[r28127]

[r28010]

[r27970]

[r27693]

remove the (mis-)feature of arch_get_irn_reg_req_out working on mode_T nodes and retruning no_reqs in this case

[r27648]

[r27277]

[r26535]

[r26468]

  (We have a separate get_input, and get_output callback for now).
  This should make the code faster for now and is a first step towards
  changing the interface to query register constraints on the mode_T node
  itself instead of the Proj nodes.
- Handle middleend node constraints and stuff in benode.c instead of in each
  backend
- Remove irn_class_branch we already had is_cfop in the middleend
- Fix a bunch of bugs/problems in the process

[r26320]

[r26317]

  by a mostly O(1) access
- simpler memory freeing

[r23161]

[r23160]

- link resource is used until scheduler ends, honor that

[r23159]

[r23158]

[r23157]

- reserve the IRN_LINK resource

[r23156]

	Node flags and node registers are stored in a generic backend_info struct now
	instead of every part of the backend doing custom (and slow) stuff

[r23142]

[r22754]

Remove the unused parameter const arch_env_t *env from arch_irn_get_flags(), arch_irn_is() and arch_irn_consider_in_reg_alloc().

[r22706]

[r22702]

[r22563]

  compiletime options like NDEBUG (as it is public API!)
- renamed some statevs (append a prefix for their module)

[r22329]

[r22168]

[r20392]

[r19628]

[r19563]

[r19071]

[r19065]

[r17829]

[r17143]

[r15831]

[r15759]

[r15684]

[r15665]

- actually use the calculated costs for scheduling
- calculate the register use of nodes

[r15664]

Drastically lower the runtime of the strong normal form theorem scheduler by not putting nodes, which are in the schedule list already, in the list again.

[r15608]

[r15567]

[r15565]