Hi Riccardo,
On 03/06/14 21:58, Riccardo Murri wrote:
Hi all,
On 3 June 2014 20:13, Jack Perdue <[email protected]> wrote:
Is there a formula/heuristic that y'all
use to determine the version numbers for
the toolchains?
I was about to ask the same question :-) I find the current numbering
scheme completely opaque.
How about using a "year.month" scheme, e.g., goolf-14.06 would be the
GCC+OpenMPI+... that are current during June 2014?
That sounds reasonable, but might introduce artificial limits too, I think.
Toolchain elements aren't always updated; sometimes we need to go back
to older versions for testing/benchmarking/compatibility reasons.
How would you version a toolchain that combines GCC v4.9 (new) and
OpenMPI v1.6 (old) with this scheme?
As I mentioned in the other mail in this thread, there's no magic
solution that makes sense and can hold up in the (near) future I'm afraid.
How could you possible 'encode' 5-6 different software versions into a
single (sensible) version number without losing information?
One thing we will be doing across the different VSC (short for Flemish
Supercomputer Centre) sites in the Flanders region of Belgium is to use
a scheme like "intel/2014a", "intel/2014b" for (ictce) toolchains.
The plan is to agree on a toolchain definition to use in a 6-month
period, and use that as much as possible to install software with.
The main reason for the customized naming is to not confuse users with
this 'weird' mnemonic 'ictce' and magic (almost meaningless) toolchain
versions.
The idea is to basically pick an existing and somewhat established ictce
version for this, and using a custom module naming scheme to map e.g.
"ictce/6.2.5" to "intel/2014b". That way, we can contribute the
easyconfig files we come up with, without creating yet another toolchain...
regards,
Kenneth
