The AArch64 feature-flag code is currently limited to a maximum
of 64 features. This patch reworks it so that the limit can be
increased more easily. The basic idea is:
(1) Turn the ARM_FEATURE_FOO macros into an enum, with the enum
counting bit positions.
(2) Make the feature-list macros take an array index argument
(currently always 0). The macros then return the
aarch64_feature_set contents for that array index.
An N-element array would then be initialised as:
{ MACRO (0), ..., MACRO (N - 1) }
(3) Provide convenience macros for initialising an
aarch64_feature_set for:
- a single feature
- a list of individual features
- an architecture version
- an architecture version + a list of additional features
(2) and (3) use the preprocessor to generate static initialisers.
The main restriction was that uses of the same preprocessor macro
cannot be nested. So if a macro wants to do something for N individual
arguments, it needs to use a chain of N macros to do it. There then
needs to be a way of deriving N, as a preprocessor token suitable for
pasting.
The easiest way of doing that was to precede each list of features
by the number of features in the list. So an aarch64_feature_set
initialiser for three features A, B and C would be written:
AARCH64_FEATURES (3, A, B, C)
This scheme makes it difficult to keep AARCH64_FEATURE_CRYPTO as a
synonym for SHA2+AES, so the patch expands the former to the latter.
Historically, flags and variables relating to architectural revisions
for the A-profile architecture omitted the trailing `A' such that, for
example, assembling for `-march=armv8.4-a' set the `AARCH64_ARCH_V8_4'
flag in the assembler.
This leads to some ambiguity, since Binutils also targets the
R-profile Arm architecture. Therefore, it seems prudent to have
everything associated with the A-profile cores end in `A' and likewise
`R' for the R-profile. Referring back to the example above, the flag
set for `-march=armv8.4-a' is better characterized if labeled
`AARCH64_ARCH_V8_4A'.
The only exception to the rule of appending `A' to variables is found
in the handling of the `AARCH64_FEATURE_V8' macro, as it is the
baseline from which ALL processors derive and should therefore be left
unchanged.
In reflecting the `ARM' architectural nomenclature choices, where we
have `ARM_ARCH_V8A' and `ARM_ARCH_V8R', the choice is made to not have
an underscore separating the numerical revision number and the
A/R-profile indicator suffix. This has meant that renaming of
R-profile related flags and variables was warranted, thus going from
`.*_[vV]8_[rR]' to `.*_[vV]8[rR]'.
Finally, this is more in line with conventions within GCC and adds consistency
across the toolchain.
gas/ChangeLog:
* gas/config/tc-aarch64.c:
(aarch64_cpus): Reference to arch feature macros updated.
(aarch64_archs): Likewise.
include/ChangeLog:
* include/opcode/aarch64.h:
(AARCH64_FEATURE_V8A): Updated name: V8_A -> V8A.
(AARCH64_FEATURE_V8_1A): A-suffix added.
(AARCH64_FEATURE_V8_2A): Likewise.
(AARCH64_FEATURE_V8_3A): Likewise.
(AARCH64_FEATURE_V8_4A): Likewise.
(AARCH64_FEATURE_V8_5A): Likewise.
(AARCH64_FEATURE_V8_6A): Likewise.
(AARCH64_FEATURE_V8_7A): Likewise.
(AARCH64_FEATURE_V8_8A):Likewise.
(AARCH64_FEATURE_V9A): Likewise.
(AARCH64_FEATURE_V8R): Updated name: V8_R -> V8R.
(AARCH64_ARCH_V8A_FEATURES): Updated name: V8_A -> V8A.
(AARCH64_ARCH_V8_1A_FEATURES): A-suffix added.
(AARCH64_ARCH_V8_2A_FEATURES): Likewise.
(AARCH64_ARCH_V8_3A_FEATURES): Likewise.
(AARCH64_ARCH_V8_4A_FEATURES): Likewise.
(AARCH64_ARCH_V8_5A_FEATURES): Likewise.
(AARCH64_ARCH_V8_6A_FEATURES): Likewise.
(AARCH64_ARCH_V8_7A_FEATURES): Likewise.
(AARCH64_ARCH_V8_8A_FEATURES): Likewise.
(AARCH64_ARCH_V9A_FEATURES): Likewise.
(AARCH64_ARCH_V9_1A_FEATURES): Likewise.
(AARCH64_ARCH_V9_2A_FEATURES): Likewise.
(AARCH64_ARCH_V9_3A_FEATURES): Likewise.
(AARCH64_ARCH_V8A): Updated name: V8_A -> V8A.
(AARCH64_ARCH_V8_1A): A-suffix added.
(AARCH64_ARCH_V8_2A): Likewise.
(AARCH64_ARCH_V8_3A): Likewise.
(AARCH64_ARCH_V8_4A): Likewise.
(AARCH64_ARCH_V8_5A): Likewise.
(AARCH64_ARCH_V8_6A): Likewise.
(AARCH64_ARCH_V8_7A): Likewise.
(AARCH64_ARCH_V8_8A): Likewise.
(AARCH64_ARCH_V9A): Likewise.
(AARCH64_ARCH_V9_1A): Likewise.
(AARCH64_ARCH_V9_2A): Likewise.
(AARCH64_ARCH_V9_3A): Likewise.
(AARCH64_ARCH_V8_R): Updated name: V8_R -> V8R.
opcodes/ChangeLog:
* opcodes/aarch64-opc.c (SR_V8A): Updated name: V8_A -> V8A.
(SR_V8_1A): A-suffix added.
(SR_V8_2A): Likewise.
(SR_V8_3A): Likewise.
(SR_V8_4A): Likewise.
(SR_V8_6A): Likewise.
(SR_V8_7A): Likewise.
(SR_V8_8A): Likewise.
(aarch64_sys_regs): Reference to arch feature macros updated.
(aarch64_pstatefields): Reference to arch feature macros updated.
(aarch64_sys_ins_reg_supported_p): Reference to arch feature macros
updated.
* opcodes/aarch64-tbl.h:
(aarch64_feature_v8_2a): a-suffix added.
(aarch64_feature_v8_3a): Likewise.
(aarch64_feature_fp_v8_3a): Likewise.
(aarch64_feature_v8_4a): Likewise.
(aarch64_feature_fp_16_v8_2a): Likewise.
(aarch64_feature_v8_5a): Likewise.
(aarch64_feature_v8_6a): Likewise.
(aarch64_feature_v8_7a): Likewise.
(aarch64_feature_v8r): Updated name: v8_r-> v8r.
(ARMV8R): Updated name: V8_R-> V8R.
(ARMV8_2A): A-suffix added.
(ARMV8_3A): Likewise.
(FP_V8_3A): Likewise.
(ARMV8_4A): Likewise.
(FP_F16_V8_2A): Likewise.
(ARMV8_5): Likewise.
(ARMV8_6A): Likewise.
(ARMV8_6A_SVE): Likewise.
(ARMV8_7A): Likewise.
(V8_2A_INSN): `A' added to macro symbol.
(V8_3A_INSN): Likewise.
(V8_4A_INSN): Likewise.
(FP16_V8_2A_INSN): Likewise.
(V8_5A_INSN): Likewise.
(V8_6A_INSN): Likewise.
(V8_7A_INSN): Likewise.
(V8R_INSN): Updated name: V8_R-> V8R.
This patch adds the RPRFM (range prefetch) instruction.
It was introduced as part of SME2, but it belongs to the
prefetch hint space and so doesn't require any specific
ISA flags.
The aarch64_rprfmop_array initialiser (deliberately) only
fills in the leading non-null elements.
This patch adds the SVE FDOT, SDOT and UDOT instructions,
which are available when FEAT_SME2 is implemented. The patch
also reorders the existing SVE_Zm3_22_INDEX to keep the
operands numerically sorted.
There are two instruction formats here:
- SQRSHR, SQRSHRU and UQRSHR, which operate on lists of two
or four registers.
- SQRSHRN, SQRSHRUN and UQRSHRN, which operate on lists of
four registers.
These are the first SME2 instructions to have immediate operands.
The patch makes sure that, when parsing SME2 instructions with
immediate operands, the new predicate-as-counter registers are
parsed as registers rather than as #-less immediates.
There are two instruction formats here:
- SQCVT, SQCVTU and UQCVT, which operate on lists of two or
four registers.
- SQCVTN, SQCVTUN and UQCVTN, which operate on lists of
four registers.
SMLALL, SMLSLL, UMLALL and UMLSLL have the same format.
USMLALL and SUMLALL allow the same operand types as those
instructions, except that SUMLALL does not have the multi-vector
x multi-vector forms (which would be redundant with USMLALL).
The {BF,F,S,U}MLAL and {BF,F,S,U}MLSL instructions share the same
encoding. They are the first instance of a ZA (as opposed to ZA tile)
operand having a range of offsets. As with ZA tiles, the expected
range size is encoded in the operand-specific data field.
Add support for the SME2 ADD. SUB, FADD and FSUB instructions.
SUB and FSUB have the same form as ADD and FADD, except that
ADD also has a 2-operand accumulating form.
The 64-bit ADD/SUB instructions require FEAT_SME_I16I64 and the
64-bit FADD/FSUB instructions require FEAT_SME_F64F64.
These are the first instructions to have tied register list
operands, as opposed to tied single registers.
The parse_operands change prevents unsuffixed Z registers (width==-1)
from being treated as though they had an Advanced SIMD-style suffix
(.4s etc.). It means that:
Error: expected element type rather than vector type at operand 2 -- `add za\.s\[w8,0\],{z0-z1}'
becomes:
Error: missing type suffix at operand 2 -- `add za\.s\[w8,0\],{z0-z1}'
SME2 adds lookup table instructions for quantisation. They use
a new lookup table register called ZT0.
LUTI2 takes an unsuffixed SVE vector index of the form Zn[<imm>],
which is the first time that this syntax has been used.
Implementation-wise, the main things to note here are:
- the WHILE* instructions have forms that return a pair of predicate
registers. This is the first time that we've had lists of predicate
registers, and they wrap around after register 15 rather than after
register 31.
- the predicate-as-counter WHILE* instructions have a fourth operand
that specifies the vector length. We can treat this as an enumeration,
except that immediate values aren't allowed.
- PEXT takes an unsuffixed predicate index of the form PN<n>[<imm>].
This is the first instance of a vector/predicate index having
no suffix.
SME2 adds LD1 and ST1 variants for lists of 2 and 4 registers.
The registers can be consecutive or strided. In the strided case,
2-register lists have a stride of 8, starting at register x0xxx.
4-register lists have a stride of 4, starting at register x00xx.
The instructions are predicated on a predicate-as-counter register in
the range pn8-pn15. Although we already had register fields with upper
bounds of 7 and 15, this is the first plain register operand to have a
nonzero lower bound. The patch uses the operand-specific data field
to record the minimum value, rather than having separate inserters
and extractors for each lower bound. This in turn required adding
an extra bit to the field.
SME2 defines new MOVA instructions for moving multiple registers
to and from ZA. As with SME, the instructions are also available
through MOV aliases.
One notable feature of these instructions (and many other SME2
instructions) is that some register lists must start at a multiple
of the list's size. The patch uses the general error "start register
out of range" when this constraint isn't met, rather than an error
specifically about multiples. This ensures that the error is
consistent between these simple consecutive lists and later
strided lists, for which the requirements aren't a simple multiple.
SME2 adds a new format for the existing SVE predicate registers:
predicates as counters rather than predicates as masks. In assembly
code, operands that interpret predicates as counters are written
pn<N> rather than p<N>.
This patch adds support for these registers and extends some
existing instructions to support them. Since the new forms
are just a programmer convenience, there's no need to make them
more restrictive than the earlier predicate-as-mask forms.
Some SME2 instructions operate on a range of consecutive ZA vectors.
This is indicated by syntax such as:
za[<Wv>, <imml>:<immh>]
Like with the earlier vgx2 and vgx4 support, we get better error
messages if the parser allows all ZA indices to have a range.
We can then reject invalid cases during constraint checking.
Many SME2 instructions operate on groups of 2 or 4 ZA vectors.
This is indicated by adding a "vgx2" or "vgx4" group size to the
ZA index. The group size is optional in assembly but preferred
for disassembly.
There is not a binary distinction between mnemonics that have
group sizes and mnemonics that don't, nor between mnemonics that
take vgx2 and mnemonics that take vgx4. We therefore get better
error messages if we allow any ZA index to have a group size
during parsing, and wait until constraint checking to reject
invalid sizes.
A quirk of the way errors are reported means that if an instruction
is wrong both in its qualifiers and its use of a group size, we'll
print suggested alternative instructions that also have an incorrect
group size. But that's a general property that also applies to
things like out-of-range immediates. It's also not obviously the
wrong thing to do. We need to be relatively confident that we're
looking at the right opcode before reporting detailed operand-specific
errors, so doing qualifier checking first seems resonable.
SME2 adds various new fields that are similar to
AARCH64_OPND_SME_ZA_array, but are distinguished by the size of
their offset fields. This patch adds _off4 to the name of the
field that we already have.
Until now, binutils has supported register ranges such
as { v0.4s - v3.4s } as an unofficial shorthand for
{ v0.4s, v1.4s, v2.4s, v3.4s }. The SME2 ISA embraces this form
and makes it the preferred disassembly. It also embraces wrapped
lists such as { z31.s - z2.s }, which is something that binutils
didn't previously allow.
The range form was already binutils's preferred disassembly for 3- and
4-register lists. This patch prefers it for 2-register lists too.
The patch also adds support for wrap-around.
SME2 has instructions that accept strided register lists,
such as { z0.s, z4.s, z8.s, z12.s }. The purpose of this
patch is to extend binutils to support such lists.
The parsing code already had (unused) support for strides of 2.
The idea here is instead to accept all strides during parsing
and reject invalid strides during constraint checking.
The SME2 instructions that accept strided operands also have
non-strided forms. The errors about invalid strides therefore
take a bitmask of acceptable strides, which allows multiple
possibilities to be summed up in a single message.
I've tried to update all code that handles register lists.
Some FLD_imm* suffixes used a counting scheme such as FLD_immN,
FLD_immN_2, FLD_immN_3, etc., while others used the lsb as the
suffix. The latter seems more mnemonic, and was a big help
in doing the SME2 work.
Similarly, the _10 suffix on FLD_SME_size_10 was nonobvious.
Presumably it indicated a 2-bit field, but it actually starts
in bit 22.
Quite a lot of SME2 instructions have an opcode bit that selects
between 32-bit and 64-bit forms of an instruction, with the 32-bit
forms being part of base SME2 and with the 64-bit forms being part
of an optional extension. It's nevertheless useful to have a single
opcode entry for both forms since (a) that matches the ISA definition
and (b) it tends to improve error reporting.
This patch therefore adds a libopcodes function called
aarch64_cpu_supports_inst_p that tests whether the target
supports a particular instruction. In future it will depend
on internal libopcodes routines.
If an instruction has invalid qualifiers, GAS would report the
error against the final opcode entry that got to the qualifier-
checking stage. It seems better to report the error against
the opcode entry that had the closest match, just like we
pick the closest match within an opcode entry for the
"did you mean this?" message.
This patch adds the number of invalid operands as an
argument to AARCH64_OPDE_INVALID_VARIANT and then picks the
AARCH64_OPDE_INVALID_VARIANT with the lowest argument.
AARCH64_OPDE_REG_LIST took a single operand that specified the
expected number of registers. However, there are quite a few
SME2 instructions that have both 2-register forms and (separate)
4-register forms. If the user tries to use a 3-register list,
it isn't obvious which opcode entry they meant. Saying that we
expect 2 registers and saying that we expect 4 registers would
both be wrong.
This patch therefore switches the operand to a bitfield. If a
AARCH64_OPDE_REG_LIST is reported against multiple opcode entries,
the patch ORs up the expected lengths.
This has no user-visible effect yet. A later patch adds more error
strings, alongside tests that use them.
SVE register lists were classified as SVE_REG, since there had been
no particular reason to separate them out. However, some SME2
instructions have tied register list operands, and so we need to
distinguish registers and register lists when checking whether two
operands match.
Also, the register list operands used a general error message,
even though we already have a dedicated error code for register
lists that are the wrong length.
libopcodes currently reports out-of-range registers as a general
AARCH64_OPDE_OTHER_ERROR. However, this means that each register
range needs its own hard-coded string, which is a bit cumbersome
if the range is determined programmatically. This patch therefore
adds a dedicated error type for out-of-range errors.
In SME, the vector select register had to be in the range
w12-w15, so it made sense to enforce that during parsing.
However, SME2 adds instructions for which the range is
w8-w11 instead.
This patch therefore moves the range check from the parsing
stage to the constraint-checking stage.
Also, the previous error used a capitalised range W12-W15,
whereas other register range errors used lowercase ranges
like p0-p7. A quick internal poll showed a preference for
the lowercase form, so the patch uses that.
The patch uses "selection register" rather than "vector
select register" so that the terminology extends more
naturally to PSEL.
This patch moves the range checks on ZA vector select offsets from
gas to libopcodes. Doing the checks there means that the error
messages contain the expected range. It also fits in better
with the error severity scheme, which becomes important later.
(This is because out-of-range indices are treated as more severe than
syntax errors, on the basis that parsing must have succeeded if we get
to the point of checking the completed opcode.)
The patch also adds a new check_za_access function for checking
ZA accesses. That's a bit over the top for one offset check, but the
function becomes more complex with later patches.
sme-9-illegal.s checked for an invalid .q suffix using:
psel p1, p15, p3.q[w15]
but this is doubly invalid because it misses the immediate part
of the index. The patch keeps that test but adds another with
a zero index, so that .q is the only thing wrong.
The aarch64-tbl.h change includes neatening up the backslash
positions.
A later patch moves the range checking for ZA vector select
offsets from gas to libopcodes. That in turn requires the
immediate field to be big enough to support all parsed values.
This shouldn't be a particularly size-sensitive structure,
so there should be no memory problems with doing this.
za_tile_vector is also used for indexing ZA as a whole, rather than
just for indexing tiles. The former is more common than the latter
in SME2, so this patch generalises the name to "indexed_za".
The patch also names the associated structure, so that later patches
can reuse it during parsing.
In the register-index forms of PRFM, the unallocated prefetch opcodes
24-31 have been reused for the encoding of the new RPRFM instruction.
The PRFM opcode space is now capped at 23 for these forms. The other
forms of PRFM are unaffected.
The newer update-copyright.py fixes file encoding too, removing cr/lf
on binutils/bfdtest2.c and ld/testsuite/ld-cygwin/exe-export.exp, and
embedded cr in binutils/testsuite/binutils-all/ar.exp string match.
Current F_STRICT qualifier checking is enforced after the fact
rather than as part of the match. This makes it impossible to
have, e.g.:
QLF2(S_D, S_D)
QLF2(S_D, NIL)
in the same list.
opcodes/
* aarch64-opc.c (aarch64_find_best_match): Handle F_STRICT here
rather than...
(match_operands_qualifier): ...here.
This commit enables disassembler styling for AArch64. After this
commit it is possible to have objdump style AArch64 disassembler
output (using --disassembler-color option). Once the required GDB
patches are merged, GDB will also style the disassembler output.
The changes to support styling are mostly split between two files
opcodes/aarch64-dis.c and opcodes/aarch64-opc.c.
The entry point for the AArch64 disassembler can be found in
aarch64-dis.c, this file handles printing the instruction mnemonics,
and assembler directives (e.g. '.byte', '.word', etc). Some operands,
mostly relating to assembler directives are also printed from this
file. This commit changes all of this to pass through suitable
styling information.
However, for most "normal" instructions, the instruction operands are
printed using a two step process. From aarch64-dis.c, in the
print_operands function, the function aarch64_print_operand is called,
this function is in aarch64-opc.c, and converts an instruction operand
into a string. Then, back in print_operands (aarch64-dis.c), the
operand string is printed.
Unfortunately, the string returned by aarch64_print_operand can be
quite complex, it will include syntax elements, like '[' and ']', in
addition to register names and immediate values. In some cases, a
single operand will expand into what will appear (to the user) as
multiple operands separated with a ','.
This makes the task of styling more complex, all these different
components need to by styled differently, so we need to get the
styling information out of aarch64_print_operand in some way.
The solution that I propose here is similar to the solution that I
used for the i386 disassembler.
Currently, aarch64_print_operand uses snprintf to write the operand
text into a buffer provided by the caller.
What I propose is that we pass an extra argument to the
aarch64_print_operand function, this argument will be a structure, the
structure contains a callback function and some state.
When aarch64_print_operand needs to format part of its output this can
be done by using the callback function within the new structure, this
callback returns a string with special embedded markers that indicate
which mode should be used for each piece of text. Back in
aarch64-dis.c we can spot these special style markers and use this to
split the disassembler output up and apply the correct style to each
piece.
To make aarch64-opc.c clearer a series of new static functions have
been added, e.g. 'style_reg', 'style_imm', etc. Each of these
functions formats a piece of text in a different style, 'register' and
'immediate' in this case.
Here's an example taken from aarch64-opc.c of the new functions in
use:
snprintf (buf, size, "[%s, %s]!",
style_reg (styler, base),
style_imm (styler, "#%d", opnd->addr.offset.imm));
The aarch64_print_operand function is also called from the assembler
to aid in printing diagnostic messages. Right now I have no plans to
add styling to the assembler output, and so, the callback function
used in the assembler ignores the styling information and just returns
an plain string.
I've used the source files in gas/testsuite/gas/aarch64/ for testing,
and have manually gone through and checked that the styling looks
reasonable, however, I'm not an AArch64 expert, so it is possible that
the odd piece is styled incorrectly. Please point out any mistakes
I've made.
With objdump disassembler color turned off, there should be no change
in the output after this commit.
The function aarch64_print_operand (aarch64-opc.c) is responsible for
converting an instruction operand into the textual representation of
that operand.
In some cases, a comment is included in the operand representation,
though this (currently) only happens for the last operand of the
instruction.
In a future commit I would like to enable the new libopcodes styling
for AArch64, this will allow objdump and GDB[1] to syntax highlight
the disassembler output, however, having operands and comments
combined in a single string like this makes such styling harder.
In this commit, I propose to extend aarch64_print_operand to take a
second buffer. Any comments for the instruction are written into this
extra buffer. The two callers of aarch64_print_operand are then
updated to pass an extra buffer, and print any resulting comment.
In this commit no styling is added, that will come later. However, I
have adjusted the output slightly. Before this commit some comments
would be separated from the instruction operands with a tab character,
while in other cases the comment was separated with two single spaces.
After this commit I use a single tab character in all cases. This
means a few test cases needed updated. If people would prefer me to
move everyone to use the two spaces, then just let me know. Or maybe
there was a good reason why we used a mix of styles, I could probably
figure out a way to maintain the old output exactly if that is
critical.
Other than that, there should be no user visible changes after this
commit.
[1] GDB patches have not been merged yet, but have been posted to the
GDB mailing list:
https://sourceware.org/pipermail/gdb-patches/2022-June/190142.html
FEAT_RNG is an optional Armv8.5-A extension, but it can be backported
to earlier architectures as well. GAS previously made the RNG registers
conditional on having both armv8.5-a and +rng, but only +rng should be
required.
This seems to be the only feature that was handled like this.
opcodes/
* aarch64-opc.c (SR_RNG): Don't require V8_5.
gas/
* testsuite/gas/aarch64/rng-1.s, testsuite/gas/aarch64/rng-1.d: New
test.
The result of running etc/update-copyright.py --this-year, fixing all
the files whose mode is changed by the script, plus a build with
--enable-maintainer-mode --enable-cgen-maint=yes, then checking
out */po/*.pot which we don't update frequently.
The copy of cgen was with commit d1dd5fcc38ead reverted as that commit
breaks building of bfp opcodes files.
AARCH64_OPDE_EXPECTED_A_AFTER_B and AARCH64_OPDE_A_SHOULD_FOLLOW_B
are not paired with an error string, but we had an assert that the
error was nonnull. Previously this assert was testing uninitialised
memory and so could pass or fail arbitrarily.
opcodes/
* aarch64-opc.c (verify_mops_pme_sequence): Initialize the error
field to null for AARCH64_OPDE_EXPECTED_A_AFTER_B and
AARCH64_OPDE_A_SHOULD_FOLLOW_B.
* aarch64-dis.c (print_verifier_notes): Move assert.
The MOPS instructions should be used as a triple, such as:
cpyfp [x0]!, [x1]!, x2!
cpyfm [x0]!, [x1]!, x2!
cpyfe [x0]!, [x1]!, x2!
The registers should also be the same for each writeback operand.
This patch adds a warning for code that doesn't follow this rule,
along similar lines to the warning that we already emit for
invalid uses of MOVPRFX.
include/
* opcode/aarch64.h (C_SCAN_MOPS_P, C_SCAN_MOPS_M, C_SCAN_MOPS_E)
(C_SCAN_MOPS_PME): New macros.
(AARCH64_OPDE_A_SHOULD_FOLLOW_B): New aarch64_operand_error_kind.
(AARCH64_OPDE_EXPECTED_A_AFTER_B): Likewise.
(aarch64_operand_error): Make each data value a union between
an int and a string.
opcodes/
* aarch64-tbl.h (MOPS_CPY_OP1_OP2_INSN): Add scan flags.
(MOPS_SET_OP1_OP2_INSN): Likewise.
* aarch64-opc.c (set_out_of_range_error): Update after change to
aarch64_operand_error.
(set_unaligned_error, set_reg_list_error): Likewise.
(init_insn_sequence): Use a 3-instruction sequence for
MOPS P instructions.
(verify_mops_pme_sequence): New function.
(verify_constraints): Call it.
* aarch64-dis.c (print_verifier_notes): Handle
AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B.
gas/
* config/tc-aarch64.c (operand_mismatch_kind_names): Add entries
for AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B.
(operand_error_higher_severity_p): Check that
AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B
come between AARCH64_OPDE_RECOVERABLE and AARCH64_OPDE_SYNTAX_ERROR;
their relative order is not significant.
(record_operand_error_with_data): Update after change to
aarch64_operand_error.
(output_operand_error_record): Likewise. Handle
AARCH64_OPDE_A_SHOULD_FOLLOW_B and AARCH64_OPDE_EXPECTED_A_AFTER_B.
* testsuite/gas/aarch64/mops_invalid_2.s,
testsuite/gas/aarch64/mops_invalid_2.d,
testsuite/gas/aarch64/mops_invalid_2.l: New test.
This patch adds support for FEAT_MOPS, an Armv8.8-A extension
that provides memcpy and memset acceleration instructions.
I took the perhaps controversial decision to generate the individual
instruction forms using macros rather than list them out individually.
This becomes useful with a follow-on patch to check that code follows
the correct P/M/E sequence.
[https://developer.arm.com/documentation/ddi0596/2021-09/Base-Instructions?lang=en]
include/
* opcode/aarch64.h (AARCH64_FEATURE_MOPS): New macro.
(AARCH64_ARCH_V8_8): Make armv8.8-a imply AARCH64_FEATURE_MOPS.
(AARCH64_OPND_MOPS_ADDR_Rd): New aarch64_opnd.
(AARCH64_OPND_MOPS_ADDR_Rs): Likewise.
(AARCH64_OPND_MOPS_WB_Rn): Likewise.
opcodes/
* aarch64-asm.h (ins_x0_to_x30): New inserter.
* aarch64-asm.c (aarch64_ins_x0_to_x30): New function.
* aarch64-dis.h (ext_x0_to_x30): New extractor.
* aarch64-dis.c (aarch64_ext_x0_to_x30): New function.
* aarch64-tbl.h (aarch64_feature_mops): New feature set.
(aarch64_feature_mops_memtag): Likewise.
(MOPS, MOPS_MEMTAG, MOPS_INSN, MOPS_MEMTAG_INSN)
(MOPS_CPY_OP1_OP2_PME_INSN, MOPS_CPY_OP1_OP2_INSN, MOPS_CPY_OP1_INSN)
(MOPS_CPY_INSN, MOPS_SET_OP1_OP2_PME_INSN, MOPS_SET_OP1_OP2_INSN)
(MOPS_SET_INSN): New macros.
(aarch64_opcode_table): Add MOPS instructions.
(aarch64_opcode_table): Add entries for AARCH64_OPND_MOPS_ADDR_Rd,
AARCH64_OPND_MOPS_ADDR_Rs and AARCH64_OPND_MOPS_WB_Rn.
* aarch64-opc.c (aarch64_print_operand): Handle
AARCH64_OPND_MOPS_ADDR_Rd, AARCH64_OPND_MOPS_ADDR_Rs and
AARCH64_OPND_MOPS_WB_Rn.
(verify_three_different_regs): New function.
* aarch64-asm-2.c: Regenerate.
* aarch64-dis-2.c: Likewise.
* aarch64-opc-2.c: Likewise.
gas/
* doc/c-aarch64.texi: Document +mops.
* config/tc-aarch64.c (parse_x0_to_x30): New function.
(parse_operands): Handle AARCH64_OPND_MOPS_ADDR_Rd,
AARCH64_OPND_MOPS_ADDR_Rs and AARCH64_OPND_MOPS_WB_Rn.
(aarch64_features): Add "mops".
* testsuite/gas/aarch64/mops.s, testsuite/gas/aarch64/mops.d: New test.
* testsuite/gas/aarch64/mops_invalid.s,
* testsuite/gas/aarch64/mops_invalid.d,
* testsuite/gas/aarch64/mops_invalid.l: Likewise.
Armv8.8-A defines a read-only system register called id_aa64isar2_el1.
The register was previously RES0 and should therefore be accepted
at all architecture levels.
[https://developer.arm.com/documentation/ddi0595/2021-09/AArch64-Registers/ID-AA64ISAR2-EL1--AArch64-Instruction-Set-Attribute-Register-2?lang=en]
opcodes/
* aarch64-opc.c (aarch64_sys_regs): Add id_aa64isar2_el1.
gas/
* testsuite/gas/aarch64/sysreg-diagnostic.s: Test writes to
id_aa64isar2_el1.
* testsuite/gas/aarch64/sysreg-diagnostic.d: Update accordingly.
* testsuite/gas/aarch64/sysreg-diagnostic.l: Likewise.
* testsuite/gas/aarch64/sysreg.s: Test reads from
id_aa64isar2_el1.
* testsuite/gas/aarch64/sysreg.d: Update accordingly.
libopcodes has some code to check constraints across sequences
of consecutive instructions. It was added to support MOVPRFX
sequences but is going to be useful for the Armv8.8-A MOPS
feature as well.
Currently the structure has one field to record the instruction
that started a sequence and another to record the remaining
instructions in the sequence. It's more convenient for the
MOPS code if we put the instructions into a single array instead.
No functional change intended.
include/
* opcode/aarch64.h (aarch64_instr_sequence): Replace num_insns
and current_insns with num_added_insns and num_allocated_insns.
opcodes/
* aarch64-opc.c (add_insn_to_sequence): New function.
(init_insn_sequence): Update for new aarch64_instr_sequence layout.
Add the first instruction to the inst array.
(verify_constraints): Update for new aarch64_instr_sequence layout.
Don't add the last instruction to the array.
The immediate form of MSR has a 4-bit immediate field (in CRm).
However, many forms of MSR require a smaller immediate. These cases
are identified by value in operand_general_constraint_met_p,
but they're now the common case rather than the exception.
This patch therefore adds the maximum value to the sys_reg
description and gets the range from there. It also enforces
the minimum of 0, which avoids a situation in which:
msr dit, #2
would give the expected:
Error: immediate value out of range 0 to 1
whereas:
msr dit, #-1
would give:
Error: immediate value out of range 0 to 15
(from the later UIMM4 checking).
Also:
- we were reporting the first error above against the wrong operand
- TCO takes a single-bit immediate, but we previously allowed
all 16 values.
[https://developer.arm.com/documentation/ddi0596/2021-09/Base-Instructions/MSR--immediate---Move-immediate-value-to-Special-Register-?lang=en]
opcodes/
* aarch64-opc.h (F_REG_MAX_VALUE, F_GET_REG_MAX_VALUE): New macros.
* aarch64-opc.c (operand_general_constraint_met_p): Read the
maximum MSR immediate value from aarch64_pstatefields.
(aarch64_pstatefields): Add the maximum immediate value
for each register.
gas/
* testsuite/gas/aarch64/sysreg-4.s: Use an immediate value of 1
rather than 8 for the TCO test.
* testsuite/gas/aarch64/sysreg-4.d: Update accordingly.
* testsuite/gas/aarch64/armv8_2-a-illegal.l: Fix operand number
in MSR immediate error messages.
* testsuite/gas/aarch64/diagnostic.l: Likewise.
* testsuite/gas/aarch64/pan-illegal.l: Likewise.
* testsuite/gas/aarch64/ssbs-illegal1.l: Likewise.
* testsuite/gas/aarch64/illegal-sysreg-4b.s,
* testsuite/gas/aarch64/illegal-sysreg-4b.d,
* testsuite/gas/aarch64/illegal-sysreg-4b.l: New test.
