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: Fix PR optimization/7520

> Try this.
> [patch against active_insn_p]

Thanks, it does pass the testcase. It prevents try_optimize_cfg called by
the flow2 pass from doing a condjump around jump optimization. That is
somewhat costly: we end up with 4 basic blocks instead of 2 and only 1 dead
insn deleted instead of 6.

Isn't the patch an overkill for broken code ? Wouldn't it be less costly to
put a fake SET to the return register instead of a CLOBBER or something like
that ? I can try to do it. Otherwise, are you willing to commit the patch
nevertheless ? If so, I'll bootstrap and regtest it.

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