On Fri, Jul 31, 2015 at 10:43 AM, Jakub Jelinek <ja...@redhat.com> wrote: > Hi! > > I wonder if we shouldn't change: > > TREE_TYPE (ctx->receiver_decl) > = build_qualified_type (build_reference_type (type), TYPE_QUAL_RESTRICT); > > in fixup_child_record_type to add: > > if (is_gimple_omp_offloaded (ctx->stmt)) > type = build_qualified_type (type, TYPE_QUAL_CONST); > > before it (and perhaps also add | TYPE_QUAL_CONST to TYPE_QUAL_RESTRICT for > all > constructs). I mean, aren't all the pointers in *.omp_data_i for the target > constructs constant in the offloaded region? Perhaps that could result in > better generated code (though perhaps just the restrict helps enough). > Can you think of any case where it is not constant?
Const qualification of a pointer or reference doesn't result in any optimization. The decl you refer to has to be constant. Richard. > Jakub
