Copy propagated type info when inserting conversion.

Copy propagated type info when inserting conversion.

Use propagated type info to improve smi-checking before
double operations (CheckEitherNonSmi)
Committed: https://code.google.com/p/dart/source/detail?r=18692

[Vyacheslav Egorov (Google), 2013-02-19 14:45:49]

lgtm

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language.h:255: Value* CopyWithType() {return new
Value(definition_, reaching_type_); }
Please add a comment here that CopyWithType should only be used when one can
guarantee that it'll not be hoisted beyound the guard that guaranteed reaching
type.

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
File runtime/vm/intermediate_language_ia32.cc (right):

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language_ia32.cc:2354: __ testl(right,
Immediate(kSmiTagMask));
How about canonicalizing CHeckEitherNonSmi into CheckNonSmi instead of
duplicating its facilities here?

[Florian Schneider, 2013-02-19 15:02:06]

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language.h:255: Value* CopyWithType() {return new
Value(definition_, reaching_type_); }
On 2013/02/19 14:45:49, Vyacheslav Egorov (Google) wrote:
> Please add a comment here that CopyWithType should only be used when one can
> guarantee that it'll not be hoisted beyound the guard that guaranteed reaching
> type.

Done.

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
File runtime/vm/intermediate_language_ia32.cc (right):

https://codereview.chromium.org/12298034/diff/1/runtime/vm/intermediate_langu...
runtime/vm/intermediate_language_ia32.cc:2354: __ testl(right,
Immediate(kSmiTagMask));
On 2013/02/19 14:45:49, Vyacheslav Egorov (Google) wrote:
> How about canonicalizing CHeckEitherNonSmi into CheckNonSmi instead of
> duplicating its facilities here?

We don't have CheckNonSmi (yet).

[srdjan, 2013-02-19 16:51:47]

DBC

https://codereview.chromium.org/12298034/diff/8001/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/12298034/diff/8001/runtime/vm/intermediate_la...
runtime/vm/intermediate_language.h:227: explicit Value(Definition* definition,
CompileType* type = NULL)
Why the optional argument? We really should avoid them.