2 ! Check the fix for PR28005, in which the mechanism for dealing
3 ! with matmul (transpose (a), b) would cause wrong results for
4 ! matmul (a(i, 1:n), b(1:n, 1:n)).
6 ! Based on the original testcase contributed by
7 ! Tobias Burnus <tobias.burnus@physik.fu-berlin.de>
10 integer, parameter :: nmax
= 3
12 integer, dimension(nmax
,nmax
) :: iB
=0 , iC
=1
13 integer, dimension(nmax
,nmax
) :: iX1
=99, iX2
=99, iChk
14 iChk
= reshape((/30,66,102,36,81,126,42,96,150/),(/3,3/))
16 ! This would give 3, 3, 99
17 iB
= reshape((/1 ,3 ,0 ,2 ,5 ,0 ,0 ,0 ,0 /),(/3,3/))
18 iX1(1:n
,1) = matmul( iB(2,1:n
),iC(1:n
,1:n
) )
20 ! This would give 4, 4, 99
22 iX2(1:n
,1) = matmul( iB(2,1:n
),iC(1:n
,1:n
) )
24 ! Whereas, we should have 8, 8, 99
25 if (any (iX1(1:n
,1) .ne
. (/8, 8, 99/))) call abort ()
26 if (any (iX1
.ne
. iX2
)) call abort ()
28 ! Make sure that the fix does not break transpose temporaries.
29 iB
= reshape((/(i
, i
= 1, 9)/),(/3,3/))
32 iX1
= matmul (iX1
, iC
)
33 iX2
= matmul (transpose (iB
), iC
)
34 if (any (iX1
.ne
. iX2
)) call abort ()
35 if (any (iX1
.ne
. iChk
)) call abort ()