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: [patch] PR 37950


On Wed, Nov 12, 2008 at 2:30 PM, Zdenek Dvorak <rakdver@kam.mff.cuni.cz> wrote:
> Hi,
>
>> >> >        * tree-flow-inline.h (memory_partition): Return NULL when aliases were
>> >> >        not computed for the current function.
>> >>
>> >> It's better to clear out gimple_loaded_syms() from the moved
>> >> statements.  In fact, build_ssa_operands should probably clear
>> >> loaded/stored syms (I thought it did already).
>> >
>> > unfortunately, that is not possible -- that would give you empty
>> > gimple_loaded_syms, breaking alias analysis (unless you are willing
>> > to rescan all statement operands at the beginning of alias analysis;
>> > I have considered that, but since rescanning the operands is rather
>> > expensive, I think it is better to get the operands right the first
>> > time),
>>
>> For fixing PR38051 I will have to remove using the loaded/stored syms
>> from alias analysis.  So it may work to clean them after that.
>
> yes, this would work as well, I think.  I am a bit worried about leaving
> the statements with the incorrect (or empty) loaded/stored_syms, though;
> the passes that are before pass_build_alias do not seem to use this
> information, but that could change,

The patch I applied for PR38051 doesn't fully get rid of stored/loaded_syms,
so I think your original patch is ok.

Thanks,
Richard.

> Zdenek
>


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