Difference between revisions of "Upgrading sword"
Jump to navigation
Jump to search
import>Cen1001 |
import>Cen1001 |
||
Line 14: | Line 14: | ||
** Most of the MPI libraries (may need to keep some for compiling purposes) |
** Most of the MPI libraries (may need to keep some for compiling purposes) |
||
** The out-of-date trial compilers (Absoft, Lahey) |
** The out-of-date trial compilers (Absoft, Lahey) |
||
+ | * Software to keep |
||
+ | ** ifort in all its versions |
||
+ | ** icc in all its versions |
||
+ | ** Portland 5.0 up |
||
+ | ** Nag f95 (this will need upgrading so it won't be exactly the same) |
||
+ | ** Dalton |
||
+ | ** FFTW 3 and 2 |
||
+ | ** ifc? |
||
If anyone has things they'd like to see on sword this would be the place to ask for them. |
If anyone has things they'd like to see on sword this would be the place to ask for them. |
Revision as of 15:39, 27 April 2006
Sword needs an upgrade. I'd like to move it to SuSE 9.3. The destiny upgrade went OK, which was effectively a trial run for upgrading sword. Here are my current plans for sword:
- Remove the MPI queues. MPI should be done on Athens or the group servers
- Use rsh internally for consistency with the parallel clusters (ssh will still work though)
- I have a list of suse packages for sword .
- Make modules compulsory
- Do network installs for the compute nodes
- Software to be removed because it's not supported or no one wants it:
- GAMESS-US
- PGI 4.0
- gcc 3.2
- BLACS
- ScaLAPACK
- Most of the MPI libraries (may need to keep some for compiling purposes)
- The out-of-date trial compilers (Absoft, Lahey)
- Software to keep
- ifort in all its versions
- icc in all its versions
- Portland 5.0 up
- Nag f95 (this will need upgrading so it won't be exactly the same)
- Dalton
- FFTW 3 and 2
- ifc?
If anyone has things they'd like to see on sword this would be the place to ask for them.