https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102059
--- Comment #14 from Kewen Lin <linkw at gcc dot gnu.org> --- (In reply to Richard Biener from comment #11) > Note that x86 uses for example > > else if (caller_opts->x_ix86_fpmath != callee_opts->x_ix86_fpmath > /* If the calle doesn't use FP expressions differences in > ix86_fpmath can be ignored. We are called from FEs > for multi-versioning call optimization, so beware of > ipa_fn_summaries not available. */ > && (! ipa_fn_summaries > || ipa_fn_summaries->get (callee_node) == NULL > || ipa_fn_summaries->get (callee_node)->fp_expressions)) > ret = false; > > I wonder if we need to give targets the ability to compute IPA inline > summary bits, like "uses HTM" or "uses intrinsics for ISA X", to > selectively ignore bits that are not actually used. > Thanks for the example, it looks useful! Now the field fp_expressions is generic, one target specific summary class seems required then. And not sure if the users still have interests/senses to further try flto when they fail to get expected inlining with some basic optimization options. :)
