Switch schedule early phase to use forward propagation.

Switch schedule early phase to use forward propagation.

R=jarin@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=24923

[Jarin, 2014-10-27 20:12:52]

lgtm.

https://codereview.chromium.org/655833005/diff/1/src/compiler/scheduler.cc
File src/compiler/scheduler.cc (right):

https://codereview.chromium.org/655833005/diff/1/src/compiler/scheduler.cc#ne...
src/compiler/scheduler.cc:960: // Run the schedule late algorithm on a set of
fixed root nodes.
Do you mean schedule *early* or am I missing something?

https://codereview.chromium.org/655833005/diff/1/src/compiler/scheduler.cc#ne...
src/compiler/scheduler.cc:991: for (Node::Uses::iterator i = uses.begin(); i !=
uses.end(); ++i) {
Just a warning here: in my experience, the uses collection may contain a lot of
junk. Especially with deoptimization on, there might be a bunch of dead
framestates hanging off the uses collection. However, it possible that Ben's
recent work on pruning the graph might help here.

[Michael Starzinger, 2014-10-28 09:33:56]

https://codereview.chromium.org/655833005/diff/1/src/compiler/scheduler.cc
File src/compiler/scheduler.cc (right):

https://codereview.chromium.org/655833005/diff/1/src/compiler/scheduler.cc#ne...
src/compiler/scheduler.cc:960: // Run the schedule late algorithm on a set of
fixed root nodes.
On 2014/10/27 20:12:52, jarin wrote:
> Do you mean schedule *early* or am I missing something?

Done.

https://codereview.chromium.org/655833005/diff/1/src/compiler/scheduler.cc#ne...
src/compiler/scheduler.cc:991: for (Node::Uses::iterator i = uses.begin(); i !=
uses.end(); ++i) {
On 2014/10/27 20:12:52, jarin wrote:
> Just a warning here: in my experience, the uses collection may contain a lot
of
> junk. Especially with deoptimization on, there might be a bunch of dead
> framestates hanging off the uses collection. However, it possible that Ben's
> recent work on pruning the graph might help here.

Acknowledged. Yes, I totally agree with that. As discussed offline, Ben already
landed his control reducer that includes pruning of dead values.

[Michael Starzinger, 2014-10-28 09:34:54]

Committed patchset #2 (id:20001) manually as 24923 (presubmit successful).