Support arguments object access from inlined functions.

Support arguments object access from inlined functions. 

R=fschneider@chromium.org
TEST=test/mjsunit/compiler/inline-arguments.js

Committed: https://code.google.com/p/v8/source/detail?r=11109

[fschneider, 2012-03-22 10:30:19]

LGTM.

https://chromiumcodereview.appspot.com/9837002/diff/1017/src/ia32/lithium-cod...
File src/ia32/lithium-codegen-ia32.cc (right):

https://chromiumcodereview.appspot.com/9837002/diff/1017/src/ia32/lithium-cod...
src/ia32/lithium-codegen-ia32.cc:2547: __ lea(result, Operand(esp, -2 *
kPointerSize));
Maybe it would be nicer to have the result of arguments-elements point right to
the start of the arguments in both cases (inlined and non-inlined), instead of
having the LArgumentsAccessAt compensate for the fixed delta.

https://chromiumcodereview.appspot.com/9837002/diff/1017/test/mjsunit/compile...
File test/mjsunit/compiler/inline-arguments.js (right):

https://chromiumcodereview.appspot.com/9837002/diff/1017/test/mjsunit/compile...
test/mjsunit/compiler/inline-arguments.js:138: return toarr_inner(marker, a/2,
b/2, c/2);
Please also add a test where inlining happens inside an argument-expression:

return foo(uninlined(0), toarr_inner(marker, a/2, b/2, c/2));

[Vyacheslav Egorov (Chromium), 2012-03-22 12:24:53]

https://chromiumcodereview.appspot.com/9837002/diff/1017/src/ia32/lithium-cod...
File src/ia32/lithium-codegen-ia32.cc (right):

https://chromiumcodereview.appspot.com/9837002/diff/1017/src/ia32/lithium-cod...
src/ia32/lithium-codegen-ia32.cc:2547: __ lea(result, Operand(esp, -2 *
kPointerSize));
On 2012/03/22 10:30:19, fschneider wrote:
> Maybe it would be nicer to have the result of arguments-elements point right
to
> the start of the arguments in both cases (inlined and non-inlined), instead of
> having the LArgumentsAccessAt compensate for the fixed delta.

discussed over chat, agreed not to change it for now.

https://chromiumcodereview.appspot.com/9837002/diff/1017/test/mjsunit/compile...
File test/mjsunit/compiler/inline-arguments.js (right):

https://chromiumcodereview.appspot.com/9837002/diff/1017/test/mjsunit/compile...
test/mjsunit/compiler/inline-arguments.js:138: return toarr_inner(marker, a/2,
b/2, c/2);
On 2012/03/22 10:30:19, fschneider wrote:
> Please also add a test where inlining happens inside an argument-expression:
> 
> return foo(uninlined(0), toarr_inner(marker, a/2, b/2, c/2));

Done.