Merge BuildLoadKeyedGeneric and BuildStoreKeyedGeneric, switch on AccessType

Merge BuildLoadKeyedGeneric and BuildStoreKeyedGeneric, switch on AccessType

R=ishell@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=19228

[Toon Verwaest, 2014-02-10 14:41:20]

PTAL

[Igor Sheludko, 2014-02-10 14:47:11]

lgtm

[Toon Verwaest, 2014-02-10 14:58:31]

Committed patchset #2 manually as r19228 (presubmit successful).

[tfarina, 2014-02-10 15:01:28]

https://codereview.chromium.org/146623007/diff/40001/src/hydrogen.cc
File src/hydrogen.cc (right):

https://codereview.chromium.org/146623007/diff/40001/src/hydrogen.cc#newcode6177
src/hydrogen.cc:6177: } else {
no 'else' after return.

you can also use a ternay operator here:

return (access_type == LOAD) ? New<...>(...) : New<...>(...);

[Sven Panne, 2014-02-11 07:08:35]

Late DBC...

https://codereview.chromium.org/146623007/diff/40001/src/hydrogen.cc
File src/hydrogen.cc (right):

https://codereview.chromium.org/146623007/diff/40001/src/hydrogen.cc#newcode6177
src/hydrogen.cc:6177: } else {
On 2014/02/10 15:01:29, tfarina wrote:
> no 'else' after return.

Although I am a big fan of "comb-like" control flow, where you simply chain
together ifs with early returns for special cases without nesting them (humans
are *very* bad at mentally following nested things), this would be a wrong place
for using it IMHO: Returning early for LOAD would make it look like a special
case, which it isn't.


> you can also use a ternay operator here:
> 
> return (access_type == LOAD) ? New<...>(...) : New<...>(...);

Nope, both "branches" of a ternary have to be of the same type, inheritance
doesn't work here as one would expect. Having said that, I normally prefer
ternaries very much, but only when possible. ;-)