Add some new control instructions to the opcodes library, and a new test
for these new instructions to the assembler. The new instructions use
an instruction flag longer than any seen before (on arc), and so the max
flag length is extended to accommodate this.
gas/ChangeLog:
* config/tc-arc.h (MAX_FLAG_NAME_LENGTH): Increase to 7.
* testsuite/gas/arc/nps400-2.d: New file.
* testsuite/gas/arc/nps400-2.s: New file.
opcodes/ChangeLog:
* arc-nps400-tbl.h: Add schd, sync, and hwschd instructions.
* arc-opc.c (arc_flag_operands): Add new flags.
(arc_flag_classes): Add new classes.
This commit completes support for having multiple instructions with the
same mnemonic in non-contiguous blocks within the arc_opcodes table.
The commit adds an iterator mechanism for the arc_opcode_hash_entry
structure, which is then used in find_opcode_match to consider all
arc_opcode entries with the same mnemonic, even when these instructions
are stored in non-contiguous blocks.
I extend the comment on the arc_opcodes table to discuss how entries
within the table are organised, and to mention how instructions can be
split into multiple groups if needed, but that the table is still
searched in table order.
There should be no user visible changes after this commit.
gas/ChangeLog:
* config/tc-arc.c (struct arc_opcode_hash_entry_iterator): New
structure.
(arc_opcode_hash_entry_iterator_init): New function.
(arc_opcode_hash_entry_iterator_next): New function.
(find_opcode_match): Iterate over all arc_opcode entries
referenced by the arc_opcode_hash_entry passed in as a parameter.
opcodes/ChangeLog:
* arc-opc.c (arc_opcodes): Extend comment to discus table layout.
Building on earlier commits, this commit moves along support for having
multiple arc_opcode entries in the arc_opcodes table that have the same
mnemonic (name) field, but are not stored in a contiguous block in the
table.
In this commit we support looking up the arc_opcode_hash_entry from the
hash table, and passing this along to the find_opcode_match function,
which then finds the specific arc_opcode that we're assembling. We
still don't actually support the multiple chains of arc_opcode entries
in this commit, but the limitation is now isolated to the
find_opcode_match function.
There is no user visible change after this commit.
gas/ChangeLog:
* config/tc-arc.c (arc_find_opcode): Now returns
arc_opcode_hash_entry pointer.
(find_opcode_match): Update argument type, extract arc_opcode from
incoming arc_opcode_hash_entry.
(find_special_case_pseudo): Update return type.
(find_special_case_flag): Update return type.
(find_special_case): Update return type.
(assemble_tokens): Lookup arc_opcode_hash_entry based on
instruction mnemonic, then use find_opcode_match to identify
specific arc_opcode.
The arc assembler builds a hash table to hold references to arc_opcode
entries in the arc_opcodes table. This hash assumes that each mnemonic
will always appear in a contiguous blocks within the arc_opcodes table;
all ADD instruction will be together, all AND instructions will likewise
be together and so on.
The problem with this is that as different variations of arc are added,
then it is often more convenient to split instructions apart, so all the
base ADD instructions are together, but, variants of ADD specific to one
variation of arc are grouped with other instructions specific to that
arc variant. The current data structures don't support splitting the
instructions in this way.
This commit is a first step towards addressing this limitation. In this
commit the hash table that currently holds arc_opcode pointers directly,
instead holds a pointer to a new, intermediate, data structure. This
new data structure holds the pointer to the arc_opcode. In this way, we
can, in the future support having the intermediate structure hold
multiple pointers to different arc_opcode groups.
There should be no visible functional change after this commit.
gas/ChangeLog:
* config/tc-arc.c (struct arc_opcode_hash_entry): New structure.
(arc_find_opcode): New function.
(find_special_case_pseudo): Use arc_find_opcode.
(find_special_case_flag): Likewise.
(assemble_tokens): Likewise.
(md_begin): Build hash using struct arc_opcode_hash_entry.
An upcoming commit will add a new arc instruction flag that uses
characters that have never appeared in an arc instruction flag before.
Currently the assembler is very conservative about which characters can
or cannot appear in an instruction flag.
This commit relaxes these constraints a little. After this commit all
alpha-numeric characters are now allowed into instruction flags. This
complete set is not required for the upcoming change, however, having
this slightly larger set does not impact the assemblers ability to
correctly parse input, but does make it easier to add new flag to the
instruction table.
gas/ChangeLog:
* config/tc-arc.c (tokenize_flags): Allow greater range of
characters into flag names.
The preprocess_operands function changes the incoming list of assembler
tokens based on the assumption that the first arc_operand found will be
the same instruction class as all of the arc_operands for the same
mnemonic.
Though this assumption is probably fine, removing this assumption, and
pushing the token change down into assemble_tokens makes the code no
more complex, and might even be easier to follow.
There should be no user visible changes after this commit.
gas/ChangeLog:
* config/tc-arc.c (find_opcode_match): Handle O_symbol case, add
new de_fault label.
(preprocess_operands): Delete.
(assemble_tokens): Remove call to preprocess_operands.
PR gas/19910
* config/tc-sparc.c (sparc_ip): Report an error if the expression
inside a %-macro could not be fully parsed.
* expr.c (integer_constant): Accept and ignore U suffixes to
integers.
(operand): When a missing closing parenthesis is encountered,
report the character that was found instead.
* testsuite/gas/mips/tls-ill.l: Update expected error message.
* testsuite/gas/sparc/pr19910-1.d: New test driver.
* testsuite/gas/sparc/pr19910-1.s: New test.
* testsuite/gas/sparc/pr19910-2.l: Expected error output.
* testsuite/gas/sparc/pr19910-2.s: New test.
* testsuite/gas/sparc/sparc.exp: Run the new tests.
* config/tc-msp430.c (msp430_operands): Check for a NOP preceding
an EINT instruction. Warn/fix as necessary.
* testsuite/gas/msp430/bad.s: Add test of EINT without preceding NOP.
* testsuite/gas/msp430/bad.l: Update expected messages.
In tc-arc.h we redefine constants that are also defined in the
include/opcode/arc.h header file. The problem is that changing one
without changing the other resulting in strange crashes.
We could comment both locations to stress the importance of remembering
to update the other location too, or we could just include the
opcode/arc.h header file into the assembler.
Given that other targets include their opcode/ARCH.h header file, and
only having a single definition, where possible, is usually the safest
solution, that's what I switch too with this commit.
gas/ChangeLog:
* config/tc-arc.h: Include 'opcode/arc.h'.
(MAX_INSN_ARGS): Delete.
(MAX_INSN_FLGS): Delete.
Providing declarations in tc.h points out that alpha wasn't properly marking
FLT_CHARS as const. We can also get rid of the confusing redefinition of
X_CHARS to mmix_x_chars. Finally we can get rid of some now redundant
declarations of these constants.
gas/ChangeLog:
2016-04-03 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-alpha.c: Const qualify FLT_CHARS.
* config/atof-ieee.c: Remove declarations of FLT_CHARS and EXP_CHARS.
* config/tc-cris.h: Likewise.
* expr.c: Likewise.
* config/tc-mmix.c (md_atof): Adjust comment.
* config/tc-mmix.h: Stop defining FLT_CHARS and EXP_CHARS as macros.
* tc.h: Declare FLT_CHARS and EXP_CHARS.
It needs a few more things to be const now.
gas/ChangeLog:
2016-04-03 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-score.c (s3_gen_reloc): Add const qualifiers.
* config/tc-score7.c (s7_gen_reloc): Likewise.
It is only ever assigned values in the enum, and it is passed to functions that
expect the argument's type to be the enum.
gas/ChangeLog:
2016-04-03 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-arm.c (do_t_branch): Change the type of reloc to
bfd_reloc_code_real_type.
Its not used for anything outside of md_assemble () so it doesn't need to be
extern. While we are there we can replace free () and xmalloc () with
XRESIZEVEC which should be faster.
gas/ChangeLog:
2016-04-03 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/bfin-parse.y (current_inputline): Remove definition.
* config/tc-bfin.c (md_assemble): Simplify use of current_inputline.
Instead of canonicalizing the argument we can just use a case insensitive
compare.
gas/ChangeLog:
2016-04-03 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-avr.c (md_parse_option): Use strcasecmp () to compare
strings.
Small spelling mistake in a #define, fixed in this commit.
gas/ChangeLog:
* config/tc-arc.h (MAX_FLAG_NAME_LENGHT): Rename to...
(MAX_FLAG_NAME_LENGTH): ...this.
(struct arc_flags): Update to use MAX_FLAG_NAME_LENGTH.
* config/tc-arc.c (tokenize_flags): Likewise.
Xtensa uses it in several macros passing it a literal string, so its convenient
for the argument type to be const char *. However some of the code in
symbols.c seems to assume tc_canonicalize_symbol_name () will return a non
const pointer, and some other target's implementations even modify the
argument, so it seems best to return a char * which means casting away const on
the argument when we return it instead of another string.
gas/ChangeLog:
2016-03-31 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-xtensa.c (struct rename_section_struct): Make old_name
const.
(xtensa_section_rename): Make argument type const char *.
* config/tc-xtensa.h (xtensa_section_rename): Adjust.
parse_ldconst () takes a char ** as a in / out argument, and sometimes points
args[0] to a constant string. Then in some cases after parse_ldconst ()
returns md_assemble () twiddles the contents of arg[0]. So it seems like
it would take some work to avoid these casts, and its not really clear
that work is worth it.
gas/ChangeLog:
2016-03-31 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-i960.c (parse_ldconst): Cast to char * when assigning to
args[0].
For these targets its not clear how md_assemble can usefully be split up so
that part can take const char *. There is also a fair number of targets that
need md_assemble () to take a char *, so we can't easily make the argument
const. So since there isn't many callers it seems easiest to just add a couple
of casts.
gas/ChangeLog:
2016-03-31 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-m32c.c (m32c_md_end): cast the argument to md_assemble to
char *.
(m32c_indirect_operand): Likewise.
* config/tc-nds32.c (do_pseudo_b): Likewise.
(do_pseudo_bal): Likewise.
(do_pseudo_ls_bhw): Likewise.
get_symbol_name () returns a char * in a out arg, which means we need to cast
to assign a literal to the variable passed to get_symbol_name (). It seems
like better APIs than get_symbol_name () could be provided, but that seems like
a fair amount of work so just casting seems to be the betterthing to do for
now.
gas/ChangeLog:
2016-03-31 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-ia64.c (md_assemble): Add temporary variable to pass to
get_symbol_name ().
* config/tc-sparc.c (s_register): Cast a literal to char * in
assignment.
various places either directly or indirectly set input_line_pointer to point at
a literal. Currently lots of places modify the string input_line_pointer
points at, so making it const isn't easy. Since most if not all of these
places assign to input_line_pointer to parse an expression it would probably be
best to add ways to generate and deal with expressions that doesn't involve
parsing strings, but for now adding some casts seems easiest.
gas/ChangeLog:
2016-03-31 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-i960.c (parse_expr): Cast to char * when assigning to
input_line_pointer.
* config/tc-m32r.c (expand_debug_syms): Likewise.
* config/tc-msp430.c (msp430_dstoperand): Likewise.
* config/tc-z80.c (md_begin): Likewise.
* stabs.c (stabs_generate_asm_func): Likewise.
these places define char arrays containing values greater than 0x80 which
doesn't fit in an 8 bit signed char, but does fit in an unsigned one.
gas/ChangeLog:
2016-03-31 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-aarch64.c (aarch64_handle_align): Make the type of some
variables unsigned char[].
* config/tc-alpha.c (alpha_handle_align): Likewise.
* config/tc-arm.c (arm_handle_align): Likewise.
* config/tc-z80.c: Likewise.
This name describes what the variable means slightly better, and the enum with
two values that is only used for this one variable is kind of silly.
gas/ChangeLog:
2016-03-25 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* as.c (parse_args): Adjust.
* as.h (flag_size_check): Rename to flag_allow_nonconst_size.
* config/obj-elf.c (elf_frob_symbol): Adjust.
In the SPARC V9 (and later) versions of the SPARC specification, the
section C.1.1 "Register Names" specifies that:
"asr_reg. An asr_reg is an Ancillary State Register name. It may have
one of the following values:
%asr16-%asr31"
The rationale of having this restriction was that the registers from 16
to 31 are reserved to implementations, and are therefore "non-V9". It
also assumes that the existing ASR registers in the range 0..31 will
have their own names such as %y, that can be used to access such
registers.
However, this is problematic. When a new ASR register is introduced,
such as %mcdper a.k.a. %asr14, it is useful to be able to use %asr14 in
order to not depend on the latest version of the assembler.
The Solaris assembler is lax and allows to assembly instructions
referring to %asr0 to %asr31. This patch makes the GNU assembler to
mimic that behavior.
gas/ChangeLog:
2016-03-24 Jose E. Marchesi <jose.marchesi@oracle.com>
* config/tc-sparc.c (sparc_ip): Remove the V9 restriction on ASR
registers to be in the 16..31 range.
frag_var () assigns its last argument to frag::fr_opcode, and it turns out
some targets modify the string that points to. However it appears niether the
generic code or the microblaze code modifies what fr_opcode points to, so this
code should be safe. So we unfortunately need to cast to char * when passing
an argument to frag_var () but otherwise microblaze can itself point to these
strings with const char *.
gas/ChangeLog:
2016-03-24 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-microblaze.c (md_assemble): Cast opc to char * when calling
frag_var ().
I'm not sure the string it returns is particularly useful, or better than the
string returned by other atof implementations on failure, but given the others
return a localized string it seems like this one should too.
gas/ChangeLog:
2016-03-24 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-visium.c (md_atof): Localize the string returned on
failure.
if the condition is true then we know that str already points to a'\0' in the
string passed to the function. Since we know the latter part of the function
doesn't modify that string, and str already points to a null byte there's no
point in changing str to point to a literal empty string.
gas/ChangeLog:
2016-03-22 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-pdp11.c (md_assemble): Remove useless if and assignment to
str.
This lets us avoid assigning a literal to a char *, and perhaps more
importantly makes it clearer what is going on here.
gas/ChangeLog:
2016-03-22 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-sparc.c (sparc_regname_to_dw2regnum): Replace strchr ()
call with a switch.
When parsing the operand instruction flags we don't currently detect the
case where multiple flags are provided from the same class set, these
will be accepted and the bit values merged together, resulting in the
wrong instruction being assembled. For example:
adc.n.eq r0,r0,r2
Will assemble without error, yet, upon disassembly, the instruction will
actually be:
adc.c r0,r0,r2
In a later commit the concept of required flags will be introduced.
Required flags are just like normal instruction flags, except that they
must be present for the instruction to match. Adding this will allow
for simpler instructions in the instruction table, and allow for more
sharing of operand extraction and insertion functions.
To solve both of the above issues (multiple flags being invalid, and
required flags), this commit reworks the flag class mechanism.
Currently the flag class is never used. Each instruction can reference
multiple flag classes, each flag class has a class type and a set of
flags. However, at present, the class type is never used. The current
values identify the type of instruction that the flag will be used in,
but this is not required information.
Instead, this commit discards the old flag classes, and introduces 3 new
classes. The first F_CLASS_NONE, is just a NULL marker value, and is
only used in the NULL marker flag class. The other two flag classes are
F_FLAG_OPTIONAL, and F_FLAG_REQUIRED.
The class F_FLAG_OPTIONAL has the property that at most one of the flags
in the flag set for that class must be present in the instruction. The
"at most" one means that no flags being present is fine.
The class F_FLAG_REQUIRED is not currently used, but will be soon. With
this class, exactly one of the flags from this class must be present in
the instruction. If the flag class contains a single flag, then of
course that flag must be present. However, if the flag class contained
two or more, then one, and only one of them must be present.
gas/ChangeLog:
* config/tc-arc.c (find_opcode_match): Move lnflg, and i
declarations to start of block. Reset code on all flags before
attempting to match them. Handle multiple hits on the same flag.
Handle flag class.
* testsuite/gas/arc/asm-errors.d: New file.
* testsuite/gas/arc/asm-errors.err: New file.
* testsuite/gas/arc/asm-errors.s: New file.
include/ChangeLog:
* opcode/arc.h (flag_class_t): Remove all old flag classes, add 3
new classes instead.
opcodes/ChangeLog:
* arc-opc.c (arc_flag_classes): Convert all flag classes to use
the new class enum values.
This commit introduces the nps400 machine type as a variant of arc.
There's a new flag in the assembler to select this machine type. All
other changes are just adding handling of the new machine type into the
relevant places.
The nps400 is an arc700 variant with some vendor specific instructions
added into the instruction set. This commit does not add any of the new
instructions, this is just laying the groundwork for future commits.
However, in preparation for these new instructions a new opcode define for
nps400 has been added to include/opcode/arc.h, this new opcode define is
used in the assembler and disassembler along with the existing define
for arc700 such that when assembling and disassembling for nps400 the
user will have access to all arc700 instructions and all the nps400
vendor extension instructions.
bfd/ChangeLog:
* archures.c (bfd_mach_arc_nps400): Define.
* bfd-in2.h: Regenerate.
* cpu-arc.c (arch_info_struct): New entry for nps400, renumber
some existing entries to make space.
* elf32-arc.c (arc_elf_object_p): Add nps400 case.
(arc_elf_final_write_processing): Likewise.
binutils/ChangeLog:
* readelf.c (decode_ARC_machine_flags): Handle nps400.
gas/ChangeLog:
* config/tc-arc.c (cpu_types): Add nps400 entry.
(check_zol): Handle nps400.
include/ChangeLog:
* elf/arc.h (E_ARC_MACH_NPS400): Define.
* opcode/arc.h (ARC_OPCODE_NPS400): Define.
opcodes/ChangeLog:
* arc-dis.c (print_insn_arc): Handle nps400.
The constant EF_ARC_CPU_GENERIC is defined in the include/elf/arc.h
file, and is used in a few places in binutils, however, this constant
should never make it into the elf header flags; we always set a valid
cpu type in the assembler, which should then be copied over during
linking.
There are some non-gnu arc compilers that don't write an architecture
type into the e_flags field, instead leaving the field as 0, which is
the EF_ARC_CPU_GENERIC value. This non-gnu compiler uses the machine
type to distinguish between the old and newer arc architectures, setting
the machine type to EM_ARC_COMPACT for old arc600, arc601, and arc700
architectures, while using EM_ARC_COMPACT2 for newer arcem and archs
architectures.
Previously when displaying the machine flags for an older EM_ARC_COMPACT
machine, if the e_flags had not been filled in, then we relied on the
default case statement to display the message "Generic ARCompact", while
in the EM_ARC_COMPACT2 case we specifically handled EF_ARC_CPU_GENERIC
to print "ARC Generic", leaving the default case to print a message
about unrecognised cpu flag.
After this commit EF_ARC_CPU_GENERIC has been removed, for both machine
types EM_ARC_COMPACT and EM_ARC_COMPACT2 we now rely on the default case
statement to handle the situation where the e_flags has not been filled
in. The message displayed is now "Unknown ARCompact" (for older arc
architectures) and "Unknown ARC" (for the newer architectures). The
switch from "Generic" to "Unknown" in the message string is for clarity,
calling the file "Generic" can give the impression that the file is
compiled for a common sub-set of the architectures, and would therefore
run on any type of machine (or at least any type of new or old machine
depending on if the machine type is ARC or ARCv2). However, this was
not what "Generic" meant, it really meant "Unknown", so that's what we
now say.
As part of the merging of the readelf flag reading code, I have unified
the strings used in displaying the ELF ABI. This means that for older
arc machines (arc600, arc601, and arc700) the string used for the
original ABI, and ABIv2 have changed, the current ABIv3 remains the
same. For the newer architectures (arcem and archs) the abi strings
remain unchanged in all cases.
bfd/ChangeLog:
* elf32-arc.c (arc_elf_print_private_bfd_data): Remove use of
EF_ARC_CPU_GENERIC.
(arc_elf_final_write_processing): Don't bother setting cpu field
in e_flags, this will have been set elsewhere.
binutils/ChangeLog:
* readelf.c (get_machine_flags): Move arc processing into...
(decode_ARC_machine_flags): ... new function. Remove use of
EF_ARC_CPU_GENERIC, change default case from "generic arc" to
"unknown arc". Merged ABI printing between two machine types.
gas/ChangeLog:
* config/tc-arc.c (arc_select_cpu): Remove use of
EF_ARC_CPU_GENERIC.
include/ChangeLog:
* elf/arc.h (EF_ARC_CPU_GENERIC): Delete. Update related comment.
This commit restructures the selection of the default cpu/mach so that
the choice is made from md_begin (if the user has not provided a command
line choice). This will reduce the amount of change needed in a later
patch.
At the request of Synopsys, the default architecture changes to ARC700
from this commit, previously the default was a non-existent
super-architecture that contained all instructions from all arc
variants. There's some clean up associated with removing the default
merged architecture, and a small test fix now that the default is
ARC700.
binutils/ChangeLog:
* testsuite/binutils-all/objdump.exp (cpus_expected): Add ARC700
to the architecture list.
gas/ChangeLog:
* config/tc-arc.c (arc_target): Delay initialisation until
arc_select_cpu.
(arc_target_name): Likewise.
(arc_features): Likewise.
(arc_mach_type): Likewise.
(cpu_types): Remove "all" entry.
(arc_select_cpu): New function, most of the content is from...
(md_parse_option): ... here. Call new arc_select_cpu.
(md_begin): Call arc_select_cpu if needed, default is now arc700.
include/ChangeLog:
* opcode/arc.h (ARC_OPCODE_BASE): Delete.
opcodes/ChangeLog:
* arc-opc.c (BASE): Delete.
