Speedup synchronization

This is inspired by what you reported on the Optane. By reverting the way we loop, we gain around 20% on cosma7 in this function call. It's not game changing but still worth the change.

On EAGLE-25 with one node (28 threads) and the Intel compiler with the usual flags we go from ~510ms per call to ~400ms.

Happy to hear any thoughts you may have on this.

changed the description

@pdraper if you have some time, any thoughts on this would be appreciated.

Seems like a good improvement and the code is clearer, so looks great to me.

Thanks. I'll merge then?

Well, I haven't tested it, but yes go ahead.

oh no rush. I'll wait for your analysis.

added 123 commits

• c94570a7...59adf48a - 122 commits from branch master
• b57b3b38 - Merge remote-tracking branch 'origin/master' into speedup_synchronization

Compare with previous version

So I ran a couple of single node EAGLE_50/.1 jobs on COSMA7, current master and this branch and can see a 15% improvement in the time taken for a drift_all. As advertised. What I didn't expect is to see that the fastest steps were 3 times slower on this branch compared to master (checked once more to see if it was a bad node, but it is real).

Just merged in master, since this was getting quite old and tried once more and this is now as fast... Looks like this is all from print_task_counts. Teach me to use verbosity.

Anyway all looks good, if lost in the noise!

Thanks. Indeed I was not hoping this would create a new revolution.

I'm happy to merge, you ready as well?

All good for me.

merged

mentioned in commit 82502aa5