Improve range analysis for modulo operations.

Improve range analysis for modulo operations.

R=mstarzinger@chromium.org

Committed: http://code.google.com/p/v8/source/detail?r=14884

[Michael Starzinger, 2013-05-29 11:59:54]

LGTM with two comments.

https://codereview.chromium.org/15952004/diff/1/src/hydrogen-instructions.cc
File src/hydrogen-instructions.cc (right):

https://codereview.chromium.org/15952004/diff/1/src/hydrogen-instructions.cc#...
src/hydrogen-instructions.cc:1815: // The magnitude of the modulus is bounded by
the right operand.
A comment why we need to go through negtive-absolutes here would be nice. After
discussing it offline I assume it's because of overflow at or near kMinInt.

https://codereview.chromium.org/15952004/diff/1/src/hydrogen-instructions.cc#...
src/hydrogen-instructions.cc:1816: int32_t positive_bound =
-Min(NegAbs(b->lower()), NegAbs(b->upper())) - 1;
Not completely sure but isn't there a parenthesis missing around the whole
expression after the "-" operator?

[Sven Panne, 2013-05-29 12:41:32]

As discussed offline, I've removed the incorrect intersection. We could do
something more clever, taking the left operand's range into account, but I'll
leave this for another CL (if I have time ;-). As it is, some range is better
than no range at all...

[Sven Panne, 2013-05-29 12:42:12]

Committed patchset #2 manually as r14884 (presubmit successful).