Don't preparse large files to find boundaries of lazy functions.

Don't preparse large files to find boundaries of lazy functions.

Instead use the preparser inline to parse only the lazy function
bodies.

This is still disabled for small files. 
More measurements are needed to determine if lazy-compiling small
sources is worth it.

Committed: http://code.google.com/p/v8/source/detail?r=10066

[Lasse Reichstein, 2011-11-24 11:47:22]

PTAL

[Lasse Reichstein, 2011-11-24 11:51:08]

Adding Yang since Jakob is pretty busy right now.

[Yang, 2011-11-24 14:28:41]

LGTM with comment.

http://codereview.chromium.org/8662037/diff/1/src/parser.cc
File src/parser.cc (right):

http://codereview.chromium.org/8662037/diff/1/src/parser.cc#newcode4106
src/parser.cc:4106: // compile if we do not have preparser data for the
function.
Comment inaccurate. If I'm not mistaken, we also compile lazily if the script
size is large enough.

[Lasse Reichstein, 2011-11-24 21:16:47]

http://codereview.chromium.org/8662037/diff/1/src/parser.cc
File src/parser.cc (right):

http://codereview.chromium.org/8662037/diff/1/src/parser.cc#newcode4106
src/parser.cc:4106: // compile if we do not have preparser data for the
function.
That was actually correct before - if the script was large enough, we created
pre-parser data, so at this point we would have it.

This CL changes that. We now don't create preparse data ahead of time, but
instead call the preparser only on the body of the function - if we are allowed
to compile lazily (some calls to Parse expects a full AST afterwards).

So the comment is outdated now. Will change.