https://community.intel.com/t5/Intel-Fortran-Compiler/Issue-with-OMP-CANCEL-construct/m-p/1747699#M178860 <P>Using the latest ifx compiler (2026.0.0) on Windows 11 in Debug mode, the following code hangs fairly regularly (even when running with a modest number of threads):</P><LI-CODE lang="fortran">PROGRAM P USE OMP_LIB IMPLICIT NONE (TYPE, EXTERNAL) INTEGER :: I, SUM, VERSION ! Check if the cancellation feature is activated (a requirement for the CANCEL construct to work). IF (.NOT. OMP_GET_CANCELLATION()) THEN WRITE(*, *) 'The CANCEL feature is not activated. Set the environment variable OMP_CANCELLATION to true.' STOP END IF ! This will fail (that is, the code will hang) randomly, when running the code ! multiple times. !$OMP PARALLEL WRITE(*, *) OMP_GET_THREAD_NUM(), '/', OMP_GET_NUM_THREADS() !$OMP DO REDUCTION(+:SUM) DO I = 1, 10000 SUM = SUM + I !$OMP CANCEL DO IF (I == 5) END DO !$OMP END DO !$OMP END PARALLEL WRITE(*, *) SUM END PROGRAM P</LI-CODE><P>I do not see any restrictions associated with the use of the REDUCTION clause with CANCEL constructs in the OpenMP specification. It does look like a bug.</P> Tue, 12 May 2026 18:45:30 GMT OP1 2026-05-12T18:45:30Z https://community.intel.com/t5/Intel-Fortran-Compiler/Issue-with-OMP-CANCEL-construct/m-p/1747699#M178860 <P>Using the latest ifx compiler (2026.0.0) on Windows 11 in Debug mode, the following code hangs fairly regularly (even when running with a modest number of threads):</P><LI-CODE lang="fortran">PROGRAM P USE OMP_LIB IMPLICIT NONE (TYPE, EXTERNAL) INTEGER :: I, SUM, VERSION ! Check if the cancellation feature is activated (a requirement for the CANCEL construct to work). IF (.NOT. OMP_GET_CANCELLATION()) THEN WRITE(*, *) 'The CANCEL feature is not activated. Set the environment variable OMP_CANCELLATION to true.' STOP END IF ! This will fail (that is, the code will hang) randomly, when running the code ! multiple times. !$OMP PARALLEL WRITE(*, *) OMP_GET_THREAD_NUM(), '/', OMP_GET_NUM_THREADS() !$OMP DO REDUCTION(+:SUM) DO I = 1, 10000 SUM = SUM + I !$OMP CANCEL DO IF (I == 5) END DO !$OMP END DO !$OMP END PARALLEL WRITE(*, *) SUM END PROGRAM P</LI-CODE><P>I do not see any restrictions associated with the use of the REDUCTION clause with CANCEL constructs in the OpenMP specification. It does look like a bug.</P> Tue, 12 May 2026 18:45:30 GMT https://community.intel.com/t5/Intel-Fortran-Compiler/Issue-with-OMP-CANCEL-construct/m-p/1747699#M178860 OP1 2026-05-12T18:45:30Z