[PATCH resubmit] RTL loop versioning (SMS improvements 1/2)

[Mostafa Hagog]

This patch already passed regression and bootstrap on i686-pc-linux-gnu and
powerpc-apple-darwin, this is a resubmit after addressing comments by
Roger. Is this ok for mainline ?
I have rerun the bootstrap and it passed on both platforms and now I am
rerunning the regression is this necessary before commit ?

2005-03-30 Mostafa Hagog <mustafa@il.ibm.com>

      * cfghooks.c (lv_flush_pending_stmts,
      cfg_hook_duplicate_loop_to_header_edge, extract_cond_bb_edges,
      lv_adjust_loop_header_phi, lv_add_condition_to_bb): New.
      * cfghooks.h (cfg_hook_duplicate_loop_to_header_edge,
      lv_add_condition_to_bb,
      lv_adjust_loop_header_phi, extract_cond_bb_edges,
      flush_pending_stmts): New in cfg_hooks structure.
      (cfg_hook_duplicate_loop_to_header_edge, lv_flush_pending_stmts,
      extract_cond_bb_edges, lv_adjust_loop_header_phi,
      lv_add_condition_to_bb): New declarations.
      * cfgloop.h (duplicate_loop_to_header_edge): Change return type to
      bool.
      (loop_version): Declare.
      * cfgloopmanip.c (cfghooks.h): Include.
      (duplicate_loop_to_header_edge): Change return type to bool.
      (loop_version, lv_adjust_loop_entry_edge): Move here.
      * cfgrtl.c (cfgloop.h): Include.
      (rtl_verify_flow_info_1): Fix.
      (rtl_lv_add_condition_to_bb, rtl_extract_cond_bb_edges): New.
      (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add hooks to
      initialization.
      * tree-cfg.c (tree_lv_adjust_loop_header_phi,
      tree_lv_add_condition_to_bb): New.
      (tree_cfg_hooks): Add new hooks to initialization.
      * tree-ssa-loop-manip.c (lv_adjust_loop_header_phi,
      lv_adjust_loop_entry_edge, tree_ssa_loop_version): Remove.

Thanks,
Mostafa.


(See attached file: loop_version_hook_9.patch)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: loop_version_hook_9.patch
Type: application/octet-stream
Size: 28950 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20050330/b95dfd47/attachment.obj>