gnu/packages/mpi.scm

   1 ;;; GNU Guix --- Functional package management for GNU
   2 ;;; Copyright © 2014, 2015, 2018, 2019 Eric Bavier <bavier@member.fsf.org>
   3 ;;; Copyright © 2014, 2015, 2016, 2017, 2018, 2019 Ludovic Courtès <ludo@gnu.org>
   4 ;;; Copyright © 2014 Ian Denhardt <ian@zenhack.net>
   5 ;;; Copyright © 2016 Andreas Enge <andreas@enge.fr>
   6 ;;; Copyright © 2017 Dave Love <fx@gnu.org>
   7 ;;; Copyright © 2017 Efraim Flashner <efraim@flashner.co.il>
   8 ;;; Copyright © 2018, 2019, 2020 Tobias Geerinckx-Rice <me@tobias.gr>
   9 ;;; Copyright © 2018 Paul Garlick <pgarlick@tourbillion-technology.com>
  10 ;;; Copyright © 2019 Ricardo Wurmus <rekado@elephly.net>
  11 ;;;
  12 ;;; This file is part of GNU Guix.
  13 ;;;
  14 ;;; GNU Guix is free software; you can redistribute it and/or modify it
  15 ;;; under the terms of the GNU General Public License as published by
  16 ;;; the Free Software Foundation; either version 3 of the License, or (at
  17 ;;; your option) any later version.
  18 ;;;
  19 ;;; GNU Guix is distributed in the hope that it will be useful, but
  20 ;;; WITHOUT ANY WARRANTY; without even the implied warranty of
  21 ;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  22 ;;; GNU General Public License for more details.
  23 ;;;
  24 ;;; You should have received a copy of the GNU General Public License
  25 ;;; along with GNU Guix.  If not, see <http://www.gnu.org/licenses/>.
  26
  27 (define-module (gnu packages mpi)
  28   #:use-module (guix packages)
  29   #:use-module ((guix licenses) #:prefix license:)
  30   #:use-module (guix download)
  31   #:use-module (guix utils)
  32   #:use-module (guix deprecation)
  33   #:use-module (guix build-system gnu)
  34   #:use-module (guix build-system python)
  35   #:use-module (gnu packages)
  36   #:use-module (gnu packages base)
  37   #:use-module (gnu packages compression)
  38   #:use-module (gnu packages fabric-management)
  39   #:use-module (gnu packages gcc)
  40   #:use-module (gnu packages java)
  41   #:use-module (gnu packages libevent)
  42   #:use-module (gnu packages linux)
  43   #:use-module (gnu packages pciutils)
  44   #:use-module (gnu packages xorg)
  45   #:use-module (gnu packages gtk)
  46   #:use-module (gnu packages xml)
  47   #:use-module (gnu packages perl)
  48   #:use-module (gnu packages ncurses)
  49   #:use-module (gnu packages parallel)
  50   #:use-module (gnu packages pkg-config)
  51   #:use-module (gnu packages valgrind)
  52   #:use-module (srfi srfi-1)
  53   #:use-module (ice-9 match))
  54
  55 (define-public hwloc-1
  56   ;; Note: For now we keep 1.x as the default because many packages have yet
  57   ;; to migrate to 2.0.
  58   (package
  59     (name "hwloc")
  60     (version "1.11.12")
  61     (source (origin
  62               (method url-fetch)
  63               (uri (string-append "https://www.open-mpi.org/software/hwloc/v"
  64                                   (version-major+minor version)
  65                                   "/downloads/hwloc-" version ".tar.bz2"))
  66               (sha256
  67                (base32
  68                 "0za1b9lvrm3rhn0lrxja5f64r0aq1qs4m0pxn1ji2mbi8ndppyyx"))))
  69     (build-system gnu-build-system)
  70     (outputs '("out"           ;'lstopo' & co., depends on Cairo, libx11, etc.
  71                "lib"           ;small closure
  72                "debug"))
  73     (inputs
  74      `(("libx11" ,libx11)
  75        ("cairo" ,cairo)
  76        ("ncurses" ,ncurses)
  77        ("expat" ,expat)
  78        ,@(if (not (string-prefix? "armhf"
  79                                   (or (%current-target-system)
  80                                       (%current-system))))
  81              `(("numactl" ,numactl))
  82              '())))
  83     (propagated-inputs
  84      ;; hwloc.pc lists it in 'Requires.private'.
  85      `(("libpciaccess" ,libpciaccess)))
  86     (native-inputs
  87      `(("pkg-config" ,pkg-config)))
  88     (arguments
  89      `(#:configure-flags '("--localstatedir=/var")
  90        #:phases
  91        (modify-phases %standard-phases
  92          (add-before 'check 'skip-linux-libnuma-test
  93            (lambda _
  94              ;; Arrange to skip 'tests/linux-libnuma', which fails on some
  95              ;; machines: <https://github.com/open-mpi/hwloc/issues/213>.
  96              (substitute* "tests/linux-libnuma.c"
  97                (("numa_available\\(\\)")
  98                 "-1"))
  99              #t))
 100          (add-after 'install 'refine-libnuma
 101            ;; Give -L arguments for libraries to avoid propagation
 102            (lambda* (#:key inputs outputs #:allow-other-keys)
 103              (let ((out  (assoc-ref outputs "lib"))
 104                    (numa (assoc-ref inputs "numactl")))
 105                (substitute* (map (lambda (f) (string-append out "/" f))
 106                                  '("lib/pkgconfig/hwloc.pc" "lib/libhwloc.la"))
 107                  (("-lnuma" lib)
 108                   (string-append "-L" numa "/lib " lib))))))
 109          (add-after 'install 'avoid-circular-references
 110            (lambda* (#:key outputs #:allow-other-keys)
 111              (let ((lib (assoc-ref outputs "lib")))
 112                ;; Suppress the 'prefix=' and 'exec_prefix=' lines so that the
 113                ;; "lib" output doesn't refer to "out".
 114                (substitute* (string-append lib "/lib/pkgconfig/hwloc.pc")
 115                  (("^.*prefix=.*$")
 116                   ""))
 117                #t))))))
 118     (home-page "https://www.open-mpi.org/projects/hwloc/")
 119     (synopsis "Abstraction of hardware architectures")
 120     (description
 121      "hwloc provides a portable abstraction (across OS,
 122 versions, architectures, ...) of the hierarchical topology of modern
 123 architectures, including NUMA memory nodes, sockets, shared caches, cores and
 124 simultaneous multithreading.  It also gathers various attributes such as cache
 125 and memory information.  It primarily aims at helping high-performance
 126 computing applications with gathering information about the hardware so as to
 127 exploit it accordingly and efficiently.
 128
 129 hwloc may display the topology in multiple convenient formats.  It also offers
 130 a powerful programming interface to gather information about the hardware,
 131 bind processes, and much more.")
 132     (license license:bsd-3)))
 133
 134 (define-public hwloc-2
 135   ;; Note: 2.0 isn't the default yet, see above.
 136   (package
 137     (inherit hwloc-1)
 138     (version "2.2.0")
 139     (source (origin
 140               (method url-fetch)
 141               (uri (string-append "https://www.open-mpi.org/software/hwloc/v"
 142                                   (version-major+minor version)
 143                                   "/downloads/hwloc-" version ".tar.bz2"))
 144               (sha256
 145                (base32
 146                 "0li27a3lnmb77qxpijj0kpblz32wmqd3b386sypq8ar7vy9vhw5f"))))
 147
 148     ;; libnuma is no longer needed.
 149     (inputs (alist-delete "numactl" (package-inputs hwloc-1)))
 150     (arguments
 151      (substitute-keyword-arguments (package-arguments hwloc-1)
 152        ((#:phases phases)
 153         `(modify-phases ,phases
 154            (replace 'skip-linux-libnuma-test
 155              (lambda _
 156                ;; Arrange to skip 'tests/hwloc/linux-libnuma', which fails on
 157                ;; some machines: <https://github.com/open-mpi/hwloc/issues/213>.
 158                (substitute* "tests/hwloc/linux-libnuma.c"
 159                  (("numa_available\\(\\)")
 160                   "-1"))
 161                #t))
 162            (add-before 'check 'skip-test-that-fails-on-qemu
 163              (lambda _
 164                ;; Skip test that fails on emulated hardware due to QEMU bug:
 165                ;; <https://bugs.gnu.org/40342>.
 166                (substitute* "tests/hwloc/hwloc_get_last_cpu_location.c"
 167                  (("hwloc_topology_init" all)
 168                   (string-append "exit (77);\n" all)))
 169                #t))))))))
 170
 171 (define-deprecated hwloc-2.0 hwloc-2)
 172
 173 (define-public hwloc
 174   ;; The latest stable series of hwloc.
 175   hwloc-2)
 176
 177 (define-public openmpi
 178   (package
 179     (name "openmpi")
 180     (version "4.0.3")
 181     (source
 182      (origin
 183       (method url-fetch)
 184       (uri (string-append "https://www.open-mpi.org/software/ompi/v"
 185                           (version-major+minor version)
 186                           "/downloads/openmpi-" version ".tar.bz2"))
 187       (sha256
 188        (base32 "00zxcw99gr5n693cmcmn4f6a47vx1ywna895p0x7p163v37gw0hl"))
 189       (patches (search-patches "openmpi-mtl-priorities.patch"))))
 190     (build-system gnu-build-system)
 191     (inputs
 192      `(("hwloc" ,hwloc-2 "lib")
 193        ("gfortran" ,gfortran)
 194        ("libfabric" ,libfabric)
 195        ("libevent" ,libevent)
 196        ("opensm" ,opensm)
 197        ,@(if (and (not (%current-target-system))
 198                   (member (%current-system) (package-supported-systems psm)))
 199              `(("psm" ,psm))
 200              '())
 201        ,@(if (and (not (%current-target-system))
 202                   (member (%current-system) (package-supported-systems psm2)))
 203              `(("psm2" ,psm2))
 204              '())
 205        ,@(if (and (not (%current-target-system))
 206                   (member (%current-system) (package-supported-systems ucx)))
 207              `(("ucx" ,ucx))
 208              '())
 209        ("rdma-core" ,rdma-core)
 210        ("valgrind" ,valgrind)
 211        ("slurm" ,slurm)))              ;for PMI support (launching via "srun")
 212     (native-inputs
 213      `(("pkg-config" ,pkg-config)
 214        ("perl" ,perl)))
 215     (outputs '("out" "debug"))
 216     (arguments
 217      `(#:configure-flags `("--enable-mpi-ext=affinity" ;cr doesn't work
 218                            "--enable-memchecker"
 219                            "--with-sge"
 220                            "--with-valgrind"
 221                            "--with-hwloc=external"
 222                            "--with-libevent"
 223
 224                            ;; InfiniBand support
 225                            "--enable-openib-control-hdr-padding"
 226                            "--enable-openib-dynamic-sl"
 227                            "--enable-openib-udcm"
 228                            "--enable-openib-rdmacm"
 229                            "--enable-openib-rdmacm-ibaddr"
 230
 231                            ;; Enable support for SLURM's Process Manager
 232                            ;; Interface (PMI).
 233                            ,(string-append "--with-pmi="
 234                                            (assoc-ref %build-inputs "slurm")))
 235        #:phases (modify-phases %standard-phases
 236                   ;; opensm is needed for InfiniBand support.
 237                   (add-after 'unpack 'find-opensm-headers
 238                     (lambda* (#:key inputs #:allow-other-keys)
 239                       (setenv "C_INCLUDE_PATH"
 240                               (string-append (assoc-ref inputs "opensm")
 241                                              "/include/infiniband"))
 242                       (setenv "CPLUS_INCLUDE_PATH"
 243                               (string-append (assoc-ref inputs "opensm")
 244                                              "/include/infiniband"))
 245                       #t))
 246                   (add-before 'build 'remove-absolute
 247                     (lambda _
 248                       ;; Remove compiler absolute file names (OPAL_FC_ABSOLUTE
 249                       ;; etc.) to reduce the closure size.  See
 250                       ;; <https://lists.gnu.org/archive/html/guix-devel/2017-07/msg00388.html>
 251                       ;; and
 252                       ;; <https://www.mail-archive.com/users@lists.open-mpi.org//msg31397.html>.
 253                       (substitute* '("orte/tools/orte-info/param.c"
 254                                      "oshmem/tools/oshmem_info/param.c"
 255                                      "ompi/tools/ompi_info/param.c")
 256                         (("_ABSOLUTE") ""))
 257                       ;; Avoid valgrind (which pulls in gdb etc.).
 258                       (substitute*
 259                           '("./ompi/mca/io/romio321/src/io_romio321_component.c")
 260                         (("MCA_io_romio321_COMPLETE_CONFIGURE_FLAGS")
 261                          "\"[elided to reduce closure]\""))
 262                       #t))
 263                   (add-before 'build 'scrub-timestamps ;reproducibility
 264                     (lambda _
 265                       (substitute* '("ompi/tools/ompi_info/param.c"
 266                                      "orte/tools/orte-info/param.c"
 267                                      "oshmem/tools/oshmem_info/param.c")
 268                         ((".*(Built|Configured) on.*") ""))
 269                       #t))
 270                   (add-after 'install 'remove-logs ;reproducibility
 271                     (lambda* (#:key outputs #:allow-other-keys)
 272                       (let ((out (assoc-ref outputs "out")))
 273                         (for-each delete-file (find-files out "config.log"))
 274                         #t))))))
 275     (home-page "https://www.open-mpi.org")
 276     (synopsis "MPI-3 implementation")
 277     (description
 278      "The Open MPI Project is an MPI-3 implementation that is developed and
 279 maintained by a consortium of academic, research, and industry partners.  Open
 280 MPI is therefore able to combine the expertise, technologies, and resources
 281 from all across the High Performance Computing community in order to build the
 282 best MPI library available.  Open MPI offers advantages for system and
 283 software vendors, application developers and computer science researchers.")
 284     ;; See file://LICENSE
 285     (license license:bsd-2)))
 286
 287 ;; TODO: javadoc files contain timestamps.
 288 (define-public java-openmpi
 289   (package (inherit openmpi)
 290     (name "java-openmpi")
 291     (inputs
 292      `(("openmpi" ,openmpi)
 293        ,@(package-inputs openmpi)))
 294     (native-inputs
 295      `(("jdk" ,openjdk11 "jdk")
 296        ("zip" ,(@ (gnu packages compression) zip))
 297        ,@(package-native-inputs openmpi)))
 298     (outputs '("out"))
 299     (arguments
 300      `(#:modules ((guix build gnu-build-system)
 301                   ((guix build ant-build-system) #:prefix ant:)
 302                   (guix build utils))
 303        #:imported-modules ((guix build ant-build-system)
 304                            (guix build syscalls)
 305                            ,@%gnu-build-system-modules)
 306        ,@(substitute-keyword-arguments (package-arguments openmpi)
 307            ((#:configure-flags flags)
 308             `(cons "--enable-mpi-java" ,flags))
 309            ((#:make-flags flags ''())
 310             `(append '("-C" "ompi/mpi/java")
 311                      ,flags))
 312            ((#:phases phases)
 313             `(modify-phases ,phases
 314                ;; We could provide the location of the JDK in the configure
 315                ;; flags, but since the configure flags are embedded in the
 316                ;; info binaries that would leave a reference to the JDK in
 317                ;; the "out" output.  To avoid this we set JAVA_HOME.
 318                (add-after 'unpack 'set-JAVA_HOME
 319                  (lambda* (#:key inputs #:allow-other-keys)
 320                    (setenv "JAVA_HOME" (assoc-ref inputs "jdk"))
 321                    #t))
 322                (add-after 'unpack 'link-with-existing-mpi-libraries
 323                  (lambda* (#:key inputs #:allow-other-keys)
 324                    (substitute* "ompi/mpi/java/c/Makefile.in"
 325                      (("\\$\\(top_builddir\\)/ompi/lib@OMPI_LIBMPI_NAME@.la")
 326                       (string-append (assoc-ref inputs "openmpi") "/lib/libmpi.la")))
 327                    #t))
 328                (add-after 'install 'strip-jar-timestamps
 329                  (assoc-ref ant:%standard-phases 'strip-jar-timestamps)))))))
 330     (synopsis "Java bindings for MPI")))
 331
 332 (define-public openmpi-thread-multiple
 333   (package
 334     (inherit openmpi)
 335     (name "openmpi-thread-multiple")
 336     (arguments
 337      (substitute-keyword-arguments (package-arguments openmpi)
 338        ((#:configure-flags flags)
 339         `(cons "--enable-mpi-thread-multiple" ,flags))))
 340     (description " This version of Open@tie{}MPI has an implementation of
 341 @code{MPI_Init_thread} that provides @code{MPI_THREAD_MULTIPLE}.  This won't
 342 work correctly with all transports (such as @code{openib}), and the
 343 performance is generally worse than the vanilla @code{openmpi} package, which
 344 only provides @code{MPI_THREAD_FUNNELED}.")))
 345
 346 ;;; Build phase to be used for packages that execute MPI code.
 347 (define-public %openmpi-setup
 348   '(lambda _
 349      ;; By default, running the test suite would fail because 'ssh' could not
 350      ;; be found in $PATH.  Define this variable to placate Open MPI without
 351      ;; adding a dependency on OpenSSH (the agent isn't used anyway.)
 352      (setenv "OMPI_MCA_plm_rsh_agent" (which "false"))
 353      ;; Allow oversubscription in case there are less physical cores available
 354      ;; in the build environment than the package wants while testing.
 355      (setenv "OMPI_MCA_rmaps_base_mapping_policy" "core:OVERSUBSCRIBE")
 356
 357      ;; UCX sometimes outputs uninteresting warnings such as:
 358      ;;
 359      ;;   mpool.c:38   UCX  WARN  object 0x7ffff44fffc0 was not returned to mpool ucp_am_bufs
 360      ;;
 361      ;; These in turn leads to failures of test suites that capture and
 362      ;; compare stdout, such as that of 'hdf5-parallel-openmpi'.  Thus, tell
 363      ;; UCX to not emit those warnings.
 364      (setenv "UCX_LOG_LEVEL" "error")
 365      #t))
 366
 367 (define-public python-mpi4py
 368   (package
 369     (name "python-mpi4py")
 370     (version "3.0.3")
 371     (source
 372      (origin
 373        (method url-fetch)
 374        (uri (pypi-uri "mpi4py" version))
 375        (sha256
 376         (base32 "07ssbhssv27rrjx1c5vd3vsr31vay5d8xcf4zh9yblcyidn72b81"))))
 377     (build-system python-build-system)
 378     (arguments
 379      `(#:phases
 380        (modify-phases %standard-phases
 381          (add-after 'build 'mpi-setup
 382            ,%openmpi-setup)
 383          (add-before 'check 'pre-check
 384            (lambda _
 385              ;; Skip BaseTestSpawn class (causes error 'ompi_dpm_dyn_init()
 386              ;; failed --> Returned "Unreachable"' in chroot environment).
 387              (substitute* "test/test_spawn.py"
 388                (("unittest.skipMPI\\('openmpi\\(<3.0.0\\)'\\)")
 389                 "unittest.skipMPI('openmpi')"))
 390              #t)))))
 391     (inputs
 392      `(("openmpi" ,openmpi)))
 393     (home-page "https://bitbucket.org/mpi4py/mpi4py/")
 394     (synopsis "Python bindings for the Message Passing Interface standard")
 395     (description "MPI for Python (mpi4py) provides bindings of the Message
 396 Passing Interface (MPI) standard for the Python programming language, allowing
 397 any Python program to exploit multiple processors.
 398
 399 mpi4py is constructed on top of the MPI-1/MPI-2 specification and provides an
 400 object oriented interface which closely follows MPI-2 C++ bindings.  It
 401 supports point-to-point and collective communications of any picklable Python
 402 object as well as optimized communications of Python objects (such as NumPy
 403 arrays) that expose a buffer interface.")
 404     (license license:bsd-3)))
 405
 406 (define-public mpich
 407   (package
 408     (name "mpich")
 409     (version "3.3.2")
 410     (source (origin
 411               (method url-fetch)
 412               (uri (string-append "http://www.mpich.org/static/downloads/"
 413                                   version "/mpich-" version ".tar.gz"))
 414               (sha256
 415                (base32
 416                 "1farz5zfx4cd0c3a0wb9pgfypzw0xxql1j1294z1sxslga1ziyjb"))))
 417     (build-system gnu-build-system)
 418     (inputs
 419      `(("zlib" ,zlib)
 420        ("hwloc" ,hwloc-2 "lib")
 421        ("slurm" ,slurm)
 422        ,@(if (and (not (%current-target-system))
 423                   (member (%current-system) (package-supported-systems ucx)))
 424              `(("ucx" ,ucx))
 425              '())))
 426     (native-inputs
 427      `(("perl" ,perl)
 428        ("which" ,which)
 429        ("gfortran" ,gfortran)))
 430     (outputs '("out" "debug"))
 431     (arguments
 432      `(#:configure-flags
 433        (list "--disable-silent-rules"             ;let's see what's happening
 434              "--enable-debuginfo"
 435              ;; "--with-device=ch4:ucx" ; --with-device=ch4:ofi segfaults in tests
 436              (string-append "--with-hwloc-prefix="
 437                             (assoc-ref %build-inputs "hwloc"))
 438
 439              ,@(if (assoc "ucx" (package-inputs this-package))
 440                    `((string-append "--with-ucx="
 441                                     (assoc-ref %build-inputs "ucx")))
 442                    '()))
 443
 444        #:phases (modify-phases %standard-phases
 445                   (add-after 'unpack 'patch-sources
 446                     (lambda _
 447                       (substitute* "./maint/gen_subcfg_m4"
 448                         (("/usr/bin/env") (which "env")))
 449                       (substitute* "src/glue/romio/all_romio_symbols"
 450                         (("/usr/bin/env") (which "env")))
 451                       (substitute* (find-files "." "buildiface")
 452                         (("/usr/bin/env") (which "env")))
 453                       (substitute* "maint/extracterrmsgs"
 454                         (("/usr/bin/env") (which "env")))
 455                       (substitute* (find-files "." "f77tof90")
 456                         (("/usr/bin/env") (which "env")))
 457                       (substitute* (find-files "." "\\.sh$")
 458                         (("/bin/sh") (which "sh")))
 459                       #t))
 460                   (add-before 'configure 'fix-makefile
 461                     (lambda _
 462                       ;; Remove "@hwloclib@" from 'pmpi_convenience_libs'.
 463                       ;; This fixes "No rule to make target '-lhwloc', needed
 464                       ;; by 'lib/libmpi.la'".
 465                       (substitute* "Makefile.in"
 466                         (("^pmpi_convenience_libs = (.*) @hwloclib@ (.*)$" _
 467                           before after)
 468                          (string-append "pmpi_convenience_libs = "
 469                                         before " " after)))
 470                       #t)))))
 471     (home-page "https://www.mpich.org/")
 472     (synopsis "Implementation of the Message Passing Interface (MPI)")
 473     (description
 474      "MPICH is a high-performance and portable implementation of the Message
 475 Passing Interface (MPI) standard (MPI-1, MPI-2 and MPI-3).  MPICH provides an
 476 MPI implementation that efficiently supports different computation and
 477 communication platforms including commodity clusters, high-speed networks (10
 478 Gigabit Ethernet, InfiniBand, Myrinet, Quadrics), and proprietary high-end
 479 computing systems (Blue Gene, Cray).  It enables research in MPI through a
 480 modular framework for other derived implementations.")
 481     (license license:bsd-2)))