(init_opcode_tables): Delete cpu_type_map, call
arc_get_opcode_mach instead.
(arc_common): Delete non-ELF stuff. Require leading '.' in .data.
(md_apply_fix): Delete non-ELF stuff.
Disable special handling of limm values, treat them normally.
(get_arc_exp_reloc_type): New argument `data_p'. Callers updated.
(md_assemble): All limm operands must be output by gas, the opcode
insertion routines cannot be used (they don't do anything and we're
using implicit addends now so we need to output something).
* config/tc-arc.h (FAKE_LABEL_NAME): Define.
(comment_chars): Add ';'.
(arc_mach_type, mach_type_specified, cpu_tables_init_p): New globals.
(md_parse_option): Delete support for -mmult. Add -mcpu=xxx.
(md_begin): Current ARCs are little endian.
Call bfd_set_arch_mach to set the cpu type.
(init_opcode_tables): New function.
(md_begin): Ignore suffixes and registers not supported by cpu.
(md_assemble): Initialize opcode tables here.
Ignore opcodes not supported by selected cpu.
Always ask for more memory in one piece.
(arc_cpu): New function.
(md_numbers_to_chars): Support both endians (will probably be needed
eventually anyway).
(md_apply_fix): Likewise.
to table entry. All callers changed.
(md_shortopts): New argument -mmult.
(md_parse_option): Recognize it.
(md_show_usage): Print it.
(md_begin): If -mmult, say we have the new multiply and shift insns.
(md_assemble): Fix reloc generation for 32 bit immediate values.
Clean up modifier handling. Try next insn if fake operand says there's
an error. Update to new syntax of insertion/extraction fns.
Recognize predefined aux regs. Don't abort if insertion fn missing.
