Implement postfix indexed increment.

Implement postfix indexed increment.

Chosen because of its wartiness.  A pair of new computations are introduced
to allow the IL to express duplication of temporaries and storing to
temporaries.  These will be translated away by the optimizing compiler, but
allow the instruction stream to represent exactly the deoptimization state
we are intereseted in (with at most one deopt point per instruction).

Translation of:

test(a, i) {
  return a[i]++;
}

is

 0: [target]
    t0 <-#0
    t1 <-LoadLocal(a)
    t2 <-LoadLocal(i)
    t3 <-CopyTemp(-1)
    t4 <-CopyTemp(-1)
    t3 <-InstanceCall([], t3, t4)
    SetTemp(-3)
    t4 <-#1
    t3 <-InstanceCall(+, t3, t4)
    InstanceCall([]=, t1, t2, t3)
    return t0

R=srdjan@google.com
BUG=
TEST=

Committed: https://code.google.com/p/dart/source/detail?r=5012

[Kevin Millikin (Google), 2012-03-05 15:58:36]

This has the advantage that the load, add/subtract, and store are all normal
instructions.

That can be used to clean up the weird indexed store (with the reintroduction of
the "placeholder") if we want.

Obviously we are starting to see a lot of duplicated code (for things like
loads, stores, binary ops).  That should be cleaned up separately.

Also, I don't intend that CopyTemp and SetTemp use implicit relative addressing.
 It's much nicer to use temp indexes, but that requires some kind of stack
height tracking that we don't currently have.

[srdjan, 2012-03-05 21:44:37]

LGTM in order to move forward and reach our weekly goal.

(note the brief comment about node->id(), something that we wil
l be computing differently in the new compiler)

The suggested solution seems complex, but it seems to work. Note that we plan to
 get rid of IncrOpIndexedNode, IncrOpInstancedFieldNode and
IncrOpStaticFieldNode, in which case that code and nodes could be removed.

https://chromiumcodereview.appspot.com/9601011/diff/1/runtime/vm/flow_graph_b...
File runtime/vm/flow_graph_builder.cc (right):

https://chromiumcodereview.appspot.com/9601011/diff/1/runtime/vm/flow_graph_b...
runtime/vm/flow_graph_builder.cc:233: new InstanceCallComp(node->id(),
node->token_index(), name,
FYI: In the old compiler node->id() is used for deoptimization purposes (deopt
points identifier) and for inline cache data mapping (code -> node). The new
compiler will probably have its own id-s, i.e., we may want to skip passing and
storing the AST's node-ids. It is OK to still pass them around, so that we can
compare and generate code that has the id-s added.