Always load the JavaScript builtins code entry from the JavaScript...

Always load the JavaScript builtins code entry from the JavaScript
function instead of baking in the address of the first one that we see
in code.

This removes the need for fixups processing and makes the stubs safe
when there is no natives cache and therefore multiple versions of the
builtin functions.


Committed: http://code.google.com/p/v8/source/detail?r=3832

[Erik Corry, 2010-02-10 17:22:48]

LGTM!

(Got to love a change that removes so much code!)

http://codereview.chromium.org/594009/diff/1/12
File src/arm/macro-assembler-arm.cc (right):

http://codereview.chromium.org/594009/diff/1/12#newcode1254
src/arm/macro-assembler-arm.cc:1254: ldr(r1, FieldMemOperand(target,
builtins_offset));
Seems like you might as well just use r1 throughout instead of having a hidden
scratch register here.

http://codereview.chromium.org/594009/diff/1/12#newcode1257
src/arm/macro-assembler-arm.cc:1257: ldr(target, FieldMemOperand(target,
SharedFunctionInfo::kCodeOffset));
A chain of 5 dependent loads!  If this turns out to be slow we should consider
putting the builtins array in the root array instead of accessing it through the
context.  Since you showed it's no slower on Intel perhaps it doesn't matter.

[Mads Ager (chromium), 2010-02-11 07:58:30]

http://codereview.chromium.org/594009/diff/1/12
File src/arm/macro-assembler-arm.cc (right):

http://codereview.chromium.org/594009/diff/1/12#newcode1254
src/arm/macro-assembler-arm.cc:1254: ldr(r1, FieldMemOperand(target,
builtins_offset));
On 2010/02/10 17:22:49, Erik Corry wrote:
> Seems like you might as well just use r1 throughout instead of having a hidden
> scratch register here.

Yes, done on all platforms.