https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127845
Bug ID: 127845
Summary: FINDLOC with DIM and MASK returns wrong results
Product: gcc
Version: 16.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: federico.perini at gmail dot com
Target Milestone: ---
findloc(x, value, dim=d, mask=m) misses matching values, and can report false
ones, once the result has more than one line in the second dimension.
In the test case, element (2,3) of the result is 0, but x(2,3,2) = 3.0 with
m(2,3,2) = .true., so it should be 2. The call goes to _gfortran_mfindloc1_r8
(seen with -S). It is possible that the mask of the wrong line is used
(mfindloc1_* never advances the mask on carry).
Test program:
program findloc_dim_mask
implicit none
real(8) :: x(2,3,2)
logical :: m(2,3,2)
integer :: i, j, k, l
l = 0
do k = 1, 2; do j = 1, 3; do i = 1, 2
x(i,j,k) = real(mod(l*7,5)+1, 8); m(i,j,k) = mod(l,3) /= 1; l = l + 1
end do; end do; end do
do k = 1, 2
print '(a,i0,a,i0,a)', 'x(:,:,', k, ') m(:,:,', k, ')'
do i = 1, 2
print '(3f4.1,6x,3l2)', x(i,:,k), m(i,:,k)
end do
end do
print '(a)', 'findloc(x, 3.0d0, dim=3, mask=m), a 2x3 array, got:'
block
integer :: r(2,3)
r = findloc(x, 3.0d0, dim=3, mask=m)
do i = 1, 2
print '(4x,3i2)', r(i,:)
end do
end block
print '(a)', ' expected:'
print '(4x,3i2)', 2, 0, 0
print '(4x,3i2)', 0, 0, 2
print '(a)', ' (x(2,3,2) = 3.0 and m(2,3,2) = T, so element (2,3) is 2)'
end program findloc_dim_mask
It seems like all versions of gfortran have this issue (ifx and flang report
the expected result, check on Compiler Explorer at
https://godbolt.org/z/cf5M8va7n).
Thank you!
Federico
