* sh-protos.h (binary_float_operator): Remove declaration.

	* sh-protos.h (binary_float_operator): Remove declaration.
	(sh_expand_unop_v2sf, sh_expand_binop_v2sf): Declare.
	* sh.c (print_operand, case 'N'): Check against CONST0_RTX.
	(unary_float_operator, sh_expand_unop_v2sf): New functions.
	(sh_expand_binop_v2sf): Likewise.
	(zero_vec_operand): Delete.
	(SH_BLTIN_UDI): New builtin shared signature define.  Renumbered
	all non-shared ones.
	(bdesc): Change all the mextr builtins to use SH_BLTIN_UDI.
	Enable nsb and byterev.
	* sh.h (CONDITIONAL_REGISTER_USAGE): Initialize DF_HI_REGS.
	(HARD_REGNO_MODE_OK): Allow TImode in fp regs.  Allow V2SFmode
	in general regs.
	(enum reg_class, REG_CLASS_NAMES, REG_CLASS_CONTENTS): Add DF_HI_REGS.
	(SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.  Remove clause for
	immediate operands.
	(SECONDARY_INPUT_RELOAD_CLASS): Add clause for immediate operands.
	Add DF_HI_REGS.
	(CLASS_CANNOT_CHANGE_MODE, CLASS_CANNOT_CHANGE_MODE_P): Allow
	lowpart fp regs - only for big endian for now.
	(LEGITIMATE_CONSTANT_P): Don't allow non-zero float vectors
	when FPU is in use.
	(EXTRA_CONTRAINT_U): Check against CONST0_RTX.
	(LOAD_EXTEND_OP): NIL for SImode.
	(REGISTER_MOVE_COST): Add DF_HI_REGS.  Const for moves between
	general and fp registers is 4.
	PREDICATE_CODES: Amend binary_float_operator entry.
	Remove zero_vec_operand.  Add unary_float_operator.
	* sh.md (udivsi3_i4_media): Use truncate instead of paradoxical
	subreg SET_DEST.
	(truncdisi2, truncdihi2, movv2sf): Allow memory destinations.
	(truncdiqi2): Do sign extension.
	(movsi_media, movdi_media): Allow to use r63 to an fp register.
	(movdf_media, movsf_media): Likewise.
	(movv2sf_i, movv2sf_i+1): Don't use f{ld,st}.p or SUBREGS.
	Collapse to one define_insn_and_split.  Allow immediate sources.
	(addv2sf3, subv2sf3, mulv2sf3, divv2sf3): New patterns.
	(movv4sf_i): Allow immediate sources.  Use simplify_gen_subreg.
	(movv4sf): Allow immediate sources.
	(movsf_media_nofpu+1): Don't split moves to FP registers.
	(unary_sf_op, binary_sf_op, mshflo_w_x, concat_v2sf): New patterns.
	(movv8qi_i+3): Check against CONST0_RTX.
	(mextr1, mextr2. mextr3. mextr4, mextr5, mextr6, mextr7): Use DImode
	for input and output operands.  Fix argument 3 to gen_mextr_rl.
	(mmul23_wl, mmul01_wl, mmulsum_wq_i): s/const_vector/parallel/
	(msad_ubq_i, mshf4_b, mshf0_b, mshf4_l, mshf0_l, mshf4_w): Likewise.
	(mshf0_w, fipr, ftrv): Likewise.
	(mshfhi_l_di): Now insn_and_split.  Can handle FP regs.

git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@55528 138bc75d-0d04-0410-961f-82ee72b054a4