|
Size: 5454
Comment:
|
Size: 6526
Comment:
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 2: | Line 2: |
This page, which is primarily maintained by David Kirkby, describes the current status of Sage on the various versions (Solaris 10 and Open Solaris) and on various hardware (SPARC and x86). |
|
| Line 5: | Line 7: |
| This page, which is primarily maintained by David Kirkby, describes the current status of Sage on Solaris on the various versions (Solaris 10 and Open Solaris) and on various hardware (SPARC and x86). | |
| Line 12: | Line 13: |
| * 2 x 1167 MHz Coolthreads processors | * 2 x 1167 MHz T2+ Coolthreads processors |
| Line 26: | Line 27: |
| Sage will run in a zone, so if you want to make a public access server, with negligible security risks, then you can do so. There are a few problems with the build on Solaris 10, all of which we hope to resolve. | |
| Line 27: | Line 29: |
| === Solaris 10 on SPARC x86 hardware. === | * Only GCC >= 4.0.1 can be used. Sun Studio can not be used. Supporting Sun Studio is an aim, but it's unlikely to happen soon. There are too pieces of code in Sage that assume GCC and use GCC compiler flags. * Sage currently only builds in 32-bit mode on Solaris 10. * Sage has a large test suite, which is run regularly by Sage developers. Not all the tests pass on Solaris 10. |
| Line 29: | Line 33: |
| This has not been tested, but would have a high probability of building. However, it possible that | === Solaris 10 on x86 hardware. === This has not been tested, but would have a high probability of working. It is likely to suffer the same issues as Sage does on SPARC hardware. === Open Solaris (also known as Solaris 11) on SPARC hardware === This has not been tested, but will almost certainly not work. We are afraid to say, nobody currently developing Sage runs OpenSolaris on SPARC hardware. === Open Solaris (also known as Solaris 11) on x86 hardware === Unfortunately, as of December 2009, the latest release of Sage will not build on Open Solaris. We are actively working on resolving this. |
Running Sage on Sun's Solaris operating system.
This page, which is primarily maintained by David Kirkby, describes the current status of Sage on the various versions (Solaris 10 and Open Solaris) and on various hardware (SPARC and x86).
Sage will build on Sun's Solaris operating system - you can in fact try out Sage running on a Sun T5240 Coolthreads server, kindly donated by Sun Microsystems, at http://t2nb.math.washington.edu:8000/
Solaris 10 on SPARC hardware.
Sage 4.3 or later should build and run properly on any SPARC hardware which runs Solaris 10. Sage has been built on several Solaris 10 SPARC systems, ranging from old machines to current models and from the first release of Solaris 10 (03/2005 edition) to the latest (05/2009 edition). So Sage should build on your Solaris 10 SPARC machine. Specific hardware/software combinations on which Sage has been built are
Sun T5240 CoolThreads server with:
- 2 x 1167 MHz T2+ Coolthreads processors
- 16 cores
- 128 threads
- 32 GB RAM
- Solaris 10 update 6 (05/2009 edition).
- A Sun Blade 2000 workstation with:
- 2 x 1200 MHz UltraSPARC III+ processors.
- 8 GB RAM
- Solaris 10 update 7 (05/2009 release) which at the time of writing (December 2009) is the latest release of Solaris 10.
- Sun Netra T1 server with:
- 1 x 500 MHz UltraSPARC-IIe processor
- 1536 MB RAM
- Solaris 10 03/2005 (the very first Solaris 10 release).
Sage will run in a zone, so if you want to make a public access server, with negligible security risks, then you can do so. There are a few problems with the build on Solaris 10, all of which we hope to resolve.
Only GCC >= 4.0.1 can be used. Sun Studio can not be used. Supporting Sun Studio is an aim, but it's unlikely to happen soon. There are too pieces of code in Sage that assume GCC and use GCC compiler flags.
- Sage currently only builds in 32-bit mode on Solaris 10.
- Sage has a large test suite, which is run regularly by Sage developers. Not all the tests pass on Solaris 10.
Solaris 10 on x86 hardware.
This has not been tested, but would have a high probability of working. It is likely to suffer the same issues as Sage does on SPARC hardware.
Open Solaris (also known as Solaris 11) on SPARC hardware
This has not been tested, but will almost certainly not work. We are afraid to say, nobody currently developing Sage runs OpenSolaris on SPARC hardware.
Open Solaris (also known as Solaris 11) on x86 hardware
Unfortunately, as of December 2009, the latest release of Sage will not build on Open Solaris. We are actively working on resolving this.
Sage on Sun's SPARC hardware
SPARC status
As of the 4th August 2009, several older copies of Sage have been built. The most recent is:
- Sage 4.1.1.rc0 (a developer release) was built on a Sun Ultra 80 running Solaris 10 update 7. For this, gcc/g++/gfortran versions were 4.4.0. The GNU compilers were built to use the Sun linker (/usr/ccs/bin/ld) and the Sun assembler (/usr/ccs/bin/as).
- Older versions of Sage have been built on a Sun Blade 2000 and a Sun Blade 2500.
SPARC Issues
ECL which is an implementation of the Common Lisp language is used inside Sage for Maxima. The version of ECL in Sage will not build on Solaris SPARC. Neither will version 9.4.1 of ECL, which is the latest as I write. However, the ECL developers have fixed the issue in the latest CVS. A new release of ECL is expected before the 9th August. This will then be integrated into Sage. Before this, an experimental package of ecl is available, which solves this issue. http://sage.math.washington.edu/home/kirkby/Solaris-fixes/ecl-9.4.1.p0/ecl-9.4.1.p0.spkg
- Maxima will not build. This is because of a bug in Maxima, which has been noticed for the first time in 35 years. It was only noticed when the newest ECL detected the error. A new release of Maxima will soon fix this, but for now, I bypassed the problem by touching the file sage-4.1.1.rc0/spkg/installed/maxima-5.16.3.p2 This of course means Maxima will not function.
The notebook() did not allow a web browser to connect at http://localhost:8000
Building Sage on Solaris SPARC
Solaris SPARC support is almost there, and things will change quickly. So ask on the sage-devel list for the most up to date information. If you want to try to build Sage on Solaris, the following is suggested. This assumes the version of Sage is 4.1.1.rc0. However, use the LATEST version, not 4.1.1.rc0.
Ensure you have a recent gcc. 4.2.4 is definately new enough. Some older versions may not be. The gcc included in Solaris /usr/sfw/bin/gcc is often not built to support Fortran (check with 'gcc -v') In this case, you will need to either build a later gcc from source code, or get a version with support for Fortran from Blastwave or SunFreeware.
If gcc is configured to use the Sun linker and assembler (/usr/ccs/bin/ld & /usr/ccs/bin/as), ensure that /usr/ccs/bin is in your path before any directory containing a copy of the GNU assembler or linker. i.e. ensure that 'which ld' returns /usr/ccs/bin/ld and not something like /usr/local/bin/ld.
- Conversely, if your gcc was configured to use the GNU linker and assembler, ensure that the GNU tools are in the path before /usr/ccs/bin/. The reason for this is that the Sage source files check the linker and assembler versions, and expect these to be the same as what is used by gcc when compiling.
Unpack download the file http://sage.math.washington.edu/home/kirkby/Solaris-fixes/ecl-9.4.1.p0/ecl-9.4.1.p0.spkg to the directory sage-4.1.1.rc0/spkg/standard (update the version of course).
Touch the file sage-4.1.1.rc0/spkg/installed/maxima-5.16.3.p2
- Type make, and expect to wait some hours!
- Eventually Sage will fail with a message indicating problems in the file sage-4.1.1.rc0/local/include/pari/paripriv.h. Edit that file manually, and comment out lines 258, 259 and 428.
Type make again, and Sage should build.
Sage on Solaris using Intel or AMD x86/x64 CPUs
Sage has been built on Solaris x86 / OpenSolaris. Someone else will have to report successful builds and any notes that will help people.
An issue with MPIR not detecting the correct linker has been seen on one computer running the November 2008 edition of OpenSolaris, yet others running the same version have not had this problem.