OpenMP target (offloading) question
Dear experts, sorry if this is a stupid question, but I was playing with offloading for the nvptx-none target and found different behavior between e.g. gfortran-10 on OpenSuse and the Nvidia compiler (nvfortran) for the attached code. With "nvfortran -mp=multicore offload-test.f90" the code prints: 2.002000.000 s1:1001000. s2:1001000. With "/usr/bin/gfortran-10 -fopenmp -foffload=nvptx-none offload-test.f90": 2. 2000.0 s1: 1001000.00 s2: 0. The core difference between the evaluations s1 and s2 is: s1: !$omp target data map(a,s) !$omp target teams reduction(+:s) map(s) do i = 1, n s = s + a(i) end do !$omp end target teams !$omp end target data s2: !$omp target data map(a,s) !$omp target teams reduction(+:s) do i = 1, n s = s + a(i) end do !$omp end target teams !$omp end target data I was assuming that the map clause in the reduction should not be necessary, but the result seems to tell me that either I am wrong (and gfortran is right), or nvfortran is wrong. With OpenACC this seems to be different; at least a simple example I tried with the reduction within an !$acc data ... !$acc end data did not show unexpected behavior. Can anybody tell me that I am wrong (and point me to the right place in the OpenMP standard), or should I open a PR? Thanks Harald program p5 implicit none integer :: i, n = 1000 real :: s real, allocatable :: a(:) allocate (a(n)) do i = 1, n a(i) = 2*i end do print *, a(1),a(n) call s1 () print *, "s1:", s call s2 () print *, "s2:", s contains subroutine s1 () integer :: i s = 0. !$omp target data map(a,s) !$omp target teams reduction(+:s) map(s) do i = 1, n s = s + a(i) end do !$omp end target teams !$omp end target data end subroutine s1 ! subroutine s2 () integer :: i s = 0. !$omp target data map(a,s) !$omp target teams reduction(+:s) do i = 1, n s = s + a(i) end do !$omp end target teams !$omp end target data end subroutine s2 end program
[PATCH] PR fortran/99840 - [8/9/10/11 Regression] ICE in gfc_simplify_matmul, at fortran/simplify.c:4777
Dear all,
the simplification of the TRANSPOSE of a zero-sized array would lead to
an ICE if the result was used in a subsequent simplification of a MATMUL.
The reason was the lack of the proper initialization of the shape, which
is mpz_t. Use mpz_init_set instead of mpz_set.
Regtested on x86_64-pc-linux-gnu. OK for mainline?
Since this is a regression, backport to all affected branches?
Thanks,
Harald
PR fortran/99840 - ICE in gfc_simplify_matmul, at fortran/simplify.c:4777
The simplification of the transposition of a constant array shall properly
initialize and set the shape of the result.
gcc/fortran/ChangeLog:
PR fortran/99840
* simplify.c (gfc_simplify_transpose): Properly initialize
resulting shape.
gcc/testsuite/ChangeLog:
PR fortran/99840
* gfortran.dg/transpose_5.f90: New test.
diff --git a/gcc/fortran/simplify.c b/gcc/fortran/simplify.c
index 388aca7c38c..c27b47aa98f 100644
--- a/gcc/fortran/simplify.c
+++ b/gcc/fortran/simplify.c
@@ -8123,8 +8123,8 @@ gfc_simplify_transpose (gfc_expr *matrix)
&matrix->where);
result->rank = 2;
result->shape = gfc_get_shape (result->rank);
- mpz_set (result->shape[0], matrix->shape[1]);
- mpz_set (result->shape[1], matrix->shape[0]);
+ mpz_init_set (result->shape[0], matrix->shape[1]);
+ mpz_init_set (result->shape[1], matrix->shape[0]);
if (matrix->ts.type == BT_CHARACTER)
result->ts.u.cl = matrix->ts.u.cl;
diff --git a/gcc/testsuite/gfortran.dg/transpose_5.f90 b/gcc/testsuite/gfortran.dg/transpose_5.f90
new file mode 100644
index 000..682b1c8552b
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/transpose_5.f90
@@ -0,0 +1,8 @@
+! { dg-do compiler }
+! { dg-options "-O2" }
+! PR fortran/99840 - ICE in gfc_simplify_matmul, at fortran/simplify.c:4777
+program p
+ integer, parameter :: x(0,0) = 0
+ integer :: y(0,0)
+ y = matmul (x, transpose(x))
+end
Re: [PATCH] PR fortran/99840 - [8/9/10/11 Regression] ICE in gfc_simplify_matmul, at fortran/simplify.c:4777
Yes OK for trunk and affected branches. Thanks, Jerry On 3/31/21 2:08 PM, Harald Anlauf via Fortran wrote: Dear all, the simplification of the TRANSPOSE of a zero-sized array would lead to an ICE if the result was used in a subsequent simplification of a MATMUL. The reason was the lack of the proper initialization of the shape, which is mpz_t. Use mpz_init_set instead of mpz_set. Regtested on x86_64-pc-linux-gnu. OK for mainline? Since this is a regression, backport to all affected branches? Thanks, Harald PR fortran/99840 - ICE in gfc_simplify_matmul, at fortran/simplify.c:4777 The simplification of the transposition of a constant array shall properly initialize and set the shape of the result. gcc/fortran/ChangeLog: PR fortran/99840 * simplify.c (gfc_simplify_transpose): Properly initialize resulting shape. gcc/testsuite/ChangeLog: PR fortran/99840 * gfortran.dg/transpose_5.f90: New test.
