Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
28958 Discussions

DO CONCURRENT crash with OPENMP with Linux ifort 2024.2

kay-diederichs
New Contributor I
213 Views

The latest ifort 2024.2 on Linux produces the output 

 22.19008 37.32648 39.86396
173.5668 235.1864 234.4165
133.3740 193.3915 193.9088
forrtl: severe (174): SIGSEGV, segmentation fault occurred
forrtl: severe (174): SIGSEGV, segmentation fault occurred
forrtl: severe (174): SIGSEGV, segmentation fault occurred
forrtl: severe (174): SIGSEGV, segmentation fault occurred

with the following simple code. It prints the correct result when using the array assignment, but crashes when calculating the same thing with DO CONCURRENT. Or am I doing something wrong?

Compilation is just

ifort -qopenmp test.f90

INTEGER i,nd
REAL a0(3,3),b(3,3),vc(3,1000),vd(3,1000)

CALL RANDOM_NUMBER(vd)
vd=vd*100
CALL RANDOM_NUMBER(a0)
CALL RANDOM_NUMBER(b)

vc=MATMUL(a0,REAL(NINT(MATMUL(b,vd))))
DO i=1,3
PRINT*,vc(:,i)
END DO

DO CONCURRENT(i=1:nd)
vc(:,i)=MATMUL(a0,REAL(NINT(MATMUL(b,vd(:,i)))))
ENDDO
DO i=1,3
PRINT*,vc(:,i)
END DO

END

A colleague tells me that with his 2019 ifort there is no bug. Any insight?

Thanks, Kay

0 Kudos
2 Replies
Devorah_H_Intel
Moderator
193 Views
0 Kudos
kay-diederichs
New Contributor I
181 Views

This bug is not in ifx 2025.0 . But ifx has other bugs that prevent us from using it. As soon as I have such minimal reproducers as the one I posted above for ifort, I will post them here.

As long as it not an official policy that ifort is no longer discussed here in this forum, I think it makes sense to post bug reports about it. 

0 Kudos
Reply