Porting r10023 and r10054 to arm (pointer cache for external strings).

Porting r10023 and r10054 to arm (pointer cache for external strings).

Committed: http://code.google.com/p/v8/source/detail?r=10059

[Yang, 2011-11-23 14:36:01]

Please take a look.

The first patch set merely refactors so that both the full-codegen version and
the crankshaft version of charCodeAt use the same piece of code now generated by
StringCharLoadGenerator::Generate in codegen-arm.cc

The second patch set includes the new external string handling.

[Rico, 2011-11-24 06:55:14]

LGMT with comments addressed

http://codereview.chromium.org/8682010/diff/1001/src/arm/code-stubs-arm.cc
File src/arm/code-stubs-arm.cc (right):

http://codereview.chromium.org/8682010/diff/1001/src/arm/code-stubs-arm.cc#ne...
src/arm/code-stubs-arm.cc:5072: masm, object_, index_, result_, &call_runtime_);
move masm up and put parameters on individual lines

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc#newco...
src/arm/codegen-arm.cc:332: // Handle external strings.
Any reason for not just having this be fall-through code after the
check_sequential branching and then just negating the branch to jump over? It
does not seem intuitive to me that this is placed up here (if you move it,
please also do so on ia32 and x64)

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc#newco...
src/arm/codegen-arm.cc:358: // Handle conses.
Handle cons strings?

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc#newco...
src/arm/codegen-arm.cc:379: // the adjusted index to the runtime function.
Please fix comment, we don't do this anymore. Please also update this on the
other platforms

[Yang, 2011-11-24 11:15:11]

Landed with r10059.

http://codereview.chromium.org/8682010/diff/1001/src/arm/code-stubs-arm.cc
File src/arm/code-stubs-arm.cc (right):

http://codereview.chromium.org/8682010/diff/1001/src/arm/code-stubs-arm.cc#ne...
src/arm/code-stubs-arm.cc:5072: masm, object_, index_, result_, &call_runtime_);
On 2011/11/24 06:55:15, Rico wrote:
> move masm up and put parameters on individual lines

Done.

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc
File src/arm/codegen-arm.cc (right):

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc#newco...
src/arm/codegen-arm.cc:332: // Handle external strings.
On 2011/11/24 06:55:15, Rico wrote:
> Any reason for not just having this be fall-through code after the
> check_sequential branching and then just negating the branch to jump over? It
> does not seem intuitive to me that this is placed up here (if you move it,
> please also do so on ia32 and x64)

Done.

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc#newco...
src/arm/codegen-arm.cc:358: // Handle conses.
On 2011/11/24 06:55:15, Rico wrote:
> Handle cons strings?

Done.

http://codereview.chromium.org/8682010/diff/1001/src/arm/codegen-arm.cc#newco...
src/arm/codegen-arm.cc:379: // the adjusted index to the runtime function.
On 2011/11/24 06:55:15, Rico wrote:
> Please fix comment, we don't do this anymore. Please also update this on the
> other platforms

Done.