coldfire moveb instruction to not allow an address register as
destination. Although the documentation does not indicate that
this is invalid, experiments uncovered unexpected behavior.
Added a comment explaining the situation. Thanks to Andreas
Schwab for pointing this out to me.
entries are presorted so that entries with the same mnemonic are
adjacent to each other in the table. Sort the entries for each
instruction so that this is true.
function up into several smaller ones and arranged for
the instruction printing function to be callable recursively
to print vector instructions that have both a load and a
math instruction packed into a single opcode.
* tic80-opc.c (tic80_opcodes): Expand comment for vld opcode
to explain why it comes after the other vector opcodes.
* m68k-opc.c (m68k_opcodes): add b, w, or l specifier to coldfire
move insns to handle immediate operands.
From Andreas Schwab:
* m68k-opc.c (m68k_opcodes): Delete duplicate entry for "cmpil".
New macros for building vector instruction opcodes.
(tic80_opcodes): Remove all uses of FMT_SI, FMT_REG, and
FMT_LI, which were unused. The field is now a flags field.
Remove some opcodes that are possible, but illegal, such
as long immediate instructions with doubles for immediate
values. Add "vadd" and "vld" instructions.
the order more logical. Move the shift alias instructions ("rotl",
"shl", "ins", "rotr", "extu", "exts", "srl", and "sra" to be
interspersed with the regular sr.x and sl.x instructions. Add
and test new instruction opcodes for "sl", "sli", "sr", "sri", "st",
"sub", "subu", "swcr", and "trap".
as floats.
* tic80-opc.c (SPFI): Add single precision floating point
immediate operand type.
(ROTATE): Add rotate operand type for shifts.
(ENDMASK): Add for shifts.
(n): Macro for the 'n' bit.
(i): Macro for the 'i' bit.
(PD): Macro for the 'PD' field.
(P2): Macro for the 'P2' field.
(P1): Macro for the 'P1' field.
(tic80_operands): Add entries for "exts", "extu", "fadd",
"fcmp", and "fdiv".
(print_insn_tic80): If R_SCALED then print ":s" modifier for operand.
* tic80-opc.c (REG0, REG22, REG27, SSOFF, LSOFF): Names
changed to REG_0, REG_22, REG_DEST, OFF_SS, OFF_SL respectively.
(SICR, LICR, REGM_SI, REGM_LI): Names changed to CR_SI, CR_LI,
REG_BASE_M_SI, REG_BASE_M_LI respectively.
(REG_SCALED, LSI_SCALED): New operand types.
(E): New macro for 'E' bit at bit 27.
(tic80_opcodes): Add and test dld, dld.u, dst, estop, and etrap
opcodes, including the various size flavors (b,h,w,d) for
the direct load and store instructions.
in an instruction.
* tic80-dis.c (print_insn_tic80): Change comma and paren handling.
Use M_SI and M_LI macros to check for ":m" modifier for GPR operands.
* tic80-opc.c (tic80_operands): Add REGM_SI and REGM_LI operands.
(F, M_REG, M_LI, M_SI, SZ_REG, SZ_LI, SZ_SI, D, S): New bit-twiddlers.
(MASK_LI_M, MASK_SI_M, MASK_REG_M): Remove and replace in opcode
masks with "MASK_* & ~M_*" to get the M bit reset.
(tic80_opcodes): Add bsr, bsr.a, cmnd, cmp, dcachec, and dcachef.
correctly. Add support for printing TIC80_OPERAND_BITNUM and
TIC80_OPERAND_CC, and TIC80_OPERAND_CR operands in symbolic
form.
* tic80-opc.c (tic80_operands): Add SSOFF, LSOFF, BITNUM,
CC, SICR, and LICR table entries.
(tic80_opcodes): Add and test "nop", "br", "bbo", "bbz",
"bcnd", and "brcr" opcodes.
actual fields (no shift field).
* sparc-opc.c (sparc_opcodes): Document why this cannot be "const".
* tic80-dis.c (print_insn_tic80): Replace abort stub with a
partial implementation, work in progress.
* tic80-opc.c (tic80_operands): Begin construction operands table.
(tic80_opcodes): Continue populating opcodes table and start
filling in the operand indices.
(tic80_num_opcodes): Add this.
* disassemble.c (ARCH_tic80): Define if ARCH_all is defined.
(disassembler): Add bfd_arch_tic80 support to set disassemble
to print_insn_tic80.
* tic80-dis.c (print_insn_tic80): Add stub.
* mips16-opc.c: New file.
* mips-dis.c: Include "elf-bfd.h" and "elf/mips.h".
(mips16_reg_names): New static array.
(print_insn_big_mips): Use print_insn_mips16 in 16 bit mode or
after seeing a 16 bit symbol.
(print_insn_little_mips): Likewise.
(print_insn_mips16): New static function.
(print_mips16_insn_arg): New static function.
* mips-opc.c: Add jalx instruction.
* Makefile.in (mips16-opc.o): New target.
* configure.in: Use mips16-opc.o for bfd_mips_arch.
* configure: Rebuild.
specifier of the effective-address operand in immediate forms of
arithmetic instructions. The specifier for the immediate operand
notes how and where the constant will be stored.
addresses symbolically if possible.
* mn10300-opc.c: Distinguish between absolute memory addresses,
pc-relative offsets & random immediates.
More disassembler work.
* mn10300-opc.c (mn10300_opcodes): Fix masks on several insns.
Selects opcodes & consumes bytes. Breaks badly if given data instead of
code. No operands yet.
in MN10300_OPERAND_SPLIT operands for how many bits
appear in the basic insn word. Add IMM32_HIGH24,
IMM32_HIGH24_LOWSHIFT8, IMM8E_SHIFT8.
(mn10300_opcodes): Use new operands as needed.
Support for everything in the basic instruction manual (yippie!)
operands. Update opcode table as appropriate.
(IMM32): Add MN10300_OPERAND_SPLIT flag.
(mn10300_opcodes): Fix single bit error in mov imm32,dn insn.
Cleaning up a little.
Attempting to insert most 32bit operands.
And a bug found by assembler testsuite.
operands (for indexed load/stores). Fix bitpos for DI
operand. Add SN8N_SHIFT8, IMM8_SHIFT8, and D16_SHIFT for the
few instructions that insert immediates/displacements in the
middle of the instruction. Add IMM8E for 8 bit immediate in
the extended part of an instruction.
(mn10300_operands): Use new opcodes as appropriate.
Opcode table changes so we can correctly insert everything except
32bit operands.
a data/address register that appears in register field 0
and register field 1.
(mn10300_opcodes): Use DN01 and AN01 for mov/cmp imm8,DN/AN
Hacking Matsushita again. Yippie!
* alpha-dis.c (print_insn_alpha): Use new NOPAL mask for
standard disassembly.
* alpha-opc.c (alpha_operands): Rearrange flags slot.
(alpha_opcodes): Add new BWX, CIX, and MAX instructions.
Recategorize PALcode instructions.
field for movhu instruction.
Bug found by gas testsuite.
* v850-dis.c (disassemble): For V850_OPERAND_SIGNED operands,
cast value to "long" not "signed long" to keep hpux10
compiler quiet.
Found in an attempt to build the v850 on hpux10 with the HP
compiler.
Fix various typos. Add "PAREN" operand.
(MEM, MEM2): Define.
(mn10300_opcodes): Surround all memory addresses with "PAREN"
operands. Fix several typos.
Should parse all opcodes in the instruction specification, except the
"user extension instructions".
(mn10300_operands): Rough cut. Enough to parse "mov" instructions
at this time.
(mn10300_opcodes): Break opcode format out into its own field.
Update many operand fields to deal with signed vs unsigned
issues. Fix one or two typos in the "mov" instruction
opcode, mask and/or operand fields.
Checkpointing today's work. Matsushita.
']' characters into the output stream.
* v850-opc.c (v850_opcodes: Remove size field from all opcodes.
Add "memop" field to all opcodes (for the disassembler).
Reorder opcodes so that "nop" comes before "mov" and "jr"
comes before "jarl".
Should give us a functional disassembler.
(v850_sreg_names, v850_cc_names): Likewise.
(disassemble): Very rough cut at printing operands (unformatted).
One step at a time.
* v850-opc.c (BOP_MASK): Fix.
(v850_opcodes): Fix mask for jarl and jr.
Bugs exposed by disassembler testing.
* Makefile.in Remove v850 references, they're not needed here
and they weren't being sanitized away.
* configure.in: Add v850-dis.o when building v850 toolchains.
* configure: Rebuilt.
* disassemble.c (disassembler): Call v850 disassembler.
Skeleton support for V850 disassembler.
(insert_d8_6, extract_d8_6): New functions.
(v850_operands): Rename D7S to D7; operand for D7 is unsigned.
Rename D8 to D8_7, use {insert,extract}_d8_7 routines.
Add D8_6.
(IF4A, IF4B): Use "D7" instead of "D7S".
(IF4C, IF4D): Use "D8_7" instead of "D8".
(IF4E, IF4F): New. Use "D8_6".
(v850_opcodes): Use IF4A/IF4B for sld.b/sst.b. Use IF4C/IF4D for
sld.h/sst.h. Use IF4E/IF4F for sld.w/sst.w.
So we can assemble sst/sld instructions correctly.
(v850_operands): Change D16 to D16_15, use special insert/extract
routines. New new D16 that uses the generic insert/extract code.
(IF7A, IF7B): Use D16_15.
(IF7C, IF7D): New. Use D16.
(v850_opcodes): Use IF7C and IF7D for ld.b and st.b.
Add D8 for 8-bit unsigned field in short load/store insns.
(IF4A, IF4D): These both need two registers.
(IF4C, IF4D): Define. Use 8-bit unsigned field.
(v850_opcodes): For "sld.h", "sld.w", "sst.h", "sst.w", use
IF4C & IF4D. For "trap" use I5U, not I5. Add IF1 operand
for "ldsr" and "stsr".
* v850-opc.c (v850_operands): 3-bit immediate for bit insns
is unsigned.
Fixing up the parser again.
* arm-opc.h: (arm_opcodes): Added halfword and sign-extension
memory transfer instructions. Add new format string entries %h and %s.
* arm-dis.c: (print_insn_arm): Provide decoding of the new
formats %h and %s.
(print_insn_alpha_vms): Remove.
(print_insn_alpha): Make globally visible. Chose the register
names based on info->flavour.
* disassemble.c: Always return print_insn_alpha for the alpha.
* d10v-opc.c: Changes to support signed and unsigned numbers.
All instructions with the same name that have long and short forms
now end in ".l" or ".s". Divs added.
* d10v-dis.c: Changes to support signed and unsigned numbers.
* m68k-opc.c (m68k_opcodes): Make opcode masks for the ColdFire
move ccr/sr insns more strict so that the disassembler only
selects them when the addressing mode is data register.
Mon Jul 22 11:25:24 1996 Martin M. Hunt <hunt@pizza.cygnus.com>
* d10v-opc.c (pre_defined_registers): Declare.
* d10v-dis.c (print_operand): Now uses pre_defined_registers
to pick a better name for the registers.
end-sanitize-d10v
* alpha-dis.c (print_insn_alpha): No longer the user-visible
print routine. Take new regnames and cpumask arguments.
Kill the environment variable nonsense.
(print_insn_alpha_osf): New function. Do OSF/1 style regnames.
(print_insn_alpha_vms): New function. Do VMS style regnames.
* disassemble.c (disassembler): Test bfd flavour to pick
between OSF and VMS routines. Default to OSF.
the 8086 instruction set.
* i386-dis.c: General cleanups. Make most things static. Add
prototypes. Get rid of static variables aflags and dflags. Pass
them as args (to almost everything).
to just "mode".
start-sanitize-h8s
* disassemble.c (disassembler): Handle H8/S.
* h8300-dis.c (print_insn_h8300s): New function for H8/S.
end-sanitize-h8s
Even more H8/S goo.
not "abs", which may be needed for the absolute in something
like btst #0,@10:8. Print L_3 immediates separately from other
immediates. Change ABSMOV reference to ABS8MEM.
One day we'll actually disassemble btst #0,@10:8 correctly... But not
yet. hmse.
(current_arch_mask): New static global.
(compute_arch_mask): New static function.
(print_insn_sparc): Delete sparc_v9_p. New static local
current_mach. Resort opcode table if current_mach changes.
Generalize "insn not supported" test.
(compare_opcodes): Prefer supported opcodes to nonsupported ones.
Delete test for v9/!v9.
* sparc-opc.c (MASK_*): Use SPARC_OPCODE_ARCH_MASK.
(v6notlet): Define.
(brfc): Split into CBR and FBR for coprocessor/fp branches.
(brfcx): Renamed to FBRX.
(condfc): Renamed to CONDFC. Pass v6notlet to CBR (standard
coprocessor mnemonics are not supported on the sparclet).
(condf): Renamed to CONDF.
(SLCBCC2): Delete F_ALIAS flag.
* configure: Rebuild.
* Makefile.in (SHLIB_DEP): New variable.
(LIBIBERTY_LISTS, BFD_LIST): New variables.
(stamp-piclist): Depend upon LIBIBERTY_LISTS and BFD_LIST. If
COMMON_SHLIB, add them to piclist with appropriate modifications.
($(SHLIB)): Depend upon $(SHLIB_DEP). Don't check COMMON_SHLIB
here: just use piclist.
* i386-dis.c (onebyte_has_modrm): New static array.
(twobyte_has_modrm): New static array.
(print_insn_i386): Only fetch the mod/reg/rm byte if it is needed.
(reg): Add HX instructions.
start-sanitize-i960xl
The HX instructions are the XL instructions, so this just involves
arranges for them to not be sanitized.
end-sanitize-i960xl
Alan Modra <alan@spri.levels.unisa.edu.au>:
* configure.in: Add AC_ARG_ENABLE for shared and commonbfdlib.
New substitutions: ALLLIBS, PICFLAG, SHLIB, SHLIB_CC,
SHLIB_CFLAGS, COMMON_SHLIB, SHLINK.
* configure: Rebuild.
* Makefile.in (ALLLIBS): New variable.
(PICFLAG, SHLIB, SHLIB_CC, SHLIB_CFLAGS): New variables.
(COMMON_SHLIB, SHLINK): New variables.
(.c.o): If PICFLAG is set, compile twice, once PIC, once normal.
(STAGESTUFF): Remove variable.
(all): Depend upon $(ALLLIBS) rather than $(TARGETLIB).
(stamp-piclist, piclist): New targets.
($(SHLIB), $(SHLINK)): New targets.
($(OFILES)): Depend upon stamp-picdir.
(disassemble.o): Build twice if PICFLAG is set.
(MOSTLYCLEAN): Add pic/*.o.
(clean): Remove $(SHLIB), $(SHLINK), piclist, and stamp-piclist.
(distclean): Remove pic and stamp-picdir.
(install): Install shared libraries.
(stamp-picdir): New target.
If DISASM_RAW_INSN, print insn in hex. Handle v9a as opcode
architecture.
(print_insn_sparc64): Deleted.
* disassemble.c (disassembler, case bfd_arch_sparc): Always use
print_insn_sparc.
* alpha-opc.h (alpha_insn_set): VAX floating point opcode was
incorrectly defined as 0x16 when it should be 0x15.
(FLOAT_FORMAT_MASK): function code is 11 bits, not just 7 bits!
(alpha_insn_set): added cvtst and cvttq float ops. Also added
excb (exception barrier) which is defined in the Alpha
Architecture Handbook version 2.
* alpha-dis.c (print_insn_alpha): Fixed special-case decoding for
OPERATE_FORMAT_CODE type instructions. The bug caused mulq to be
disassembled as or, for example.
* m68k-dis.c (NEXTSINGLE): Change i to unsigned int.
(NEXTDOUBLE): Likewise.
(print_insn_m68k): Don't match fmoveml if there is more than one
register in the list.
(print_insn_arg): Handle a place of '8' for a type of 'L'.
* m68k-dis.c (print_insn_m68k): Recognize all two-word instructions that take
no args by looking at the match mask.
(print_insn_arg): Always print "%" before register names.
[case 'c']: Use "nc" for the no-cache case, as recognized by gas.
[case '_']: Don't print "@#" before address.
[case 'J']: Use "%s" as format string, not register name.
[case 'B']: Treat place == 'C' like 'l' and 'L'.
* alpha-opc.h (MEMORY_FUNCTION_FORMAT_MASK): added.
(alpha_insn_set): added definitions for VAX floating point
instructions (Unix compilers don't generate these, but handcoded
assembly might still use them).
* alpha-dis.c (print_insn_alpha): added support for disassembling
the miscellaneous instructions in the Alpha instruction set.
which use '0', '1', and '2' instead. Specify the proper size for
a pmove immediate operand. Correct the pmovefd patterns to be
moves to a register, not from a register.
* m68k-dis.c (print_insn_arg): Replace 'P' with '0', '1', '2'.
Subsitute CFLAGS and AR. Call AC_PROG_INSTALL. Don't substitute
host_makefile_frag or frags.
* aclocal.m4: New file.
* configure: Rebuild.
* Makefile.in (INSTALL): Set to @INSTALL@.
(INSTALL_PROGRAM): Set to @INSTALL_PROGRAM@.
(INSTALL_DATA): Set to @INSTALL_DATA@.
(AR): Set to @AR@.
(AR_FLAGS): Set to rc rather than qc.
(CC): Define as @CC@.
(CFLAGS): Set to @CFLAGS@.
(@host_makefile_frag@): Remove.
(config.status): Remove dependency upon @frags@.
(reg_names): Likewise.
(print_insn_arg): Don't explicitly print % before register names.
Add % before register names in static array names. In case 'r',
print data registers as `@(Dn)', not `Dn@'. When printing a
memory address, don't print @# before it.
(print_indexed): Change base_disp and outer_disp from int to
bfd_vma. Print using MIT syntax, not mutant invalid Motorola
syntax. Sign extend 8 byte displacement correctly.
(print_base): Print using MIT syntax. Print zpc when appropriate.
Change parameter disp from int to bfd_vma.
F_REG_M, FPSCR_M, FPSCR_N, FPUL_M and FPUL_N.
* sh-opc.h (sh_arg_type): Add new operand types.
(sh_table): Add new opcodes from SH3E Floating Point ISA.
sh3e stuff. Sanitized out for now.
(opcode_bits, opcode_hash_table, sparc64_p): New variables.
(opcodes_initialized): Renamed from opcodes_sorted.
(build_hash_table): New function.
(is_delayed_branch): Use hash table.
(print_insn): Renamed from print_insn_sparc, made static.
Build and use hash table.
(print_insn_sparc, print_insn_sparc64): New functions.
(compare_opcodes): If !sparc64, move sparc64 opcodes to end,
and vice-versa if sparc64.
* sparc-opc.c (all non-v9 insns): Use flag F_NOTV9 instead of F_ALIAS.
* mips-opc.c (L1): Define.
(mips_opcodes): Add R4010 instructions: flushi, flushd, flushid,
addciu, madd, maddu, ffc, ffs, msub, msubu, selsi, selsr, waiti,
and wb.
Tue Jul 11 11:49:49 1995 Ian Lance Taylor <ian@cygnus.com>
* mips-opc.c (mips_opcodes): For the move pseudo-op, prefer daddu
if ISA 3 and addu otherwise, replacing or, since some MIPS chips
have multiple add units but only a single logical unit.
bfd_vma to unsigned long, because _print_insn_mips expects an unsigned long,
and that might be fewer words of argument storage (e.g., if bfd_vma is long
long on a 32-bit machine).
(print_insn_big_mips): Likewise with bfd_getb32 value.
(_print_insn_mips): Now static.
Call arc_get_opcode_mach to map bfd mach number to opcode value.
(print_insn_*): Pass bfd mach number, not opcode version.
* arc-opc.c (arc_get_opcode_mach): New function.
* alpha-opc.h (OSF_ASMCODE): define print pal-code names as defined in App C of
the Alpha Architecture Reference Manual
* alpha-dis.c: cleaned up output print stylized code forms as defined in App
A.4.3 of the Alpha Architecture Reference Manual
(print_insn_arc_{host,graphics,audio}): Likewise.
(print_insn): Add prototype.
Delete "+ 4" addition to relative branch address.
(arc_get_disassembler): New arg `big_p'. Return little or big
print fn accordingly.
* arc-opc.c (arc_opcode_init_tables): Init arc_operand_map once.
(arc_opcode_supported): Use ARC_OPCODE_CPU to ignore byte order.
(arc_opval_supported): Likewise.
* disassemble.c (disassembler): Pass big endian flag to
arc_get_disassembler.
<edelsohn@npac.syr.edu>.
(powerpc_operands): Add operands SPRBAT and SPRG. Split TBR out of
SPR.
(FXM_MASK): Define.
(insert_tbr): New static function.
(extract_tbr): New static function.
(XFXFXM_MASK, XFXM): Define.
(XSPRBAT_MASK, XSPRG_MASK): Define.
(powerpc_opcodes): Add instructions to access special registers by
name. Add mtcr and mftbu.
All uses changed.
(extraction fns): Insn argument now array of two words. Return pointer
to value's table entry. All uses changed.
(arc_opcode_lookup_suffix): Exported for arc-dis.c.
(insert_multshift, extract_multshift): New fns.
(arc_operands): Add support for cache bypass suffix. Add support for
predefined aux regs. Modifier bits moved to flags field.
(arc_opcodes): Likewise.
Add mul/mulu/shift insns. Syntax of zero/sign extension insns changed.
New insn rlc. Update to syntax in programmer's manual.
(arc_reg_names): Fix typo in lp_count. Add predefined aux regs.
(arc_suffixes): New synonyms lo,hs for cs,cc. New suffix for cache
bypass.
(arc_opcode_init_tables): New argument to indicate cpu type.
(insert_reg): Handle predefined aux regs.
(extract_reg): Likewise.
(lookup_register): New fn.
* arc-dis.c (arc_condition_codes): Deleted.
(print_insn_arc): Handle insns with 32 bit immediate constants better.
Clean up modifier handling. Handle predefined aux regs.
