This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Trivial cleanup of uses of flag_lto || flag_whopr


Hi Diego, thanks for thinking about this.

Note that flag_generate_lto is not equivalent to flag_lto||flag_whopr.
 In particular, flag_lto and flag_generate_lto are typically not set
at the same time.  I haven't checked at these spots, but in principle
this change does not seem sound.

Yes, I hadn't spotted that lto/lto-lang.c also sets flag_generate_lto. However if flag_lto or flag_whopr is set then flag_generate_lto will be set (see opts.c). As far as I can see, all the places that check flag_lto || flag_whopr actually want flag_generate_lto (with the possible exception of fortran/options.c), so I still think the change is correct. The reason I would like this change is that my plugin does an additional type of LTO, and all the places that do flag_lto || flag_whopr should do flag_lto || flag_whopr || flag_baldrick. Rather than do this or fake it by setting flag_lto or flag_whopr, I would rather be able to simply set flag_generate_lto, which in any case seems to be the right thing to do (but which currently does not have all the right consequences).

Ciao,

Duncan.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]