2003-06-23 H.J. Lu <hongjiu.lu@intel.com>
* gas/config/tc-i386.c (md_assemble): Support Intel Precott New
Instructions.
* gas/config/tc-i386.h (CpuPNI): New.
(CpuUnknownFlags): Add CpuPNI.
gas/testsuite/
2003-06-23 H.J. Lu <hongjiu.lu@intel.com>
* gas/i386/i386.exp: Add prescott.
* gas/i386/prescott.d: New file.
* gas/i386/prescott.s: Likewise.
include/opcode/
2003-06-23 H.J. Lu <hongjiu.lu@intel.com>
* i386.h (i386_optab): Support Intel Precott New Instructions.
opcodes/
2003-06-23 H.J. Lu <hongjiu.lu@intel.com>
* i386-dis.c (PNI_Fixup): New. Fix up "mwait" and "monitor" in
Intel Precott New Instructions.
(PREGRP27): New. Added for "addsubpd" and "addsubps".
(PREGRP28): New. Added for "haddpd" and "haddps".
(PREGRP29): New. Added for "hsubpd" and "hsubps".
(PREGRP30): New. Added for "movsldup" and "movddup".
(PREGRP31): New. Added for "movshdup" and "movhpd".
(PREGRP32): New. Added for "lddqu".
(dis386_twobyte): Use PREGRP30 to replace the "movlpX" entry.
Use PREGRP31 to replace the "movhpX" entry. Use PREGRP28 for
entry 0x7c. Use PREGRP29 for entry 0x7d. Use PREGRP27 for
entry 0xd0. Use PREGRP32 for entry 0xf0.
(twobyte_has_modrm): Updated.
(twobyte_uses_SSE_prefix): Likewise.
(grps): Use PNI_Fixup in the "sidtQ" entry.
(prefix_user_table): Add PREGRP27, PREGRP28, PREGRP29, PREGRP30,
PREGRP31 and PREGRP32.
(float_mem): Use "fisttp{l||l|}" in entry 1 in opcode 0xdb.
Use "fisttpll" in entry 1 in opcode 0xdd.
Use "fisttp" in entry 1 in opcode 0xdf.
special handling for n32 ABI.
(macro): Likewise.
* gas/mips/elf-rel-got-n32.d: Remove special handling for n32 ABI.
* gas/mips/elf-rel-xgot-n32.d: Likewise.
* gas/mips/jal-newabi.d: Likewise.
* ld-mips-elf/elf-rel-got-n32.d: Remove special handling for n32 ABI.
* ld-mips-elf/elf-rel-xgot-n32.d: Likewise.
now, not only a warning. Add some more checks to detect invalid
registers.
(get_operand): For CLASS_IR remember register size in mode struct.
(get_specific): Handle new CLASS_IRO type. Add register size
checks for CLASS_IR and CLASS_IRO.
(md_apply_fix3): Fix undefined usage of buf.
(FPU_DEFAULT, case TE_NetBSD): Default to FPU_ARCH_VFP for ELF,
FPU_ARCH_FPA for AOUT.
(md_begin): Don't try to guess the floating point architecture from
the CPU if the OS ABI (Linux, NetBSD) mandates a particular form.
(elf_mips_howto_table_rel): Use it.
(gprel32_with_gp): Move prototype.
(mips_elf_hi16_reloc): Check for ! BSF_LOCAL instead of zero addend.
Use mips_elf_generic_reloc.
(mips_elf_got16_reloc): Check for ! BSF_LOCAL instead of zero addend.
Code cleanup.
(_bfd_mips_elf32_gprel16_reloc): Check for ! BSF_LOCAL instead of
zero addend.
(mips_elf_gprel32_reloc): Likewise. Use the same GP assignment logic
as in the other *_gprel*_reloc functions.
(gprel32_with_gp): Handle partial_inplace properly.
(mips32_64bit_reloc): Use mips_elf_generic_reloc.
(mips16_gprel_reloc): Check for ! BSF_LOCAL instead of zero addend.
Do addend handling directly instead of calling
_bfd_mips_elf_gprel16_with_gp. Handle partial_inplace properly.
* elf64-mips.c (mips_elf64_hi16_reloc): Check for ! BSF_LOCAL instead
of zero addend. Handle partial_inplace properly.
(mips_elf64_got16_reloc): Check for ! BSF_LOCAL instead of zero
addend.
(mips_elf64_gprel16_reloc): Likewise.
(mips_elf64_literal_reloc): Likewise.
(mips_elf64_gprel32_reloc): Likewise. Use the same GP assignment
logic as in the other *_gprel*_reloc functions. Handle
partial_inplace properly.
(mips_elf64_shift6_reloc): Check for ! BSF_LOCAL instead of zero
addend. Handle partial_inplace properly.
(mips16_gprel_reloc): Likewise. Do addend handling directly instead
of calling _bfd_mips_elf_gprel16_with_gp.
* elfn32-mips.c (mips_elf_got16_reloc): Check for BSF_LOCAL.
(mips_elf_gprel32_reloc): Check for ! BSF_LOCAL instead
of zero addend.
(mips_elf_shift6_reloc): Handle partial_inplace properly.
(mips16_gprel_reloc): Likewise. Do addend handling directly instead
of calling _bfd_mips_elf_gprel16_with_gp.
* elfxx-mips.c (_bfd_mips_elf_gprel16_with_gp): Handle
partial_inplace properly. Fix wrong addend handling. Fix overflow
check.
(_bfd_mips_elf_sign_extend): Renamed from mips_elf_sign_extend and
exported.
(mips_elf_calculate_relocation): Use _bfd_mips_elf_sign_extend.
(_bfd_mips_elf_relocate_section): Likewise.
(mips_elf_create_dynamic_relocation): Update sec_info_type access.
* elfxx-mips.h (_bfd_mips_relax_section): Fix prototype declaration.
(_bfd_mips_elf_sign_extend): New prototype.
* config/tc-mips.c (md_pcrel_from): Return actual pcrel address.
(md_apply_fix3): Ignore non-special relocations. Remove superfluous
exceptions from size assert. Remove most of the addend fixup
specialcasing. Remove value, use valP directly. simplify fx_addnumber
handling. Remove zero addend specialcases.
(tc_gen_reloc): Use appropriate value for reloc2 addend. Remove
the addend fixup specialcase.
* config/tc-mips.h (MD_APPLY_SYM_VALUE): Define as 0.
* NEWS: Updated for the new -n option for the i386 assembler.
* config/tc-i386.c (optimize_align_code): New.
(md_shortopts): Add 'n'.
(md_parse_option): Handle 'n'.
(md_show_usage): Add '-n'.
* config/tc-i386.h (optimize_align_code): Declared.
(md_do_align): Optimize code alignment only if optimize_align_code
is not 0.
* doc/as.texinfo: Add the new -n option.
* doc/c-i386.texi: Document the new -n option.
* config/tc-h8sx.c (get_specific): Distinguish h8h from h8s ops.
(build_bytes): Ditto.
2003-06-05 Richard Sandiford <rsandifo@redhat.com>
* config/tc-h8sx.c (DMODE): Remove.
(colonmod24): Don't choose a default if the operand is a 16-bit
constant integer.
(fix_operand_size): New function.
(md_assemble): Use it to choose between @(d:2, ERn) and @(d:16,ERn).
Adjust @(d:2,ERn) operands before choosing the specific opcodes.
From Bernd Schmidt <bernds@redhat.com>
and Michael Snyder <msnyder@redhat.com>
and Alexandre Oliva <aoliva@redhat.com>
* config/tc-h8300.c: Add insns and addressing modes for h8300sx.
* config/tc-h8300.h: Ditto.
fixp's. Don't relax overflow checking for partial_inplace relocations.
Use the actual relocation type in combined relocs, not just the type
of the first one.
(macro_build_jalr): Use actual relocation size for new fix.
(s_cpsetup, s_gpdword): Likewise.
Alexandre Oliva <aoliva@redhat.com>
* elfxx-mips.c (_bfd_mips_elf_merge_private_bfd_data): Adjust
pic tests, change to warning.
(_bfd_mips_elf_final_link): Remove EF_MIPS_CPIC flag setting.
2003-05-07 Eric Christopher <echristo@redhat.com>
* config/tc-mips.c (mips_abicalls): New variable.
(md_parse_option): Use.
(s_option): Ditto.
(s_abicalls): Ditto.
(mips_elf_final_processing): Set EF_MIPS_PIC and
EF_MIPS_CPIC dependent on above.
* config/tc-i860.c (MAX_FIXUPS): Define.
(struct i860_fi fi[]): New struct.
(struct i860_it the_insn): Add above as member and move fields
exp, reloc, pcrel and fup into i860_fi.
(md_assemble): Replace all instances of exp, reloc, pcrel
and fup with fi[].exp, fi[].reloc, fi[].pcrel, fi[].fup.
Add a loop to possibly emit multiple fix-ups for each insn.
(i860_process_insn): Likewise.
(i860_get_expression): Likewise.
(md_apply_fix3): Use a bitwise check for OP_IMM_U5, not equality.
2003-05-03 H.J. Lu <hjl@gnu.org>
* config/obj-elf.c (obj_elf_parse_section_letters): Make it a
fatal error for unknown section attribute.
* config/tc-alpha.c (alpha_elf_section_letter): Return -1 for
unknown section attribute.
* config/tc-ia64.c (ia64_elf_section_letter): Likewise.
* config/tc-ppc.c (ppc_section_letter): Likewise.
* config/tc-ia64.c (ia64_elf_section_letter): Handle 'o'.
(ia64_elf_section_type): Accept "unwind".
gas/testsuite/
2003-05-03 H.J. Lu <hjl@gnu.org>
* gas/ia64/ia64.exp: Add unwind.
* gas/ia64/unwind.s: New. Test the new section attribute 'o'
and the new section type "unwind".
* gas/ia64/unwind.d: Likewise.
2003-05-01 H.J. Lu <hjl@gnu.org>
* config/tc-ia64.h (tc_canonicalize_section_name): New.
* config/obj-elf.c (obj_elf_section_name): Call
tc_canonicalize_section_name if it is defined.
gas/testsuite/
2003-05-01 H.J. Lu <hjl@gnu.org>
* gas/ia64/ia64.exp: Add secname.
* gas/ia64/secname.s: New. Test the trailing '#' in section
name.
* gas/ia64/secname.d: Likewise.
* config/tc-z8k.c: Add 2003 to copyright message.
Fold s_segm() and s_unseg() into one function s_segm(parm) which
decides by the parameter.
(md_begin): Don't set linkrelax. Only set Z8002 default if no
command line argument was given to select the intended
architecure.
(get_interrupt_operand): Warn if NOP type code is emitted.
(newfix): New parameter 'size', forward it to 'fix_new_exp'.
(apply_fix): Call newfix with additional 'size' parameter.
(build_bytes): Remove unused variable 'nib'. Detect overflow in
4 bit immediate arguments.
(md_longopts): Add 'linkrelax' option.
(md_parse_option): Adapt to new s_segm function. Set 'linkrelax'
variable when 'linkrelax' command line option is specified.
(md_show_usage): Display 'linkrelax' option.
(md_apply_fix3): Fix cases R_IMM4L, R_JR, and R_IMM8. Add cases
R_CALLR and R_REL16.
* config/tc-z8k.h: Undef WARN_SIGNED_OVERFLOW_WORD.
2003-04-30 H.J. Lu <hjl@gnu.org>
* config/tc-ia64.c (ia64_number_to_chars): New function pointer.
(ia64_float_to_chars): Likewise.
(dot_byteorder): Set target_big_endian, ia64_number_to_chars
and ia64_float_to_chars by tc_segment_info_data.endian from
the current segment if byteorder == -1.
(md_begin): Call dot_byteorder to set target_big_endian.
(md_atof): Call ia64_float_to_chars to convert floating point.
(ia64_float_to_chars_bigendian): New function.
(ia64_float_to_chars_littleendian): Likewise.
(ia64_elf_section_change_hook): Likewise.
* config/tc-ia64.h (ia64_number_to_chars): New.
(md_number_to_chars): Changed to (*ia64_number_to_chars)
(ia64_elf_section_change_hook): New.
(md_elf_section_change_hook): Defined.
(ia64_segment_info_type): New struct.
(TC_SEGMENT_INFO_TYPE): Defined.
gas/testsuite/
2003-04-30 H.J. Lu <hjl@gnu.org>
* gas/ia64/ia64.exp: Add order.
* gas/ia64/order.s: New file.
* gas/ia64/order.d: Likewise.
* config/tc-mips.c: Use signed add for n32 address arithmetic.
(append_insn): When filling delay slots with instructions
that have fixups that tc_gen_reloc might consider modifyable
in variant frags, start a new frag.
(load_address): Generate GOT_DISP with of without offset
depending on whether symbol is local. For -xgot, use
GOT_PAGE/GOT_OFST or GOT_HI16/GOT_LO16.
(macro) <M_DLA_AB, M_LA_AB>: Likewise.
<M_JAL_A>: In NewABI, use CALL16 or GOT_DISP for small got,
CALL_HI16/CALL_LO16 or GOT_PAGE/GOT_OFST for big got.
<ld_st>: In NewABI with small got, always use
GOT_PAGE/GOT_OFST, with the latter in the load/store
instruction. With big got, use GOT_HI16/GOT_LO16 or
GOT_PAGE/GOT_OFST.
(tc_gen_reloc): Adjust variant frags with GOT_DISP in NewABI.
Add tc_frag_data.tc_fr_offset to addends. Decay CALL16,
GOT_OFST and GOT_DISP to GOT_DISP in NewABI.
(md_convert_frag): Use memmove for safe copying of overlapping
regions.
constant address that Alexandre took out by accident. Reject
64-bit addresses that are not sign extensions of 32 bits only if
we don't support 64-bit address constants.
(M6811_OP_PAGE_ADDR): New internal define.
(get_operand): New modifier %page and %addr to obtain page and
address part of a far-function.
(fixup8): Use BFD_RELOC_M68HC11_PAGE for a %page modifier; don't
complain on overflow for the BFD_RELOC_M68HC11_PAGE and truncation
relocs.
(fixup16): Use BFD_RELOC_M68HC11_LO16 for a %addr modifier.
(find_opcode): Add comment.
(md_estimate_size_before_relax): Force relocation of
STATE_UNDEXED_OFFSET types when the symbol is not absolute.
(tc_m68hc11_fix_adjustable): Check for BFD_RELOC_M68HC11_LO16
instead of BFD_RELOC_LO16; temporarily make the BFD_RELOC_32
on the symbol itself so that DWARF2 strings are merged correctly.
2003-04-02 Chris Demetriou <cgd@broadcom.com>
* config/tc-mips.c (macro2): Adjust implementation of
M_ULH, M_ULHU, M_ULW, and M_ULD so that they work properly
in the case where the source and destination registers
are the same.
[ gas/testsuite/ChangeLog ]
2003-04-02 Chris Demetriou <cgd@broadcom.com>
* gas/mips/ulh.d: Adjust for ulh and ulhu macro assembly changes.
* gas/mips/mips.exp: Define new "gpr_ilocks" architecture
property, and add it to mips2 (and later) chips and r3900.
* gas/mips/uld2.s: New test source file.
* gas/mips/ulh2.s: Likewise.
* gas/mips/ulw2.s: Likewise.
* gas/mips/uld2.l: New test stderr listing.
* gas/mips/ulh2.l: Likewise.
* gas/mips/ulw2.l: Likewise.
* gas/mips/uld2-eb.d: New test.
* gas/mips/uld2-el.d: Likewise.
* gas/mips/ulh2-eb.d: Likewise.
* gas/mips/ulh2-el.d: Likewise.
* gas/mips/ulw2-eb-ilocks.d: Likewise.
* gas/mips/ulw2-eb.d: Likewise.
* gas/mips/ulw2-el-ilocks.d: Likewise.
* gas/mips/ulw2-el.d: Likewise.
* gas/mips/mips.exp: Run new tests for appropriate architectures.
(current_arch_mask): Rename to current_mode_mask.
(current_arch_requested): Remove variable.
(current_cpu): New variable.
(init_default_arch): Set defaults values for s390_arch_size,
current_mode_mask and current_cpu.
(md_parse_option): New options -mesa, -mzarch and -march={g5,g6,z900}.
(md_begin): Replace current_arch_mask by current_cpu.
(md_assemble): Adapt check and error message to current_mode_mask and
current_cpu.
gcc:
* config/sh/sh.h (EXTRA_SPECS): Add subtarget_asm_relax_spec and
subtarget_asm_isa_spec.
(SUBTARGET_ASM_RELAX_SPEC, SUBTARGET_ASM_ISA_SPEC): Define.
(ASM_SPEC): Define as SH_ASM_SPEC.
(SH_ASM_SPEC): New; take the role of ASM_SPEC, but safe from svr4.h.
Use subtarget_asm_relax_spec and subtarget_asm_isa_spec.
* config/sh/elf.h (ASM_SPEC): Use SH_ASM_SPEC.
(SUBTARGET_ASM_ISA_SPEC): Undef / define.
gcc/testsuite:
gcc.dg/sh-relax.c: New test.
include/elf:
* sh.h (EF_SH_MERGE_MACH): Make sure SH2E & SH3/SH3E merge to SH3E,
and SH2E & SH4 merge to SH4, not SH2E.
gas:
* config/tc-sh.c (sh_dsp): Replace with preset_target_arch.
(md_begin): Use preset_target_arch.
(md_longopts): Make isa option unconditional.
(md_parse_option): Make OPTION_DSP and OPTION_ISA sh4 / any
set preset_target_arch.
(md_apply_fix3): If BFD_ASSEMBLER, adjust SWITCH_TABLE fixups
by -S_GET_VALUE (fixP->fx_subsy).
(tc_gen_reloc): For SWITCH_TABLE fixups, the symbol is fixp->fx_subsy,
and the addend is 0.
Adjust addend of R_SH_IND12W relocations by fixp->fx_offset - 4.
* config/tc-sh.h (TC_FORCE_RELOCATION_SUB_LOCAL): Define.
bfd:
elf32-sh.c (sh_elf_howto_tab): Make R_SH_IND12W into an ordinary
relocation (no special function), and make it non-partial_inplace.
(sh_elf_relax_section): When creating a bsr, use a consistent value
no matter if the symbol is extern or not; set addend to -4.
Don't swap load / non-load instructions for SH4.
(sh_elf_relax_delete_bytes): In R_SH_IND12W case, check the offset
rather than if the symbol is external to determine if adjusting the
offset makes sense. Adjust the addend too if appropriate.
(sh_elf_relocate_section): In R_SH_IND12W, don't fiddle with the
relocation.
