1 2004-08-16 Ben Elliston <bje@au.ibm.com>
3 * basic-block.h: Include "errors.h".
4 (ENABLE_CHECKING): Define to 0 if not already defined.
5 (FOR_EACH_EDGE): Reimplement.
6 (END_FOR_EACH_EDGE): New.
7 * Update all callers: Eliminate explicit iterator variables, use
8 FOR_EACH_EDGE and END_FOR_EACH_EDGE macros.
10 2004-08-12 Ben Elliston <bje@au.ibm.com>
12 * basic-block.h (FOR_EACH_PRED_EDGE, FOR_EACH_SUCC_EDGE): Remove.
13 * Update all callers to use FOR_EACH_EDGE instead.
15 2004-08-12 Ben Elliston <bje@au.ibm.com>
17 * basic-block.h (EDGE_PRED_COUNT, EDGE_SUCC_COUNT): Remove.
18 * Update all callers to use EDGE_COUNT instead.
20 2004-07-31 Steven Bosscher <stevenb@suse.de>
22 * cfg.c (remove_edge): Use VEC_unordered_remove.
23 (redirect_edge_succ): Likewise.
24 (redirect_edge_pred): Likewise.
25 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
27 * dominance.c (calc_idoms): Initialize ix to -1 when computing
28 post dominance and the block has fake exit edges.
30 * cfgcleanup.c (try_forward_edges): Revert previous commit.
32 * version.c: Add edge-vector-branch special version and merge tag.
34 2004-07-30 Steven Bosscher <stevenb@suse.de>
36 * cfgexpand.c (expand_block): Decrement the iterator if
37 an edge is removed so that we re-examine the current vector slot.
38 * tree-ssa-dce.c (remove_dead_stmt): Do not remove the first one but
41 2004-07-30 Andrew Pinski <apinski@apple.com>
43 * cfgcleanup.c (try_forward_edges): Go back to the first edge after
45 * tree-cfg.c (disband_implicit_edges): Remove unused ix.
46 (thread_jumps): Decrement the iterator if an edge is removed
47 so that we re-examine the current vector slot.
49 2004-07-29 Ben Elliston <bje@au.ibm.com>
51 * basic-block.h (FALLTHRU_EDGE): Use EDGE_SUCC.
52 (BRANCH_EDGE): Likewise.
53 (EDGE_CRITICAL_P): Use EDGE_{PRED,SUCC}_COUNT.
54 (FOR_EACH_PRED_EDGE, FOR_EACH_SUCC_EDGE): New.
55 (EDGE_PRED_COUNT, EDGE_SUCC_COUNT): Likewise.
56 (EDGE_PRED, EDGE_SUCC): Likewise.
57 * bb-reorder.c, bt-load.c, cfg.c, cfganal.c, cfgbuild.c,
58 cfgcleanup.c, cfgexpand.c, cfghooks.c, cfglayout.c, cfgloop.c,
59 cfgloopanal.c, cfgloopmanip.c, cfgrtl.c, cse.c, df.c, dominance.c,
60 domwalk.c, except.c, final.c, flow.c, function.c, gcse.c,
61 gdbinit.in, global.c, graph.c, ifcvt.c, lcm.c, loop-doloop.c,
62 loop-init.c, loop-invariant.c, loop-iv.c, loop-unroll.c,
63 loop-unswitch.c, modulo-sched.c, predict.c, profile.c, ra-build.c,
64 ra-rewrite.c, ra.c, recog.c, reg-stack.c, regrename.c, reload1.c,
65 sbitmap.c, sched-ebb.c, sched-rgn.c, tracer.c, tree-cfg.c,
66 tree-into-ssa.c, tree-mudflap.c, tree-outof-ssa.c,
67 tree-pretty-print.c, tree-sra.c, tree-ssa-ccp.c, tree-ssa-dce.c,
68 tree-ssa-dom.c, tree-ssa-live.c, tree-ssa-loop-ch.c,
69 tree-ssa-loop-im.c, tree-ssa-loop-niter.c, tree-ssa-phiopt.c,
70 tree-ssa-pre.c, tree-ssa.c, tree-tailcall.c, var-tracking.c,
71 config/frv/frv.c, config/i386/i386.c, config/ia64/ia64.c,
72 config/rs6000/rs6000.c: Update all callers.
74 2004-07-26 Ben Elliston <bje@au.ibm.com>
76 * bb-reorder.c (connect_traces): Use distinct iterator variables
77 when nesting FOR_EACH_EDGE loops (ix vs. ix2).
79 2004-07-26 Andrew Pinski <pinskia@physics.uc.edu>
81 * bb-reorder.c (fix_up_fall_thru_edges): Fix up when there is no edge
82 and when only one exists.
83 (fix_crossing_conditional_branches): Likewise.
85 * cfgcleanup.c (outgoing_edges_match): Speedup loop and use FOR_EACH_EDGE.
86 (try_crossjump_bb): Speed up by checking the EDGE_COUNT before loop and
87 remove the check in the loop.
88 * cfghooks.c (delete_basic_block): Do not use FOR_EACH_EDGE when remove
89 all edges but a loop of EDGE_COUNT.
90 (merge_blocks): Likewise.
91 (duplicate_block): Use EDGE_COUNT.
92 * cfgrtl.c (purge_dead_edges): Adjust the iterator
93 when removing an edge.
94 * dominance.c (calc_idoms): Do not initialize ix as
96 * tree-ssa-dce.c (remove_dead_phis): Do not use FOR_EACH_EDGE when remove
97 all edges but a loop of EDGE_COUNT.
98 * tree-ssa-loop-ch.c (duplicate_blocks): Make sure we found the other edge.
100 2004-07-26 Steven Bosscher <stevenb@suse.de>
102 * cfgcleanup.c (outgoing_edges_match): Test for exactly two edges,
104 * tree-cfg.c (cleanup_control_expr_graph): Adjust the iterator
105 when removing an edge.
106 (tree_purge_dead_eh_edges): Likewise.
108 2004-07-22 Andrew Pinski <pinskia@physics.uc.edu>
110 * cfganal.c (add_noreturn_fake_exit_edges): Use EDGE_COUNT.
111 * cfgcleanup.c (try_optimize_cfg): Likewise.
112 * cfglayout.c (fixup_reorder_chain): Likewise.
113 * ra-build.c (live_in): Likewise.
114 * sched-rgn.c (build_control_flow): Likewise.
115 * tree-ssa-pre.c (compute_antic_aux): Likewise.
117 * cfgloop.c (flow_loop_pre_header_scan): Use EDGE_0 of the pred.
118 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Initialize found.
119 * dominance.c (calc_idoms): Initialize ix.
120 * ra-build.c (live_in): Initialize e.
121 * tree-mudflap.c (mf_build_check_statement_for): Use EDGE_0 of the pred.
122 * tree-ssa-loop-ch.c (duplicate_blocks): Move defintion of ix up.
124 2004-07-22 Ben Elliston <bje@au.ibm.com>
126 * cfgcleanup.c (outgoing_edges_match): Clean up search of outgoing
127 edges. Don't test e1 and e2 for non-null values--check the edge
129 * tree-ssa-dce.c (remove_dead_stmt): Remove erroneous VEC_pop.
131 2004-07-22 Steven Bosscher <stevenb@suse.de>
133 * ifcvt.c (find_if_case_1): Use then_bb->pred, not succ. Typo.
135 2004-07-22 Ben Elliston <bje@au.ibm.com>
137 * cfg.c (redirect_edge_succ): Use FOR_EACH_EDGE for consistency.
138 (redirect_edge_pred): Likewise.
139 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
140 * tree-ssa-loop-ch.c (duplicate_blocks): Likewise.
142 2004-07-22 Steven Bosscher <stevenb@suse.de>
144 * var-tracking.c (vt_stack_adjustments): Fix stack push thinko.
146 2004-07-22 Andrew Pinski <apinski@apple.com>
148 * basic-block.h (FOR_EACH_EDGE): Set e at the begining too.
149 * gcse.c (bypass_block): e2 is the iterator for the edges.
150 * ifcvt.c (find_cond_trap): Use EDGE_COUNT.
151 (block_has_only_trap): Likewise.
152 (find_if_case_2): Set else_edge and not else_succ.
153 Use else_succ when we used to.
154 * sched-rgn.c (find_rgns): Use EDGE_COUNT.
155 * tree-cfg.c (make_ctrl_stmt_edges): Likewise.
156 (make_goto_expr_edges): Likewise.
157 (thread_jumps): Likewise.
158 (execute_warn_function_return): Likewise.
159 * tree-ssa-dom.c (redirect_edges_and_update_ssa_graph):
160 Check the number of edges of e before setting e again.
162 2004-07-22 Ben Elliston <bje@au.ibm.com>
164 * ifcvt.c (find_if_block): Check else_succ edge count.
165 * gcse.c (remove_reachable_equiv_notes): Assign NULL to `act' if
166 the iterator overruns the end of the vector.
168 2004-07-22 Ben Elliston <bje@au.ibm.com>
170 * except.c (connect_post_landing_pads): Use EDGE_COUNT.
171 * ifcvt.c (find_if_case_2): Only access edge 0 if the edge count
173 * cfghooks.c (delete_basic_block): Truncate unconditionally as it
174 is now safe to do so.
176 2004-07-21 Ben Elliston <bje@au.ibm.com>
178 * cfganal.c (remove_fake_predecessors): Decrement the iterator if
179 an edge is removed so that we re-examine the current vector slot.
180 * except.c (finish_eh_generation): Likewise.
181 * cfgexpand.c (expand_block): Likewise.
182 * cfghooks.c (delete_basic_block): Likewise.
183 (merge_blocks): Likewise.
184 * profile.c (compute_branch_probabilities): Use EDGE_COUNT, not
185 a FOR_EACH_EDGE loop with an increment in the loop body.
186 * cfgloopmanip.c (fix_irreducible_loops): Use EDGE_COUNT.
187 * bb-reorder.c (copy_bb_p): Directly test if edge count exceeds 8.
188 * ifcvt.c (merge_if_block): Use EDGE_COUNT for clarity.
190 2004-07-21 Ben Elliston <bje@au.ibm.com>
192 * cfg.c (clear_edges): Truncate the bb->succ, bb->pred vectors.
193 Likewise for EXIT_BLOCK_PTR->succ and EXIT_BLOCK_PTR->pred.
195 2004-07-20 Ben Elliston <bje@au.ibm.com>
197 * config/ia64/ia64.c (ia64_expand_prologue): Iterate over edge
198 vector instead of traversing a linked list.
199 * config/rs6000/rs6000.c (rs6000_emit_prologue): Use EDGE_COUNT to
200 count elements rather than testing the basic_block->pred member.
201 * config/frv/frv.c (frv_ifcvt_modify_tests): Use edge accessor.
203 2004-07-20 Ben Elliston <bje@au.ibm.com>
205 * Ben to add a large ChangeLog entry soon!