This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [hammer] Fix PR/11559
Josef Zlomek <zlomj9am@artax.karlin.mff.cuni.cz> writes:
>> > this patch is a fix for PR/11559.
>>
>> VRP is your child, feel free to commit it after
>> bootstrapping/regtesting.
>>
>> > Bootstrap i686 in progress.
>> >
>> > Josef
>> >
>> > 2003-07-17 Josef Zlomek <zlomekj@suse.cz>
>> >
>> > PR/11559:
>> > * vrp.c (process_ranges_lt): If both operands are registers and their
>> > ranges are not constant or empty do not shorten the ranges.
>> > Minor speedup.
>> > (process_ranges_ltu): Likewise.
>> > (dump_operation_list): New function.
>> > (debug_operation_list): New function.
>>
>> Please commit the testcase also.
>> Can we commit this to mainline, also? This would make merging
>> easier...
>
> VRP is not in mainline.
But the testcase could be added there so that we do not forget to
merge the testcase iff VRP is merged up. I don't know what the rule
for this is in general regarding branches...
>> > HASH_REG (op0), NO_INSERT);
>> > if (slot0e)
>> > r0e = *slot0e;
>> > +
>> > + /* If the type of one range is not signed int we do not know how
>> > + to shorten ranges so we leave them as they are. */
>> > + if (r0t && (r0t->type & RANGE_SIGNED_INT) == 0)
>> > + return;
>>
>> You repeat this quite often. Isn't it better to use a macro for this?
>
> Maybe. I used copy & paste ;-)
;-)
Have a look whether the code would look better with a macro or not,
Andreas
--
Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
SuSE Linux AG, Deutschherrnstr. 15-19, 90429 Nürnberg, Germany
GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126