On Mon, Sep 6, 2021 at 2:30 PM Jakub Jelinek <ja...@redhat.com> wrote: > > On Mon, Sep 06, 2021 at 02:18:59PM +0200, Richard Biener wrote: > > On Mon, Sep 6, 2021 at 1:15 PM Jakub Jelinek <ja...@redhat.com> wrote: > > > > > > On Mon, Sep 06, 2021 at 07:15:41PM +0800, Hongtao Liu wrote: > > > > > So what about finish_options then? > > > > > default_options_optimization has only a single caller that then calls > > > > > read_cmdline_options and then finish_options. > > > > in finish_options > > > > (gdb) p opts_set->x_flag_tree_loop_vectorize > > > > $37 = 1 > > > > with -O2 -ftree-loop-vectorize, > > > > > > > > but > > > > 1000 if (opts->x_dump_base_name > > > > (gdb) p opts_set->x_flag_tree_loop_vectorize > > > > $38 = 0 > > > > for -O2 -ftree-vectorize??? > > > > > > > > Any magic for ftree-vectorize w/ EnabledBy??? > > > > > > I guess a way to get this working would be: > > > ; Alias to enable both -ftree-loop-vectorize and -ftree-slp-vectorize. > > > ftree-vectorize > > > -Common Optimization > > > +Common Var(flag_tree_vectorize) Optimization > > > Enable vectorization on trees. > > > > > > and then you can test both > > > opts_set->x_flag_tree_vectorize > > > and > > > opts_set->x_flag_tree_loop_vectorize > > > > Or make EnabledBy have set opts_set-> as well. > > That would change a little bit what *_set->x_* means. > > If user has explicit -O2, we have *_set->x_optimize set, > but don't enable it on all the suboptions that are implicitly > enabled because explicit -O2 has been used. > And isn't EnabledBy practically the same? If user writes > -Wall explicitly which implicitly enables hundreds of warning > options, do we want just Wall or also all the other options > marked as explicit? E.g. backend code than can't easily differentiate > between -Wwhatever and -Wall implying -Wwhatever and e.g. changing that > if implicit only.
Hmm, good point. Richard. > > Jakub >
