Add documentation for the MIPS assembler's -march=from-abi command line option

[binutils-gdb.git] / gas / doc / c-mips.texi

@c Copyright (C) 1991-2023 Free Software Foundation, Inc.
@c This is part of the GAS manual.
@c For copying conditions, see the file as.texinfo.
@ifset GENERIC
@page
@node MIPS-Dependent
@chapter MIPS Dependent Features
@end ifset
@ifclear GENERIC
@node Machine Dependencies
@chapter MIPS Dependent Features
@end ifclear

@cindex MIPS processor
@sc{gnu} @code{@value{AS}} for MIPS architectures supports several
different MIPS processors, and MIPS ISA levels I through V, MIPS32,
and MIPS64.  For information about the MIPS instruction set, see
@cite{MIPS RISC Architecture}, by Kane and Heindrich (Prentice-Hall).
For an overview of MIPS assembly conventions, see ``Appendix D:
Assembly Language Programming'' in the same work.

@menu
* MIPS Options::        Assembler options
* MIPS Macros::         High-level assembly macros
* MIPS Symbol Sizes::   Directives to override the size of symbols
* MIPS Small Data::     Controlling the use of small data accesses
* MIPS ISA::            Directives to override the ISA level
* MIPS assembly options:: Directives to control code generation
* MIPS autoextend::     Directives for extending MIPS 16 bit instructions
* MIPS insn::           Directive to mark data as an instruction
* MIPS FP ABIs::        Marking which FP ABI is in use
* MIPS NaN Encodings::  Directives to record which NaN encoding is being used
* MIPS Option Stack::   Directives to save and restore options
* MIPS ASE Instruction Generation Overrides:: Directives to control
                        generation of MIPS ASE instructions
* MIPS Floating-Point:: Directives to override floating-point options
* MIPS Syntax::         MIPS specific syntactical considerations
@end menu

@node MIPS Options
@section Assembler options

The MIPS configurations of @sc{gnu} @code{@value{AS}} support these
special options:

@table @code
@cindex @code{-G} option (MIPS)
@item -G @var{num}
Set the ``small data'' limit to @var{n} bytes.  The default limit is 8 bytes.
@xref{MIPS Small Data,, Controlling the use of small data accesses}.

@cindex @code{-EB} option (MIPS)
@cindex @code{-EL} option (MIPS)
@cindex MIPS big-endian output
@cindex MIPS little-endian output
@cindex big-endian output, MIPS
@cindex little-endian output, MIPS
@item -EB
@itemx -EL
Any MIPS configuration of @code{@value{AS}} can select big-endian or
little-endian output at run time (unlike the other @sc{gnu} development
tools, which must be configured for one or the other).  Use @samp{-EB}
to select big-endian output, and @samp{-EL} for little-endian.

@item -KPIC
@cindex PIC selection, MIPS
@cindex @option{-KPIC} option, MIPS
Generate SVR4-style PIC.  This option tells the assembler to generate
SVR4-style position-independent macro expansions.  It also tells the
assembler to mark the output file as PIC.

@item -mvxworks-pic
@cindex @option{-mvxworks-pic} option, MIPS
Generate VxWorks PIC.  This option tells the assembler to generate
VxWorks-style position-independent macro expansions.

@cindex MIPS architecture options
@item -mips1
@itemx -mips2
@itemx -mips3
@itemx -mips4
@itemx -mips5
@itemx -mips32
@itemx -mips32r2
@itemx -mips32r3
@itemx -mips32r5
@itemx -mips32r6
@itemx -mips64
@itemx -mips64r2
@itemx -mips64r3
@itemx -mips64r5
@itemx -mips64r6
Generate code for a particular MIPS Instruction Set Architecture level.
@samp{-mips1} corresponds to the R2000 and R3000 processors,
@samp{-mips2} to the R6000 processor, @samp{-mips3} to the
R4000 processor, and @samp{-mips4} to the R8000 and R10000 processors.
@samp{-mips5}, @samp{-mips32}, @samp{-mips32r2}, @samp{-mips32r3},
@samp{-mips32r5}, @samp{-mips32r6}, @samp{-mips64}, @samp{-mips64r2},
@samp{-mips64r3}, @samp{-mips64r5}, and @samp{-mips64r6} correspond to
generic MIPS V, MIPS32, MIPS32 Release 2, MIPS32 Release 3, MIPS32
Release 5, MIPS32 Release 6, MIPS64, and MIPS64 Release 2, MIPS64
Release 3, MIPS64 Release 5, and MIPS64 Release 6 ISA processors,
respectively.  You can also switch instruction sets during the assembly;
see @ref{MIPS ISA, Directives to override the ISA level}.

@item -mgp32
@itemx -mfp32
Some macros have different expansions for 32-bit and 64-bit registers.
The register sizes are normally inferred from the ISA and ABI, but these
flags force a certain group of registers to be treated as 32 bits wide at
all times.  @samp{-mgp32} controls the size of general-purpose registers
and @samp{-mfp32} controls the size of floating-point registers.

The @code{.set gp=32} and @code{.set fp=32} directives allow the size
of registers to be changed for parts of an object. The default value is
restored by @code{.set gp=default} and @code{.set fp=default}.

On some MIPS variants there is a 32-bit mode flag; when this flag is
set, 64-bit instructions generate a trap.  Also, some 32-bit OSes only
save the 32-bit registers on a context switch, so it is essential never
to use the 64-bit registers.

@item -mgp64
@itemx -mfp64
Assume that 64-bit registers are available.  This is provided in the
interests of symmetry with @samp{-mgp32} and @samp{-mfp32}.

The @code{.set gp=64} and @code{.set fp=64} directives allow the size
of registers to be changed for parts of an object. The default value is
restored by @code{.set gp=default} and @code{.set fp=default}.

@item -mfpxx
Make no assumptions about whether 32-bit or 64-bit floating-point
registers are available. This is provided to support having modules
compatible with either @samp{-mfp32} or @samp{-mfp64}. This option can
only be used with MIPS II and above.

The @code{.set fp=xx} directive allows a part of an object to be marked
as not making assumptions about 32-bit or 64-bit FP registers.  The
default value is restored by @code{.set fp=default}.

@item -modd-spreg
@itemx -mno-odd-spreg
Enable use of floating-point operations on odd-numbered single-precision
registers when supported by the ISA.  @samp{-mfpxx} implies
@samp{-mno-odd-spreg}, otherwise the default is @samp{-modd-spreg}

@item -mips16
@itemx -no-mips16
Generate code for the MIPS 16 processor.  This is equivalent to putting
@code{.module mips16} at the start of the assembly file.  @samp{-no-mips16}
turns off this option.

@item -mmips16e2
@itemx -mno-mips16e2
Enable the use of MIPS16e2 instructions in MIPS16 mode.  This is equivalent
to putting @code{.module mips16e2} at the start of the assembly file.
@samp{-mno-mips16e2} turns off this option.

@item -mmicromips
@itemx -mno-micromips
Generate code for the microMIPS processor.  This is equivalent to putting
@code{.module micromips} at the start of the assembly file.
@samp{-mno-micromips} turns off this option.  This is equivalent to putting
@code{.module nomicromips} at the start of the assembly file.

@item -msmartmips
@itemx -mno-smartmips
Enables the SmartMIPS extensions to the MIPS32 instruction set, which
provides a number of new instructions which target smartcard and
cryptographic applications.  This is equivalent to putting
@code{.module smartmips} at the start of the assembly file.
@samp{-mno-smartmips} turns off this option.

@item -mips3d
@itemx -no-mips3d
Generate code for the MIPS-3D Application Specific Extension.
This tells the assembler to accept MIPS-3D instructions.
@samp{-no-mips3d} turns off this option.

@item -mdmx
@itemx -no-mdmx
Generate code for the MDMX Application Specific Extension.
This tells the assembler to accept MDMX instructions.
@samp{-no-mdmx} turns off this option.

@item -mdsp
@itemx -mno-dsp
Generate code for the DSP Release 1 Application Specific Extension.
This tells the assembler to accept DSP Release 1 instructions.
@samp{-mno-dsp} turns off this option.

@item -mdspr2
@itemx -mno-dspr2
Generate code for the DSP Release 2 Application Specific Extension.
This option implies @samp{-mdsp}.
This tells the assembler to accept DSP Release 2 instructions.
@samp{-mno-dspr2} turns off this option.

@item -mdspr3
@itemx -mno-dspr3
Generate code for the DSP Release 3 Application Specific Extension.
This option implies @samp{-mdsp} and @samp{-mdspr2}.
This tells the assembler to accept DSP Release 3 instructions.
@samp{-mno-dspr3} turns off this option.

@item -mmt
@itemx -mno-mt
Generate code for the MT Application Specific Extension.
This tells the assembler to accept MT instructions.
@samp{-mno-mt} turns off this option.

@item -mmcu
@itemx -mno-mcu
Generate code for the MCU Application Specific Extension.
This tells the assembler to accept MCU instructions.
@samp{-mno-mcu} turns off this option.

@item -mmsa
@itemx -mno-msa
Generate code for the MIPS SIMD Architecture Extension.
This tells the assembler to accept MSA instructions.
@samp{-mno-msa} turns off this option.

@item -mxpa
@itemx -mno-xpa
Generate code for the MIPS eXtended Physical Address (XPA) Extension.
This tells the assembler to accept XPA instructions.
@samp{-mno-xpa} turns off this option.

@item -mvirt
@itemx -mno-virt
Generate code for the Virtualization Application Specific Extension.
This tells the assembler to accept Virtualization instructions.
@samp{-mno-virt} turns off this option.

@item -mcrc
@itemx -mno-crc
Generate code for the cyclic redundancy check (CRC) Application Specific
Extension.  This tells the assembler to accept CRC instructions.
@samp{-mno-crc} turns off this option.

@item -mginv
@itemx -mno-ginv
Generate code for the Global INValidate (GINV) Application Specific
Extension.  This tells the assembler to accept GINV instructions.
@samp{-mno-ginv} turns off this option.

@item -mloongson-mmi
@itemx -mno-loongson-mmi
Generate code for the Loongson MultiMedia extensions Instructions (MMI)
Application Specific Extension.  This tells the assembler to accept MMI
instructions.
@samp{-mno-loongson-mmi} turns off this option.

@item -mloongson-cam
@itemx -mno-loongson-cam
Generate code for the Loongson Content Address Memory (CAM)
Application Specific Extension.  This tells the assembler to accept CAM
instructions.
@samp{-mno-loongson-cam} turns off this option.

@item -mloongson-ext
@itemx -mno-loongson-ext
Generate code for the Loongson EXTensions (EXT) instructions
Application Specific Extension.  This tells the assembler to accept EXT
instructions.
@samp{-mno-loongson-ext} turns off this option.

@item -mloongson-ext2
@itemx -mno-loongson-ext2
Generate code for the Loongson EXTensions R2 (EXT2) instructions
Application Specific Extension.  This tells the assembler to accept EXT2
instructions.
@samp{-mno-loongson-ext2} turns off this option.

@item -minsn32
@itemx -mno-insn32
Only use 32-bit instruction encodings when generating code for the
microMIPS processor.  This option inhibits the use of any 16-bit
instructions.  This is equivalent to putting @code{.set insn32} at
the start of the assembly file.  @samp{-mno-insn32} turns off this
option.  This is equivalent to putting @code{.set noinsn32} at the
start of the assembly file.  By default @samp{-mno-insn32} is
selected, allowing all instructions to be used.

@item -mfix7000
@itemx -mno-fix7000
Cause nops to be inserted if the read of the destination register
of an mfhi or mflo instruction occurs in the following two instructions.

@item -mfix-rm7000
@itemx -mno-fix-rm7000
Cause nops to be inserted if a dmult or dmultu instruction is
followed by a load instruction.

@item -mfix-loongson2f-jump
@itemx -mno-fix-loongson2f-jump
Eliminate instruction fetch from outside 256M region to work around the
Loongson2F @samp{jump} instructions.  Without it, under extreme cases,
the kernel may crash.  The issue has been solved in latest processor
batches, but this fix has no side effect to them.

@item -mfix-loongson2f-nop
@itemx -mno-fix-loongson2f-nop
Replace nops by @code{or at,at,zero} to work around the Loongson2F
@samp{nop} errata.  Without it, under extreme cases, the CPU might
deadlock.  The issue has been solved in later Loongson2F batches, but
this fix has no side effect to them.

@item -mfix-loongson3-llsc
@itemx -mno-fix-loongson3-llsc
Insert @samp{sync} before @samp{ll} and @samp{lld} to work around
Loongson3 LLSC errata.  Without it, under extrame cases, the CPU might
deadlock. The default can be controlled by the
@option{--enable-mips-fix-loongson3-llsc=[yes|no]} configure option.

@item -mfix-vr4120
@itemx -mno-fix-vr4120
Insert nops to work around certain VR4120 errata.  This option is
intended to be used on GCC-generated code: it is not designed to catch
all problems in hand-written assembler code.

@item -mfix-vr4130
@itemx -mno-fix-vr4130
Insert nops to work around the VR4130 @samp{mflo}/@samp{mfhi} errata.

@item -mfix-24k
@itemx -mno-fix-24k
Insert nops to work around the 24K @samp{eret}/@samp{deret} errata.

@item -mfix-cn63xxp1
@itemx -mno-fix-cn63xxp1
Replace @code{pref} hints 0 - 4 and 6 - 24 with hint 28 to work around
certain CN63XXP1 errata.

@item -mfix-r5900
@itemx -mno-fix-r5900
Do not attempt to schedule the preceding instruction into the delay slot
of a branch instruction placed at the end of a short loop of six
instructions or fewer and always schedule a @code{nop} instruction there
instead.  The short loop bug under certain conditions causes loops to
execute only once or twice, due to a hardware bug in the R5900 chip.

@item -m4010
@itemx -no-m4010
Generate code for the LSI R4010 chip.  This tells the assembler to
accept the R4010-specific instructions (@samp{addciu}, @samp{ffc},
etc.), and to not schedule @samp{nop} instructions around accesses to
the @samp{HI} and @samp{LO} registers.  @samp{-no-m4010} turns off this
option.

@item -m4650
@itemx -no-m4650
Generate code for the MIPS R4650 chip.  This tells the assembler to accept
the @samp{mad} and @samp{madu} instruction, and to not schedule @samp{nop}
instructions around accesses to the @samp{HI} and @samp{LO} registers.
@samp{-no-m4650} turns off this option.

@item -m3900
@itemx -no-m3900
@itemx -m4100
@itemx -no-m4100
For each option @samp{-m@var{nnnn}}, generate code for the MIPS
R@var{nnnn} chip.  This tells the assembler to accept instructions
specific to that chip, and to schedule for that chip's hazards.

@item -march=@var{cpu}
Generate code for a particular MIPS CPU.  It is exactly equivalent to
@samp{-m@var{cpu}}, except that there are more value of @var{cpu}
understood.  Valid @var{cpu} value are:

@quotation
2000,
3000,
3900,
4000,
4010,
4100,
4111,
vr4120,
vr4130,
vr4181,
4300,
4400,
4600,
4650,
5000,
rm5200,
rm5230,
rm5231,
rm5261,
rm5721,
vr5400,
vr5500,
6000,
rm7000,
8000,
rm9000,
10000,
12000,
14000,
16000,
4kc,
4km,
4kp,
4ksc,
4kec,
4kem,
4kep,
4ksd,
m4k,
m4kp,
m14k,
m14kc,
m14ke,
m14kec,
24kc,
24kf2_1,
24kf,
24kf1_1,
24kec,
24kef2_1,
24kef,
24kef1_1,
34kc,
34kf2_1,
34kf,
34kf1_1,
34kn,
74kc,
74kf2_1,
74kf,
74kf1_1,
74kf3_2,
1004kc,
1004kf2_1,
1004kf,
1004kf1_1,
interaptiv,
interaptiv-mr2,
m5100,
m5101,
p5600,
5kc,
5kf,
20kc,
25kf,
sb1,
sb1a,
i6400,
i6500,
p6600,
loongson2e,
loongson2f,
gs464,
gs464e,
gs264e,
octeon,
octeon+,
octeon2,
octeon3,
xlr,
xlp
@end quotation

For compatibility reasons, @samp{@var{n}x} and @samp{@var{b}fx} are
accepted as synonyms for @samp{@var{n}f1_1}.  These values are
deprecated.

In addition the special name @samp{from-abi} can be used, in which
case the assembler will select an architecture suitable for whichever
ABI has been selected, either via the @option{-mabi=} command line
option or the built in default.

@item -mtune=@var{cpu}
Schedule and tune for a particular MIPS CPU.  Valid @var{cpu} values are
identical to @samp{-march=@var{cpu}}.

@item -mabi=@var{abi}
Record which ABI the source code uses.  The recognized arguments
are: @samp{32}, @samp{n32}, @samp{o64}, @samp{64} and @samp{eabi}.

@item -msym32
@itemx -mno-sym32
@cindex -msym32
@cindex -mno-sym32
Equivalent to adding @code{.set sym32} or @code{.set nosym32} to
the beginning of the assembler input.  @xref{MIPS Symbol Sizes}.

@cindex @code{-nocpp} ignored (MIPS)
@item -nocpp
This option is ignored.  It is accepted for command-line compatibility with
other assemblers, which use it to turn off C style preprocessing.  With
@sc{gnu} @code{@value{AS}}, there is no need for @samp{-nocpp}, because the
@sc{gnu} assembler itself never runs the C preprocessor.

@item -msoft-float
@itemx -mhard-float
Disable or enable floating-point instructions.  Note that by default
floating-point instructions are always allowed even with CPU targets
that don't have support for these instructions.

@item -msingle-float
@itemx -mdouble-float
Disable or enable double-precision floating-point operations.  Note
that by default double-precision floating-point operations are always
allowed even with CPU targets that don't have support for these
operations.

@item --construct-floats
@itemx --no-construct-floats
The @code{--no-construct-floats} option disables the construction of
double width floating point constants by loading the two halves of the
value into the two single width floating point registers that make up
the double width register.  This feature is useful if the processor
support the FR bit in its status  register, and this bit is known (by
the programmer) to be set.  This bit prevents the aliasing of the double
width register by the single width registers.

By default @code{--construct-floats} is selected, allowing construction
of these floating point constants.

@item --relax-branch
@itemx --no-relax-branch
The @samp{--relax-branch} option enables the relaxation of out-of-range
branches.  Any branches whose target cannot be reached directly are
converted to a small instruction sequence including an inverse-condition
branch to the physically next instruction, and a jump to the original
target is inserted between the two instructions.  In PIC code the jump
will involve further instructions for address calculation.

The @code{BC1ANY2F}, @code{BC1ANY2T}, @code{BC1ANY4F}, @code{BC1ANY4T},
@code{BPOSGE32} and @code{BPOSGE64} instructions are excluded from
relaxation, because they have no complementing counterparts.  They could
be relaxed with the use of a longer sequence involving another branch,
however this has not been implemented and if their target turns out of
reach, they produce an error even if branch relaxation is enabled.

Also no MIPS16 branches are ever relaxed.

By default @samp{--no-relax-branch} is selected, causing any out-of-range
branches to produce an error.

@item -mignore-branch-isa
@itemx -mno-ignore-branch-isa
Ignore branch checks for invalid transitions between ISA modes.

The semantics of branches does not provide for an ISA mode switch, so in
most cases the ISA mode a branch has been encoded for has to be the same
as the ISA mode of the branch's target label.  If the ISA modes do not
match, then such a branch, if taken, will cause the ISA mode to remain
unchanged and instructions that follow will be executed in the wrong ISA
mode causing the program to misbehave or crash.

In the case of the @code{BAL} instruction it may be possible to relax
it to an equivalent @code{JALX} instruction so that the ISA mode is
switched at the run time as required.  For other branches no relaxation
is possible and therefore GAS has checks implemented that verify in
branch assembly that the two ISA modes match, and report an error
otherwise so that the problem with code can be diagnosed at the assembly
time rather than at the run time.

However some assembly code, including generated code produced by some
versions of GCC, may incorrectly include branches to data labels, which
appear to require a mode switch but are either dead or immediately
followed by valid instructions encoded for the same ISA the branch has
been encoded for.  While not strictly correct at the source level such
code will execute as intended, so to help with these cases
@samp{-mignore-branch-isa} is supported which disables ISA mode checks
for branches.

By default @samp{-mno-ignore-branch-isa} is selected, causing any invalid
branch requiring a transition between ISA modes to produce an error.

@cindex @option{-mnan=} command-line option, MIPS
@item -mnan=@var{encoding}
This option indicates whether the source code uses the IEEE 2008
NaN encoding (@option{-mnan=2008}) or the original MIPS encoding
(@option{-mnan=legacy}).  It is equivalent to adding a @code{.nan}
directive to the beginning of the source file.  @xref{MIPS NaN Encodings}.

@option{-mnan=legacy} is the default if no @option{-mnan} option or
@code{.nan} directive is used.

@item --trap
@itemx --no-break
@c FIXME!  (1) reflect these options (next item too) in option summaries;
@c         (2) stop teasing, say _which_ instructions expanded _how_.
@code{@value{AS}} automatically macro expands certain division and
multiplication instructions to check for overflow and division by zero.  This
option causes @code{@value{AS}} to generate code to take a trap exception
rather than a break exception when an error is detected.  The trap instructions
are only supported at Instruction Set Architecture level 2 and higher.

@item --break
@itemx --no-trap
Generate code to take a break exception rather than a trap exception when an
error is detected.  This is the default.

@item -mpdr
@itemx -mno-pdr
Control generation of @code{.pdr} sections.  Off by default on IRIX, on
elsewhere.

@item -mshared
@itemx -mno-shared
When generating code using the Unix calling conventions (selected by
@samp{-KPIC} or @samp{-mcall_shared}), gas will normally generate code
which can go into a shared library.  The @samp{-mno-shared} option
tells gas to generate code which uses the calling convention, but can
not go into a shared library.  The resulting code is slightly more
efficient.  This option only affects the handling of the
@samp{.cpload} and @samp{.cpsetup} pseudo-ops.
@end table

@node MIPS Macros
@section High-level assembly macros

MIPS assemblers have traditionally provided a wider range of
instructions than the MIPS architecture itself.  These extra
instructions are usually referred to as ``macro'' instructions
@footnote{The term ``macro'' is somewhat overloaded here, since
these macros have no relation to those defined by @code{.macro},
@pxref{Macro,, @code{.macro}}.}.

Some MIPS macro instructions extend an underlying architectural instruction
while others are entirely new.  An example of the former type is @code{and},
which allows the third operand to be either a register or an arbitrary
immediate value.  Examples of the latter type include @code{bgt}, which
branches to the third operand when the first operand is greater than
the second operand, and @code{ulh}, which implements an unaligned
2-byte load.

One of the most common extensions provided by macros is to expand
memory offsets to the full address range (32 or 64 bits) and to allow
symbolic offsets such as @samp{my_data + 4} to be used in place of
integer constants.  For example, the architectural instruction
@code{lbu} allows only a signed 16-bit offset, whereas the macro
@code{lbu} allows code such as @samp{lbu $4,array+32769($5)}.
The implementation of these symbolic offsets depends on several factors,
such as whether the assembler is generating SVR4-style PIC (selected by
@option{-KPIC}, @pxref{MIPS Options,, Assembler options}), the size of symbols
(@pxref{MIPS Symbol Sizes,, Directives to override the size of symbols}),
and the small data limit (@pxref{MIPS Small Data,, Controlling the use
of small data accesses}).

@kindex @code{.set macro}
@kindex @code{.set nomacro}
Sometimes it is undesirable to have one assembly instruction expand
to several machine instructions.  The directive @code{.set nomacro}
tells the assembler to warn when this happens.  @code{.set macro}
restores the default behavior.

@cindex @code{at} register, MIPS
@kindex @code{.set at=@var{reg}}
Some macro instructions need a temporary register to store intermediate
results.  This register is usually @code{$1}, also known as @code{$at},
but it can be changed to any core register @var{reg} using
@code{.set at=@var{reg}}.  Note that @code{$at} always refers
to @code{$1} regardless of which register is being used as the
temporary register.

@kindex @code{.set at}
@kindex @code{.set noat}
Implicit uses of the temporary register in macros could interfere with
explicit uses in the assembly code.  The assembler therefore warns
whenever it sees an explicit use of the temporary register.  The directive
@code{.set noat} silences this warning while @code{.set at} restores
the default behavior.  It is safe to use @code{.set noat} while
@code{.set nomacro} is in effect since single-instruction macros
never need a temporary register.

Note that while the @sc{gnu} assembler provides these macros for compatibility,
it does not make any attempt to optimize them with the surrounding code.

@node MIPS Symbol Sizes
@section Directives to override the size of symbols

@kindex @code{.set sym32}
@kindex @code{.set nosym32}
The n64 ABI allows symbols to have any 64-bit value.  Although this
provides a great deal of flexibility, it means that some macros have
much longer expansions than their 32-bit counterparts.  For example,
the non-PIC expansion of @samp{dla $4,sym} is usually:

@smallexample
lui     $4,%highest(sym)
lui     $1,%hi(sym)
daddiu  $4,$4,%higher(sym)
daddiu  $1,$1,%lo(sym)
dsll32  $4,$4,0
daddu   $4,$4,$1
@end smallexample

whereas the 32-bit expansion is simply:

@smallexample
lui     $4,%hi(sym)
daddiu  $4,$4,%lo(sym)
@end smallexample

n64 code is sometimes constructed in such a way that all symbolic
constants are known to have 32-bit values, and in such cases, it's
preferable to use the 32-bit expansion instead of the 64-bit
expansion.

You can use the @code{.set sym32} directive to tell the assembler
that, from this point on, all expressions of the form
@samp{@var{symbol}} or @samp{@var{symbol} + @var{offset}}
have 32-bit values.  For example:

@smallexample
.set sym32
dla     $4,sym
lw      $4,sym+16
sw      $4,sym+0x8000($4)
@end smallexample

will cause the assembler to treat @samp{sym}, @code{sym+16} and
@code{sym+0x8000} as 32-bit values.  The handling of non-symbolic
addresses is not affected.

The directive @code{.set nosym32} ends a @code{.set sym32} block and
reverts to the normal behavior.  It is also possible to change the
symbol size using the command-line options @option{-msym32} and
@option{-mno-sym32}.

These options and directives are always accepted, but at present,
they have no effect for anything other than n64.

@node MIPS Small Data
@section Controlling the use of small data accesses

@c This section deliberately glosses over the possibility of using -G
@c in SVR4-style PIC, as could be done on IRIX.  We don't support that.
@cindex small data, MIPS
@cindex @code{gp} register, MIPS
It often takes several instructions to load the address of a symbol.
For example, when @samp{addr} is a 32-bit symbol, the non-PIC expansion
of @samp{dla $4,addr} is usually:

@smallexample
lui     $4,%hi(addr)
daddiu  $4,$4,%lo(addr)
@end smallexample

The sequence is much longer when @samp{addr} is a 64-bit symbol.
@xref{MIPS Symbol Sizes,, Directives to override the size of symbols}.

In order to cut down on this overhead, most embedded MIPS systems
set aside a 64-kilobyte ``small data'' area and guarantee that all
data of size @var{n} and smaller will be placed in that area.
The limit @var{n} is passed to both the assembler and the linker
using the command-line option @option{-G @var{n}}, @pxref{MIPS Options,,
Assembler options}.  Note that the same value of @var{n} must be used
when linking and when assembling all input files to the link; any
inconsistency could cause a relocation overflow error.

The size of an object in the @code{.bss} section is set by the
@code{.comm} or @code{.lcomm} directive that defines it.  The size of
an external object may be set with the @code{.extern} directive.  For
example, @samp{.extern sym,4} declares that the object at @code{sym}
is 4 bytes in length, while leaving @code{sym} otherwise undefined.

When no @option{-G} option is given, the default limit is 8 bytes.
The option @option{-G 0} prevents any data from being automatically
classified as small.

It is also possible to mark specific objects as small by putting them
in the special sections @code{.sdata} and @code{.sbss}, which are
``small'' counterparts of @code{.data} and @code{.bss} respectively.
The toolchain will treat such data as small regardless of the
@option{-G} setting.

On startup, systems that support a small data area are expected to
initialize register @code{$28}, also known as @code{$gp}, in such a
way that small data can be accessed using a 16-bit offset from that
register.  For example, when @samp{addr} is small data,
the @samp{dla $4,addr} instruction above is equivalent to:

@smallexample
daddiu  $4,$28,%gp_rel(addr)
@end smallexample

Small data is not supported for SVR4-style PIC.

@node MIPS ISA
@section Directives to override the ISA level

@cindex MIPS ISA override
@kindex @code{.set mips@var{n}}
@sc{gnu} @code{@value{AS}} supports an additional directive to change
the MIPS Instruction Set Architecture level on the fly: @code{.set
mips@var{n}}.  @var{n} should be a number from 0 to 5, or 32, 32r2, 32r3,
32r5, 32r6, 64, 64r2, 64r3, 64r5 or 64r6.
The values other than 0 make the assembler accept instructions
for the corresponding ISA level, from that point on in the
assembly.  @code{.set mips@var{n}} affects not only which instructions
are permitted, but also how certain macros are expanded.  @code{.set
mips0} restores the ISA level to its original level: either the
level you selected with command-line options, or the default for your
configuration.  You can use this feature to permit specific MIPS III
instructions while assembling in 32 bit mode.  Use this directive with
care!

@cindex MIPS CPU override
@kindex @code{.set arch=@var{cpu}}
The @code{.set arch=@var{cpu}} directive provides even finer control.
It changes the effective CPU target and allows the assembler to use
instructions specific to a particular CPU.  All CPUs supported by the
@samp{-march} command-line option are also selectable by this directive.
The original value is restored by @code{.set arch=default}.

The directive @code{.set mips16} puts the assembler into MIPS 16 mode,
in which it will assemble instructions for the MIPS 16 processor.  Use
@code{.set nomips16} to return to normal 32 bit mode.

Traditional MIPS assemblers do not support this directive.

The directive @code{.set micromips} puts the assembler into microMIPS mode,
in which it will assemble instructions for the microMIPS processor.  Use
@code{.set nomicromips} to return to normal 32 bit mode.

Traditional MIPS assemblers do not support this directive.

@node MIPS assembly options
@section Directives to control code generation

@cindex MIPS directives to override command-line options
@kindex @code{.module}
The @code{.module} directive allows command-line options to be set directly
from assembly.  The format of the directive matches the @code{.set}
directive but only those options which are relevant to a whole module are
supported.  The effect of a @code{.module} directive is the same as the
corresponding command-line option.  Where @code{.set} directives support
returning to a default then the @code{.module} directives do not as they
define the defaults.

These module-level directives must appear first in assembly.

Traditional MIPS assemblers do not support this directive.

@cindex MIPS 32-bit microMIPS instruction generation override
@kindex @code{.set insn32}
@kindex @code{.set noinsn32}
The directive @code{.set insn32} makes the assembler only use 32-bit
instruction encodings when generating code for the microMIPS processor.
This directive inhibits the use of any 16-bit instructions from that
point on in the assembly.  The @code{.set noinsn32} directive allows
16-bit instructions to be accepted.

Traditional MIPS assemblers do not support this directive.

@node MIPS autoextend
@section Directives for extending MIPS 16 bit instructions

@kindex @code{.set autoextend}
@kindex @code{.set noautoextend}
By default, MIPS 16 instructions are automatically extended to 32 bits
when necessary.  The directive @code{.set noautoextend} will turn this
off.  When @code{.set noautoextend} is in effect, any 32 bit instruction
must be explicitly extended with the @code{.e} modifier (e.g.,
@code{li.e $4,1000}).  The directive @code{.set autoextend} may be used
to once again automatically extend instructions when necessary.

This directive is only meaningful when in MIPS 16 mode.  Traditional
MIPS assemblers do not support this directive.

@node MIPS insn
@section Directive to mark data as an instruction

@kindex @code{.insn}
The @code{.insn} directive tells @code{@value{AS}} that the following
data is actually instructions.  This makes a difference in MIPS 16 and
microMIPS modes: when loading the address of a label which precedes
instructions, @code{@value{AS}} automatically adds 1 to the value, so
that jumping to the loaded address will do the right thing.

@kindex @code{.global}
The @code{.global} and @code{.globl} directives supported by
@code{@value{AS}} will by default mark the symbol as pointing to a
region of data not code.  This means that, for example, any
instructions following such a symbol will not be disassembled by
@code{objdump} as it will regard them as data.  To change this
behavior an optional section name can be placed after the symbol name
in the @code{.global} directive.  If this section exists and is known
to be a code section, then the symbol will be marked as pointing at
code not data.  Ie the syntax for the directive is:

  @code{.global @var{symbol}[ @var{section}][, @var{symbol}[ @var{section}]] ...},

Here is a short example:

@example
        .global foo .text, bar, baz .data
foo:
        nop
bar:
        .word 0x0
baz:
        .word 0x1

@end example

@node MIPS FP ABIs
@section Directives to control the FP ABI
@menu
* MIPS FP ABI History::                History of FP ABIs
* MIPS FP ABI Variants::               Supported FP ABIs
* MIPS FP ABI Selection::              Automatic selection of FP ABI
* MIPS FP ABI Compatibility::          Linking different FP ABI variants
@end menu

@node MIPS FP ABI History
@subsection History of FP ABIs
@cindex @code{.gnu_attribute 4, @var{n}} directive, MIPS
@cindex @code{.gnu_attribute Tag_GNU_MIPS_ABI_FP, @var{n}} directive, MIPS
The MIPS ABIs support a variety of different floating-point extensions
where calling-convention and register sizes vary for floating-point data.
The extensions exist to support a wide variety of optional architecture
features.  The resulting ABI variants are generally incompatible with each
other and must be tracked carefully.

Traditionally the use of an explicit @code{.gnu_attribute 4, @var{n}}
directive is used to indicate which ABI is in use by a specific module.
It was then left to the user to ensure that command-line options and the
selected ABI were compatible with some potential for inconsistencies.

@node MIPS FP ABI Variants
@subsection Supported FP ABIs
The supported floating-point ABI variants are:

@table @code
@item 0 - No floating-point
This variant is used to indicate that floating-point is not used within
the module at all and therefore has no impact on the ABI.  This is the
default.

@item 1 - Double-precision
This variant indicates that double-precision support is used.  For 64-bit
ABIs this means that 64-bit wide floating-point registers are required.
For 32-bit ABIs this means that 32-bit wide floating-point registers are
required and double-precision operations use pairs of registers.

@item 2 - Single-precision
This variant indicates that single-precision support is used.  Double
precision operations will be supported via soft-float routines.

@item 3 - Soft-float
This variant indicates that although floating-point support is used all
operations are emulated in software.  This means the ABI is modified to
pass all floating-point data in general-purpose registers.

@item 4 - Deprecated
This variant existed as an initial attempt at supporting 64-bit wide
floating-point registers for O32 ABI on a MIPS32r2 CPU.  This has been
superseded by 5, 6 and 7.

@item 5 - Double-precision 32-bit CPU, 32-bit or 64-bit FPU
This variant is used by 32-bit ABIs to indicate that the floating-point
code in the module has been designed to operate correctly with either
32-bit wide or 64-bit wide floating-point registers.  Double-precision
support is used.  Only O32 currently supports this variant and requires
a minimum architecture of MIPS II.

@item 6 - Double-precision 32-bit FPU, 64-bit FPU
This variant is used by 32-bit ABIs to indicate that the floating-point
code in the module requires 64-bit wide floating-point registers.
Double-precision support is used.  Only O32 currently supports this
variant and requires a minimum architecture of MIPS32r2.

@item 7 - Double-precision compat 32-bit FPU, 64-bit FPU
This variant is used by 32-bit ABIs to indicate that the floating-point
code in the module requires 64-bit wide floating-point registers.
Double-precision support is used.  This differs from the previous ABI
as it restricts use of odd-numbered single-precision registers.  Only
O32 currently supports this variant and requires a minimum architecture
of MIPS32r2.
@end table

@node MIPS FP ABI Selection
@subsection Automatic selection of FP ABI
@cindex @code{.module fp=@var{nn}} directive, MIPS
In order to simplify and add safety to the process of selecting the
correct floating-point ABI, the assembler will automatically infer the
correct @code{.gnu_attribute 4, @var{n}} directive based on command-line
options and @code{.module} overrides.  Where an explicit
@code{.gnu_attribute 4, @var{n}} directive has been seen then a warning
will be raised if it does not match an inferred setting.

The floating-point ABI is inferred as follows.  If @samp{-msoft-float}
has been used the module will be marked as soft-float.  If
@samp{-msingle-float} has been used then the module will be marked as
single-precision.  The remaining ABIs are then selected based
on the FP register width.  Double-precision is selected if the width
of GP and FP registers match and the special double-precision variants
for 32-bit ABIs are then selected depending on @samp{-mfpxx},
@samp{-mfp64} and @samp{-mno-odd-spreg}.

@node MIPS FP ABI Compatibility
@subsection Linking different FP ABI variants
Modules using the default FP ABI (no floating-point) can be linked with
any other (singular) FP ABI variant.

Special compatibility support exists for O32 with the four
double-precision FP ABI variants.  The @samp{-mfpxx} FP ABI is specifically
designed to be compatible with the standard double-precision ABI and the
@samp{-mfp64} FP ABIs.  This makes it desirable for O32 modules to be
built as @samp{-mfpxx} to ensure the maximum compatibility with other
modules produced for more specific needs.  The only FP ABIs which cannot
be linked together are the standard double-precision ABI and the full
@samp{-mfp64} ABI with @samp{-modd-spreg}.

@node MIPS NaN Encodings
@section Directives to record which NaN encoding is being used

@cindex MIPS IEEE 754 NaN data encoding selection
@cindex @code{.nan} directive, MIPS
The IEEE 754 floating-point standard defines two types of not-a-number
(NaN) data: ``signalling'' NaNs and ``quiet'' NaNs.  The original version
of the standard did not specify how these two types should be
distinguished.  Most implementations followed the i387 model, in which
the first bit of the significand is set for quiet NaNs and clear for
signalling NaNs.  However, the original MIPS implementation assigned the
opposite meaning to the bit, so that it was set for signalling NaNs and
clear for quiet NaNs.

The 2008 revision of the standard formally suggested the i387 choice
and as from Sep 2012 the current release of the MIPS architecture
therefore optionally supports that form.  Code that uses one NaN encoding
would usually be incompatible with code that uses the other NaN encoding,
so MIPS ELF objects have a flag (@code{EF_MIPS_NAN2008}) to record which
encoding is being used.

Assembly files can use the @code{.nan} directive to select between the
two encodings.  @samp{.nan 2008} says that the assembly file uses the
IEEE 754-2008 encoding while @samp{.nan legacy} says that the file uses
the original MIPS encoding.  If several @code{.nan} directives are given,
the final setting is the one that is used.

The command-line options @option{-mnan=legacy} and @option{-mnan=2008}
can be used instead of @samp{.nan legacy} and @samp{.nan 2008}
respectively.  However, any @code{.nan} directive overrides the
command-line setting.

@samp{.nan legacy} is the default if no @code{.nan} directive or
@option{-mnan} option is given.

Note that @sc{gnu} @code{@value{AS}} does not produce NaNs itself and
therefore these directives do not affect code generation.  They simply
control the setting of the @code{EF_MIPS_NAN2008} flag.

Traditional MIPS assemblers do not support these directives.

@node MIPS Option Stack
@section Directives to save and restore options

@cindex MIPS option stack
@kindex @code{.set push}
@kindex @code{.set pop}
The directives @code{.set push} and @code{.set pop} may be used to save
and restore the current settings for all the options which are
controlled by @code{.set}.  The @code{.set push} directive saves the
current settings on a stack.  The @code{.set pop} directive pops the
stack and restores the settings.

These directives can be useful inside an macro which must change an
option such as the ISA level or instruction reordering but does not want
to change the state of the code which invoked the macro.

Traditional MIPS assemblers do not support these directives.

@node MIPS ASE Instruction Generation Overrides
@section Directives to control generation of MIPS ASE instructions

@cindex MIPS MIPS-3D instruction generation override
@kindex @code{.set mips3d}
@kindex @code{.set nomips3d}
The directive @code{.set mips3d} makes the assembler accept instructions
from the MIPS-3D Application Specific Extension from that point on
in the assembly.  The @code{.set nomips3d} directive prevents MIPS-3D
instructions from being accepted.

@cindex SmartMIPS instruction generation override
@kindex @code{.set smartmips}
@kindex @code{.set nosmartmips}
The directive @code{.set smartmips} makes the assembler accept
instructions from the SmartMIPS Application Specific Extension to the
MIPS32 ISA from that point on in the assembly.  The
@code{.set nosmartmips} directive prevents SmartMIPS instructions from
being accepted.

@cindex MIPS MDMX instruction generation override
@kindex @code{.set mdmx}
@kindex @code{.set nomdmx}
The directive @code{.set mdmx} makes the assembler accept instructions
from the MDMX Application Specific Extension from that point on
in the assembly.  The @code{.set nomdmx} directive prevents MDMX
instructions from being accepted.

@cindex MIPS DSP Release 1 instruction generation override
@kindex @code{.set dsp}
@kindex @code{.set nodsp}
The directive @code{.set dsp} makes the assembler accept instructions
from the DSP Release 1 Application Specific Extension from that point
on in the assembly.  The @code{.set nodsp} directive prevents DSP
Release 1 instructions from being accepted.

@cindex MIPS DSP Release 2 instruction generation override
@kindex @code{.set dspr2}
@kindex @code{.set nodspr2}
The directive @code{.set dspr2} makes the assembler accept instructions
from the DSP Release 2 Application Specific Extension from that point
on in the assembly.  This directive implies @code{.set dsp}.  The
@code{.set nodspr2} directive prevents DSP Release 2 instructions from
being accepted.

@cindex MIPS DSP Release 3 instruction generation override
@kindex @code{.set dspr3}
@kindex @code{.set nodspr3}
The directive @code{.set dspr3} makes the assembler accept instructions
from the DSP Release 3 Application Specific Extension from that point
on in the assembly.  This directive implies @code{.set dsp} and
@code{.set dspr2}.  The @code{.set nodspr3} directive prevents DSP
Release 3 instructions from being accepted.

@cindex MIPS MT instruction generation override
@kindex @code{.set mt}
@kindex @code{.set nomt}
The directive @code{.set mt} makes the assembler accept instructions
from the MT Application Specific Extension from that point on
in the assembly.  The @code{.set nomt} directive prevents MT
instructions from being accepted.

@cindex MIPS MCU instruction generation override
@kindex @code{.set mcu}
@kindex @code{.set nomcu}
The directive @code{.set mcu} makes the assembler accept instructions
from the MCU Application Specific Extension from that point on
in the assembly.  The @code{.set nomcu} directive prevents MCU
instructions from being accepted.

@cindex MIPS SIMD Architecture instruction generation override
@kindex @code{.set msa}
@kindex @code{.set nomsa}
The directive @code{.set msa} makes the assembler accept instructions
from the MIPS SIMD Architecture Extension from that point on
in the assembly.  The @code{.set nomsa} directive prevents MSA
instructions from being accepted.

@cindex Virtualization instruction generation override
@kindex @code{.set virt}
@kindex @code{.set novirt}
The directive @code{.set virt} makes the assembler accept instructions
from the Virtualization Application Specific Extension from that point
on in the assembly.  The @code{.set novirt} directive prevents Virtualization
instructions from being accepted.

@cindex MIPS eXtended Physical Address (XPA) instruction generation override
@kindex @code{.set xpa}
@kindex @code{.set noxpa}
The directive @code{.set xpa} makes the assembler accept instructions
from the XPA Extension from that point on in the assembly.  The 
@code{.set noxpa} directive prevents XPA instructions from being accepted.

@cindex MIPS16e2 instruction generation override
@kindex @code{.set mips16e2}
@kindex @code{.set nomips16e2}
The directive @code{.set mips16e2} makes the assembler accept instructions
from the MIPS16e2 Application Specific Extension from that point on in the
assembly, whenever in MIPS16 mode.  The @code{.set nomips16e2} directive
prevents MIPS16e2 instructions from being accepted, in MIPS16 mode.  Neither
directive affects the state of MIPS16 mode being active itself which has
separate controls.

@cindex MIPS cyclic redundancy check (CRC) instruction generation override
@kindex @code{.set crc}
@kindex @code{.set nocrc}
The directive @code{.set crc} makes the assembler accept instructions
from the CRC Extension from that point on in the assembly.  The
@code{.set nocrc} directive prevents CRC instructions from being accepted.

@cindex MIPS Global INValidate (GINV) instruction generation override
@kindex @code{.set ginv}
@kindex @code{.set noginv}
The directive @code{.set ginv} makes the assembler accept instructions
from the GINV Extension from that point on in the assembly.  The
@code{.set noginv} directive prevents GINV instructions from being accepted.

@cindex Loongson MultiMedia extensions Instructions (MMI) generation override
@kindex @code{.set loongson-mmi}
@kindex @code{.set noloongson-mmi}
The directive @code{.set loongson-mmi} makes the assembler accept
instructions from the MMI Extension from that point on in the assembly.
The @code{.set noloongson-mmi} directive prevents MMI instructions from
being accepted.

@cindex Loongson Content Address Memory (CAM) generation override
@kindex @code{.set loongson-cam}
@kindex @code{.set noloongson-cam}
The directive @code{.set loongson-cam} makes the assembler accept
instructions from the Loongson CAM from that point on in the assembly.
The @code{.set noloongson-cam} directive prevents Loongson CAM instructions
from being accepted.

@cindex Loongson EXTensions (EXT) instructions generation override
@kindex @code{.set loongson-ext}
@kindex @code{.set noloongson-ext}
The directive @code{.set loongson-ext} makes the assembler accept
instructions from the Loongson EXT from that point on in the assembly.
The @code{.set noloongson-ext} directive prevents Loongson EXT instructions
from being accepted.

@cindex Loongson EXTensions R2 (EXT2) instructions generation override
@kindex @code{.set loongson-ext2}
@kindex @code{.set noloongson-ext2}
The directive @code{.set loongson-ext2} makes the assembler accept
instructions from the Loongson EXT2 from that point on in the assembly.
This directive implies @code{.set loognson-ext}.
The @code{.set noloongson-ext2} directive prevents Loongson EXT2 instructions
from being accepted.

Traditional MIPS assemblers do not support these directives.

@node MIPS Floating-Point
@section Directives to override floating-point options

@cindex Disable floating-point instructions
@kindex @code{.set softfloat}
@kindex @code{.set hardfloat}
The directives @code{.set softfloat} and @code{.set hardfloat} provide
finer control of disabling and enabling float-point instructions.
These directives always override the default (that hard-float
instructions are accepted) or the command-line options
(@samp{-msoft-float} and @samp{-mhard-float}).

@cindex Disable single-precision floating-point operations
@kindex @code{.set singlefloat}
@kindex @code{.set doublefloat}
The directives @code{.set singlefloat} and @code{.set doublefloat}
provide finer control of disabling and enabling double-precision
float-point operations.  These directives always override the default
(that double-precision operations are accepted) or the command-line
options (@samp{-msingle-float} and @samp{-mdouble-float}).

Traditional MIPS assemblers do not support these directives.

@node MIPS Syntax
@section Syntactical considerations for the MIPS assembler
@menu
* MIPS-Chars::                Special Characters
@end menu

@node MIPS-Chars
@subsection Special Characters

@cindex line comment character, MIPS
@cindex MIPS line comment character
The presence of a @samp{#} on a line indicates the start of a comment
that extends to the end of the current line.

If a @samp{#} appears as the first character of a line, the whole line
is treated as a comment, but in this case the line can also be a
logical line number directive (@pxref{Comments}) or a
preprocessor control command (@pxref{Preprocessing}).

@cindex line separator, MIPS
@cindex statement separator, MIPS
@cindex MIPS line separator
The @samp{;} character can be used to separate statements on the same
line.