Intrinsify random nextState, improve perfromance significantly.

Intrinsify random nextState, improve perfromance significantly.

R=zra@google.com

Committed: https://code.google.com/p/dart/source/detail?r=22912

[zra, 2013-05-20 23:38:17]

https://codereview.chromium.org/15502004/diff/9001/runtime/lib/math_patch.dart
File runtime/lib/math_patch.dart (right):

https://codereview.chromium.org/15502004/diff/9001/runtime/lib/math_patch.dar...
runtime/lib/math_patch.dart:71: int _nextState() {
return type should be void?

https://codereview.chromium.org/15502004/diff/9001/runtime/vm/intrinsifier_ia...
File runtime/vm/intrinsifier_ia32.cc (right):

https://codereview.chromium.org/15502004/diff/9001/runtime/vm/intrinsifier_ia...
runtime/vm/intrinsifier_ia32.cc:1450: ASSERT(a_int_value <= 0xFFFFFFFF);
I'm not sure what this ASSERT and the next line are trying to accomplish, but
the test in the ASSERT will be a signed comparison, which I'm guessing is not
what you want. Maybe use Utils::IsInt or Utils::IsUint as appropriate in the
ASSERT?

[srdjan, 2013-05-20 23:48:25]

Thanks

https://codereview.chromium.org/15502004/diff/9001/runtime/lib/math_patch.dart
File runtime/lib/math_patch.dart (right):

https://codereview.chromium.org/15502004/diff/9001/runtime/lib/math_patch.dar...
runtime/lib/math_patch.dart:71: int _nextState() {
On 2013/05/20 23:38:17, zra wrote:
> return type should be void?

Done.

https://codereview.chromium.org/15502004/diff/9001/runtime/vm/intrinsifier_ia...
File runtime/vm/intrinsifier_ia32.cc (right):

https://codereview.chromium.org/15502004/diff/9001/runtime/vm/intrinsifier_ia...
runtime/vm/intrinsifier_ia32.cc:1450: ASSERT(a_int_value <= 0xFFFFFFFF);
On 2013/05/20 23:38:17, zra wrote:
> I'm not sure what this ASSERT and the next line are trying to accomplish, but
> the test in the ASSERT will be a signed comparison, which I'm guessing is not
> what you want. Maybe use Utils::IsInt or Utils::IsUint as appropriate in the
> ASSERT?

Using Utils::IsUint(32, a_int_value);
Since immediate accepts only int32, I am converting it. It is a mask so only
bits matter.

[zra, 2013-05-20 23:55:09]

lgtm

[srdjan, 2013-05-20 23:59:42]

Committed patchset #4 manually as r22912 (presubmit successful).