Add type check tracing (--trace_type_checks). Inline type checks with class types that have only ...

Add type check tracing (--trace_type_checks). Inline type checks with class types that have only Dynamic type arguments (or none) and with List<Dynamic> interface.
Committed: https://code.google.com/p/dart/source/detail?r=1372

[regis, 2011-11-09 01:47:53]

Just curious, what is the speedup brought by this change?

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc
File runtime/lib/error.cc (right):

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc#newcode218
runtime/lib/error.cc:218: if (FLAG_trace_type_checks) {
Are you planning on leaving this flag in or was it just to help you debug
inlining of type checks?

If we keep it, we should print all type checks, otherwise it is misleading.

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc#newcode224
runtime/lib/error.cc:224: == Class::kNoTypeArguments) ||
cls.HasTypeArguments() ?

And why only print raw types?

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc#newcode228
runtime/lib/error.cc:228: OS::Print("TypeCheck <%s> vs <%s>\n",
The chosen print syntax with "<" and ">" is unfortunate :-)

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
File runtime/vm/code_generator_ia32.cc (right):

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1544: // A class equality check is only
applicable to a non-parameterized class.
// A class equality check is only applicable with a dst type of a
non-parameterized class or with a raw dst type of a parameterized class.

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1545: if (dst_has_type_arguments) {
dst_class_has_type_arguments

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1546: const TypeArguments& arguments =
arguments -> dst_type_arguments.

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1548: bool no_type_arguments =
arguments.IsNull() ||
Why the local? Make it const if you want to keep it:
const bool is_raw_dst_type = dst_type_arguments.IsNull() ...

[srdjan, 2011-11-09 19:36:30]

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc
File runtime/lib/error.cc (right):

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc#newcode218
runtime/lib/error.cc:218: if (FLAG_trace_type_checks) {
On 2011/11/09 01:47:53, regis wrote:
> Are you planning on leaving this flag in or was it just to help you debug
> inlining of type checks?
> 
> If we keep it, we should print all type checks, otherwise it is misleading.

I plan to keep it in, removed "if (known_type)".

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc#newcode224
runtime/lib/error.cc:224: == Class::kNoTypeArguments) ||
On 2011/11/09 01:47:53, regis wrote:
> cls.HasTypeArguments() ?

Removed.

> 
> And why only print raw types?

http://codereview.chromium.org/8503034/diff/4002/runtime/lib/error.cc#newcode228
runtime/lib/error.cc:228: OS::Print("TypeCheck <%s> vs <%s>\n",
On 2011/11/09 01:47:53, regis wrote:
> The chosen print syntax with "<" and ">" is unfortunate :-)

:-). Using quotes instead.

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
File runtime/vm/code_generator_ia32.cc (right):

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1544: // A class equality check is only
applicable to a non-parameterized class.
On 2011/11/09 01:47:53, regis wrote:
> // A class equality check is only applicable with a dst type of a
> non-parameterized class or with a raw dst type of a parameterized class.

Done.

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1545: if (dst_has_type_arguments) {
On 2011/11/09 01:47:53, regis wrote:
> dst_class_has_type_arguments

Done.

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1546: const TypeArguments& arguments =
On 2011/11/09 01:47:53, regis wrote:
> arguments -> dst_type_arguments.

Done.

http://codereview.chromium.org/8503034/diff/4002/runtime/vm/code_generator_ia...
runtime/vm/code_generator_ia32.cc:1548: bool no_type_arguments =
arguments.IsNull() ||
On 2011/11/09 01:47:53, regis wrote:
> Why the local? Make it const if you want to keep it:
> const bool is_raw_dst_type = dst_type_arguments.IsNull() ...

Made it a const. Giving it a name makes clear what we are testing.

[regis, 2011-11-09 19:43:38]

LGTM

5% are nice to take.