00:04 <hkaiser> gnikunj[m]: sure, one of the LSU ITS systems

00:04 <gnikunj[m]> hkaiser: would it be possible for me to get an access to it?

00:05 <weilewei> hkaiser got a hpx+apex otf2 on Wombat: https://imgur.com/a/kx8hK5b

00:06 <weilewei> Kevin has great contribution in getting apex profiling improve over these two weeks, now it can track how many bytes allocated in GPU

00:06 <weilewei> Which is great for profiling memory usage for gpudirect stuff later

00:07 <hkaiser> \o/

00:08 <hkaiser> gnikunj[m]: apply for an account here: https://allocations.loni.org/login_request.php, put myself down as the sponsor

00:09 <hkaiser> use the cct email, if possible

00:13 <gnikunj[m]> hkaiser: what do I add in office phone number and office address? your details?

00:13 <hkaiser> yes

00:15 <gnikunj[m]> cool

00:18 <gnikunj[m]> hkaiser: done

00:19 <hkaiser> gnikunj[m]: they will contact me first

00:19 <gnikunj[m]> alright. I'll wait until it is allocated.

00:19 <hkaiser> I'll give you access to an allocation and you can use qbc.loni.org

00:25 <gnikunj[m]> hkaiser: nice!

00:26 <weilewei> hkaiser what if we can't get Cori GPU to work before paper submission?

00:27 <hkaiser> don't panic

00:27 <weilewei> It seems Rostam Diablo node is similar to Cori GPU node...

00:27 <weilewei> ok

00:50 <gnikunj[m]> hkaiser: In which repository should I dump the code I used for my arm paper?

00:50 <gnikunj[m]> vectorized 2d stencil code i.e.

00:50 <hkaiser> I can create one

00:51 <gnikunj[m]> ohh, I thought you would want me to add it to the tutorial repo

00:51 <gnikunj[m]> where we have heller's code of 2d stencil

00:51 <gnikunj[m]> https://github.com/STEllAR-GROUP/tutorials/tree/master/examples/03_stencil

00:57 <gnikunj[m]> hkaiser: how do I make std::rotate work with vector<id_type>?

00:58 <hkaiser> gnikunj[m]: that would work as well, why don't you create a PR then?

00:58 <gnikunj[m]> sure, I'll add one to the tutorial one. Should be easier for others to navigate

01:02 <gnikunj[m]> ok std::rotate works, my bad

01:11 <gnikunj[m]> hkaiser: I have a 1d stencil implementation as well. I added it to hpx/examples/1d_stencil. Do you want me to add it to tutorial as well?

01:11 <hkaiser> that rather belongs to the 1d_stencil examples, wouldn't you think?

01:13 <gnikunj[m]> yes

01:14 <gnikunj[m]> that's why i added it there, do you want me to add another directory in tutorials for 1d stencil is what I'm asking

01:20 <hkaiser> nah, just put it in there

01:20 <gnikunj[m]> cool

01:56 <gnikunj[m]> hkaiser: https://github.com/STEllAR-GROUP/tutorials/pull/6

01:56 <gnikunj[m]> the tests won't pass most likely. NSIMD is not installed on the testing scripts

14:23 <hkaiser> gnikunj[m]: your account has been approved

14:34 <zao> \o/

14:49 <rori> Can we make our website http://stellar-group.org using SSL? Having password to protect editing doesn't make a lot of sense since they are sent in clear text (cf wireshark on http requests) ^^'

14:49 <hkaiser> rori: I thought it was using https :/

14:49 <hkaiser> sure, I'll look into it

14:50 <rori> tks! ;)

14:51 <hkaiser> rori: https works, it doesn't redirect http however

14:54 <rori> ok so I guess the only weird thing is the automatic wordpress link to create a password for a new account gives you the http link, but it should be easy to change in the settings

14:59 <hkaiser> I'll talk to the provider, let's see what they suggest - I can't find how to enforce https for this domain

14:59 <hkaiser> I found it for hpx.stellar-group.org, however

14:59 <rori> np :)

15:07 <hkaiser> rori: I'll keep you in the loop, I have opened a ticket with the provider, let's see what they say

15:08 <rori> tks!

15:15 <hkaiser> rori: done now

16:52 <tiagofg[m]> Hello everyone

16:52 <tiagofg[m]> hkaiser: in order to continue the conversation in the class hierarchy, I have an issue.

16:52 <gonidelis[m]> hello

16:52 <tiagofg[m]> I have a templated component called Agent derived from another templated component named Executor. The Executor has an int variable member named var. When I run an action belonging to Executor to return the variable var works fine. But if the component Agent is derived from another simple component called Resource, along with Executor, when I run the action to return the variable var, return always 0.

16:53 <tiagofg[m]> In other words,

16:53 <tiagofg[m]> Agent<T> : Executor<T> -> Executor component actions works fine

16:53 <tiagofg[m]> Agent<T> : Executor<T>, Resource -> Executor component actions works wrong

16:54 <tiagofg[m]> it's a strange problem and I don't know how to solve it...

16:58 <hkaiser> tiagofg[m]: I'd need to see the code to answer competently... Could you cretae a small example reproducing your issue?

16:59 <tiagofg[m]> yes no problem, I will try to simplify my example

17:03 <tiagofg[m]> I will put the code here, but you can suggest another way.

17:11 <tiagofg[m]> so far, everything works fine, nn == 78

17:14 <tiagofg[m]> when I execute the action,

17:14 <tiagofg[m]> nn==0

17:14 <tiagofg[m]> auto nn = hpx::async<action_type>(f.get()).get();

17:17 <tiagofg[m]> it fails with all actions that work with variable member of Value_Componet, whether it's integers, functions, strings, but if I remove Resource_Component, everything works fine

17:17 <gonidelis[m]> hkaiser:

17:18 <gonidelis[m]> Will tag_invoke be proposed for C++23?

17:20 <tiagofg[m]> hkaiser: I can create a file with this example with all code if it is better

17:20 <K-ballo> it's a paper, so it was already proposed

17:20 <K-ballo> https://wg21.link/p1895

17:24 <gonidelis[m]> K-ballo: duh... :q

17:28 <gonidelis[m]> hkaiser: As standard does not specify any requirements here http://eel.is/c++draft/alg.transform#itemdecl:1 . When creating the CPO should I imitate the requirements we use in the deprecated versions?

17:29 <gonidelis[m]> *But in a selective way for sure. Because, the CPO for hpx::transform will not have execution policy arguments and thus no restrictions should be added about that...

17:31 <K-ballo> fwiw the standard does specify requirements in the old algorithms, via the template param names

17:31 <K-ballo> http://eel.is/c++draft/algorithms#requirements-4

17:33 <gonidelis[m]> K-ballo: wow! thatnks! Hadn't noticed that....

17:43 <gonidelis[m]> hkaiser: How do we dispatch to the segmented algos using the CPO's? I mean ... that we do not use an `is_sgemented` tag anymore...

17:43 <gonidelis[m]> ... as far as your implementation and mikel's are concerned

17:52 <gonidelis[m]> actually mikel did use the segmented counterparts

17:52 <gonidelis[m]> https://github.com/STEllAR-GROUP/hpx/blob/0bbfb3c5c22954b324dad2f43d1fb259998e9e11/libs/algorithms/include/hpx/parallel/algorithms/for_each.hpp#L648

19:13 <hkaiser> gonidelis[m]: : I think we have not touched the way the segmented algorithms are intergated

19:14 <hkaiser> at least not yet - the plan is to try using the CPO mechanism for this as well, but that will be a second step

20:43 <gnikunj[m]> hkaiser: I opened my LONI account and it says that I don't have any allocations. What do I have to do to get one?

20:44 <hkaiser> gnikunj[m]: I did cc you on an email to Dominic this morning

20:45 <gnikunj[m]> yes, I clicked on the link and it says "You do not have any allocations at this time."

20:46 <hkaiser> Dominic will have to add you

20:47 <hkaiser> he probably misunderstood

20:47 <gnikunj[m]> aah alright. I'll wait

20:47 <hkaiser> he has one

20:47 <gnikunj[m]> meanwhile I'll do some testing on medusa

20:47 <hkaiser> sure

20:47 <gnikunj[m]> I updated my distributed benchmarks to have nodes as faulty instead of injecting errors in the code

21:38 <weilewei> Does anyone here use Cori before?

21:39 <weilewei> I got an error when trying to get an interactive node: weile@cori07:~/dev/src/spack/bin> salloc -N 1 -C haswell -q interactive -t 01:00:00salloc: error: Job request does not match any supported policy for escorisalloc: error: Job submit/allocate failed: Unspecified error

21:40 <weilewei> I just follow their instruction: https://docs.nersc.gov/jobs/interactive/

21:57 <hkaiser> weilewei: ask their helpdesk

21:57 <weilewei> hkaiser ok

22:31 <weilewei> hkaiser if I disable DCA MPI, then DCA+HPX runs fine

22:31 <weilewei> So could be some MPI setting is wrong...