man/man7/arch.7

.\" Copyright (c) 2016-2017 The FreeBSD Foundation.
.\"
.\" This documentation was created by Ed Maste under sponsorship of
.\" The FreeBSD Foundation.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\"    notice, this list of conditions and the following disclaimer in the
.\"    documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.Dd October 1, 2025
.Dt ARCH 7
.Os
.Sh NAME
.Nm arch
.Nd Architecture-specific details
.Sh DESCRIPTION
Differences between CPU architectures and platforms supported by
.Fx .
.Ss Introduction
This document is a quick reference of key ABI details of
.Fx
architecture ports.
For full details consult the processor-specific ABI supplement
documentation.
.Pp
If not explicitly mentioned, sizes are in bytes.
The architecture details in this document apply to
.Fx 13.0
and later, unless otherwise noted.
.Pp
.Fx
uses a flat address space.
Variables of types
.Vt unsigned long
and
.Vt size_t
have the same representation.
.Pp
In order to maximize compatibility with future pointer integrity mechanisms,
manipulations of pointers as integers should be performed via
.Vt uintptr_t
or
.Vt intptr_t
and no other types as these types are the only integer types where the
C standard guarantees that a pointer may be cast to it and then cast back
to the original type.
On CHERI systems,
.Vt uintptr_t
and
.Vt intptr_t
are defined as
.Vt __uintcap_t
and
.Vt __intcap_t
which represent capabilities that can be manipulated by integer operations.
Pointers should not be cast to
.Vt long ,
.Vt ptrdiff_t ,
or
.Vt size_t
if they will later be cast back to a pointer that is expected to be
dereferencable as they remain bare integer types on all architectures.
.Pp
On some architectures, e.g.,
AIM variants of
.Dv powerpc64 ,
the kernel uses a separate address space.
On other architectures, kernel and a user mode process share a
single address space.
The kernel is located at the highest addresses.
.Pp
On each architecture, the main user mode thread's stack starts near
the highest user address and grows down.
.Pp
.Fx
architecture support varies by release.
This table shows currently supported CPU architectures along with the first
.Fx
release to support each architecture.
.Bl -column -offset indent "Architecture" "Initial Release"
.It Sy Architecture Ta Sy Initial Release
.It aarch64     Ta 11.0
.It aarch64c    Ta 16.0 (planned)
.It amd64       Ta 5.1
.It armv7       Ta 12.0
.It powerpc64   Ta 9.0
.It powerpc64le Ta 13.0
.It riscv64     Ta 12.0
.It riscv64c    Ta 16.0 (planned)
.El
.Pp
Discontinued architectures are shown in the following table.
.Bl -column -offset indent "Architecture" "Initial Release" "Final Release"
.It Sy Architecture Ta Sy Initial Release Ta Sy Final Release
.It alpha       Ta 3.2   Ta 6.4
.It arm         Ta 6.0   Ta 12.4
.It armeb       Ta 8.0   Ta 11.4
.It armv6       Ta 10.0  Ta 14.x
.It ia64        Ta 5.0   Ta 10.4
.It i386        Ta 1.0   Ta 14.x
.It mips        Ta 8.0   Ta 13.5
.It mipsel      Ta 9.0   Ta 13.5
.It mipselhf    Ta 12.0  Ta 13.5
.It mipshf      Ta 12.0  Ta 13.5
.It mipsn32     Ta 9.0   Ta 13.5
.It mips64      Ta 9.0   Ta 13.5
.It mips64el    Ta 9.0   Ta 13.5
.It mips64elhf  Ta 12.0  Ta 13.5
.It mips64hf    Ta 12.0  Ta 13.5
.It pc98        Ta 2.2   Ta 11.4
.It powerpc     Ta 6.0   Ta 14.x
.It powerpcspe  Ta 12.0  Ta 14.x
.It riscv64sf   Ta 12.0  Ta 13.5
.It sparc64     Ta 5.0   Ta 12.4
.El
.Ss Type sizes
All
.Fx
architectures use some variant of the ELF (see
.Xr elf 5 )
.Sy Application Binary Interface
(ABI) for the machine processor.
Supported ABIs can be divided into three main groups:
.Bl -tag -width "Dv L64PC128"
.It Dv ILP32
.Vt int ,
.Vt intptr_t ,
.Vt long ,
and
.Vt void *
types machine representations all have 4-byte size.
.It Dv LP64
.Vt int
type machine representation uses 4 bytes,
while
.Vt intptr_t ,
.Vt long ,
and
.Vt void *
are 8 bytes.
.It Dv L64PC128
.Vt int
type machine representation uses 4 bytes.
.Vt long
type machine representation uses 8 bytes.
.Vt intptr_t
and
.Vt void *
are 16 byte capabilities.
.El
.Pp
Some machines support more than one
.Fx
ABI.
Typically these are 64-bit machines, where the
.Dq native
.Dv LP64
execution environment is accompanied by the
.Dq legacy
.Dv ILP32
environment, which was the historical 32-bit predecessor for 64-bit evolution.
Examples are:
.Bl -column -offset indent "powerpc64" "ILP32 counterpart"
.It Sy LP64        Ta Sy ILP32 counterpart
.It Dv amd64       Ta Dv i386
.It Dv powerpc64   Ta Dv powerpc
.It Dv aarch64     Ta Dv armv7
.El
.Pp
.Dv aarch64
will support execution of
.Dv armv7
binaries if the CPU implements
.Dv AArch32
execution state.
Binaries targeting
.Dv armv6
and earlier are no longer supported by
.Fx .
.Pp
Architectures with 128-bit capabilities support both a
.Dq native
.Dv L64PC128
execution environment and a
.Dv LP64
environment:
.Bl -column -offset indent "aarch64c" "LP64 counterpart"
.It Sy L64PC128    Ta Sy LP64 counterpart
.It Dv aarch64c    Ta Dv aarch64
.It Dv riscv64c    Ta Dv riscv64
.El
.Pp
On all supported architectures:
.Bl -column -offset indent "long long" "Size"
.It Sy Type Ta Sy Size
.It short Ta 2
.It int Ta 4
.It long long Ta 8
.It float Ta 4
.It double Ta 8
.El
.Pp
Integers are represented in two's complement.
Alignment of integer and pointer types is natural, that is,
the address of the variable must be congruent to zero modulo the type size.
The sole exception is that
.Dv i386
requires only 4-byte alignment for 64-bit integers.
.Pp
Machine-dependent type sizes:
.Bl -column -offset indent "Architecture" "long" "void *" "long double" "time_t"
.It Sy Architecture Ta Sy long Ta Sy void * Ta Sy long double Ta Sy time_t
.It aarch64     Ta 8 Ta  8 Ta 16 Ta 8
.It aarch64c    Ta 8 Ta 16 Ta 16 Ta 8
.It amd64       Ta 8 Ta  8 Ta 16 Ta 8
.It armv7       Ta 4 Ta  4 Ta  8 Ta 8
.It i386        Ta 4 Ta  4 Ta 12 Ta 4
.It powerpc     Ta 4 Ta  4 Ta  8 Ta 8
.It powerpcspe  Ta 4 Ta  4 Ta  8 Ta 8
.It powerpc64   Ta 8 Ta  8 Ta  8 Ta 8
.It powerpc64le Ta 8 Ta  8 Ta  8 Ta 8
.It riscv64     Ta 8 Ta  8 Ta 16 Ta 8
.It riscv64c    Ta 8 Ta 16 Ta 16 Ta 8
.El
.Pp
.Sy time_t
is 8 bytes on all supported architectures except i386.
.Ss Endianness and Char Signedness
.Bl -column -offset indent "Architecture" "Endianness" "char Signedness"
.It Sy Architecture Ta Sy Endianness Ta Sy char Signedness
.It aarch64     Ta little Ta unsigned
.It aarch64c    Ta little Ta unsigned
.It amd64       Ta little Ta   signed
.It armv7       Ta little Ta unsigned
.It i386        Ta little Ta   signed
.It powerpc     Ta big    Ta unsigned
.It powerpcspe  Ta big    Ta unsigned
.It powerpc64   Ta big    Ta unsigned
.It powerpc64le Ta little Ta unsigned
.It riscv64     Ta little Ta   signed
.It riscv64c    Ta little Ta   signed
.El
.Ss Page Size
.Bl -column -offset indent "Architecture" "Page Sizes"
.It Sy Architecture Ta Sy Page Sizes
.It aarch64     Ta 4K, 64K, 2M, 1G
.It aarch64c    Ta 4K, 64K, 2M, 1G
.It amd64       Ta 4K, 2M, 1G
.It armv7       Ta 4K, 1M
.It i386        Ta 4K, 2M (PAE), 4M
.It powerpc     Ta 4K
.It powerpcspe  Ta 4K
.It powerpc64   Ta 4K
.It powerpc64le Ta 4K
.It riscv64     Ta 4K, 2M, 1G
.It riscv64c    Ta 4K, 2M, 1G
.El
.Ss User Address Space Layout
.Bl -column -offset indent "riscv64 (Sv48)" "0x0001000000000000" "NNNU"
.It Sy Architecture Ta Sy Maximum Address Ta Sy Address Space Size
.It aarch64        Ta 0x0001000000000000 Ta 256TiB
.It aarch64c       Ta 0x0001000000000000 Ta 256TiB
.It amd64 (LA48)   Ta 0x0000800000000000 Ta 128TiB
.It amd64 (LA57)   Ta 0x0100000000000000 Ta 64PiB
.It armv7          Ta 0xbfc00000         Ta 3GiB
.It i386           Ta 0xffc00000         Ta 4GiB
.It powerpc        Ta 0xfffff000         Ta 4GiB
.It powerpcspe     Ta 0x7ffff000         Ta 2GiB
.It powerpc64      Ta 0x000fffffc0000000 Ta 4PiB
.It powerpc64le    Ta 0x000fffffc0000000 Ta 4PiB
.It riscv64 (Sv39) Ta 0x0000004000000000 Ta 256GiB
.It riscv64c (Sv39) Ta 0x0000004000000000 Ta 256GiB
.It riscv64 (Sv48) Ta 0x0000800000000000 Ta 128TiB
.It riscv64c (Sv48) Ta 0x0000800000000000 Ta 128TiB
.El
.Pp
The layout of a process' address space can be queried via the
.Dv KERN_PROC_VM_LAYOUT
.Xr sysctl 3
MIB.
.Pp
Historically, amd64 CPUs were limited to a 48-bit virtual address space.
Newer CPUs support 5-level page tables, which extend the significant bits of
addresses to 57 bits (LA57 mode).
The address space layout is determined by the CPU's support for LA57.
Setting the
.Sy vm.pmap.la57
tunable to 0 forces the system into 4-level paging mode, even on hardware that
supports 5-level paging.
In this mode, all processes get a 48-bit address space.
The
.Sy vm.pmap.prefer_la48_uva
tunable determines whether processes running on a LA57 system are limited to
a 48-bit address space by default.
Some applications make use of unused upper bits in pointer values to store
information, and thus implicitly assume they are running in LA48 mode.
To avoid breaking compatibility, all processes run in LA48 mode by default.
The
.Xr elfctl 1
utility can be used to request LA48 or LA57 mode for specific executables.
Similarly,
.Xr proccontrol 1
can be used to configure the address space layout when executing a process.
.Pp
The RISC-V specification permits 3-level (Sv39), 4-level (Sv48), and
5-level (Sv57) page tables.
Hardware is only required to implement Sv39; implementations which support
Sv48 must also support Sv39, and implementations which support Sv57 must also
support Sv48.
The
.Sy vm.pmap.mode
tunable can be used to select the layout.
.Fx
currently supports Sv39 and Sv48 and defaults to using Sv39.
.Ss Floating Point
.Bl -column -offset indent "Architecture" "float, double" "long double"
.It Sy Architecture Ta Sy float, double Ta Sy long double
.It aarch64     Ta hard Ta soft, quad precision
.It aarch64c    Ta hard Ta soft, quad precision
.It amd64       Ta hard Ta hard, 80 bit
.It armv7       Ta hard Ta hard, double precision
.It i386        Ta hard Ta hard, 80 bit
.It powerpc     Ta hard Ta hard, double precision
.It powerpcspe  Ta hard Ta hard, double precision
.It powerpc64   Ta hard Ta hard, double precision
.It powerpc64le Ta hard Ta hard, double precision
.It riscv64     Ta hard Ta hard, quad precision
.It riscv64c    Ta hard Ta hard, quad precision
.El
.Ss Default Tool Chain
.Fx
uses
.Xr clang 1
as the default compiler on all supported CPU architectures,
LLVM's
.Xr ld.lld 1
as the default linker, and
LLVM binary utilities such as
.Xr objcopy 1
and
.Xr readelf 1 .
.Ss MACHINE_ARCH vs MACHINE_CPUARCH vs MACHINE
.Dv MACHINE_CPUARCH
should be preferred in Makefiles when the generic
architecture is being tested.
.Dv MACHINE_ARCH
should be preferred when there is something specific to a particular type of
architecture where there is a choice of many, or could be a choice of many.
Use
.Dv MACHINE
when referring to the kernel, interfaces dependent on a specific type of kernel
or similar things like boot sequences.
.Bl -column -offset indent "Dv MACHINE" "Dv MACHINE_CPUARCH" "Dv MACHINE_ARCH"
.It Dv MACHINE Ta Dv MACHINE_CPUARCH Ta Dv MACHINE_ARCH
.It arm64 Ta aarch64 Ta aarch64, aarch64c
.It amd64 Ta amd64 Ta amd64
.It arm Ta arm Ta armv7
.It i386 Ta i386 Ta i386
.It powerpc Ta powerpc Ta powerpc, powerpcspe, powerpc64, powerpc64le
.It riscv Ta riscv Ta riscv64, riscv64c
.El
.Ss Predefined Macros
The compiler provides a number of predefined macros.
Some of these provide architecture-specific details and are explained below.
Other macros, including those required by the language standard, are not
included here.
.Pp
The full set of predefined macros can be obtained with this command:
.Bd -literal -offset indent
cc -x c -dM -E /dev/null
.Ed
.Pp
Common type size and endianness macros:
.Bl -column -offset indent "__SIZEOF_POINTER__" "Meaning"
.It Sy Macro Ta Sy Meaning
.It Dv __SIZEOF_LONG__ Ta size in bytes of long
.It Dv __SIZEOF_POINTER__ Ta size in bytes of intptr_t and pointers
.It Dv __SIZEOF_SIZE_T__ Ta size in bytes of size_t
.It Dv __LP64__ Ta 64-bit (8-byte) long and pointer, 32-bit (4-byte) int
.It Dv __ILP32__ Ta 32-bit (4-byte) int, long and pointer
.It Dv __CHERI__ Ta 128-bit (16-byte) capability pointer, 64-bit (8-byte) long
.It Dv BYTE_ORDER Ta Either Dv BIG_ENDIAN or Dv LITTLE_ENDIAN .
.El
.Pp
Because systems were historically either
.Dv __ILP32__
or
.Dv __LP64__
it has been common for programmers to test only one and assume the other
one in an else branch.
With the arrival of CHERI architectures, this is no longer the case.
.Dv __SIZEOF_*__
macros should be used instead.
New uses of
.Dv __ILP32__
and
.Dv __LP64__
should be avoided.
Compilers for CHERI targets do not define
.Dv __LP64__
as their pointers are 128-bit capabilities.
.Pp
Architecture-specific macros:
.Bl -column -offset indent "Architecture" "Predefined macros"
.It Sy Architecture Ta Sy Predefined macros
.It aarch64     Ta Dv __aarch64__
.It aarch64c    Ta Dv __aarch64__ , Dv __CHERI__
.It amd64       Ta Dv __amd64__ , Dv __x86_64__
.It armv7       Ta Dv __arm__ , Dv __ARM_ARCH >= 7
.It i386        Ta Dv __i386__
.It powerpc     Ta Dv __powerpc__
.It powerpcspe  Ta Dv __powerpc__ , Dv __SPE__
.It powerpc64   Ta Dv __powerpc__ , Dv __powerpc64__
.It powerpc64le Ta Dv __powerpc__ , Dv __powerpc64__
.It riscv64     Ta Dv __riscv , Dv __riscv_xlen == 64
.It riscv64c    Ta Dv __riscv , Dv __riscv_xlen == 64 , Dv __CHERI__
.El
.Pp
Compilers may define additional variants of architecture-specific macros.
The macros above are preferred for use in
.Fx .
.Ss Important Xr make 1 variables
Most of the externally settable variables are defined in the
.Xr build 7
man page.
These variables are not otherwise documented and are used extensively
in the build system.
.Bl -tag -width "MACHINE_CPUARCH"
.It Dv MACHINE
Represents the hardware platform.
This is the same as the native platform's
.Xr uname 1
.Fl m
output.
It defines both the userland / kernel interface, as well as the
bootloader / kernel interface.
It should only be used in these contexts.
Each CPU architecture may have multiple hardware platforms it supports
where
.Dv MACHINE
differs among them.
It is used to collect together all the files from
.Xr config 8
to build the kernel.
It is often the same as
.Dv MACHINE_ARCH
just as one CPU architecture can be implemented by many different
hardware platforms, one hardware platform may support multiple CPU
architecture family members, though with different binaries.
For example,
.Dv MACHINE
of i386 supported the IBM-AT hardware platform while the
.Dv MACHINE
of pc98 supported the Japanese company NEC's PC-9801 and PC-9821
hardware platforms.
Both of these hardware platforms supported only the
.Dv MACHINE_ARCH
of i386 where they shared a common ABI, except for certain kernel /
userland interfaces relating to underlying hardware platform
differences in bus architecture, device enumeration and boot interface.
Generally,
.Dv MACHINE
should only be used in src/sys and src/stand or in system imagers or
installers.
.It Dv MACHINE_ARCH
Represents the CPU processor architecture.
This is the same as the native platforms
.Xr uname 1
.Fl p
output.
It defines the CPU instruction family supported.
It may also encode a variation in the byte ordering of multi-byte
integers (endian).
It may also encode a variation in the size of the integer or pointer.
It may also encode a ISA revision.
It may also encode hard versus soft floating point ABI and usage.
It may also encode a variant ABI when the other factors do not
uniquely define the ABI.
It, along with
.Dv MACHINE ,
defines the ABI used by the system.
Generally, the plain CPU name specifies the most common (or at least
first) variant of the CPU.
This is why powerpc and powerpc64 imply 'big endian' while armv7 and aarch64
imply little endian.
If we ever were to support the so-called x32 ABI (using 32-bit
pointers on the amd64 architecture), it would most likely be encoded
as amd64-x32.
It is unfortunate that amd64 specifies the 64-bit evolution of the x86 platform
(it matches the 'first rule') as almost everybody else uses x86_64.
The
.Fx
port was so early, it predated processor name standardization after Intel joined
the market.
At the time, each OS selected its own conventions.
Backwards compatibility means it is not easy to change to the consensus name.
.It Dv MACHINE_CPUARCH
Represents the source location for a given
.Dv MACHINE_ARCH .
It is generally the common prefix for all the MACHINE_ARCH that
share the same implementation, though 'riscv' breaks this rule.
While amd64 and i386 are closely related, MACHINE_CPUARCH is not x86
for them.
The
.Fx
source base supports amd64 and i386 with two
distinct source bases living in subdirectories named amd64 and i386
(though behind the scenes there's some sharing that fits into this
framework).
.It Dv CPUTYPE
Sets the flavor of
.Dv MACHINE_ARCH
to build.
It is used to optimize the build for a specific CPU / core that the
binaries run on.
Generally, this does not change the ABI, though it can be a fine line
between optimization for specific cases.
.It Dv TARGET
Used to set
.Dv MACHINE
in the top level Makefile for cross building.
Unused outside of that scope.
It is not passed down to the rest of the build.
Makefiles outside of the top level should not use it at all (though
some have their own private copy for historical reasons).
.It Dv TARGET_ARCH
Used to set
.Dv MACHINE_ARCH
by the top level Makefile for cross building.
Like
.Dv TARGET ,
it is unused outside of that scope.
.El
.Sh SEE ALSO
.Xr elfctl 1 ,
.Xr proccontrol 1 ,
.Xr sysctl 3 ,
.Xr src.conf 5 ,
.Xr build 7 ,
.Xr simd 7
.Sh HISTORY
An
.Nm
manual page appeared in
.Fx 11.1 .