Differences between revisions 13 and 14
Revision 13 as of 2012-03-30 16:09:12
Size: 6138
Comment:
Revision 14 as of 2012-04-11 15:07:46
Size: 8938
Comment: Expanded details of triplets, linked to external specs where possible
Deletions are marked like this. Additions are marked like this.
Line 24: Line 24:
 ||<rowbgcolor="maroon" rowstyle="color: white; font-weight: bold;"style="text-align: center">multiarch name ||<style="text-align: center">syscall ABI ||<style="text-align: center">instruction set ||<style="text-align: center">endianness ||<style="text-align: center">wordsize ||<style="text-align: center">status ||<style="text-align: center">description ||
 || alpha-linux-gnu || linux || Alpha || little || 64 || port || foo bar ... ||
 || arm-linux-gnueabi || linux || ARM || little || 32 || release || [[ArmEabiPort]] ||
 || arm-linux-gnueabihf || linux || ARM || little || 32 || release || [[ArmHardFloatPort]] ||
 || hppa-linux-gnu || linux || PA-RISC || big || 64 || port || foo bar ... ||
 || i386-gnu || Hurd || IA-32 || little || 32 || port || foo bar ... ||
 || i386-linux-gnu || linux || IA-32 || little || 32 || release || foo bar ... ||
 || ia64-linux-gnu || linux || IA-64 || little || 64 || release || foo bar ... ||
 || amd64-kfreebsd-gnu || linux || x86_64 || little || 32 || release || foo bar ... ||
 || i386-kfreebsd-gnu || linux || IA-32 || little || 32 || release || foo bar ... ||
 || m68k-linux-gnu || linux || MC68000 || big || 32 || port || currently, 68020; gcc multilib could do 68040 and 68060 but is disabled as there is no infrastructure to actually make use out of it; Coldfire-MMU port's status is unclear (the decision on whether it'll share this multiarch tuple will probably be deferred until it's clear how exactly m68k and CF relate to each other, i.e. how much code sharing is done, whether it's one or two full or a full and an incomplete port, etc.) ||
 || mips-linux-gnu || linux || MIPS || big || 32 || release || foo bar ... ||
 || mipsel-linux-gnu || linux || MIPS || little || 32 || release || foo bar ... ||
 || mips-linux-gnuabin32 || linux || MIPS || big || 64 || biarch || glibc, MIPS n64 ABI ||
 || mipsel-linux-gnuabin32 || linux || MIPS || little || 64 || biarch || glibc, MIPS n64 ABI ||
 || mips64-linux-gnuabi64 || linux || MIPS64 || big || 64 || biarch || foo bar ... ||
 || mips64el-linux-gnuabi64 || linux || MIPS64 || little || 64 || biarch || foo bar ... ||
 || powerpc-linux-gnu || linux || Power || big || 32 || release || glibc, SYSV hard-float ABI (biarch is ppc64) ||
 || powerpc-linux-gnuspe || linux || Power || big || 32 || port || glibc, SYSV soft-float ABI, uses e500v2 SPE FPU ||
 || ppc64-linux-gnu || linux || Power || big || 64 || biarch || glibc, EABI, partial biarch of "powerpc" ||
 || s390-linux-gnu || linux || System/390 || big || 32 || release || foo bar ... ||
 || s390x-linux-gnu || linux || z/Architecture || big || 32 || port || foo bar ... ||
 || sparc-linux-gnu || linux || SPARC || big || 32 || release || foo bar ... ||
 || sparc64-linux-gnu || linux || SPARC || big || 64 || port || foo bar ... ||
 || x86_64-linux-gnu || linux || x86_64 || little || 64 || release || foo bar ... ||
 ||<rowbgcolor="maroon" rowstyle="color: white; font-weight: bold;"style="text-align: center">multiarch name ||<style="text-align: center">syscall ABI ||<style="text-align: center">instruction set ||<style="text-align: center">endianness ||<style="text-align: center">wordsize ||<style="text-align: center">status in Debian ||<style="text-align: center">description || <style="text-align: center">spec document(s) ||
 || alpha-linux-gnu || linux || Alpha || little || 64 || EOL || glibc, Tru64 calling standard || http://www.openwatcom.org/ftp/devel/docs/alpha%20calling%20standard.pdf ||
 || arm-linux-gnu || linux || ARM || little || 32 || EOL || Old ARM ABI (APCS). HW FPU expected with kernel emulation as fallback. DO NOT USE || http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0041c/BGBGFIDA.html||
 || arm-linux-gnueabi || linux || ARMv4t+ || little || 32 || released || New ARM EABI (AAPCS), soft-float || http://infocenter.arm.com/help/topic/com.arm.doc.ihi0042a/IHI0042A_aapcs.pdf and https://sourcery.mentor.com/GNUToolchain/kbentry39 ||
 || arm-linux-gnueabihf || linux || ARMv7+ || little || 32 || development || New ARM EABI (AAPCS), hard-float || http://infocenter.arm.com/help/topic/com.arm.doc.ihi0042a/IHI0042A_aapcs.pdf and '''XXXXX (TBD)''' ||
 || hppa-linux-gnu || linux || PA-RISC || big || 64 || EOL || 32/64-bit kernel, 32-bit userland || http://wiki.parisc-linux.org/ToolChain ||
 || i386-gnu || Hurd || IA-32 || little || 32 || development || http://wiki.debian.org/Debian_GNU/Hurd || '''XXXXX''' none I could find ||
 || i386-linux-gnu || linux || IA-32 || little || 32 || released || Intel ia32 psABI || http://www.sco.com/developers/devspecs/ and http://refspecs.linuxbase.org/elf/abi386-4.pdf ||
 || ia64-linux-gnu || linux || IA-64 || little || 64 || released || Intel ia64 psABI || http://refspecs.linuxbase.org/elf/IA64-SysV-psABI.pdf ||
 || amd64-kfreebsd-gnu || FreeBSD || x86_64 || little || 64 || released || http://wiki.debian.org/Debian_GNU/kFreeBSD || '''XXXXX''' none I could find ||
 || i386-kfreebsd-gnu || FreeBSD || IA-32 || little || 32 || released || http://wiki.debian.org/Debian_GNU/kFreeBSD || '''XXXXX''' none I could find ||
 || m68k-linux-gnu || linux || MC68000 || big || 32 || EOL || currently, 68020; gcc multilib could do 68040 and 68060 but is disabled as there is no infrastructure to actually make use out of it; Coldfire-MMU port's status is unclear (the decision on whether it'll share this multiarch tuple will probably be deferred until it's clear how exactly m68k and CF relate to each other, i.e. how much code sharing is done, whether it's one or two full or a full and an incomplete port, etc.) || '''XXXXX''' none I could find ||
 || mips-linux-gnu || linux || MIPS || big || 32 || released || MIPS o32 ABI || http://refspecs.linuxbase.org/elf/mipsabi.pdf ||
 || mipsel-linux-gnu || linux || MIPS || little || 32 || release || MIPS o32 ABI || http://refspecs.linuxbase.org/elf/mipsabi.pdf ||
 || mips-linux-gnuabin32 || linux || MIPS || big || 64 || none yet || glibc, MIPS n32 ABI (32-bit pointers) || http://techpubs.sgi.com/library/manuals/2000/007-2816-005/pdf/007-2816-005.pdf ||
 || mipsel-linux-gnuabin32 || linux || MIPS || little || 64 || none yet || glibc, MIPS n32 ABI (32-bit pointers) || http://techpubs.sgi.com/library/manuals/2000/007-2816-005/pdf/007-2816-005.pdf ||
 || mips64-linux-gnuabi64 || linux || MIPS64 || big || 64 || none yet || glibc, MIPS n32 ABI (64-bit pointers) || http://techpubs.sgi.com/library/manuals/2000/007-2816-005/pdf/007-2816-005.pdf ||
 || mips64el-linux-gnuabi64|| linux || MIPS64 || little || 64 || none yet || glibc, MIPS n32 ABI (64-bit pointers) || http://techpubs.sgi.com/library/manuals/2000/007-2816-005/pdf/007-2816-005.pdf ||
 || powerpc-linux-gnu || linux || PowerPC || big || 32 || released || glibc, SYSV hard-float ABI || http://refspecs.freestandards.org/LSB_2.1.0/LSB-Core-PPC32/index.shtml ||
 || powerpc-linux-gnuspe || linux || PowerPC || big || 32 || development || glibc, SYSV soft-float ABI, uses e500v2 SPE FPU, limited hw options http://wiki.debian.org/PowerPCSPEPort || '''XXXXX''' none I could find ||
 || ppc64-linux-gnu || linux || PowerPC || big || 64 || development || glibc, EABI || http://refspecs.freestandards.org/LSB_2.1.0/LSB-Core-PPC64/index.shtml ||
 || s390-linux-gnu || linux || System/390 || big || 32 || released || ELF, Linux syscalls, glibc 2.x. || http://refspecs.linuxbase.org/ELF/zSeries/lzsabi0_s390.html ||
 || s390x-linux-gnu || linux || z/Architecture || big || 64 || development || ELF, Linux syscalls, glibc 2.x. || http://refspecs.linuxbase.org/ELF/zSeries/lzsabi0_zSeries.html ||
 || sparc-linux-gnu || linux || SPARC || big || 32 || released || ELF, Linux syscalls, glibc 2.x. || http://www.sparc.com/standards/psABI3rd.pdf ||
 || sparc64-linux-gnu || linux || SPARCv9+ || big || 64 || development || ELF, Linux syscalls, glibc 2.x. || Initial version at http://www.sparc.com/standards/64.psabi.1.35.ps.Z, can't find anything more concrete ||
 || x86_64-linux-gnu || linux || x86_64 || little || 64 || released || ELF, Linux syscalls, glibc 2.x. || http://x86-64.org/documentation/abi.pdf ||

Multiarch Architecture Specifiers (Tuples)

This document describes the new schema for representing architecture ABIs as directory names, using normalized GNU triplets.

The abandoned proposal can be found at Multiarch/TuplesAbandoned.

This page is part of the larger effort to implement Multiarch library handling on Linux. For more information about Multiarch, see Multiarch, https://wiki.ubuntu.com/MultiarchSpec.

The problem

Multiarch specifies a method of making libraries for multiple, mutually-incompatible architectures installable on a single filesystem in a manner that ensures the same binaries can be used without modification on any system. To accomplish this, we require unique identifiers for each architecture that identifies an incompatible set of libraries that we want to be co-installed. We want these identifiers to be specified in as vendor-neutral a manner as possible, to ensure we retain maximal binary compatibility across Linux distributions.

Why not use GNU triplets?

Earlier proposals for multiarch did use GNU triplets as the components of the library path. Proof-of-concept implementations ran into two distinct cases where GNU triplets could not be used effectively in a cross-distribution manner:

  • On IA-32, the GNU triplet varies according to the precise instruction set being targeted: e.g., i486-linux-gnu, i586-linux-gnu i686-linux-gnu. Such triplets will be inconsistent over time within a single distribution as compiler defaults are changed, let alone between distributions.
  • On ARM, the canonical GNU triplet for EABI systems, whether using hard-float or soft-float, is arm-linux-gnueabi, despite the fact that hard-float and soft-float libraries use different calling conventions and can't be intermixed. The advice given by upstream GCC developers was to use the vendor field in the GNU triplet to express this difference1; however, the vendor field is private by design, making this unreliable for cross-distribution use. Work is in progress (March 2012) to push the arm-linux-gnueabihf triplet upstream.

Used solution

Use normalized GNU triplets for most tuples, and extend the GNU triplet when it is ambiguous.

The status describes if the name is used in a (to be) released version of Debian and/or Ubuntu (release), if it is a new or abandoned port on Debian Ports (port), or if it's used as a name for a Debian gcc multilib configuration (biarch).

Supporting interfaces

  • dpkg-architecture -qDEB_HOST_MULTIARCH (on dpkg based systems).
  • gcc -print-multiarch (proposed patch; currently only available in Debian wheezy (unreleased), and in Ubuntu 11.10 (oneiric) or newer.
  • To avoid divergent embedded implementations of architecture->tuple mappings in every piece of software that wants to make use of these paths, a standard commandline tool that encapsulates this mapping should be provided, similar to the config.guess and config.sub tools that are a standard part of GNU autotools. We recommend a name of lsb_architecture for this tool.