For instanceof inline check against right hand class id possible and interface String for OneByte...

For instanceof inline check against right hand class id possible and interface String for OneByteString (positive checks only). Improves frog runtime by about 20%. Future work will be based on collecting types and using that data in optimized code. Add tracing of instance of type checks.
Committed: https://code.google.com/p/dart/source/detail?r=1408

[regis, 2011-11-10 03:09:50]

LGTM with a few comments

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator.cc...
runtime/vm/code_generator.cc:333: OS::Print("InstanceOf '%s' vs '%s'\n",
You could use the result of the type test to print "is" or "is !", instead of
"vs".

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
File runtime/vm/code_generator_ia32.cc (right):

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1420: // TODO(regis): Should we still inline a
Smi type check when checking for a
I think you can remove this TODO now.

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1432: __ j(ZERO, &runtime_call,
Assembler::kNearJump);
Can't you perform this Smi check even if the type is not raw?

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1599: } else {  // dst_type has NO type
parameters.
Sorry to be picky, but "type parameters" is the wrong term. It should be "type
arguments".

[srdjan, 2011-11-10 17:24:41]

Thanks!

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator.cc...
runtime/vm/code_generator.cc:333: OS::Print("InstanceOf '%s' vs '%s'\n",
On 2011/11/10 03:09:51, regis wrote:
> You could use the result of the type test to print "is" or "is !", instead of
> "vs".

Done.

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
File runtime/vm/code_generator_ia32.cc (right):

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1420: // TODO(regis): Should we still inline a
Smi type check when checking for a
On 2011/11/10 03:09:51, regis wrote:
> I think you can remove this TODO now.

Done.

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1432: __ j(ZERO, &runtime_call,
Assembler::kNearJump);
On 2011/11/10 03:09:51, regis wrote:
> Can't you perform this Smi check even if the type is not raw?

Done.

http://codereview.chromium.org/8508035/diff/3001/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1599: } else {  // dst_type has NO type
parameters.
On 2011/11/10 03:09:51, regis wrote:
> Sorry to be picky, but "type parameters" is the wrong term. It should be "type
> arguments".

Done.