We can actually map the Dart shift right to the JavaScript one under certain conditions.

We can actually map the Dart shift right to the JavaScript one under certain conditions.

R=kasperl@google.com

Committed: https://code.google.com/p/dart/source/detail?r=30665

[floitsch, 2013-11-25 18:19:02]

For Kasper's information: we discovered that the right-shift semantics is more
complicated: if the receiver is positive it should use >>>, otherwise it should
use >> (and then make it a positive number by applying >>>).

[kasperl, 2013-11-26 08:29:29]

https://codereview.chromium.org/85663003/diff/30001/sdk/lib/_internal/compile...
File
sdk/lib/_internal/compiler/implementation/ssa/invoke_dynamic_specializers.dart
(right):

https://codereview.chromium.org/85663003/diff/30001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/ssa/invoke_dynamic_specializers.dart:216:
bool clearAllSideEffects(HInstruction instruction) {
bool -> void

https://codereview.chromium.org/85663003/diff/30001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/ssa/invoke_dynamic_specializers.dart:348:
return instruction is HBinaryBitOp;
No way of using value range information here? Add a comment that explains why or
a TODO. At least you would think that certain constants (like 1) could be
trivially positive. You may not hit that case because of constant folding, but
still...

[ngeoffray, 2013-11-26 08:36:35]

PTAL

https://codereview.chromium.org/85663003/diff/30001/sdk/lib/_internal/compile...
File
sdk/lib/_internal/compiler/implementation/ssa/invoke_dynamic_specializers.dart
(right):

https://codereview.chromium.org/85663003/diff/30001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/ssa/invoke_dynamic_specializers.dart:216:
bool clearAllSideEffects(HInstruction instruction) {
On 2013/11/26 08:29:29, kasperl wrote:
> bool -> void

Done.

https://codereview.chromium.org/85663003/diff/30001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/ssa/invoke_dynamic_specializers.dart:348:
return instruction is HBinaryBitOp;
On 2013/11/26 08:29:29, kasperl wrote:
> No way of using value range information here? Add a comment that explains why
or
> a TODO. At least you would think that certain constants (like 1) could be
> trivially positive. You may not hit that case because of constant folding, but
> still...

I have added a TODO for the range analysis. I have a follow-up CL that will
recognize positive constants.

[kasperl, 2013-11-26 09:06:16]

LGTM.

[ngeoffray, 2013-11-26 09:11:02]

Committed patchset #5 manually as r30665 (presubmit successful).