Hello Pavan, Thank you for the review of the report and feel free to add a link to your page. Could you please clarify whether MPIX_* functions are public/documented? Whether they can be used in applications? I will prepare and upload corrected version of the report tomorrow. Thank you. 07.02.2016, 20:02, "Balaji, Pavan" <[email protected]>:
Thank you, Andrey. This is great and very helpful.
One small comment -- the ABI compatibility initiative is only for MPI_ functionality (i.e., functionality that is a part of the MPI standard). So the removal of MPIX_ functionality (MPICH-specific extensions) is not considered ABI breakage in our case. If you could screen out MPIX_ functionality in your list, it might more accurately depict where we stand.
In mpich-3.2, you are right that we added a bunch of new functions. If you need some notes on that, we moved from MPI-3.0 to MPI-3.1 in that release, which added some additional functions.
Would you mind if we added this link to our ABI page?
Thanks,
О©╫О©╫-- Pavan
О©╫On Feb 7, 2016, at 1:32 AM, Ponomarenko Andrey <[email protected]> wrote:
О©╫О©╫The primary aim of the initiative is for all parties to agree on a schedule for necessary ABI changes, leading to a more stable release cycle and fewer surprises for developers. This project is not limited to just MPICH-derived implementations. Other collaborators are welcome to join and participate in the initiative for the benefit of users, application developers and system administrators alike.
О©╫Hello,
О©╫I've prepared ABI report for the recent versions of the library here: http://abi-laboratory.pro/tracker/timeline/mpich/
О©╫Also I've shared a special report for the latest MPICH 3.2 in order to navigate over the ABI structure in detail (on x86_64): http://abi-laboratory.pro/tracker/abi_view/mpich/3.2/f8b8b/symbols.html
О©╫Hope this helps the Initiative and may be Linux maintainers of the library too. The source code of the analysis tools is available here: https://github.com/lvc
О©╫Thank you.