Data-implied-do rejected for non-trivial subscript expression

[Harald1]

Hello,

the following code is accepted by NAG but rejected by ifort/ifx:

program p
  integer :: i(4)
  data (i(modulo(j,5)),j=1,4) /4*1/
  print *, i
end

I get:

% ifort ifort-data-implied-do.f90
ifort-data-implied-do.f90(3): error #7829: This expression cannot be evaluated.   [MODULO]
  data (i(modulo(j,5)),j=1,4) /4*1/
----------^
compilation aborted for ifort-data-implied-do.f90 (code 1)

NAG warns in -f2003 mode:

% nagfor ifort-data-implied-do.f90 -f2003
NAG Fortran Compiler Release 7.1(Hanzomon) Build 7101
Extension(F2008): ifort-data-implied-do.f90, line 3: Function reference in DATA-implied-DO
[NAG Fortran Compiler normal termination, 1 warning]

but stays silent with -f2008 or -f2018 and works as expected.

Note that a trivial implied do works, such as:

  data (i(j),j=1,4) /4*1/

[jimdempseyatthecove]

Try using the intrinsic function MOD.

Jim Dempsey

[Steve_Lionel]

---

@jimdempseyatthecove wrote:

Try using the intrinsic function MOD.

MODULO is also an intrinsic function, subtly different from MOD.

[jimdempseyatthecove]

>>MODULO is also an intrinsic function, subtly different from MOD.

Yes, but using MOD, might avoid the ICE and attain the desired effect.

Jim Dempsey

[Harald1]

Well, I also tried MOD and INT, and all are rejected (so rejects-valid, no ICE).

Which ifort version are you using?

[Barbara_P_Intel · Accepted solution]

I filed a bug report on both ifx and ifort for this, CMPLRLLVM-50916.

@Harald1, I'm curious. Did you run into this issue with an application or do you just like to test the bounds of a compiler? Either way, keep them coming! These improve the Intel compiler for other developers.

[Harald1]

Hi Barbara,

I am performing checks on a few things that sometimes touch more recent Fortran features, and for the (usually more reliable) reference I use Intel and NAG.

From time to time there is interesting fallout, which I report here...