Simplify AST ID generation.

Simplify AST ID generation.

Every CompilationInfo has an ID generator now, and it is never
reset/copied/assigned. Simplified FunctionState.

R=mstarzinger@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=24481

[Sven Panne, 2014-10-08 13:08:48]

svenpanne@chromium.org changed reviewers:
+ mstarzinger@chromium.org

[wingo, 2014-10-08 13:34:45]

wingo@igalia.com changed reviewers:
+ wingo@igalia.com

[wingo, 2014-10-08 13:34:46]

https://codereview.chromium.org/633373003/diff/20001/src/ast.h
File src/ast.h (right):

https://codereview.chromium.org/633373003/diff/20001/src/ast.h#newcode187
src/ast.h:187: IdGen() : id_(BailoutId::FirstUsable().ToInt()) {}
What is the plan with bailout ids -- still a per-function locally unique
incrementing integer?  For parameter destructuring of arrow functions I will
have to transplant expressions from one function to another.  I'm finally
starting on that now:
https://docs.google.com/document/d/1iJWvuakeKsPJOeA-RJ-4bM3Jha5iewQjxKsylgzHE....
 The plan was to see if we could add a post-pass to assign bailout ids, to see
how that would impact perf, because when transplanting you have to preserve
local uniqueness; but is that still a valid strategy?

[Dmitry Lomov (no reviews), 2014-10-08 18:12:49]

On 2014/10/08 13:34:46, wingo wrote:
> https://codereview.chromium.org/633373003/diff/20001/src/ast.h
> File src/ast.h (right):
> 
> https://codereview.chromium.org/633373003/diff/20001/src/ast.h#newcode187
> src/ast.h:187: IdGen() : id_(BailoutId::FirstUsable().ToInt()) {}
> What is the plan with bailout ids -- still a per-function locally unique
> incrementing integer?  For parameter destructuring of arrow functions I will
> have to transplant expressions from one function to another.  I'm finally
> starting on that now:
>
https://docs.google.com/document/d/1iJWvuakeKsPJOeA-RJ-4bM3Jha5iewQjxKsylgzHE....
>  The plan was to see if we could add a post-pass to assign bailout ids, to see
> how that would impact perf, because when transplanting you have to preserve
> local uniqueness; but is that still a valid strategy?

Yes, this is still valid. You still need to guarantee uniqueness, but it does
not have to fill the consecutive range (never had to; I do not know where we
have got the idea that it does)

[Michael Starzinger, 2014-10-09 08:13:09]

LGTM.

[Sven Panne, 2014-10-09 08:16:24]

Committed patchset #3 (id:40001) manually as 24481 (presubmit successful).