Intrinsify InternalByteArray's get_length and [].

Intrinsify InternalByteArray's get_length and []. 
Disallow intrinsification of non-corelib methods. Recognize private classes for intrinsification.
Committed: https://code.google.com/p/dart/source/detail?r=4171

[srdjan, 2012-02-10 14:36:27]

Note: indexed setter will follow later. Inlined code (optimized codegen) will
follow later.

[siva, 2012-02-11 01:39:29]

LGTM

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32.cc
File runtime/vm/intrinsifier_ia32.cc (right):

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:77: V(_ByteArrayBase, [],
ByteArrayBase_getIndexed)                              \
The more interesting ones are the getInt8, getUint8, getInt16, getUint16,
getInt32 ..... getFloar32(), getFloar64() etc. functions. Will you also be
intrinsifying those?

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:333: // Handles only class InternalByteArray.
Would you at some point also implement a similar intrinsification for external
byte arrays?

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:368: __ SmiTag(EAX);
Perhaps a comment that values stored in the byte array are regular values and
hence we need to SmiTag them on access.

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:1224: }                                         
                                \
This piece of code here seems too big to repeat for every function being
instrisified, why not make it a function and call it here in the macro...

The function could return true in which case you return destination(assembler)
otherwise keep going.

[srdjan, 2012-02-12 06:52:24]

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32.cc
File runtime/vm/intrinsifier_ia32.cc (right):

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:77: V(_ByteArrayBase, [],
ByteArrayBase_getIndexed)                              \
On 2012/02/11 01:39:29, asiva wrote:
> The more interesting ones are the getInt8, getUint8, getInt16, getUint16,
> getInt32 ..... getFloar32(), getFloar64() etc. functions. Will you also be
> intrinsifying those?

Do we have existing code/benchmarks that are already using those methods?
I picked the two methods above because I know that Peter is using them in the
scanner.
But, yes, all the mentioned methods will be intrinsified.

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:333: // Handles only class InternalByteArray.
On 2012/02/11 01:39:29, asiva wrote:
> Would you at some point also implement a similar intrinsification for external
> byte arrays?

Yes.

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:368: __ SmiTag(EAX);
On 2012/02/11 01:39:29, asiva wrote:
> Perhaps a comment that values stored in the byte array are regular values and
> hence we need to SmiTag them on access.

Done.

http://codereview.chromium.org/9383004/diff/3001/runtime/vm/intrinsifier_ia32...
runtime/vm/intrinsifier_ia32.cc:1224: }                                         
                                \
On 2012/02/11 01:39:29, asiva wrote:
> This piece of code here seems too big to repeat for every function being
> instrisified, why not make it a function and call it here in the macro...
> 
> The function could return true in which case you return destination(assembler)
> otherwise keep going.

Good point. Done.