Simplify interface to optimized code map lookup.

Simplify interface to optimized code map lookup.

This is one step torwards extracting an OptimizedCodeMap out from the
SharedFunctionInfo in order to have a more flexible implementation.

R=bmeurer@chromium.org, jarin@chromium.org

Committed: https://chromium.googlesource.com/v8/v8/+/112f1973415bba4a6afc887a009fabe5bb4ba236

[Jarin, 2015-06-24 14:01:53]

jarin@chromium.org changed reviewers:
+ jarin@chromium.org

[Jarin, 2015-06-24 14:01:54]

drive-by lgtm.

https://codereview.chromium.org/1205783003/diff/1/src/factory.cc
File src/factory.cc (right):

https://codereview.chromium.org/1205783003/diff/1/src/factory.cc#newcode1390
src/factory.cc:1390: if (cached.literals == nullptr && !info->bound()) {
Any reason why not to say something like

...
} else if (!info->bound()) {
  // Maybe if you like, DCHECK(cached.literals == nullptr);
  ...
}

[Michael Starzinger, 2015-06-24 14:36:42]

https://codereview.chromium.org/1205783003/diff/1/src/factory.cc
File src/factory.cc (right):

https://codereview.chromium.org/1205783003/diff/1/src/factory.cc#newcode1390
src/factory.cc:1390: if (cached.literals == nullptr && !info->bound()) {
On 2015/06/24 14:01:53, jarin wrote:
> Any reason why not to say something like
> 
> ...
> } else if (!info->bound()) {
>   // Maybe if you like, DCHECK(cached.literals == nullptr);
>   ...
> }

I would like to get to an interface where the result of the lookup can be
contain a code object but no literals array (i.e. { code=<some-code>,
literals=nullptr}) to indicate that the code object can be used, but a new
literals array needs to be allocated.

This would happen when we have optimized code that can be shared across
native-contexts. The code can be shared, but the literals array cannot, because
the boilerplate objects always are specific to one native-context.

WDYT?

[Benedikt Meurer, 2015-06-25 04:12:14]

lgtm

[Jarin, 2015-06-25 07:37:05]

On 2015/06/24 14:36:42, Michael Starzinger wrote:
> https://codereview.chromium.org/1205783003/diff/1/src/factory.cc
> File src/factory.cc (right):
> 
> https://codereview.chromium.org/1205783003/diff/1/src/factory.cc#newcode1390
> src/factory.cc:1390: if (cached.literals == nullptr && !info->bound()) {
> On 2015/06/24 14:01:53, jarin wrote:
> > Any reason why not to say something like
> > 
> > ...
> > } else if (!info->bound()) {
> >   // Maybe if you like, DCHECK(cached.literals == nullptr);
> >   ...
> > }
> 
> I would like to get to an interface where the result of the lookup can be
> contain a code object but no literals array (i.e. { code=<some-code>,
> literals=nullptr}) to indicate that the code object can be used, but a new
> literals array needs to be allocated.
> 
> This would happen when we have optimized code that can be shared across
> native-contexts. The code can be shared, but the literals array cannot,
because
> the boilerplate objects always are specific to one native-context.
> 
> WDYT?

Makes sense.

[Michael Starzinger, 2015-06-25 07:52:23]

The CQ bit was checked by mstarzinger@chromium.org

[Michael Starzinger, 2015-06-25 08:28:54]

Committed patchset #1 (id:1) manually as
112f1973415bba4a6afc887a009fabe5bb4ba236 (presubmit successful).

[jochen (gone - plz use gerrit), 2015-07-13 07:11:49]

jochen@chromium.org changed reviewers:
+ jochen@chromium.org

[jochen (gone - plz use gerrit), 2015-07-13 07:11:50]

https://codereview.chromium.org/1205783003/diff/1/src/objects.cc
File src/objects.cc (right):

https://codereview.chromium.org/1205783003/diff/1/src/objects.cc#newcode10610
src/objects.cc:10610: return {Code::cast(optimized_code_map->get(i +
kCachedCodeOffset)),
fyi, this syntax is forbidden by chromium's C++ style guide because MSVS does
not create correct code for it

[Michael Starzinger, 2015-07-13 07:43:31]

https://codereview.chromium.org/1205783003/diff/1/src/objects.cc
File src/objects.cc (right):

https://codereview.chromium.org/1205783003/diff/1/src/objects.cc#newcode10610
src/objects.cc:10610: return {Code::cast(optimized_code_map->get(i +
kCachedCodeOffset)),
On 2015/07/13 07:11:50, jochen wrote:
> fyi, this syntax is forbidden by chromium's C++ style guide because MSVS does
> not create correct code for it

Can we haz presubmit check then?