Do not directly load smi constants larger than a 16 bit payload on ia32.

Do not directly load smi constants larger than a 16 bit payload on ia32.

Instead of emitting a large smi constant directly as an immediate, xor the
smi with a random 32-bit cookie at compile time, emit code to move that into
a register and then to xor the register with the cookie.

1. In unoptimized code constants are always loaded into a register in
   ConstantInstr::EmitNativeCode.  Change it to load safely.

2. In optimized code constants loaded into a register for use by an
   instruction are loaded by ParallelMoveResolver::EmitMove.  Change it to
   load safely.

3. In optimized code some instructions allow constants as inputs.  These are
   reflected with an input location of Location::RegisterOrConstant (or
   similar) or else Location::Constant.  For the former allocate a register
   if the constant is unsafe, and for the latter check for safety before
   allowing the constant input.

BUG=https://code.google.com/p/dart/issues/detail?id=1742
R=iposva@google.com, srdjan@google.com

Committed: https://code.google.com/p/dart/source/detail?r=29439

[Kevin Millikin (Google), 2013-10-28 13:35:46]

The intent of this change is to intercept all immediates with more than 16 bits
of payload.  This change to the compiler should be sufficient.  16 is the length
used by V8.

Note that V8 does not do this for optimized code.  I'm still waiting for full
performance numbers, I'll update as soon as they're in.

How do you want to generate the cookie?

[srdjan, 2013-10-28 23:06:49]

LGTM, assuming the cookie computation will be changed not to be constant.

https://codereview.chromium.org/48743002/diff/1/runtime/vm/assembler_ia32.h
File runtime/vm/assembler_ia32.h (right):

https://codereview.chromium.org/48743002/diff/1/runtime/vm/assembler_ia32.h#n...
runtime/vm/assembler_ia32.h:640: // If 'object' is a Smi, xor it with a
per-assembler cookie value to
If 'object' is unsafe Smi.

[Ivan Posva, 2013-10-29 05:58:39]

LGTM

-Ivan

P.S. Please assign the bug to me that we can wire up a proper cookie value.

[Kevin Millikin (Google), 2013-10-29 12:23:20]

Committed patchset #1 manually as r29439 (presubmit successful).