Allow uint32 value on optimized frames if they are consumed by safe operations.

Allow uint32 value on optimized frames if they are consumed by safe operations.

Safe operations are those that either do not observe unsignedness or have special support for uint32 values:

- all binary bitwise operations: they perform ToInt32 on inputs;
- >> and << shifts: they perform ToInt32 on left hand side and ToUint32 on right hand side;
- >>> shift: it performs ToUint32 on both inputs;
- stores to integer external arrays (not pixel, float or double ones): these stores are "bitwise";
- HChange: special support added for conversions of uint32 values to double and tagged values;
- HSimulate: special support added for deoptimization with uint32 values in registers and stack slots;
- HPhi: phis that have only safe uses and only uint32 operands are uint32 themselves. 

BUG=v8:2097
TEST=test/mjsunit/compiler/uint32.js

Committed: https://code.google.com/p/v8/source/detail?r=12367

[Vyacheslav Egorov (Google), 2012-07-17 12:23:45]

This is request for comments (not for commit, lacks full ARM/x64 support).

But it passes all tests and is pretty close to what I would like to commit at
the end.

[Vyacheslav Egorov (Google), 2012-08-15 12:42:48]

Hi again,

The change has been cleaned up and extended.

Ports to ia32 and x64 are added.

I would like to land it before it rots :-)

Can you review it?

[Massi, 2012-08-16 13:57:55]

http://codereview.chromium.org/10778029/diff/6001/src/deoptimizer.h
File src/deoptimizer.h (right):

http://codereview.chromium.org/10778029/diff/6001/src/deoptimizer.h#newcode681
src/deoptimizer.h:681: if (value <= static_cast<uint32_t>(Smi::kMaxValue)) {
Maybe having a Smi::IsValidUint(uint32_t value) method to mirror the existing
Smi::IsValid and a Factory::NewNumberFromUInt method mirroring
Factory::NewNumberFromInt would make the code a bit more readable (the INT32 and
UINT32 cases would be coded in a similar way but calling different methods).

[danno, 2012-08-21 13:55:26]

Fundamentally LGTM. However, if there is anyway to beef up the tests to make
sure that all of the cases in the uint32 tracking algorithm are covered, please
add them.

http://codereview.chromium.org/10778029/diff/6001/test/mjsunit/compiler/uint3...
File test/mjsunit/compiler/uint32.js (right):

http://codereview.chromium.org/10778029/diff/6001/test/mjsunit/compiler/uint3...
test/mjsunit/compiler/uint32.js:120: assertEquals(K4 | 0, NonUint32Phi(false,
K3, K4));
Please add tests for phi usages of phis.