This optimization is unsafe and breaks Load-Store-optimization.

It is not enough that both loads are connected by the memory edge,
if one Load has an exception-edge, they cannot be parallelized, because
this would destroy exception order.
load-store-opt handle this case and many others, but expect all
Loads in order yet.
opt_parallelize_mem() does this optimization in a wider range.

[r28385]

All leaders of a partition must have the same opcode, which is the pair
of <irn_opcode, ir_mode> in Firm.

[r28384]

[r28383]

[r28382]

[r28381]

[r28379]

[r28378]

Was used by the (removed) ILP scheduler. Dead code now.

[r28377]

[r28375]

[r28374]

[r28373]

[r28372]

[r28371]

[r28369]

There seems to be no "open" methods at all (at least now, this concept seems
to be dead).

[r28368]

[r28367]

Check for mode int on several places where Div nodes are handled:
- fixed arch_dep_replace_div_by_const() (works for int only)
- fixed constant folding for Div nodes (different for float and int)
- fixed transform_node_Div() (some code pathes allowed only for int)
- fixed comment and one missing check in Div mode verifycation

[r28366]

[r28365]

[r28364]

[r28361]

[r28346]

[r28345]

[r28344]

[r28343]

[r28342]

[r28341]

[r28340]

[r28339]

[r28338]

[r28337]

[r28336]

[r28334]

Add a hint about the infamous pn_Cmp_Lg/Ne mixup in the assertion message of verify_node_Proj_Cmp().

[r28333]

[r28332]

[r28331]

[r28330]

[r28329]

Additionally
- switched the type of a visitor counter to ir_visited_t
- add some doxygen docu

[r28328]

Additionally:
- rename some variables to reflect its content
- add some doxygen comments

[r28327]

[r28310]