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] Fix PR middle-end/28690, indexed load/store performance + reload bug


On Tue, 2006-12-05 at 06:20 -0800, Ian Lance Taylor wrote:
> Yes, sorry about that.  We shouldn't use this.
> 
> And in fact I realized that if you fix LEGITIMIZE_MEMORY_ADDRESS, you
> probably don't need this patch anyhow.

If you look at the PR, you'll see we already tried riding that pony.
The problem with LEGITIMIZE_MEMORY_ADDRESS is that it's only called
for addresses that don't look legitimate.  There are test cases in
the PR where the address already looks legitimate (ie, REG + REG)
so we don't bother calling LEGITIMIZE_MEMORY_ADDRESS and we never
transform the address to the PEG_POINTER reg is first.  That's what
led us back to modifying swap_commutative_operands_p.

Peter




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