);
>>>         MPI_Comm_size(MPI_COMM_WORLD, &nranks);
>>>
>>>         int DATA;
>>>
>>>         #pragma omp parallel
>>>         #pragma omp single
>>>         {
>>>                 for (int t = 0; t < TIMESTEPS; ++t) {
>>>                         for (int r = 0; r < nranks; ++r) {
>>>                                 for (int b = 0; b < BLOCKS; ++b) {
>>>                                         #pragma omp task depend(in: 
>>> DATA)
>>>                                         { }
>>>                                 }
>>>                         }
>>>
>>>                         #pragma omp task depend(inout: DATA)
>>>                         {}
>>>                 }
>>>                 #pragma omp taskwait
>>>         }
>>>
>>>         MPI_Finalize();
>>>
>>> }
>>>
>>> llvm_project debug build, commitaafdeeade8d
>>> MPICH Version: 3.3a2
>>> MPICH Release date: Sun Nov 13 09:12:11 MST 2016
>>>
>>> $ MPICH_CC=clang mpicc -fopenmp t1.c -o t1
>>> $ for i in {1..100}; do mpiexec.hydra -n 4 ./t1; done
>>>
>>>
>>
>> http://bsc.es/disclaimer
>> _______________________________________________
>> Openmp-dev mailing list
>> Openmp-dev at lists.llvm.org
>> https://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev
>
> _______________________________________________
> Openmp-dev mailing list
> Openmp-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/openmp-dev/attachments/20200526/2b9cb9e0/attachment-0001.html>


More information about the Openmp-dev mailing list