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: RFA: fix execute/20010119-1.c failures on sh-elf


Hi Joern,
> 2003-05-16  J"orn Rennecke <joern.rennecke@superh.com>
>
>        * gcse.c (cprop_jump): If substitution fails using the pattern
>        of SETCC, try to use a REG_EQUAL / REG_EQUIV note.

You might also be interested in a similar patch to do this that I
posted back on 9th March entitled "REG_EQUAL notes on cond_jumps":
http://gcc.gnu.org/ml/gcc-patches/2003-03/msg00824.html

This not only inspects the REG_EQUAL notes on both the conditional
branch and the condition code setter as you do in your patch, but
also updates or adds a REG_EQUAL note on the conditional jump if
a substitution can be made, but the resulting instruction isn't
recognized (which allows further optimizations).

That patch fixes execute/20010119-1.c on several platforms.  I
wonder if you could check it also solves the problem on sh-elf?


May I just repeat that not only have I asked for reviews of this
patch several times, but its also been requested by David Anglin,
DJ Delorie and Kazu Hirata.  There have even been GNATS PRs filed
since the fix was posted, and now you've expended time and effort
tracking the source of the problem, to duplicate a near identical
fix.

Such are the problems with contributing to a voluntary project.

Roger
--


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