new code release of Intel(R) OpenMP* Runtime Library at openmprtl.org

Exactly how does the new code release of Intel® OpenMP* Runtime
Library at openmprtl.org, announced at http://planet.clang.org,
compare to what is currently committed in openmp trunk at llvm.org?
         Jack

   Exactly how does the new code release of Intel® OpenMP* Runtime Library at openmprtl.org,
announced at http://planet.clang.org, compare with what is currently committed in openmp trunk
at llvm.org?

If you really want *exactly*, then diff is your friend and will be more accurate and verbose than I can possibly be. :slight_smile:

So, recognizing that this is not "exactly"...

The link included in that planet clang article (which points to https://www.openmprtl.org/news/new-code-release-4 ) gives you the release notes, which tell you the functional aspects.

My note to this list yesterday explained what is happening, and that this is not a fork; all of the code in that release will be appearing here as a series of patches, probably in the New Year, now, since many people have either already left on vacation, or leave real soon now.

So, if the idea that Intel is developing code and not showing it to you every day upsets you, you can forget that you ever saw the OpenMPrtl.org code and ignore it completely. Alternately you can think of it as a preview of developments that will be arriving here as functional patch-sets for test and review soon.

HTH

-- Jim

James Cownie <james.h.cownie@intel.com>
SSG/DPD/TCAR (Technical Computing, Analyzers and Runtimes)
Tel: +44 117 9071438

Hi Jim,

Is there a chance of Intel moving their active development work into the LLVM OpenMP runtime project? That is, Intel is clearly doing large chunks of work behind closed doors and then trying to split things into smaller logical patches after the fact. Why not start with developing the patches in the community in the first place? Intel would effectively get free code reviews for their work, the community would have an opportunity to participate in the development process, and the quality of the code base and commits would improve. It would be significantly lower net effort than trying to split up openmprtl.org snapshot deltas into logical patches. Community contribution would also be easier if the major development focus was directly here in the LLVM project.

  • Steven

Is there a chance of Intel moving their active development work into the LLVM OpenMP runtime project? That is, Intel is clearly doing large chunks of work behind closed doors and then trying to split things into smaller logical patches after the fact. Why not start with developing the patches in the community in the first place? Intel would effectively get free code reviews for their work, the community would have an opportunity to participate in the development process, and the quality of the code base and commits would improve.

I would like to move in that direction, but changing the development approach and getting management approval for such a change is non-trivial.

It would be significantly lower net effort than trying to split up openmprtl.org snapshot deltas into logical patches.

Actually, it doesn’t work like that, rather we have a set of feature level patches from functional internal commits, so no splitting should be required and it should be reasonably automatic (at least, so Johnny tells me, and he implemented it so should know!)

Community contribution would also be easier if the major development focus was directly here in the LLVM project.

I don’t really follow the logic of that. Community contribution here has been light, but what contributions there have been have been integrated relatively fast (certainly faster than contributions of OpenMP support into the clang mainline

FWIW my view is that there are probably about 50 people in the world (certainly <100) who really care about OpenMP runtime implementation and that some reasonable proportion of those aren’t interested in either open-source or LLVM, so our “total available market” is somewhere between 25 and 50 people of whom maybe 10 are actively doing runtime development. I don’t, therefore, expect a huge community or a lot of contributions…

– Jim

James Cownie james.h.cownie@intel.com
SSG/DPD/TCAR (Technical Computing, Analyzers and Runtimes)

Tel: +44 117 9071438