Implement fast load and call of arguments in the presence of eval....

Implement fast load and call of arguments in the presence of eval.

Load the arguments object from the context if there are no extensions
objects on the way. Then load the argument with a keyed load ic.


Committed: http://code.google.com/p/v8/source/detail?r=4633

[Mads Ager (chromium), 2010-05-10 11:34:17]

The code for loads in the presence of eval is duplicated in the code generators
currently. In order to fix that I need to play with the virtual frame and I'll
do that as a follow-up change.

[Erik Corry, 2010-05-10 11:49:58]

LGTM

http://codereview.chromium.org/2033004/diff/1/3
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/2033004/diff/1/3#newcode4734
src/ia32/codegen-ia32.cc:4734: NOT_INSIDE_TYPEOF,
How do we suddenly know we are not in a typeof?

http://codereview.chromium.org/2033004/diff/1/5
File test/mjsunit/arguments-load-across-eval.js (right):

http://codereview.chromium.org/2033004/diff/1/5#newcode77
test/mjsunit/arguments-load-across-eval.js:77: f();
If you look at x and h here, which one do you get, and shouldn't we test that
too?

[Mads Ager (chromium), 2010-05-10 12:20:46]

http://codereview.chromium.org/2033004/diff/1/3
File src/ia32/codegen-ia32.cc (right):

http://codereview.chromium.org/2033004/diff/1/3#newcode4734
src/ia32/codegen-ia32.cc:4734: NOT_INSIDE_TYPEOF,
On 2010/05/10 11:49:58, Erik Corry wrote:
> How do we suddenly know we are not in a typeof?

Whoops, we don't! Well spotted. Will add regression test.

http://codereview.chromium.org/2033004/diff/1/5
File test/mjsunit/arguments-load-across-eval.js (right):

http://codereview.chromium.org/2033004/diff/1/5#newcode77
test/mjsunit/arguments-load-across-eval.js:77: f();
On 2010/05/10 11:49:58, Erik Corry wrote:
> If you look at x and h here, which one do you get, and shouldn't we test that
> too?

You get x == 1 and h() == 2. I have added the test.