Contents
Getting Sage to build and pass all tests on Skynet
It is very important to read
/usr/local/README.background_jobs
on the skynet machines, otherwise your builds will take a long time due to long-running CPU intensive tasks. That README file tells you how to stop those CPU intensive tasks.
Current tarballs are here:
http://sage.math.washington.edu/home/release/
Current versions of Sage should build without problem on the skynet machines cicero, cleo, eno, flavius, fulvia, iras, lena, mark, mark2, sextus and taurus.
Here is information about the skynet machines. For the linux boxes, this was obtained using the script at http://trac.sagemath.org/sage_trac/ticket/8048. For the Solaris 10 machines (fulvia, mark and mark2), it was obtained using the commands listed in the description.
Cicero
basic info
uname -a: Linux cicero 2.6.32.12-115.fc12.i686.PAE #1 SMP Fri Apr 30 20:14:08 UTC 2010 i686 i686 i386 GNU/Linux
OS/system info
/etc/*-release: not found! (This is generally not a problem.) /etc/issue: Fedora release 12 (Constantine)'\r'Kernel \r on an \m (\l)'\r''\r' version: Linux version 2.6.32.12-115.fc12.i686.PAE ([email protected]) (gcc version 4.4.3 20100127 (Red Hat 4.4.3-4) (GCC) ) #1 SMP Fri Apr 30 20:14:08 UTC 2010 OS bitlength: 32-bit (may not reflect hardware capabilities)
CPU info
model name: Intel(R) Pentium(R) 4 CPU 2.66GHz CPU Count: 1 cores per cpu: either "1" or not directly available from /proc/cpuinfo total number of threads: 1 64-bit capable: either "no" or not directly available from /proc/cpuinfo
memory info
MemTotal: 1025144 kB MemFree: 436416 kB SwapCached: 2944 kB SwapTotal: 2064376 kB SwapFree: 2021352 kB
misc info
gcc (GCC) 4.5.0
Cleo
basic info
uname -a: Linux cleo 2.6.18-128.1.1.el5 #1 SMP Mon Jan 26 13:57:09 EST 2009 ia64 ia64 ia64 GNU/Linux
OS/system info
/etc/*-release: not found! (This is generally not a problem.) /etc/issue: Red Hat Enterprise Linux Server release 5.3 (Tikanga)'\r'Kernel \r on an \m'\r''\r' version: Linux version 2.6.18-128.1.1.el5 ([email protected]) (gcc version 4.1.2 20080704 (Red Hat 4.1.2-44)) #1 SMP Mon Jan 26 13:57:09 EST 2009 OS bitlength: 64-bit
CPU info
vendor: GenuineIntel family: Itanium 2 arch: IA-64 CPU Count: 2 cores per cpu: 2 total number of threads: 4 64-bit capable: yes (IA-64)
memory info
MemTotal: 4061952 kB MemFree: 1663840 kB SwapCached: 0 kB SwapTotal: 8388576 kB SwapFree: 8388224 kB
misc info
gcc (GCC) 4.5.0
Eno
basic info
uname -a: Linux eno 2.6.32.12-115.fc12.x86_64 #1 SMP Fri Apr 30 19:46:25 UTC 2010 x86_64 x86_64 x86_64 GNU/Linux
OS/system info
/etc/*-release: not found! (This is generally not a problem.) /etc/issue: Fedora release 12 (Constantine)'\r'Kernel \r on an \m (\l)'\r''\r' version: Linux version 2.6.32.12-115.fc12.x86_64 ([email protected]) (gcc version 4.4.3 20100127 (Red Hat 4.4.3-4) (GCC) ) #1 SMP Fri Apr 30 19:46:25 UTC 2010 OS bitlength: 64-bit
CPU info
model name: Intel(R) Xeon(R) CPU E5345 @ 2.33GHz CPU Count: 2 cores per cpu: 4 total number of threads: 8 64-bit capable: yes (long-mode)
memory info
MemTotal: 16467944 kB MemFree: 2018052 kB SwapCached: 25784 kB SwapTotal: 18579448 kB SwapFree: 18534260 kB
misc info
gcc (GCC) 4.5.0
Flavius
basic info
uname -a: Linux flavius 2.6.32.12-115.fc12.x86_64 #1 SMP Fri Apr 30 19:46:25 UTC 2010 x86_64 x86_64 x86_64 GNU/Linux
OS/system info
/etc/*-release: not found! (This is generally not a problem.) /etc/issue: Fedora release 12 (Constantine)'\r'Kernel \r on an \m (\l)'\r''\r' version: Linux version 2.6.32.12-115.fc12.x86_64 ([email protected]) (gcc version 4.4.3 20100127 (Red Hat 4.4.3-4) (GCC) ) #1 SMP Fri Apr 30 19:46:25 UTC 2010 OS bitlength: 64-bit
CPU info
model name: AMD Opteron(tm) Processor 248 CPU Count: either "1" or not directly available from /proc/cpuinfo cores per cpu: either "1" or not directly available from /proc/cpuinfo total number of threads: 2 64-bit capable: yes (long-mode)
memory info
MemTotal: 8778376 kB MemFree: 1348220 kB SwapCached: 100 kB SwapTotal: 2031608 kB SwapFree: 2030804 kB
misc info
gcc (GCC) 4.5.0
Fulvia
uname -a: SunOS fulvia 5.10 Generic_127128-11 i86pc i386 i86pc cat /etc/release: Solaris 10 5/08 s10x_u5wos_10 X86 CPU count: 1 physical CPU (according to /usr/sbin/psrinfo -p) 4 threads (?) (according to /usr/sbin/psrinfo) /usr/sbin/psrinfo -v: The i386 processor operates at 2400 MHz, and has an i387 compatible floating point processor. /usr/sbin/prtconf | grep Memory: Memory size: 8126 Megabytes
To build Sage on fulvia, set the following variables:
export SAGE_PORT='yes' export PATH=/home/palmieri/fulvia/bin:/usr/local/gcc-4.5.1/x86_64-SunOS-core2-sun-ld/bin:$PATH export LD_LIBRARY_PATH="/usr/local/gcc-4.5.1/x86_64-SunOS-core2-sun-ld/lib/amd64:/usr/local/gcc-4.5.1/x86_64-SunOS-core2-sun-ld/lib" export SAGE_FORTRAN="/usr/local/gcc-4.5.1/x86_64-SunOS-core2-sun-ld/bin/gfortran" export SAGE_FORTRAN_LIB="/usr/local/gcc-4.5.1/x86_64-SunOS-core2-sun-ld/lib/libgfortran.so"
(The PATH includes /home/palmieri/fulvia/bin to make sure that GNU tar is present and first in the path.) You might also want to do
export MAKE='make -j2' (or 'make -j3' or 'make -j4'?) export SAGE_PARALLEL_SPKG_BUILD=yes
If you want to try a 64-bit build, then in addition to the above, set
export SAGE64='yes' export SAGE_FORTRAN_LIB="/usr/local/gcc-4.5.1/x86_64-SunOS-core2-sun-ld/lib/amd64/libgfortran.so"
and continue as above. (The second setting replaces the earlier 32-bit setting.) As of this writing, the build will certainly fail when building R. You can bypass this by touching the file SAGE_ROOT/spkg/installed/r-2.10.1.p2 and then restarting the build. (This issue is being tracked at http://trac.sagemath.org/sage_trac/ticket/9040.) However, whilst it's possible to create a 64-bit version of Sage, it will fail to run properly, and will almost certainly crash either at startup, or whenever one tries to do something non-trivial. Some, and possibly all of the problems are caused by libraries having text relocations. These include
Cliquer http://trac.sagemath.org/sage_trac/ticket/9833 & http://trac.sagemath.org/sage_trac/ticket/9871
Maxima http://trac.sagemath.org/sage_trac/ticket/9099 (almost certainly caused by ECL's text relocations).
Iras
basic info
uname -a: Linux iras 2.6.16.46-0.12-default #1 SMP Thu May 17 14:00:09 UTC 2007 ia64 ia64 ia64 GNU/Linux
OS/system info
/etc/*-release: LSB_VERSION="core-2.0-noarch:core-3.0-noarch:core-2.0-ia64:core-3.0-ia64" /etc/issue: '\r'Welcome to SUSE Linux Enterprise Server 10 SP1 (ia64) - Kernel \r (\l).'\r''\r''\r' version: Linux version 2.6.16.46-0.12-default (geeko@buildhost) (gcc version 4.1.2 20070115 (prerelease) (SUSE Linux)) #1 SMP Thu May 17 14:00:09 UTC 2007 OS bitlength: 64-bit
CPU info
vendor: GenuineIntel family: 32 arch: IA-64 CPU Count: 2 cores per cpu: 2 total number of threads: 4 64-bit capable: yes (IA-64)
memory info
MemTotal: 3998256 kB MemFree: 1549984 kB SwapCached: 0 kB SwapTotal: 2097120 kB SwapFree: 2096592 kB
misc info
gcc (GCC) 4.5.0
Lena
basic info
uname -a: Linux lena 2.6.32.12-115.fc12.x86_64 #1 SMP Fri Apr 30 19:46:25 UTC 2010 x86_64 x86_64 x86_64 GNU/Linux
OS/system info
/etc/*-release: not found! (This is generally not a problem.) /etc/issue: Fedora release 12 (Constantine)'\r'Kernel \r on an \m (\l)'\r''\r' version: Linux version 2.6.32.12-115.fc12.x86_64 ([email protected]) (gcc version 4.4.3 20100127 (Red Hat 4.4.3-4) (GCC) ) #1 SMP Fri Apr 30 19:46:25 UTC 2010 OS bitlength: 64-bit
CPU info
model name: AMD Phenom(tm) II X4 940 Processor CPU Count: 1 cores per cpu: 4 total number of threads: 4 64-bit capable: yes (long-mode)
memory info
MemTotal: 16340044 kB MemFree: 1629512 kB SwapCached: 0 kB SwapTotal: 18350072 kB SwapFree: 18350072 kB
misc info
gcc (GCC) 4.5.0
Mark
uname -a: SunOS mark 5.10 Generic_127111-01 sun4u sparc SUNW,Sun-Blade-2500 cat /etc/release: Solaris 10 1/06 s10s_u1wos_19a SPARC CPU count: 2 physical CPUs (according to /usr/sbin/psrinfo -p) 2 threads (according to /usr/sbin/psrinfo) /usr/sbin/psrinfo -v: The sparcv9 processor operates at 1280 MHz, and has a sparcv9 floating point processor. /usr/sbin/prtconf | grep Memory: Memory size: 4096 Megabytes
As of version 4.5.2.rc1, Sage should build on mark without trouble (although it takes a long time). For this to work, you need to set the following environment variables:
export PATH=/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/bin:$PATH export LD_LIBRARY_PATH=/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/lib:/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/lib/sparcv9:/usr/local/flex-2.5.35/sparc-SunOS-ultrasparc3-gcc-4.4.3/lib export SAGE_FORTRAN=/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/bin/gfortran export SAGE_FORTRAN_LIB=/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/lib/libgfortran.so
If you want to try a 64-bit build, then in addition to the above, set
export SAGE64='yes' export SAGE_FORTRAN_LIB="/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/lib/sparcv9/libgfortran.so"
and continue as above. (The second setting replaces the earlier 32-bit setting.) As of this writing (13th September 2010). it is possible to build a 64-bit version of Sage on SPARC systems. which is able to do some computations, but the resulting build is very unstable. The instability is at leasat in part due to the problems described above for 'fulvia', in that some of the libraries have text relocations.
Mark2
same as "mark"
Sextus
basic info
uname -a: Linux sextus 2.6.32.12-115.fc12.x86_64 #1 SMP Fri Apr 30 19:46:25 UTC 2010 x86_64 x86_64 x86_64 GNU/Linux
OS/system info
/etc/*-release: not found! (This is generally not a problem.) /etc/issue: Fedora release 12 (Constantine)'\r'Kernel \r on an \m (\l)'\r''\r' version: Linux version 2.6.32.12-115.fc12.x86_64 ([email protected]) (gcc version 4.4.3 20100127 (Red Hat 4.4.3-4) (GCC) ) #1 SMP Fri Apr 30 19:46:25 UTC 2010 OS bitlength: 64-bit
CPU info
model name: Intel(R) Xeon(TM) CPU 3.60GHz CPU Count: 1 cores per cpu: 1 total number of threads: 2 64-bit capable: yes (long-mode)
memory info
MemTotal: 1022612 kB MemFree: 617012 kB SwapCached: 3752 kB SwapTotal: 16777208 kB SwapFree: 16764052 kB
misc info
gcc (GCC) 4.5.0
Silius
> date
Fri Aug 19 19:32:36 GMT 2011
> uname -a
Linux silius 2.6.32.43-0.4-ppc64 #1 SMP 2011-07-14 14:47:44 +0200 ppc64 ppc64 ppc64 GNU/Linux
> cat /etc/issue
Welcome to SUSE Linux Enterprise Server 11 SP1 (ppc64) - Kernel \r (\l).
> cat /proc/cpuinfo
processor : 0
cpu : POWER7 (architected), altivec supported
clock : 3000.000000MHz
revision : 2.1 (pvr 003f 0201)
processor : 4
cpu : POWER7 (architected), altivec supported
clock : 3000.000000MHz
revision : 2.1 (pvr 003f 0201)
...
processor : 60
cpu : POWER7 (architected), altivec supported
clock : 3000.000000MHz
revision : 2.1 (pvr 003f 0201)
timebase : 512000000
platform : pSeries
model : IBM,8233-E8B
machine : CHRP IBM,8233-E8B
> free
total used free shared buffers cached
Mem: 32211328 22843008 9368320 0 369280 17518592
-/+ buffers/cache: 4955136 27256192
Swap: 2104320 0 2104320
Taurus
basic info
uname -a: Linux taurus 2.6.32.12-115.fc12.x86_64 #1 SMP Fri Apr 30 19:46:25 UTC 2010 x86_64 x86_64 x86_64 GNU/Linux
OS/system info
/etc/*-release: not found! (This is generally not a problem.) /etc/issue: Fedora release 12 (Constantine)'\r'Kernel \r on an \m (\l)'\r''\r' version: Linux version 2.6.32.12-115.fc12.x86_64 ([email protected]) (gcc version 4.4.3 20100127 (Red Hat 4.4.3-4) (GCC) ) #1 SMP Fri Apr 30 19:46:25 UTC 2010 OS bitlength: 64-bit
CPU info
model name: Intel(R) Xeon(R) CPU X5570 @ 2.93GHz CPU Count: 2 cores per cpu: 4 total number of threads: 16 64-bit capable: yes (long-mode)
memory info
MemTotal: 12318660 kB MemFree: 3820936 kB SwapCached: 14372 kB SwapTotal: 14417912 kB SwapFree: 14377820 kB
misc info
gcc (GCC) 4.5.0
Some completely unreliable timings
Here is how long it took to build Sage 4.5.2.alpha1 (not including the documentation) on the various skynet machines. These were all built with SAGE_PARALLEL_SPKG_BUILD set to 'yes' and MAKE='make -jN', with N replaced by the number of cores on the machine in question.
cicero |
332 minutes |
cleo |
186 minutes |
eno |
46 minutes |
flavius |
128 minutes |
fulvia |
This needs to be measured |
iras |
190 minutes |
lena |
53 minutes |
mark |
> 8 hours |
sextus |
195 minutes |
taurus |
36 minutes |