From: Paul Thomas Date: Thu, 2 Feb 2012 21:20:14 +0000 (+0000) Subject: re PR fortran/52012 (Wrong-code with realloc on assignment and RESHAPE w/ ORDER=) X-Git-Tag: releases/gcc-4.7.0~596 X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=7de7ae1841e599746f1974ac507bb1d51483c47c;p=thirdparty%2Fgcc.git re PR fortran/52012 (Wrong-code with realloc on assignment and RESHAPE w/ ORDER=) 2012-02-02 Paul Thomas PR fortran/52012 * trans-expr.c (fcncall_realloc_result): If variable shape is correct, retain the bounds, whatever they are. 2012-02-02 Paul Thomas PR fortran/52012 * gfortran.dg/realloc_on_assign_11.f90: New test. From-SVN: r183849 --- diff --git a/gcc/fortran/ChangeLog b/gcc/fortran/ChangeLog index 7f8cc0606842..459e4e4c2808 100644 --- a/gcc/fortran/ChangeLog +++ b/gcc/fortran/ChangeLog @@ -1,3 +1,9 @@ +2012-02-02 Paul Thomas + + PR fortran/52012 + * trans-expr.c (fcncall_realloc_result): If variable shape is + correct, retain the bounds, whatever they are. + 2012-02-02 Tobias Burnus PR fortran/52093 diff --git a/gcc/fortran/trans-expr.c b/gcc/fortran/trans-expr.c index 4574c8ea8657..b0fc79cbd55a 100644 --- a/gcc/fortran/trans-expr.c +++ b/gcc/fortran/trans-expr.c @@ -6276,7 +6276,7 @@ realloc_lhs_loop_for_fcn_call (gfc_se *se, locus *where, gfc_ss **ss, } -/* For Assignment to a reallocatable lhs from intrinsic functions, +/* For assignment to a reallocatable lhs from intrinsic functions, replace the se.expr (ie. the result) with a temporary descriptor. Null the data field so that the library allocates space for the result. Free the data of the original descriptor after the function, @@ -6290,44 +6290,88 @@ fcncall_realloc_result (gfc_se *se, int rank) tree res_desc; tree tmp; tree offset; + tree zero_cond; int n; /* Use the allocation done by the library. Substitute the lhs descriptor with a copy, whose data field is nulled.*/ desc = build_fold_indirect_ref_loc (input_location, se->expr); + /* Unallocated, the descriptor does not have a dtype. */ tmp = gfc_conv_descriptor_dtype (desc); gfc_add_modify (&se->pre, tmp, gfc_get_dtype (TREE_TYPE (desc))); + res_desc = gfc_evaluate_now (desc, &se->pre); gfc_conv_descriptor_data_set (&se->pre, res_desc, null_pointer_node); se->expr = gfc_build_addr_expr (TREE_TYPE (se->expr), res_desc); - /* Free the lhs after the function call and copy the result to + /* Free the lhs after the function call and copy the result data to the lhs descriptor. */ tmp = gfc_conv_descriptor_data_get (desc); + zero_cond = fold_build2_loc (input_location, EQ_EXPR, + boolean_type_node, tmp, + build_int_cst (TREE_TYPE (tmp), 0)); + zero_cond = gfc_evaluate_now (zero_cond, &se->post); tmp = gfc_call_free (fold_convert (pvoid_type_node, tmp)); gfc_add_expr_to_block (&se->post, tmp); - gfc_add_modify (&se->post, desc, res_desc); - offset = gfc_index_zero_node; + tmp = gfc_conv_descriptor_data_get (res_desc); + gfc_conv_descriptor_data_set (&se->post, desc, tmp); - /* Now reset the bounds from zero based to unity based and set the - offset accordingly. */ + /* Check that the shapes are the same between lhs and expression. */ + for (n = 0 ; n < rank; n++) + { + tree tmp1; + tmp = gfc_conv_descriptor_lbound_get (desc, gfc_rank_cst[n]); + tmp1 = gfc_conv_descriptor_lbound_get (res_desc, gfc_rank_cst[n]); + tmp = fold_build2_loc (input_location, MINUS_EXPR, + gfc_array_index_type, tmp, tmp1); + tmp1 = gfc_conv_descriptor_ubound_get (desc, gfc_rank_cst[n]); + tmp = fold_build2_loc (input_location, MINUS_EXPR, + gfc_array_index_type, tmp, tmp1); + tmp1 = gfc_conv_descriptor_ubound_get (res_desc, gfc_rank_cst[n]); + tmp = fold_build2_loc (input_location, PLUS_EXPR, + gfc_array_index_type, tmp, tmp1); + tmp = fold_build2_loc (input_location, NE_EXPR, + boolean_type_node, tmp, + gfc_index_zero_node); + tmp = gfc_evaluate_now (tmp, &se->post); + zero_cond = fold_build2_loc (input_location, TRUTH_OR_EXPR, + boolean_type_node, tmp, + zero_cond); + } + + /* 'zero_cond' being true is equal to lhs not being allocated or the + shapes being different. */ + zero_cond = gfc_evaluate_now (zero_cond, &se->post); + + /* Now reset the bounds returned from the function call to bounds based + on the lhs lbounds, except where the lhs is not allocated or the shapes + of 'variable and 'expr' are different. Set the offset accordingly. */ + offset = gfc_index_zero_node; for (n = 0 ; n < rank; n++) { - tmp = gfc_conv_descriptor_ubound_get (desc, gfc_rank_cst[n]); + tree lbound; + + lbound = gfc_conv_descriptor_lbound_get (desc, gfc_rank_cst[n]); + lbound = fold_build3_loc (input_location, COND_EXPR, + gfc_array_index_type, zero_cond, + gfc_index_one_node, lbound); + lbound = gfc_evaluate_now (lbound, &se->post); + + tmp = gfc_conv_descriptor_ubound_get (res_desc, gfc_rank_cst[n]); tmp = fold_build2_loc (input_location, PLUS_EXPR, - gfc_array_index_type, - tmp, gfc_index_one_node); + gfc_array_index_type, tmp, lbound); gfc_conv_descriptor_lbound_set (&se->post, desc, - gfc_rank_cst[n], - gfc_index_one_node); + gfc_rank_cst[n], lbound); gfc_conv_descriptor_ubound_set (&se->post, desc, gfc_rank_cst[n], tmp); - /* Accumulate the offset. Since all lbounds are unity, offset - is just minus the sum of the strides. */ + /* Accumulate the offset. */ tmp = gfc_conv_descriptor_stride_get (desc, gfc_rank_cst[n]); + tmp = fold_build2_loc (input_location, MULT_EXPR, + gfc_array_index_type, + lbound, tmp); offset = fold_build2_loc (input_location, MINUS_EXPR, gfc_array_index_type, offset, tmp); diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog index 9c0b3b8ced83..025480442b9c 100644 --- a/gcc/testsuite/ChangeLog +++ b/gcc/testsuite/ChangeLog @@ -1,3 +1,8 @@ +2012-02-02 Paul Thomas + + PR fortran/52012 + * gfortran.dg/realloc_on_assign_11.f90: New test. + 2012-02-02 Tobias Burnus PR fortran/52093 diff --git a/gcc/testsuite/gfortran.dg/realloc_on_assign_11.f90 b/gcc/testsuite/gfortran.dg/realloc_on_assign_11.f90 new file mode 100644 index 000000000000..ab96bb9deaf8 --- /dev/null +++ b/gcc/testsuite/gfortran.dg/realloc_on_assign_11.f90 @@ -0,0 +1,36 @@ +! { dg-do run } +! PR52012 - tests of automatic reallocation on assignment for variable = array_intrinsic +! +! Contributed by Tobias Burnus and Dominique Dhumieres +! + integer, allocatable :: a(:), b(:), e(:,:) + integer :: c(1:5,1:5), d(1:5,1:5) + allocate(b(3)) + b = [1,2,3] + +! Shape conforms so bounds follow allocation. + allocate (a(7:9)) + a = reshape( b, shape=[size(b)]) + if (any ([lbound(a), ubound(a), size(a), shape (a)] .ne. [7,9,3,3])) call abort + + deallocate (a) +! 'a' not allocated so lbound defaults to 1. + a = reshape( b, shape=[size(b)]) + if (any ([lbound(a), ubound(a), size(a), shape (a)] .ne. [1,3,3,3])) call abort + + deallocate (a) +! Shape conforms so bounds follow allocation. + allocate (a(0:0)) + a(0) = 1 + if (any ([lbound(a), ubound(a), size(a), shape (a)] .ne. [0,0,1,1])) call abort + +! 'a' not allocated so lbound defaults to 1. + e = matmul (c(2:5,:), d(:, 3:4)) + if (any ([lbound(e), ubound(e), size(e), shape (e)] .ne. [1,1,4,2,8,4,2])) call abort + deallocate (e) + +! Shape conforms so bounds follow allocation. + allocate (e(4:7, 11:12)) + e = matmul (c(2:5,:), d(:, 3:4)) + if (any ([lbound(e), ubound(e), size(e), shape (e)] .ne. [4,11,7,12,8,4,2])) call abort +end