egcs/gcc gcse.c toplev.c ChangeLog.RA ra.c con ...
matz@gcc.gnu.org
matz@gcc.gnu.org
Tue Aug 28 10:20:00 GMT 2001
CVSROOT: /cvs/gcc
Module name: egcs
Branch: new-regalloc-branch
Changes by: matz@gcc.gnu.org 2001-08-28 10:20:33
Modified files:
gcc : gcse.c toplev.c ChangeLog.RA ra.c
gcc/config/i386: i386.h
Log message:
Try the third time:
* gcse.c (store_killed_in_insn): Even pure/const calls kill
argument stack slots, and all things aliased to
CALL_INSN_FUNCTION_USAGE.
* toplev.c (rest_of_compilation): Preserve basic blocks in
delete_trivially_dead_insns(), initialize
rebuild_label_notes_after_reload.
* i386/i386.h (LIMIT_RELOAD_CLASS): Also limit LEGACY_REGS.
* ra.c (NO_REMAT): New macro.
(struct df *df): Staticify.
(ra_max_regno, ra_reg_renumber): New.
(struct undef_table_s): Also handle index 15.
(undef_to_bitmap): case 0xF: remove, default: only handle
lengths as powers of two.
(live_out_1): Remove the x86 hack.
(freeze): Call freeze_moves() again.
(emit_colors): Initialize ra_reg_renumber[] instead of reg_renumber[].
(init_ra): Init ra_reg_renumber.
(rematerializable): Return 0 if NO_REMAT.
(setup_renumber): New...
(reg_alloc): ... call it. Call update_equiv_regs() after allocation.
Don't calculate DU- and UD-chains if NO_REMAT.
Preserve basic blocks over delete_trivially_dead_insns().
Call allocate_reg_info() to not allocate reg_renumber.
Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/egcs/gcc/gcse.c.diff?cvsroot=gcc&only_with_tag=new-regalloc-branch&r1=1.113.2.6&r2=1.113.2.7
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/egcs/gcc/toplev.c.diff?cvsroot=gcc&only_with_tag=new-regalloc-branch&r1=1.418.2.9&r2=1.418.2.10
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/egcs/gcc/ChangeLog.RA.diff?cvsroot=gcc&only_with_tag=new-regalloc-branch&r1=1.1.2.47&r2=1.1.2.48
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/egcs/gcc/ra.c.diff?cvsroot=gcc&only_with_tag=new-regalloc-branch&r1=1.1.2.33&r2=1.1.2.34
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/egcs/gcc/config/i386/i386.h.diff?cvsroot=gcc&only_with_tag=new-regalloc-branch&r1=1.146.2.3&r2=1.146.2.4
More information about the Gcc-cvs
mailing list