13:34 <weilewei> hkaiser ms[m] @auriane meeting now

13:34 <hkaiser> weilewei: can't connect

13:34 <ms[m]> 👍️

13:34 <weilewei> hkaiser ok, no worries

14:09 <hkaiser> ms[m]: now it's telling me that master and the new branch are identical, maybe it's just the github webview that's behind...

14:10 <hkaiser> strange

14:13 <gnikunj[m]> hkaiser: I'm done with a basic layout of the paper. I've made several changes to the previous version that makes it look more appealing as a research. Also, I've added our design ideas for distributed resilience and why we chose the current distributed design.

14:13 <hkaiser> ok, good now :/

14:14 <hkaiser> gnikunj[m]: ok, let me have a look

14:14 <diehlpk_work> hkaiser, Which is the parcel colasing paper?

14:15 <hkaiser> diehlpk_work: Bibek Wagle, Mohammad Alaul Haque Monil, Kevin Huck, Allen D. Malony, Adrian Serio, and Hartmut Kaiser: Runtime Adaptive Task Inlining on Asynchronous Multitasking Runtime Systems, In Proceedings of the 48th International Conference on Parallel Processing (ICPP 2019). ACM, New York, NY, USA, Article 76, 10 pages. doi: 10.1145/3337821.3337915

14:15 <diehlpk_work> Ok, thanks

14:15 <diehlpk_work> This was missing for Bita's paper

14:16 <ms[m]> hkaiser: strange indeed...

14:17 <diehlpk_work> hkaiser, I have read the updated JOSS paper and it reads good

14:17 <diehlpk_work> Have you poked David to read it again?

14:23 <gnikunj[m]> hkaiser: btw I have an idea for distributed 1d stencil that will not involve large data movement. I'll try to implement it tonight. Otherwise, I'll contact Zahra tomorrow.

14:28 <diehlpk_work> gnikunj[m], Bibek read the paper and made some remarks. Can you have a look?

14:28 <gnikunj[m]> diehlpk_work: on it

14:29 <diehlpk_work> I can take care of the abstract and conclusion

14:30 <gnikunj[m]> diehlpk_work: I think he added remarks only in abstract and conclusion. Others are todos that I added.

14:31 <diehlpk_work> gnikunj[m], What would you show in the table with the perf reuslts?

14:31 <gnikunj[m]> diehlpk_work: just the counter numbers. In case the reviewers are picky about why we wrote it silently in a paragraph and not add a table that clearly shows the numbers.

14:32 <gnikunj[m]> I have no clue if that will happen or not though. That's why I added that remark.

14:32 <diehlpk_work> We have not much space left. So would be a small tbale

14:33 <diehlpk_work> I think, I will extend the abstract, because it is very short

14:33 <gnikunj[m]> It will be similar to Table III. I had to add Table III as I rely on all numbers that I list.

14:33 <diehlpk_work> Sure, please add it

14:33 <gnikunj[m]> alright, I'll add one for Rpi4 then

14:34 <diehlpk_work> I will work this morning on a different paper, but can work on tthis later today

14:34 <gnikunj[m]> ok, thanks. I'm working on resilience work today. I'll add the table tomorrow morning (my time).

14:36 <hkaiser> diehlpk_work: I wanted to wait for you to read it over before pinging David

14:37 <diehlpk_work> hkaiser, Ok, I am happy with it. Let me do one more spell and grammar check and I will ping him

14:38 <hkaiser> diehlpk_work: thanks, Katie would be glad to read it over as well

14:38 <hkaiser> (so she told me)

14:38 <diehlpk_work> hkaiser, I was thinking that katie will do the final read

14:39 <diehlpk_work> Let us get it approved from Dave and she can read the final version before we submit the paper

14:39 <hkaiser> ok

14:39 <diehlpk_work> He might wants to see more changes

15:25 <jbjnr> Testing. Finally given up completely on the matrix riot thingy and installed hexchat on the laptop. Just seeing if it auto logs me in as the right person etc

15:30 <gnikunj[m]> hkaiser: I've sent the email to Adrian, Jackson and Keita

15:30 <hkaiser> nod, good

15:30 <hkaiser> thanks

15:30 <gnikunj[m]> I need to send an email to the organizers now. How do I frame the email?

15:33 <hkaiser> gnikunj[m]: say that you still work on running benchmarks and that you think those results will significantly improve the submission

15:35 <gnikunj[m]> hkaiser: ok, let me try.

15:36 <jbjnr> I'm testing hexchat, could someone please mention my nickname so I can see if I get an alert. Thanks

15:37 <rori> jbjnr

15:37 <gnikunj[m]> jbjnr: does it work?

15:37 <jbjnr> yes. it worked, thanks

15:38 <rori> sorry latency ^^'

15:39 <gnikunj[m]> jbjnr: you shifted to hexchat at a time when I switched to element ;-)

15:40 <ms[m]> jbjnr: ping

15:40 <ms[m]> element has a nice sidebar with mentions nowadays, but don't bother switching back... ;)

15:50 <hkaiser> weilewei: I can't login to zoom, maintenance...

15:51 <jbjnr> what is element? I used hexchat on my windows machine, but that was permanently on, so I always saw logs and mentions. My laptop is only on during working hours and runs linux. I didn't know hexchat had a linux version, so I installed it now, but will happily use another if it works well.

15:51 <zao> hkaiser: Bit of an outage today with SSO and stuff - https://status.zoom.us/incidents/1z2lrf4nrv8p

15:52 <gnikunj[m]> hkaiser: email to organizers sent. Now I'll work on the paper.

15:52 <jbjnr> I wanted to use riot because it kepts logs when you were offline, but it turns out to be so rubbish I gace up

15:52 <ms[m]> element is just riot renamed

15:52 <zao> jbjnr: Mate of mine was losing days of chat as it appeared to be connected but didn't actually put chat through, did you have that?

15:52 <hkaiser> weilewei: worked now

15:52 <gnikunj[m]> jbjnr: element is a new name for riot. The UI is pretty much the same with a few tweaks here and there.

15:52 <hkaiser> zao: yah

15:58 <jbjnr> zao, not so much that as it never notified me of when I was mentioned and when it did it was impossible to find the message in the history because the volume of messages now seems so high that scrolling through hundreds of papges looking for a mention was too annoying

18:34 <gonidelis[m]> final gsoc evaluations are on

18:34 <gonidelis[m]> ;)

18:34 <K-ballo> gsoc finished?

18:34 <K-ballo> is the summer over?

18:34 <gonidelis[m]> it's the last week

18:35 <gonidelis[m]> so we kinda fill these evaluations

18:35 <gonidelis[m]> summer is over in one week

18:35 <K-ballo> odd

18:35 <K-ballo> winter still has another month down here

18:35 <gonidelis[m]> what thing?

18:37 <gonidelis[m]> well september tends to be quite summery

18:38 <gonidelis[m]> do you like have cold down there?

18:55 <K-ballo> yeah, though the worse should already be behind

20:04 <weilewei> hkaiser meeting now

21:47 <weilewei> hkaiser I changed the affinity code in std thread in DCA, it is same slow as original code. HPX is achieves 20% speedup

22:33 <hkaiser> weilewei: nod, as expected

22:33 <hkaiser> you can mention this in the paper

22:34 <weilewei> hkaiser what should I say?

22:35 <hkaiser> that even if the dca code is modified to use the same affinity settings a HPX (which requires explicit changes) the performance is not improved - i.e. th affinity settings do not cause the speedup of HPX

22:36 <weilewei> I see, let me add that