Port keyed interceptor stubs to ARM and x64.

Port keyed interceptor stubs to ARM and x64.

Committed: http://code.google.com/p/v8/source/detail?r=3874

[antonm, 2010-02-15 08:55:44]

Mads,

may you have a look?

[Mads Ager (chromium), 2010-02-15 11:04:21]

LGTM

http://codereview.chromium.org/600123/diff/1/3
File src/x64/ic-x64.cc (right):

http://codereview.chromium.org/600123/diff/1/3#newcode588
src/x64/ic-x64.cc:588: // Check that the receiver isn't a smi.
Not sure that the order of these checks matter much, but they are reversed here
compared to the ARM version.  Would be nice to be consistent.  It is probably
most likely that the index is not a smi, so maybe check that first on all
platforms?

[antonm, 2010-02-15 11:12:11]

http://codereview.chromium.org/600123/diff/1/3
File src/x64/ic-x64.cc (right):

http://codereview.chromium.org/600123/diff/1/3#newcode588
src/x64/ic-x64.cc:588: // Check that the receiver isn't a smi.
On 2010/02/15 11:04:21, Mads Ager wrote:
> Not sure that the order of these checks matter much, but they are reversed
here
> compared to the ARM version.  Would be nice to be consistent.  It is probably
> most likely that the index is not a smi, so maybe check that first on all
> platforms?

I just followed the sequence from KeyedLoadIC::GenerateGeneric for the
corresponding platforms.

What would you prefer: keep things as is now or adjust both generic and
interceptor cases?

[Mads Ager (chromium), 2010-02-15 11:30:12]

On 2010/02/15 11:12:11, antonm wrote:
> http://codereview.chromium.org/600123/diff/1/3
> File src/x64/ic-x64.cc (right):
> 
> http://codereview.chromium.org/600123/diff/1/3#newcode588
> src/x64/ic-x64.cc:588: // Check that the receiver isn't a smi.
> On 2010/02/15 11:04:21, Mads Ager wrote:
> > Not sure that the order of these checks matter much, but they are reversed
> here
> > compared to the ARM version.  Would be nice to be consistent.  It is
probably
> > most likely that the index is not a smi, so maybe check that first on all
> > platforms?
> 
> I just followed the sequence from KeyedLoadIC::GenerateGeneric for the
> corresponding platforms.
> 
> What would you prefer: keep things as is now or adjust both generic and
> interceptor cases?

It probably makes very little difference, but I like the platforms to be
consistent.

[antonm, 2010-02-15 11:57:33]

I changed the order in GenerateGeneric and GenerateIndexedInterceptor.  Let's
hope it won't affect performance.

yours,
anton.

On 2010/02/15 11:30:12, Mads Ager wrote:
> On 2010/02/15 11:12:11, antonm wrote:
> > http://codereview.chromium.org/600123/diff/1/3
> > File src/x64/ic-x64.cc (right):
> > 
> > http://codereview.chromium.org/600123/diff/1/3#newcode588
> > src/x64/ic-x64.cc:588: // Check that the receiver isn't a smi.
> > On 2010/02/15 11:04:21, Mads Ager wrote:
> > > Not sure that the order of these checks matter much, but they are reversed
> > here
> > > compared to the ARM version.  Would be nice to be consistent.  It is
> probably
> > > most likely that the index is not a smi, so maybe check that first on all
> > > platforms?
> > 
> > I just followed the sequence from KeyedLoadIC::GenerateGeneric for the
> > corresponding platforms.
> > 
> > What would you prefer: keep things as is now or adjust both generic and
> > interceptor cases?
> 
> It probably makes very little difference, but I like the platforms to be
> consistent.