This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Ping: Turn off unit-at-a-time for java in a less hacky way
- From: Andrew Haley <aph at redhat dot com>
- To: Richard Sandiford <richard at codesourcery dot com>
- Cc: gcc-patches at gcc dot gnu dot org, jh at suse dot com
- Date: Wed, 1 Feb 2006 17:36:40 +0000
- Subject: Re: Ping: Turn off unit-at-a-time for java in a less hacky way
- References: <877j8fasug.fsf@talisman.home>
Richard Sandiford writes:
> This is a ping for Jan's patch:
>
> http://gcc.gnu.org/ml/gcc-patches/2005-08/msg00877.html
>
> which still applies cleanly to mainline. I've bootstrapped & regression
> tested it on i686-pc-linux-gnu.
>
> [The comment in java/lang.c has an overly-long line and isn't really
> appropriate now. Perhaps we should just say:
>
> /* Unit at a time is disabled for Java because it is considered
> too expensive. */
> no_unit_at_a_time_default = 1;
> ]
>
> In summary, the problem is that the java compiler overrides
> flag_unit_at_a_time in end_java_method() (i.e. part way through
> compilation):
>
> flag_unit_at_a_time = 0;
>
> Jan says that this crept in as part of the tree-ssa merge.
> The consensus in the thread above seemed to be that
> -funit-at-a-time was still too expensive for Java, and that
> -fno-unit-at-a-time should remain the default.
>
> However, end_java_method doesn't seem like the best place to enforce
> this rule. For one thing, it defeats any option processing code that
> relies on flag_unit_at_a_time being accurate (and I'm about to post a
> patch with such code, which is why I'm suddenly interested).
>
> Jan wanted the flag_unit_at_a_time setting to be sorted out during
> the options processing phase and suggested the patch above.
> Is it OK for mainline?
Fine by me.
Andrew.