DPS8M Simulator
DPS8M Homepage DPS8M Releases DPS8M Documentation Blog Mailing Lists Slack DPS8M RSS Notifications DPS8M GitLab Repository Toggle Theme Toggle Theme Toggle Theme

Source Compilation

The simulator is distributed in various forms, including an easy-to-build source code distribution, which can be built simply via make on most systems.

The following sections document only some platform-specific details necessary to build the simulator, and are not intended to be a general reference or to replace the full documentation.

Review the complete DPS8M Omnibus Documentation for additional details.


General Information

  • Expert users may wish to build the simulator from source code to enable experimental features, or for reasons of performance and compatibility.

  • Building on a 64-bit platform is strongly encouraged for optimal performance.

  • The DPS8M Development Team recommends most users download a source kit distribution.
    • A source kit requires approximately 20 MiB of storage space to decompress and 40 MiB to build.

  • Advanced users may prefer to clone the git repository which contains additional tools not required for simulator operation, but useful to developers.
    • The git repository requires approximately 275 MiB of storage space to clone and 300 MiB to build.

The following sections document only platform-specific variations, and are not intended to be a general or exhaustive reference.



FreeBSD

FreeBSD prerequisites

Install the required prerequisites (using FreeBSD Packages or Ports):

  • Using FreeBSD Packages (as root):

    pkg install gmake libuv

  • Using FreeBSD Ports (as root):

    cd /usr/ports/devel/gmake/ && make install clean
cd /usr/ports/devel/libuv/ && make install clean

Standard FreeBSD compilation

  • Build the simulator (standard build) from the top-level source directory (using GNU Make):

    gmake

Optimized FreeBSD compilation

  • FreeBSD provides the Clang compiler as part of the base system. While sufficient to build the simulator, we recommend that version 12 or later of the GNU C (gcc) compiler be used for optimal performance.

  • At the time of writing, GCC 13.2 is available for FreeBSD systems and is the version of GCC currently recommended by The DPS8M Development Team.

    It can be installed via FreeBSD Packages or Ports:

    • Using FreeBSD Packages (as root):

      pkg install gcc13

    • Using FreeBSD Ports (as root):

      cd /usr/ports/lang/gcc13/ && make install clean

  • Build the simulator from the top-level source directory (using GNU Make):

    env CC="gcc13" LDFLAGS="-Wl,-rpath=/usr/local/lib/gcc13" gmake

blinkenLights2 on FreeBSD

  • To build the blinkenLights2 front-panel display, install it’s prerequisites via FreeBSD Packages or Ports:

    • Using FreeBSD Packages (as root):

      pkg install pkgconf gtk3

    • Using FreeBSD Ports (as root):

      cd /usr/ports/devel/pkgconf/ && make install clean
cd /usr/ports/x11-toolkits/gtk30/ && make install clean

  • Build blinkenLights2 from the top-level source directory (using GNU Make):

    gmake blinkenLights2

Additional FreeBSD Notes

  • When running on FreeBSD, DPS8M utilizes FreeBSD-specific atomic operations.

  • The FreeBSD-specific atomic_testandset_64 operation is currently not implemented in all versions of FreeBSD or on all platforms FreeBSD supports (e.g. powerpc64, or arm64 prior to 13.0-RELEASE).

    If you are unable to build the simulator because this atomic operation is unimplemented on your platform, specify ATOMICS="GNU" as an argument to gmake, or export this value in the shell environment before compiling.



NetBSD

  • DPS8M is fully supported on NetBSD 9.2 or later.

NetBSD prerequisites

Install the required prerequisites (using NetBSD Packages or pkgsrc):

  • Using NetBSD Packages (as root):

    pkgin in gmake libuv

  • Using pkgsrc (as root):

    cd /usr/pkgsrc/devel/gmake/ && make install clean
cd /usr/pkgsrc/devel/libuv/ && make install clean

Standard NetBSD compilation

  • Build the simulator (standard build) from the top-level source directory (using GNU Make):

    gmake

Optimized NetBSD compilation

  • NetBSD provides an older version of GCC (or Clang) as part of the base system (depending on the platform). While sufficient to build the simulator, we recommend that version 12 or later of the GNU C (gcc) compiler be used for optimal performance.

  • At the time of writing, GCC 13.2 is available for NetBSD 9 systems and is the version of GCC currently recommended by The DPS8M Development Team.

    It can be installed via Packages or pkgsrc.

    • Using NetBSD Packages (as root):

      pkgin in gcc13

    • Using pkgsrc (as root):

      cd /usr/pkgsrc/lang/gcc13/ && make install clean

  • Build the simulator from the top-level source directory (using GNU Make):

    env CC="/usr/pkg/gcc13/bin/gcc" LDFLAGS="Wl,-rpath=/usr/pkg/gcc13/lib" gmake

Compilation using Clang

  • GCC is recommended for optimal performance, but compilation using Clang (and linking using LLD, the LLVM linker) is supported.

    Both Clang and LLD can be installed via Packages or pkgsrc.

    • Using NetBSD Packages (as root):

      pkgin in clang lld

    • Using pkgsrc (as root):

      cd /usr/pkgsrc/lang/clang/ && make install clean
cd /usr/pkgsrc/devel/lld/ && make install clean

  • Build the simulator from the top-level source directory (using GNU Make):

    env CC="clang" \
  LDFLAGS="-L/usr/lib -L/usr/pkg/lib -fuse-ld=\"$(command -v ld.lld)\"" gmake

blinkenLights2 on NetBSD

  • To build the blinkenLights2 front-panel display, install it’s prerequisites via NetBSD Packages or pkgsrc:

    • Using NetBSD Packages (as root):

      pkg install pkgconf gtk3+

    • Using pkgsrc (as root):

      cd /usr/pkgsrc/devel/pkgconf/ && make install clean
cd /usr/pkgsrc/x11/gtk3/ && make install clean

  • Build blinkenLights2 from the top-level source directory (using GNU Make):

    gmake blinkenLights2


OpenBSD

  • The following instructions were verified using OpenBSD 7.5 on amd64 and arm64.

OpenBSD prerequisites

Install the required prerequisites (using OpenBSD Packages or Ports):

  • Using OpenBSD Packages (as root):

    pkg_add gmake libuv

  • Using OpenBSD Ports (as root):

    cd /usr/ports/devel/gmake/ && make install clean
cd /usr/ports/devel/libuv/ && make install clean

Standard OpenBSD compilation

  • Build the simulator (standard build) from the top-level source directory (using GNU Make):

    gmake

Optimized OpenBSD compilation

  • OpenBSD provides an older version of GCC (or Clang) as part of the base system (depending on the platform). While sufficient to build the simulator, we recommend that a recent version of the GNU assembler (gas) and version 11 or later of the GNU C (gcc) compiler be used for optimal performance.

  • At the time of writing, appropriate versions of the GNU assembler and GNU C (version 11.2) are available for OpenBSD. (In addition, LLD, the LLVM linker, may be required.) These tools have been tested and are highly recommended by The DPS8M Development Team.

    They can be installed via OpenBSD Packages or Ports:

    • Using OpenBSD Packages (as root):

      pkg_add gas gcc

    • Using OpenBSD Ports (as root):

      cd /usr/ports/devel/gas/ && make install clean
cd /usr/ports/lang/gcc/11/ && make install clean

  • LLVM LLD 16.0.6 or later is recommended for linking, even when using GCC 11.2 for compilation. LLD is the default linker on most (but not all) supported OpenBSD platforms. To determine the linker in use, examine the output of ‘ld --version’.

    If the linker identifies itself by a name other than LLD (e.g. “GNU ld” or similar), LLD must be installed via OpenBSD Packages or Ports.

    • Using OpenBSD Packages (as root):

      pkg_add llvm

    • Using OpenBSD Ports (as root):

      cd /usr/ports/devel/llvm/ && make install clean

  • Configure GCC to execute the correct assembler and/or linker:

    mkdir -p ~/.override
test -x /usr/local/bin/gas && ln -fs /usr/local/bin/gas ~/.override/as
test -x /usr/local/bin/lld && ln -fs /usr/local/bin/lld ~/.override/ld

  • Build the simulator from the top-level source directory (using GNU Make):

    env CC="egcc" CFLAGS="-B ~/.override" gmake

Compilation using Clang

  • GCC is recommended for optimal performance, but compilation using Clang is supported.
  • A version of Clang newer than the base system version may be available via the ‘llvm’ package or port.
  • Once installed, it can be used for compilation by setting the appropriate environment variables before invoking the ‘gmake’ program (i.e. ‘CC="/usr/local/bin/clang"’ and ‘LDFLAGS="-fuse-ld=lld"’).

Additional OpenBSD Notes

  • OpenBSD/luna88k is not supported.


DragonFly BSD

  • At the time of writing, DragonFly BSD 6.4.0 was current and used to verify the following instructions.

DragonFly BSD prerequisites

  • Install the required prerequisites using DragonFly BSD DPorts (as root):

    pkg install gmake libuv

Standard DragonFly BSD compilation

  • Build the simulator (standard build) from the top-level source directory (using GNU Make):

    env CFLAGS="-I/usr/local/include"  \
     LDFLAGS="-L/usr/local/lib"    \
     ATOMICS="GNU"                 \
  gmake

Optimized DragonFly BSD compilation

  • DragonFly BSD provides an older version of GCC as part of the base system. While this compiler is sufficient to build the simulator, we recommend that version 12 or later of the GNU C (gcc) compiler be used for optimal performance.

  • At the time of writing, GCC 13.1 is available for DragonFly BSD and recommended by The DPS8M Development Team.

    • GCC 13.1 may be installed using DragonFly BSD DPorts (as root):

      pkg install gcc13

  • Build the simulator from the top-level source directory (using GNU Make):

    env CC="gcc12" CFLAGS="-I/usr/local/include"                    \
    LDFLAGS="-L/usr/local/lib -Wl,-rpath=/usr/local/lib/gcc13"  \
    ATOMICS="GNU"                                               \
  gmake

Compiling using Clang

  • GCC is recommended for optimal performance, but compilation using Clang is supported.

  • At the time of writing, Clang 16 is available for DragonFly BSD and recommended by The DPS8M Development Team.

  • While some optional utilities may fail to build using Clang on DragonFly, the simulator (src/dps8/dps8) is fully tested with each DragonFly release.

    • Clang 16 may be installed using DragonFly BSD DPorts (as root):

      pkg install llvm16

  • Build the simulator from the top-level source directory (using GNU Make):

    env CC="clang16" CFLAGS="-I/usr/include -I/usr/local/include"  \
    LDFLAGS="-L/usr/lib -L/usr/local/lib -fuse-ld=lld"         \
    ATOMICS="GNU"                                              \
  gmake


Solaris

  • Ensure your Solaris installation is reasonably current.

  • The simulator can be built on Solaris using the GCC, Clang, and Oracle Developer Studio compilers.

  • GCC 11 is the recommended compiler for optimal performance on all Intel-based Solaris systems.
    • GCC 11 can be installed from the standard IPS repository via ‘pkg install gcc-11’.
    • Link-time optimization (LTO) is supported only when building with GCC version 10 or later.
    • The NO_LTO=1 build option should be specified when using earlier versions of the GCC compiler.

  • Building with Clang 11 (or later) is also supported (but not recommended due to lack of LTO support).
    • Clang 11 can be installed from the standard IPS repository, i.e. ‘pkg install clang@11 llvm@11’.

  • Building with the Oracle Developer Studio 12.6 (suncc) compiler is also supported.
    • Note that building for Solaris using the Oracle Developer Studio compiler currently requires a non-trivial amount of CFLAGS to be specified. This will be simplified in a future release of the simulator.

Solaris prerequisites

  • Install the required prerequisites from the standard IPS repository (as root):

    pkg install gnu-make gnu-binutils gnu-sed gnu-grep gnu-tar gawk \
  gnu-coreutils pkg-config libtool autoconf automake wget

Solaris compilation

The following commands will download and build a local static libuv before compiling the simulator.

If a site-provided libuv library has been installed (in the “/usr/local” prefix), the libuvrel stage of the build may be omitted.

Build libuv and the simulator from the top-level source directory (using GNU Make):

GCC

  • Build using GCC:

    • Build libuv:

      env TAR="gtar" TR="gtr" CC="gcc" gmake libuvrel

    • Build the simulator:

      env TR="gtr" CC="gcc" gmake

Clang

  • Build using Clang:

    • Build libuv:

      env TAR="gtar" NO_LTO=1 TR="gtr" CC="clang" gmake libuvrel

    • Build the simulator:

      env NO_LTO=1 TR="gtr" CC="clang" gmake

Oracle Developer Studio

  • Build using Oracle Developer Studio 12.6:

    • Build libuv:

      env TAR="gtar" NO_LTO=1 SUNPRO=1 NEED_128=1 TR="gtr" CSTD="c11"        \
    CFLAGS="-DNO_C_ELLIPSIS -Qy -xO5 -m64 -xlibmil -xCC -mt -xlibmopt  \
    -fno-semantic-interposition -xprefetch=auto -xprefetch_level=3"    \
    CC="/opt/developerstudio12.6/bin/suncc"                            \
  gmake libuvrel

    • Build the simulator:

      env NO_LTO=1 SUNPRO=1 NEED_128=1 TR="gtr" CSTD="c11"                   \
    CFLAGS="-DNO_C_ELLIPSIS -Qy -xO5 -m64 -xlibmil -xCC -mt -xlibmopt  \
    -fno-semantic-interposition -xprefetch=auto -xprefetch_level=3"    \
    CC="/opt/developerstudio12.6/bin/suncc"                            \
  gmake


OpenIndiana

  • Ensure your OpenIndiana installation is up-to-date.
    • OpenIndiana Hipster April 2024 was used to verify these instructions.

  • GCC 13.2 is currently the recommended compiler for optimal performance.
    • GCC 13.2 can be installed from the standard IPS repository via ‘pkg install developer/gcc-13’.
    • Link-time optimization (LTO) is supported only when building with GCC version 10 or later.
    • The NO_LTO=1 build option should be specified when using earlier versions of the GCC compiler.

  • Building with Clang (version 13 or later) is also supported (but not recommended due to lack of LTO support).
    • Clang 18 can be installed from the standard IPS repository via ‘pkg install developer/clang-18’.

OpenIndiana prerequisites

  • Install the required prerequisites from the standard IPS repository (as root):

    pkg install gnu-make libuv gnu-binutils gnu-coreutils

Standard OpenIndiana compilation

  • Build the simulator from the top-level source directory (using GNU Make and GCC 13):

    env CC="gcc-13" gmake

Compiling using Clang

  • Build the simulator from the top-level source directory (using GNU Make and Clang 18):

    env NO_LTO=1 CC="clang-18" gmake


AIX

  • The DPS8M Development Team recommends building with IBM Open XL C/C++ V17.1.2 (or later) or mainline Clang 18 (or later) for optimal performance. Building with GNU C is supported (but not recommended) when using GCC 10 (or later).

  • Clang 18.1.8 is the minimum recommended version of the Clang compiler on POWER8, POWER9, and Power10 systems.
    • LTO (link-time optimization) and native 128-bit integer operations are both fully supported.

Other supported compilers

  • IBM Open XL C/C++ for AIX V17.1.1 Fix Pack 6 (February 2024) is the minimum recommended version of the Open XL C/C++ V17.1.1 compiler on POWER8, POWER9, and Power10 systems.
    • LTO (link-time optimization) is supported, but 128-bit integer operations are not supported.
      • Use of the NEED_128=1 option is required when building with Open XL C/C++ V17.1.1.

  • IBM XL C/C++ for AIX V16.1.0 Fix Pack 18 (May 2024) is the minimum recommended version of the IBM XL C/C++ V16.1.0 compiler on POWER8 and POWER9 systems.
    • LTO (link-time optimization) and 128-bit integer operations are both not supported.
      • Use of the NEED_128=1 and NO_LTO=1 options are required when building with IBM XL C/C++ V16.1.0.
    • IBM XL C/C++ V16.1.0 is not recommended for Power10 systems.

  • GNU C 11.3.0 is the minimum recommended version of the GNU C compiler on POWER8, POWER9, and Power10 systems.
    • LTO (link-time optimization) and 128-bit integer operations are both not supported.
      • Use of the NEED_128=1 and NO_LTO=1 options are required when building with GNU C.
    • GCC 11 and GCC 12 can be installed from the IBM AIX Toolbox for Open Source Software repository.

  • Note that building fully-featured binaries for IBM AIX currently requires a non-trivial number of options to be specified after the gmake command, which overrides various build defaults appropriate for Linux, macOS, and BSD systems, but not IBM AIX. This will be simplified in a future release of the simulator.

AIX prerequisites

Libraries and tools

GNU C compilers

  • Install GCC 11.3:

    /opt/freeware/bin/dnf install gcc gcc11

  • Install GCC 12.3:

    /opt/freeware/bin/dnf install gcc gcc12

Clang compilers

IBM compiler support

  • IBM does not provide support for any version of GNU C through IBM AIX support cases; we recommend IBM customers with support contracts use Open XL C/C++ if possible.

AIX compilation

Build the simulator from the top-level source directory (using GNU Make):

IBM Open XL C/C++ for AIX

  • Using IBM Open XL C/C++ for AIX V17.1.2:

    env PATH="/opt/freeware/bin:${PATH}"                                      \
    CC="/opt/IBM/openxlC/17.1.2/bin/ibm-clang_r"                          \
    ATOMICS="AIX"                                                         \
    AWK="gawk"                                                            \
    OBJECT_MODE=64                                                        \
  gmake PULIBS="-lpopt"                                                   \
        LDFLAGS="-L/opt/freeware/lib -L/usr/local/lib -flto=auto -b64"    \
        LIBS="-lpthread -luv -lbsd -lm"                                   \
        CFLAGS="-flto=auto -I/opt/freeware/include -I/usr/local/include   \
                -I../simh -I../decNumber -DUSE_FLOCK=1 -DUSE_FCNTL=1      \
                -DHAVE_POPT=1 -DAIX_ATOMICS=1 -m64                        \
                -DLOCKLESS=1 -D_ALL_SOURCE -D_GNU_SOURCE -O3              \
                -U__STRICT_POSIX__ -fno-strict-aliasing -mcpu=power8"

    • When building on IBM POWER9 (or Power10) systems, ‘-mcpu=power9’ (or ‘-mcpu=power10’) should replace ‘-mcpu=power8’ in the above compiler invocation.

    • Refer to the IBM Open XL C/C++ for AIX V17.1.2 documentation for additional information.

  • Using IBM Open XL C/C++ for AIX V17.1.1:

    env PATH="/opt/freeware/bin:${PATH}"                                      \
    CC="/opt/IBM/openxlC/17.1.1/bin/ibm-clang_r"                          \
    ATOMICS="AIX"                                                         \
    AWK="gawk"                                                            \
    OBJECT_MODE=64                                                        \
    NEED_128=1                                                            \
  gmake PULIBS="-lpopt"                                                   \
        LDFLAGS="-L/opt/freeware/lib -L/usr/local/lib -flto=auto -b64"    \
        LIBS="-lpthread -luv -lbsd -lm"                                   \
        CFLAGS="-flto=auto -I/opt/freeware/include -I/usr/local/include   \
                -I../simh -I../decNumber -DUSE_FLOCK=1 -DUSE_FCNTL=1      \
                -DHAVE_POPT=1 -DNEED_128=1 -DAIX_ATOMICS=1 -m64           \
                -DLOCKLESS=1 -D_ALL_SOURCE -D_GNU_SOURCE -O3              \
                -U__STRICT_POSIX__ -fno-strict-aliasing -mcpu=power8"

    • When building on IBM POWER9 (or Power10) systems, ‘-mcpu=power9’ (or ‘-mcpu=power10’) should replace ‘-mcpu=power8’ in the above compiler invocation.

    • Refer to the IBM Open XL C/C++ for AIX V17.1.1 documentation for additional information.

Clang

  • Using mainline Clang 18.1.8:

    env PATH="/opt/llvm/bin:/opt/freeware/bin:${PATH}"                        \
    CC="/opt/llvm/bin/clang"                                              \
    ATOMICS="AIX"                                                         \
    AWK="gawk"                                                            \
    OBJECT_MODE=64                                                        \
  gmake PULIBS="-lpopt"                                                   \
        LDFLAGS="-L/opt/freeware/lib -L/usr/local/lib -flto=auto -b64"    \
        LIBS="-lpthread -luv -lbsd -lm"                                   \
        CFLAGS="-flto=auto -I/opt/freeware/include -I/usr/local/include   \
                -I../simh -I../decNumber -DUSE_FLOCK=1 -DUSE_FCNTL=1      \
                -DHAVE_POPT=1 -DAIX_ATOMICS=1 -m64                        \
                -DLOCKLESS=1 -D_ALL_SOURCE -D_GNU_SOURCE -O3              \
                -U__STRICT_POSIX__ -fno-strict-aliasing -mcpu=power8"

    • When building on IBM POWER9 (or Power10) systems, ‘-mcpu=power9’ (or ‘-mcpu=power10’) should replace ‘-mcpu=power8’ in the above compiler invocation.

    • Refer to the Clang documentation for additional information.

IBM XL C/C++ for AIX

  • Using IBM XL C/C++ for AIX V16.1.0:

    env PATH="/opt/freeware/bin:${PATH}"                                      \
    ATOMICS="AIX"                                                         \
    AWK="gawk"                                                            \
    NO_LTO=1                                                              \
    OBJECT_MODE=64                                                        \
  gmake CC="/opt/IBM/xlC/16.1.0/bin/xlc_r"                                \
        NEED_128=1                                                        \
        USE_POPT=1                                                        \
        PULIBS="-lpopt"                                                   \
        LDFLAGS="-L/opt/freeware/lib -L/usr/local/lib -b64"               \
        LIBS="-luv -lbsd -lm"                                             \
        CFLAGS="-O3 -qhot -qarch=pwr8 -qalign=natural -qtls -DUSE_POPT=1  \
                -DUSE_FLOCK=1 -DUSE_FCNTL=1 -DAIX_ATOMICS=1 -DNEED_128=1  \
                -DLOCKLESS=1 -I/opt/freeware/include -I../simh            \
                -I../decNumber -I/usr/local/include -D_GNU_SOURCE         \
                -D_ALL_SOURCE -U__STRICT_POSIX__"

    • When building on POWER9 systems, ‘-qarch=pwr9’ should replace ‘-qarch=pwr8’ in the above compiler invocation.

    • Compilation using higher optimization levels (i.e. ‘-O4’ or ‘-O5’ replacing ‘-O3 -qhot -qarch=pwr8’) and/or enabling automatic parallelization (i.e. ‘-qsmp’) may be possible, but the resulting binaries have not been benchmarked or extensively tested by The DPS8M Development Team.

    • Refer to the IBM XL C/C++ for AIX V16.1 Optimization and Tuning Guide for additional information.

GCC

  • Using GCC 11.3:

    env PATH="/opt/freeware/bin:${PATH}" CC="gcc-11" \
  ATOMICS="AIX" NO_LTO=1 gmake

  • Using GCC 12.3:

    env PATH="/opt/freeware/bin:${PATH}" CC="gcc-12" \
  ATOMICS="AIX" NO_LTO=1 gmake


Haiku

  • Ensure you are running a recent release of Haiku on a supported 64-bit platform.
    • Use the ‘SoftwareUpdater’ application to ensure your Haiku installation is up-to-date.
  • The DPS8M Development Team regularly tests the simulator using the nightly Haiku x86_64 snapshots.
    • Haiku x86_64 (hrev57700) was used to verify the following instructions.

Haiku prerequisites

The default Haiku installation includes the required header files, the recommended compiler (at the time of writing, GCC 13.2), and most of the necessary development utilities (i.e. GNU Make) required to build DPS8M. The remaining prerequisites are available via the standard package management tools.

  • Install the required prerequisites (from HaikuPorts using pkgman or the ‘HaikuDepot’ application):
    • libuv
    • libuv_devel
    • getconf

Standard Haiku compilation

  • Build the simulator (standard build) from the top-level source directory (using GNU Make):

    make

Compiling using Clang

  • Building with GCC is strongly recommended for optimal performance, but compilation using Clang is also supported (although not recommended, due to the lack of support for LTO optimization).

    • At the time of writing, Clang 17 is available from HaikuPorts (as the ‘llvm17_clang’ and ‘llvm17_lld’ packages), installable using pkgman or the ‘HaikuDepot’ application.

  • Build the simulator from the top-level source directory (using GNU Make):

    env CSTD="c11" CC="clang" CFLAGS="-fPIC" NO_LTO=1 make

Additional Haiku Notes

  • DPS8M on 32-bit Haiku platforms (i.e. x86, x86_gcc2) is not supported at this time.


GNU/Hurd

  • GCC 11 (or later) is the recommended compiler for optimal performance.
    • Compilation is also supported using Clang 11 or later.

  • Building on GNU/Hurd should be essentially the same as building on Debian GNU/Linux.


Linux

Linux compilers

  • GCC 12 or later is recommended for optimal performance on most architectures including Intel and ARM.
    • The DPS8M Development Team regularly tests and supports a wide range of Linux compilers, including Clang, AMD Optimizing C/C++ (AOCC), Arm C/C++ Compiler (ARMClang), GNU C (GCC) (version 9+), IBM Advance Toolchain for Linux, IBM XL C/C++ for Linux (XLC), IBM Open XL C/C++ for Linux (IBMClang), Intel oneAPI DPC++/C++ (ICX), NVIDIA HPC SDK C Compiler (NVC), and Oracle Developer Studio (SunCC).

  • Red Hat offers the Red Hat Developer Toolset for Red Hat Enterprise Linux and CentOS Stream, which provides up-to-date versions of GCC on a rapid release cycle, with full support.
    • The Toolset packages are also included in various “clone” distributions such as AlmaLinux. These tools are regularly tested and highly recommended by The DPS8M Development Team. Check your distribution packager manager (i.e. ‘dnf search’) for packages named ‘gcc-toolset-12’, ‘gcc-toolset-13’, or similar.

  • Canonical similarly offers two Ubuntu Toolchain PPAs, one providing GCC updates for release branches, and the other providing new GCC versions for both current and LTS releases, maintained by the Ubuntu Toolchain team.
    • For example, at the time of writing, Ubuntu 22.04 LTS ships GCC 11.3 and GCC 12.1, and the Toolchain PPAs ship GCC 12.3 and GCC 13.1. Although these packages are not formally supported by Canonical, they are regularly and successfully used by The DPS8M Development Team.
    • NOTE: Ubuntu has shipped multiple versions of the LLVM and Clang packages that are broken in various ways (incorrect library paths, missing libraries, broken LTO support, etc.). If you encounter problems building from source code using Clang on Ubuntu, try installing different package versions, wait for an update, or use GCC to build instead.

  • Intel® C++ Compiler Classic (ICC) for Linux is no longer supported for building DPS8M (as of R3.0.0):

  • Cross-compilation is supported. Popular targets including various Linux platforms, Microsoft Windows on Intel and ARM (using the MinGW-w64 and LLVM-MinGW toolchains) and Linux on POWER (using the IBM Advance Toolchain for Linux) are regularly built and tested.

Linux prerequisites

Users of some Red Hat variants may need to enable the PowerTools repository or the CodeReady Builder AppStream to install libuv:

  • RHEL 8:

    subscription-manager repos --enable \
  "codeready-builder-for-rhel-8-$(arch)-rpms"

  • CentOS Stream 8:

    dnf config-manager --set-enabled "powertools"

  • RHEL 9:

    subscription-manager repos --enable \
  "codeready-builder-for-rhel-9-$(arch)-rpms"

  • CentOS Stream 9:

    dnf config-manager --set-enabled "crb"

Install the required prerequisites using a distribution package manager:

  • Using dnf (for most rpm-based distributions) (as root):

    dnf install "@Development Tools" libuv-devel

  • Using apt (for most deb-based distributions) (as root):

    apt install build-essential libuv1-dev

Standard Linux compilation

  • Build the simulator (standard build) from the top-level source directory (using GNU Make):

    make

Alternative Linux compilation

To use a compiler other than the default it is usually sufficient to simply set the CC environment variable (if the compiler accepts command-line arguments compatible with GCC or Clang). Other compilers are supported as well, but require additional configuration.

Examples of building the simulator on Linux using various popular compilers follows:

Clang

  • Build the simulator using Clang (Clang 15 or later recommended):

    env CC="clang" make

Intel oneAPI DPC++/C++

  • Build the simulator using Intel oneAPI DPC++/C++ (ICX):

    source /opt/intel/oneapi/setvars.sh && \
env CC="icx" make

AMD Optimizing C/C++

  • Build the simulator using AMD Optimizing C/C++ (AOCC), version 4.2.0, (with AOCC-provided AMD LibM):

    export AOCCVER="4.2.0" &&                               \
export AOCLPATH="/opt/AMD/aocc-compiler-${AOCCVER}" &&  \
source ${AOCCPATH}/setenv_AOCC.sh &&                    \
env CC="clang" CFLAGS="-mllvm -vector-library=AMDLIBM"  \
    LDFLAGS="-Wno-unused-command-line-argument"         \
    LOCALLIBS="-lalm"                                   \
  make
AOCC with AMD Optimized CPU Libraries

  • Build the simulator using AMD Optimizing C/C++ (AOCC), version 4.2.0, with AMD Optimized CPU Libraries (AOCL) (AMD AOCL-LibM and AMD AOCL-LibMem), version 4.2.0:

    export AOCCVER="4.2.0" &&                                               \
export AOCCPATH="/opt/AMD/aocc-compiler-${AOCCVER}" &&                  \
export AOCLVER="4.2.0" &&                                               \
export AOCLPATH="/opt/AMD/aocl/aocl-linux-aocc-${AOCLVER}" &&           \
export LD_LIBRARY_PATH="${AOCLPATH}/aocc/lib:${LD_LIBRARY_PATH}" &&     \
source ${AOCCPATH}/setenv_AOCC.sh &&                                    \
env CC="clang" CFLAGS="-mllvm -vector-library=AMDLIBM"                  \
    LDFLAGS="-Wno-unused-command-line-argument -L${AOCLPATH}/aocc/lib"  \
    LOCALLIBS="-lalm -laocl-libmem"                                     \
  make

Oracle Developer Studio

  • Build the simulator using Oracle Developer Studio (SunCC) for Linux, version 12.6:

    env CFLAGS="-DNO_C_ELLIPSIS -Qy -xO5 -m64 -xlibmil -xCC -mt  \
            -xlibmopt -fno-semantic-interposition            \
            -xprefetch=auto -xprefetch_level=3"              \
    CC="/opt/oracle/developerstudio12.6/bin/suncc"           \
    NO_LTO=1 SUNPRO=1 NEED_128=1 CSTD="c11"                  \
  make

IBM Open XL C/C++ for Linux

  • Build the simulator using IBM Open XL C/C++ for Linux V17.1.1 Fix Pack 2 (February 2024) for Linux on POWER:

    env CFLAGS="-mcpu=power8"                         \
    CC="/opt/ibm/openxlC/17.1.1/bin/ibm-clang_r"  \
  make

  • When building on IBM POWER9 (or Power10) systems, ‘-mcpu=power9’ (or ‘-mcpu=power10’) should replace ‘-mcpu=power8’ in the above compiler invocation.

  • Refer to the IBM Open XL C/C++ for Linux V17.1.1 documentation for additional information.

IBM XL C/C++ for Linux

  • Build the simulator using IBM XL C/C++ for Linux V16.1.1 Fix Pack 15 (September 2023) for Linux on POWER:

    env CFLAGS="-qtls -qarch=pwr8"          \
    CC="/opt/ibm/xlC/16.1.1/bin/c99_r"  \
    CSTD="c11" NO_LTO=1                 \
  make

    • When building on POWER9 or higher systems, ‘-qarch=pwr9’ should replace ‘-qarch=pwr8’ in the above compiler invocation.

    • Compilation using higher optimization levels (e.g. ‘-O4’, ‘-O5’, ‘-qhot’, etc.) and/or enabling automatic parallelization (i.e. ‘-qsmp’) may be possible, but the resulting binaries have not been benchmarked or tested by The DPS8M Development Team.

NVIDIA HPC SDK C Compiler

  • Build the simulator using NVIDIA HPC SDK C Compiler (NVC), version 24.3, for Linux/x86_64 (also available for Linux/ARM64 and Linux/OpenPOWER):

    export NVCVER="24.3" &&                                                    \
export NVPLAT="Linux_x86_64" &&                                            \
export NVCPATH="/opt/nvidia/hpc_sdk/${NVPLAT}/${NVCVER}/compilers/bin" &&  \
env CFLAGS="-noswitcherror --diag_suppress=mixed_enum_type                 \
                           --diag_suppress=branch_past_initialization      \
                           --diag_suppress=set_but_not_used"               \
    CC="${NVCPATH}/nvc" NO_LTO=1                                           \
  make OPTFLAGS="-fast -O4 -Mnofprelaxed"

    • DPS8M is known to trigger bugs in some versions of the NVIDIA HPC SDK C Compiler, such as:

      NVC++-F-0000-Internal compiler error. add_cilis(): bad jmp code 1056

      If you encounter this (or similar) compiler errors, try adding ‘-Mnovect’ to ‘OPTFLAGS’ as a workaround.

Arm HPC C/C++ Compiler for Linux

The Arm HPC C/C++ Compiler for Linux with Arm Performance Libraries (also available as a component of Arm Allinea Studio) provides a packaged Clang/LLVM-based toolchain with optimized math and string libraries, validated against common ARM HPC platforms.

Note the following examples do not make use of Environment Modules and/or Lmod, commonly used to manage compiler and development tool installations in HPC environments.

If your site uses modules (i.e. module avail), loading the appropriate module is usually preferred to setting paths manually. Contact your system administrator for site-specific configuration details and recommended local compiler flags.

  • Build the simulator using the Arm HPC C/C++ Compiler for Linux (ARMClang), version 24.04, for Linux/ARM64:

    export ACFLVER="24.04" &&                                 \
export ACFLCMP="arm-linux-compiler-${ACFLVER}" &&         \
export ACFLTYP="Generic-AArch64_RHEL-9_aarch64-linux" &&  \
export ACFLPATH="/opt/arm/${ACFLCMP}_${ACFLTYP}" &&       \
export PATH="${ACFLPATH}/bin:${PATH}" &&                  \
env CFLAGS="-mcpu=native"                                 \
    CC="armclang"                                         \
  make
ACFL with Arm Performance Libraries

  • Build the simulator using the Arm HPC C/C++ Compiler for Linux (ARMClang) with the integrated Arm Performance Libraries (ArmPL), version 24.04, for Linux/ARM64:

    export ACFLVER="24.04" &&                                 \
export ACFLCMP="arm-linux-compiler-${ACFLVER}" &&         \
export ACFLTYP="Generic-AArch64_RHEL-9_aarch64-linux" &&  \
export ACFLPATH="/opt/arm/${ACFLCMP}_${ACFLTYP}" &&       \
export PATH="${ACFLPATH}:${PATH}" &&                      \
env CFLAGS="-mcpu=native -armpl"                          \
    LDFLAGS="-armpl"                                      \
    CC="armclang"                                         \
  make OPTFLAGS="-Ofast"

  • Build the simulator using the Arm HPC C/C++ Compiler for Linux (ARMClang) with the integrated Arm Performance Libraries (ArmPL), version 24.04, for Linux/ARMv8-A+SVE2 (Scalable Vector Extensions):

    export ACFLVER="24.04" &&                                 \
export ACFLCMP="arm-linux-compiler-${ACFLVER}" &&         \
export ACFLTYP="Generic-AArch64_RHEL-9_aarch64-linux" &&  \
export ACFLPATH="/opt/arm/${ACFLCMP}_${ACFLTYP}" &&       \
export PATH="${ACFLPATH}:${PATH}" &&                      \
env CFLAGS="-march=armv8-a+sve2 -mcpu=native -armpl=sve"  \
    LDFLAGS="-armpl=sve"                                  \
    CC="armclang"                                         \
  make OPTFLAGS="-Ofast"

Linux cross-compilation

The following commands will download and cross-compile a local static libuv and then cross-compile the simulator.

IBM Advance Toolchain

  • Using the IBM Advance Toolchain V17 to cross-compile for Linux/POWER:

    • Build libuv:

      env CC="/opt/at17.0/bin/powerpc64le-linux-gnu-gcc"  \
    LOCAL_CONFOPTS="--host=powerpc64le-linux-gnu"   \
    CFLAGS="-mcpu=power8 -mtune=power8"             \
  make libuvrel

    • Build the simulator:

      env CC="/opt/at17.0/bin/powerpc64le-linux-gnu-gcc"  \
    CFLAGS="-mcpu=power8 -mtune=power8"             \
  make

    • When targeting POWER9 or Power10 processors, ‘power9’ or ‘power10’ should replace ‘power8’ in the above compiler invocation.

    • The IBM Advance Toolchain versions 14, 15, 16, and 17 have been extensively tested and used for cross-compiling DPS8M.

Arm GNU Toolchain

The GNU Toolchain for the Arm Architecture (referred to as the “Arm GNU Toolchain”) provides regularly updated, high-quality, validated Linux/ARM cross-compilers running on Microsoft Windows, Linux, and macOS.

Linux/ARMv7-HF

  • Using the Arm GNU Toolchain on Linux/x86_64, version 13.2.Rel1, to cross-compile for Linux/ARMv7-HF (hardware floating point):

    • Build libuv:

      export AGTREL="arm-gnu-toolchain-13.2.rel1-x86_64" &&             \
export AGTPATH="/opt/${AGTREL}-arm-none-linux-gnueabihf/bin/" &&  \
env CC="${AGTPATH}/arm-none-linux-gnueabihf-gcc"                  \
    CFLAGS="-march=armv7-a+fp"                                    \
    LOCAL_CONFOPTS="--host=arm-none-linux-gnueabihf"              \
  make libuvrel

    • Build the simulator:

      env CC="${AGTPATH}/arm-none-linux-gnueabihf-gcc"                  \
    CFLAGS="-march=armv7-a+fp"                                    \
     NEED_128=1                                                   \
  make
Linux/ARM64

  • Using the Arm GNU Toolchain on Linux/x86_64, version 13.2.Rel1, to cross-compile for Linux/ARM64:

    • Build libuv:

      export AGTREL="arm-gnu-toolchain-13.2.rel1-x86_64" &&             \
export AGTPATH="/opt/${AGTREL}-aarch64-none-linux-gnu/bin/" &&    \
env CC="${AGTPATH}/aarch64-none-linux-gnu-gcc"                    \
    LOCAL_CONFOPTS="--host=aarch64-none-linux-gnu"                \
  make libuvrel

    • Build the simulator:

      env CC="${AGTPATH}/aarch64-none-linux-gnu-gcc" \
  make
Linux/ARM64BE

  • Using the Arm GNU Toolchain on Linux/x86_64, version 13.2.Rel1, to cross-compile for Linux/ARM64BE (big endian):

    • Build libuv:

      export AGTREL="arm-gnu-toolchain-13.2.rel1-x86_64" &&                \
export AGTPATH="/opt/${AGTREL}-aarch64_be-none-linux-gnu/bin/" &&    \
env CC="${AGTPATH}/aarch64_be-none-linux-gnu-gcc"                    \
    LOCAL_CONFOPTS="--host=aarch64_be-none-linux-gnu"                \
  make libuvrel

    • Build the simulator:

      env CC="${AGTPATH}/aarch64_be-none-linux-gnu-gcc" \
  make

Linaro GNU Toolchain

The Linaro GNU Toolchain Integration Builds provides Linux/ARM and Linux/ARM64 reference toolchains, closely tracking upstream repositories, allowing developers to easily test new compiler and processor features before the next Arm GNU Toolchain release.

Linux/ARMv7-HF

  • Using the Linaro GNU Toolchain Integration Build on Linux/x86_64, version 14.0.0-2023.06, to cross-compile for Linux/ARMv7-HF (hardware floating point):

    • Build libuv:

      export LINREL="gcc-linaro-14.0.0-2023.06-x86_64" &&            \
export LINPATH="/opt/${LINREL}_arm-linux-gnueabihf/bin/" &&    \
env CC="${LINPATH}/arm-linux-gnueabihf-gcc"                    \
    CFLAGS="-march=armv7-a+fp"                                 \
    LOCAL_CONFOPTS="--host=arm-linux-gnueabihf"                \
  make libuvrel

    • Build the simulator:

      env CC="${LINPATH}/arm-linux-gnueabihf-gcc"                    \
    CFLAGS="-march=armv7-a+fp"                                 \
    NEED_128=1                                                 \
  make
Linux/ARM64

  • Using the Linaro GNU Toolchain Integration Build on Linux/x86_64, version 14.0.0-2023.06, to cross-compile for Linux/ARM64:

    • Build libuv:

      export LINREL="gcc-linaro-14.0.0-2023.06-x86_64" &&          \
export LINPATH="/opt/${LINREL}_aarch64-linux-gnu/bin/" &&    \
env CC="${LINPATH}/aarch64-linux-gnu-gcc"                    \
    LOCAL_CONFOPTS="--host=aarch64-linux-gnu"                \
  make libuvrel

    • Build the simulator:

      env CC="${LINPATH}/aarch64-linux-gnu-gcc" \
  make

crosstool-NG

crosstool-NG is a versatile cross-toolchain generator, which can be used to generate GCC-based toolchains for a huge variety of architectures and operating systems (mainly Linux).

DPS8M is regularly built by The DPS8M Development Team for many Linux architectures using crosstool-NG generated toolchains, utilizing both the glibc and musl C libraries. The following CT-NG examples are intended to be instructive, but are by no means exhaustive.

Linux/RV64

  • Using a crosstool-NG generated GCC+musl toolchain to cross-compile for Linux/RV64 (64-bit RISC-V static binary):

    • Build libuv:

      export CTNG="riscv64-local-linux-musl" &&               \
env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    LOCAL_CONFOPTS="--host=${CTNG}"                     \
  make libuvrel

    • Build the simulator:

      env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    LDFLAGS="-static"                                   \
    LOCALLIBS="-latomic"                                \
  make
Linux/i686

  • Using a crosstool-NG generated GCC+musl toolchain to cross-compile for Linux/i686 (32-bit static binary):

    • Build libuv:

      export CTNG="i686-local-linux-musl" &&                  \
env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    LOCAL_CONFOPTS="--host=${CTNG}"                     \
  make libuvrel

    • Build the simulator:

      env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    LDFLAGS="-static"                                   \
    LOCALLIBS="-latomic"                                \
    NEED_128=1                                          \
  make
Linux/ARMv6-HF

  • Using a crosstool-NG generated GCC+glibc toolchain to cross-compile for Linux/ARMv6-HF (hardware floating point):

    • Build libuv:

      export CTNG="armv6-local-linux-gnueabihf" &&            \
env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    LOCAL_CONFOPTS="--host=${CTNG}"                     \
    CFLAGS="-march=armv6+fp"                            \
  make libuvrel

    • Build the simulator:

      env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    CFLAGS="-march=armv6+fp"                            \
    LOCALLIBS="-latomic"                                \
    NEED_128=1                                          \
  make
Linux/PPC64le

  • Using a crosstool-NG generated GCC+musl toolchain to cross-compile for Linux/PPC64le (64-bit POWER9 little endian static binary):

    • Build libuv:

      export CTNG="powerpc64le-local-linux-musl" &&           \
env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    LOCAL_CONFOPTS="--host=${CTNG}"                     \
    CFLAGS="-mcpu=power9"                               \
  make libuvrel

    • Build the simulator:

      env CC="/home/user/x-tools/${CTNG}/bin/${CTNG}-gcc"     \
    CFLAGS="-mcpu=power9"                               \
    LDFLAGS="-static"                                   \
    LOCALLIBS="-latomic"                                \
  make

Additional Linux Notes

  • Although normally handled automatically, when building on (or cross-compiling to) some 32-bit targets using a compiler lacking support for 128-bit integer types, it may be necessary to explicitly set the NEED_128=1 build option via the environment or as an argument to make.


macOS

  • Ensure you are running a supported release of macOS with current updates applied.
    • Both Intel and ARM64 systems are regularly tested by The DPS8M Development Team.

  • Xcode is required; it is strongly recommended to use the most recent release for optimal performance.
    • Intel® C++ Compiler Classic for macOS (icc) is no longer supported for building DPS8M (as of R3.0.2).
      • Intel® has retired support for ICC.
    • Building the simulator on macOS using GCC is not recommended.
  • The following instructions were verified using macOS 14.4 with Xcode 15.4 (Apple Clang 15.0.0).

macOS prerequisites

  • Homebrew is the recommended package manager for installing build prerequisites:

    brew install libuv pkg-config

  • Installation and configuration of Homebrew and its prerequisites (i.e. Xcode CLT) is outside the scope of the DPS8M documentation. Refer to the Homebrew Installation Documentation if support is required.

  • Users of other package managers (e.g. pkgsrc, MacPorts) must set the CFLAGS (e.g. ‘-I/opt/include’), LDFLAGS (e.g. ‘-L/opt/lib’), and LIBUV (e.g. ‘-luv’) environment variables appropriately.

macOS compilation

Build the simulator from the top-level source directory (using GNU Make):

  • Build using Xcode:

    make

macOS cross-compilation

The following commands will download and cross-compile a local static libuv and then cross-compile the simulator using Xcode.

You must perform a ‘make distclean’ before building for a different target.

  • Install required prerequisites (using Homebrew):

    brew install wget pkg-config autoconf automake libtool coreutils

Build the simulator from the top-level source directory (using GNU Make):

ARM64

  • Cross-compilation targeting ARM64 macOS 12:

    • Build libuv:

      make distclean &&                                \
env CFLAGS="-target arm64-apple-macos12          \
            -mmacosx-version-min=12.0"           \
    LOCAL_CONFOPTS="--host=arm64-apple-darwin"   \
  make libuvrel HOMEBREW_INC= HOMEBREW_LIB=

    • Build the simulator:

      env CFLAGS="-target arm64-apple-macos12          \
            -mmacosx-version-min=12.0"           \
    LDFLAGS="-target arm64-apple-macos12         \
             -mmacosx-version-min=12.0"          \
  make HOMEBREW_INC= HOMEBREW_LIB=

Intel

  • Cross-compilation targeting Intel macOS 10.15:

    • Build libuv:

      make distclean &&                                \
env CFLAGS="-target x86_64-apple-macos10.15      \
            -mmacosx-version-min=10.15"          \
    LOCAL_CONFOPTS="--host=x86_64-apple-darwin"  \
  make libuvrel HOMEBREW_INC= HOMEBREW_LIB=

    • Build the simulator:

      env CFLAGS="-target x86_64-apple-macos10.15      \
            -mmacosx-version-min=10.15"          \
    LDFLAGS="-target x86_64-apple-macos10.15     \
             -mmacosx-version-min=10.15"         \
  make HOMEBREW_INC= HOMEBREW_LIB=

Universal

  • The following more complex example builds a macOS Universal Binary.

    • The Universal Binary will support three architectures: ARM64, Intel, and Intel Haswell.
    • The simulator (and libuv) will be cross-compiled three times each, once for each architecture.
    • The lipo utility will be used to create the universal dps8 binary (in the top-level build directory).

  • Cross-compilation targeting ARM64, Intel, Intel Haswell (AVX2):

    • Build ARM64 libuv:

      make distclean &&                                \
env CFLAGS="-target arm64-apple-macos11          \
            -mmacosx-version-min=11.0"           \
    LOCAL_CONFOPTS="--host=arm64-apple-darwin"   \
  make libuvrel HOMEBREW_INC= HOMEBREW_LIB=

    • Build ARM64 dps8:

      env CFLAGS="-target arm64-apple-macos11          \
            -mmacosx-version-min=11.0"           \
    LDFLAGS="-target arm64-apple-macos11         \
             -mmacosx-version-min=11.0"          \
  make HOMEBREW_INC= HOMEBREW_LIB= &&            \
cp -f "src/dps8/dps8" "dps8.arm64"

    • Build Intel libuv:

      make distclean &&                                \
env CFLAGS="-target x86_64-apple-macos10.15      \
            -mmacosx-version-min=10.15"          \
    LOCAL_CONFOPTS="--host=x86_64-apple-darwin"  \
  make libuvrel HOMEBREW_INC= HOMEBREW_LIB=

    • Build Intel dps8:

      env CFLAGS="-target x86_64-apple-macos10.15      \
            -mmacosx-version-min=10.15"          \
    LDFLAGS="-target x86_64-apple-macos10.15     \
             -mmacosx-version-min=10.15"         \
  make HOMEBREW_INC= HOMEBREW_LIB= &&            \
cp -f "src/dps8/dps8" "dps8.x86_64"

    • Build Intel Haswell libuv:

      make distclean &&                                \
env CFLAGS="-target x86_64h-apple-macos10.15     \
            -mmacosx-version-min=10.15           \
            -march=haswell"                      \
    LOCAL_CONFOPTS="--host=x86_64-apple-darwin"  \
  make libuvrel HOMEBREW_INC= HOMEBREW_LIB=

    • Build Intel Haswell dps8:

      env CFLAGS="-target x86_64h-apple-macos10.15     \
            -mmacosx-version-min=10.15           \
            -march=haswell"                      \
    LDFLAGS="-target x86_64h-apple-macos10.15    \
            -mmacosx-version-min=10.15"          \
  make HOMEBREW_INC= HOMEBREW_LIB= &&            \
cp -f "src/dps8/dps8" "dps8.x86_64h"

    • Create the Universal Binary using lipo:

      lipo -create -output "dps8"                      \
  "dps8.x86_64" "dps8.x86_64h" "dps8.arm64" &&   \
make distclean && rm -f                          \
  "dps8.x86_64" "dps8.x86_64h" "dps8.arm64" &&   \
lipo "dps8"


Windows

  • Ensure you are running a supported release of Microsoft Windows on a supported platform.
    • Microsoft Windows 10 and 11 on x86_64 and i686 are regularly tested by The DPS8M Development Team.

  • Microsoft Windows supports various development and runtime environments, including Universal CRT, MinGW-w64, Cygwin, Midipix, MSYS2, and many others.
    • Care must be taken to avoid mixing incompatible libraries and tools.

  • Cross-compilation is supported. Builds targeting Microsoft Windows (MinGW and Cygwin) running on x86_64, i686, ARMv7, and ARM64 platforms are regularly cross-compiled from a variety of UNIX-like systems (using LLVM-MinGW and MinGW-GCC), and from Microsoft Windows using Cygwin.

  • Microsoft Windows also provides Linux compatibility via the Windows Subsystem for Linux (WSL).
    • Windows Subsystem for Linux users should refer to the Linux sections of the documentation.

Cygwin

  • Ensure you are running a current and up-to-date version of Cygwin.

  • Only the 64-bit version of Cygwin is regularly tested by The DPS8M Development Team.
    • Although the 32-bit version of Cygwin is not regularly tested (and not recommended due to suboptimal performance), it should work for building DPS8M (with the NEED_128=1 build option).

Cygwin prerequisites

  • Compilation problems in the Cygwin environment are often caused by incomplete or interrupted package installations, or by the installation of packages using non-standard tools (e.g. apt-cyg), resulting in missing files and dangling or missing symbolic links.

    • Before attempting to build DPS8M using Cygwin:
      1. First, update the Cygwin setup.exe application to the latest available version.
      2. Update all installed packages using the new Cygwin setup.exe application.
      3. Install the required prerequisite packages using Cygwin setup.exe:
        • autoconf
        • cmake
        • cygcheck
        • gcc
        • libtool
        • libuv1
        • libuv1-devel
        • make
        • pkg-config
        • unzip
        • wget
      4. Most importantly, invoke the cygcheck utility (i.e. cygcheck -cv) to verify the integrity of all currently installed packages and correct any problems before continuing.

Standard Cygwin compilation

  • Build the simulator from the top-level source directory (using GNU Make):

    make

Cygwin-hosted cross-compilation to MinGW

The following commands will download and cross-compile a local native libuv library and then cross-compile the simulator.

You must perform a ‘make distclean’ followed by an ‘rm -rf ${HOME}/libuv-build’ before building for a different target (or changing build flags).

In the following cross-compilation examples, the latest libuv sources (from the v1.x git branch) are used, but the current official release (available from https://libuv.org/) can also be used.

Windows i686

  • Using GCC (the Cygwin mingw64-i686-gcc-core package) to cross-compile a native 32-bit Windows executable (not depending on Cygwin):

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                \
mkdir -p "${HOME}/libuv-win32-i686" &&                           \
( cd "${HOME}/libuv-build" &&                                    \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&   \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" &&                      \
  mkdir -p "build" && cd "build" &&                              \
  cmake .. -DCMAKE_SYSTEM_NAME="Windows"                         \
           -DCMAKE_SYSTEM_VERSION="6.1"                          \
           -DCMAKE_C_COMPILER="i686-w64-mingw32-gcc"             \
           -DCMAKE_INSTALL_PREFIX="${HOME}/libuv-win32-i686" &&  \
  cmake --build . && cmake --install . )

    • Build the simulator:

      env CFLAGS="-I${HOME}/libuv-win32-i686/include -D__MINGW64__     \
            -DUSE_FLOCK=1 -DUSE_FCNTL=1"                         \
    CC="i686-w64-mingw32-gcc"                                    \
    LDFLAGS="-L${HOME}/libuv-win32-i686/lib" NEED_128=1          \
  make CROSS="MINGW64"

    • The compiled native binary will require libwinpthread-1.dll (located at /usr/i686-w64-mingw32/sys-root/mingw/bin/libwinpthread-1.dll) and libuv.dll (located at ${HOME}/libuv-win32-i686/bin/libuv.dll) at runtime.

      • It is sufficient to copy these files into the directory containing the dps8.exe binary.
Windows x86_64

  • Using GCC (the Cygwin mingw64-x86_64-gcc-core package) to cross-compile a native 64-bit Windows executable (not depending on Cygwin):

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                  \
mkdir -p "${HOME}/libuv-win32-x86_64" &&                           \
( cd "${HOME}/libuv-build" &&                                      \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&     \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" &&                        \
  mkdir -p "build" && cd "build" &&                                \
  cmake .. -DCMAKE_SYSTEM_NAME="Windows"                           \
           -DCMAKE_SYSTEM_VERSION="6.1"                            \
           -DCMAKE_C_COMPILER="x86_64-w64-mingw32-gcc"             \
           -DCMAKE_INSTALL_PREFIX="${HOME}/libuv-win32-x86_64" &&  \
  cmake --build . && cmake --install . )

    • Build the simulator:

      env CFLAGS="-I${HOME}/libuv-win32-x86_64/include -D__MINGW64__     \
            -DUSE_FLOCK=1 -DUSE_FCNTL=1"                           \
    CC="x86_64-w64-mingw32-gcc"                                    \
    LDFLAGS="-L${HOME}/libuv-win32-x86_64/lib"                     \
  make CROSS="MINGW64"

    • The compiled native binary will require libwinpthread-1.dll (located at /usr/x86_64-w64-mingw32/sys-root/mingw/bin/libwinpthread-1.dll) and libuv.dll (located at ${HOME}/libuv-win32-x86_64/bin/libuv.dll) at runtime.

      • It is sufficient to copy these files into the directory containing the dps8.exe binary.

MSYS2

  • DPS8M can be built as a native MSYS2 application without special configuration, using the “MSYS2 Environment”.

  • Cross-compilation using the MSYS2-provided MINGW32, MINGW64, UCRT64, CLANG32, CLANG64, and CLANGARM64 environments is currently untested.

Unix-hosted LLVM-MinGW Clang cross-compilation

The LLVM-MinGW Clang toolchain supports building native Windows binaries (i686, x86_64, ARMv7, and ARM64 systems) on non-Windows host systems (or using the Windows Subsystem for Linux).

The LLVM-MinGW Docker Container provides pre-built and fully configured LLVM-MinGW toolchains (including appropriate compiler symlinks) which are regularly used by The DPS8M Development Team.

In the following cross-compilation examples, the latest libuv sources (from the v1.x git branch) are used, but the current official release (available from https://libuv.org/) can also be used.

Windows i686

  • Using Clang (the LLVM-MinGW compiler) to cross-compile a local static libuv library and a native 32-bit Windows/i686 executable:

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                  \
mkdir -p "${HOME}/libuv-win32-i686" &&                             \
( cd "${HOME}/libuv-build" &&                                      \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&     \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" && sh ./autogen.sh &&     \
  ./configure --prefix="${HOME}/libuv-win32-i686"                  \
    --enable-static --disable-shared --host="i686-w64-mingw32" &&  \
  make && make install )

    • Build the simulator:

      env CC="i686-w64-mingw32-clang"                                    \
    CFLAGS="-I${HOME}/libuv-win32-i686/include -D__MINGW64__"      \
    LDFLAGS="-L${HOME}/libuv-win32-i686/lib" NEED_128=1            \
  make CROSS="MINGW64"

Windows x86_64

  • Using Clang (the LLVM-MinGW compiler) to cross-compile a local static libuv library and a native 64-bit Windows/x86_64 executable:

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                    \
mkdir -p "${HOME}/libuv-win32-x86_64" &&                             \
( cd "${HOME}/libuv-build" &&                                        \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&       \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" && sh ./autogen.sh &&       \
  ./configure --prefix="${HOME}/libuv-win32-x86_64"                  \
    --enable-static --disable-shared --host="x86_64-w64-mingw32" &&  \
  make && make install )

    • Build the simulator:

      env CC="x86_64-w64-mingw32-clang"                                    \
    CFLAGS="-I${HOME}/libuv-win32-x86_64/include -D__MINGW64__"      \
    LDFLAGS="-L${HOME}/libuv-win32-x86_64/lib"                       \
  make CROSS="MINGW64"

Windows ARMv7

  • Using Clang (the LLVM-MinGW compiler) to cross-compile a local static libuv library and a native 32-bit Windows/ARMv7 executable:

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                   \
mkdir -p "${HOME}/libuv-win32-armv7" &&                             \
( cd "${HOME}/libuv-build" &&                                       \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&      \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" && sh ./autogen.sh &&      \
  ./configure --prefix="${HOME}/libuv-win32-armv7"                  \
    --enable-static --disable-shared --host="armv7-w64-mingw32" &&  \
  make && make install )

    • Build the simulator:

      env CC="armv7-w64-mingw32-clang"                                    \
    CFLAGS="-I${HOME}/libuv-win32-armv7/include -D__MINGW64__"      \
    LDFLAGS="-L${HOME}/libuv-win32-armv7/lib"                       \
    NEED_128=1                                                      \
  make CROSS="MINGW64"

Windows ARM64

  • Using Clang (the LLVM-MinGW compiler) to cross-compile a local static libuv library and a native 64-bit Windows/ARM64 executable:

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                     \
mkdir -p "${HOME}/libuv-win32-arm64" &&                               \
( cd "${HOME}/libuv-build" &&                                         \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&        \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" && sh ./autogen.sh &&        \
  ./configure --prefix="${HOME}/libuv-win32-arm64"                    \
    --enable-static --disable-shared --host="aarch64-w64-mingw32" &&  \
  make && make install )

    • Build the simulator:

      env CC="aarch64-w64-mingw32-clang"                                    \
    CFLAGS="-I${HOME}/libuv-win32-arm64/include -D__MINGW64__"        \
    LDFLAGS="-L${HOME}/libuv-win32-arm64/lib"                         \
  make CROSS="MINGW64"

Unix-hosted MinGW-w64 GCC cross-compilation

The MinGW-w64 GCC toolchain supports building native Windows (i686 and x86_64) executables on non-Windows host systems (or Windows using the Windows Subsystem for Linux).

  • Many MinGW-w64 toolchains are available for a wide variety of host platforms and operating systems.
  • Version 9.0 is the minimum version of MinGW-w64 tested with DPS8M; version 11.0 (or later) is strongly recommended.
  • The DPS8M Development Team regularly cross-compiles Windows executables using GCC-based MinGW-w64 toolchains on Alpine Linux and Fedora Linux host systems.

In the following cross-compilation examples, the latest libuv sources (from the v1.x git branch) are used, but the current official release (available from https://libuv.org/) can also be used.

Windows i686

  • Using GCC (the MinGW-w64 compiler) to cross-compile a local static libuv library and a native 32-bit Windows/i686 executable:

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                  \
mkdir -p "${HOME}/libuv-win32-i686" &&                             \
( cd "${HOME}/libuv-build" &&                                      \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&     \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" && sh ./autogen.sh &&     \
  ./configure --prefix="${HOME}/libuv-win32-i686"                  \
    --enable-static --disable-shared --host="i686-w64-mingw32" &&  \
  make && make install )

    • Build the simulator:

      env CC="i686-w64-mingw32-gcc"                                      \
    CFLAGS="-I${HOME}/libuv-win32-i686/include                     \
            -D__MINGW64__ -pthread"                                \
    LDFLAGS="-L${HOME}/libuv-win32-i686/lib -lpthread"             \
    NEED_128=1                                                     \
  make CROSS="MINGW64"

Windows x86_64

  • Using GCC (the MinGW-w64 compiler) to cross-compile a local static libuv library and a native 64-bit Windows/x86_64 executable:

    • Build libuv:

      mkdir -p "${HOME}/libuv-build" &&                                    \
mkdir -p "${HOME}/libuv-win32-x86_64" &&                             \
( cd "${HOME}/libuv-build" &&                                        \
  wget -v "https://github.com/libuv/libuv/archive/v1.x.zip" &&       \
  unzip -xa "v1.x.zip" && cd "libuv-1.x" && sh ./autogen.sh &&       \
  ./configure --prefix="${HOME}/libuv-win32-x86_64"                  \
    --enable-static --disable-shared --host="x86_64-w64-mingw32" &&  \
  make && make install )

    • Build the simulator:

      env CC="x86_64-w64-mingw32-gcc"                                      \
    CFLAGS="-I${HOME}/libuv-win32-x86_64/include                     \
            -D__MINGW64__ -pthread"                                  \
    LDFLAGS="-L${HOME}/libuv-win32-x86_64/lib -lpthread"             \
  make CROSS="MINGW64"