This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Basic block renumbering removal, part 2


Hello.

> > 	* bb-reorder.c (make_reorder_chain_1): Modified.
> > 	* cfganal.c (can_fallthru, flow_call_edges_add,
> > 	flow_preorder_transversal_compute): Modified.
> > 	* cfgbuild.c (make_edges, find_basic_blocks, find_many_sub_basic_blocks,
> > 	find_sub_basic_blocks): Modified.
> > 	* cfgcleanup.c (try_simplify_condjump, try_optimize_cfg): Modified.
> > 	* cfglayout.c (skip_insns_after_block, fixup_reorder_chain,
> > 	fixup_fallthru_exit_predecessor, cfg_layout_redirect_edge): Modified.
> > 	* cfgrtl.c (tidy_fallthru_edges, verify_flow_info): Modified.
> > 	* combine.c (this_basic_block): Type changed to basic_block.
> > 	(combine_instructions, set_nonzero_bits_and_sign_copies, try_combine,
> > 	nonzero_bits, num_sign_bit_copies, get_last_value_validate,
> > 	get_last_value, distribute_notes, distribute_links): Modified.
> > 	* final.c (compute_alignments): Modified.
> > 	* flow.c (regno_uninitialized, regno_clobbered_at_setjmp): Modified.
> > 	* function.c (thread_prologue_and_epilogue_insns): Modified.
> > 	* gcse.c (compute_code_hoist_vbeinout): Modified.
> > 	* global.c (build_insn_chain): Modified.
> > 	* ifcvt.c (find_if_block, find_cond_trap): Modified.
> > 	* predict.c (last_basic_block_p, note_prediction_to_br_prob): Modified.
> > 	* regmove.c (regmove_optimize): Modified.
> > 	* resource.c (find_basic_block): Modified.
> > 	* sched-ebb.c (schedule_ebbs): Modified.
> > 	* ssa-dce.c (find_control_dependence, find_pdom): Modified.
> 
> Ok except
> 
> > *************** compute_code_hoist_vbeinout ()
> > - 	  if (bb != n_basic_blocks - 1)
> > + 	  if (BASIC_BLOCK (bb)->next_bb != EXIT_BLOCK_PTR)
> 
> Surely this should be
> 
> 	bb != EXIT_BLOCK_PTR->prev_bb->index
> 
> as we want to get rid of BASIC_BLOCK (N) references?

Would work too; but in fact I want to get rid of relying that (e.g.) last basic
block has number n_basic_blocks - 1, which both satisfy. It does not really
matter -- this will be changed in following (large :-( ) patch to remove
iterations over indices.

Zdenek


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]