Next step in extract_insn changes
Jeffrey A Law
law@cygnus.com
Wed Nov 4 23:04:00 GMT 1998
In message <Pine.GSO.4.02A.9810261359140.29793-100000@colt.informatik.rwth-aa
chen.de>you write:
> * recog.h (enum op_type): Define.
> (constrain_operands): Adjust prototype.
> (recog_op_type): Declare new variable.
> * recog.c (recog_op_type): New variable.
> (insn_invalid_p): Allow modifying an asm statement after reload.
> (extract_insn): Set up recog_op_type.
> (constrain_operands): Lose INSN_CODE_NUM arg. All callers changed.
> Don't compute operand types, use recog_op_type.
> Use the information computed by extract_insn instead of the previous
> method of finding it by insn code number.
> * caller-save.c (init_caller_save): Use extract_insn, not insn_extract.
> * reorg.c (fill_slots_from_thread): Likewise.
> * reload1.c (reload_as_needed): Likewise.
> (gen_reload): Likewise.
> (inc_for_reload): Likewise.
> (reload_cse_simplify_operands): Likewise.
> Use the information computed by extract_insn instead of the previous
> method of finding it by insn code number.
> * genattrtab.c (write_attr_case): Generate call to extract_insn, not
> insn_extract.
> * final.c (final_scan_insn): Use extract_insn, not insn_extract.
> Use the information computed by extract_insn instead of the previous
> method of finding it by insn code number.
> * regmove.c (find_matches): Likewise. Change meaning of the return
> value to be nonzero if the optimization can be performed, zero if
> not. All callers changed.
> Shorten some variable names to fix formatting problems.
> (regmove_optimize): Shorten some variable names to fix formatting
> problems.
> Use the information computed by extract_insn instead of the previous
> method of finding it by insn code number.
> * regclass.c (scan_one_insn): Likewise.
> (record_reg_classes): Don't compute operand types, use recog_op_type.
> * reload.c (find_reloads): Lose CONSTRAINTS1 variable; use
> recog_constraints instead.
I updated this slightly to work with the recent cleanup_operand_subreg changes
and installed the patch.
Thanks!
Jeff
More information about the Gcc-patches
mailing list