Specialize string equality for various string classes. Implement intrinsics for string equal…

Specialize string equality for various string classes. Add intrinsics for one/two-byte string equality. Improves sudoku_solver by 10%.

R=regis@google.com, zra@google.com

Committed: https://code.google.com/p/dart/source/detail?r=29266

[regis, 2013-10-24 23:00:05]

LGTM

https://codereview.chromium.org/41573003/diff/80001/runtime/lib/string_patch....
File runtime/lib/string_patch.dart (right):

https://codereview.chromium.org/41573003/diff/80001/runtime/lib/string_patch....
runtime/lib/string_patch.dart:534: return super == other;
Is this the same as calling ==(other) on super? Just curious.

[zra, 2013-10-24 23:59:46]

lgtm

https://codereview.chromium.org/41573003/diff/80001/runtime/vm/intrinsifier_a...
File runtime/vm/intrinsifier_arm.cc (right):

https://codereview.chromium.org/41573003/diff/80001/runtime/vm/intrinsifier_a...
runtime/vm/intrinsifier_arm.cc:1701: __ AddImmediate(R1, 2);
Forgot that Regis did implement scaled indexes for ARM. So, I think you could
say:

ldrh(R3, Address(R0, R2, LSL, 1))

a la ia32, but I think this code is fine as is. Maybe add a TODO for me to try
the other sequence when I start looking more deeply at performance on ARM and
MIPS.

[srdjan, 2013-10-25 00:13:17]

https://codereview.chromium.org/41573003/diff/80001/runtime/lib/string_patch....
File runtime/lib/string_patch.dart (right):

https://codereview.chromium.org/41573003/diff/80001/runtime/lib/string_patch....
runtime/lib/string_patch.dart:534: return super == other;
On 2013/10/24 23:00:05, regis wrote:
> Is this the same as calling ==(other) on super? Just curious.

Yes.

https://codereview.chromium.org/41573003/diff/80001/runtime/vm/intrinsifier_a...
File runtime/vm/intrinsifier_arm.cc (right):

https://codereview.chromium.org/41573003/diff/80001/runtime/vm/intrinsifier_a...
runtime/vm/intrinsifier_arm.cc:1701: __ AddImmediate(R1, 2);
On 2013/10/24 23:59:46, zra wrote:
> Forgot that Regis did implement scaled indexes for ARM. So, I think you could
> say:
> 
> ldrh(R3, Address(R0, R2, LSL, 1))
> 
> a la ia32, but I think this code is fine as is. Maybe add a TODO for me to try
> the other sequence when I start looking more deeply at performance on ARM and
> MIPS.

Done.

[srdjan, 2013-10-25 18:00:35]

Committed patchset #3 manually as r29266 (presubmit successful).