[r8298]

[r8296]

 - use irg_block_walk_graph() to speed up
 - build block only at critical edges: curretn block has more than 1 predecessor
  and the predecessor is a forking op
 - ignore ALL exception edges

[r8294]

[r8292]

[r8291]

[r8288]

[r8286]

[r8285]

added callback to detect Object allocation

[r8284]

mtp_property_runtime added

[r8283]

[r8282]

[r8281]

removed sink list

[r8280]

[r8279]

[r8278]

[r8265]

[r8264]

[r8263]

fixed SymConst handling

[r8262]

[r8261]

[r8260]

speed up add_compound_ent_value()

[r8259]

[r8258]

remove some depencies from current_ir_graph

[r8257]

[r8256]

[r8255]

 - current_ir_graph must be set because of out walker use
 - ignore only Muls in x86 mode

[r8254]

 - post-dominance did not restore current_ir_graph, but set local var wrong

[r8253]

[r8252]

 - when a mode_T node is enqueued, enqueue it's Projs. This
   is needed because most optimizations run on the Proj

[r8251]

[r8249]

[r8248]

[r8245]

[r8244]

[r8242]

 - When moving Stores over Phis we must ensure that the store block
   is different from the Phi-block, otherwise the post-dominance
   did NOT mean we we expect (and a Store is moved from the end of a block
   to its start)

[r8240]

[r8239]

[r8238]

add is_entity_compiler_generated

[r8237]

[r8236]