04:20 <Neeraj> any mentor are there ?

05:52 <satacker[m]> <aryan_j[m]> "I have installed HPX on my..." <- There is an example mentioned in https://satacker.github.io/blog/gsoc-2022-work which is under unit tests for the execution_base core library part of HPX. But if you get the gist of S/R you should be able to write one of your own. Apart from this one, stdexec has the same one from where I have referred to the type erased task type. libunifex have one as well IIRC.

06:14 <satacker[m]> For now, I would like to see some examples (might be trivial) that use this type erased task type and S/R and showcase HPX functionality through this example.

08:32 <HHN93> Can someone have a look at https://github.com/STEllAR-GROUP/hpx/pull/6199 , I have tried to add support for par_unseq by adding is_unseq() as a function parameter which calls the appropriate overload

08:33 <HHN93> The builds seem fail due to policy being an unused variable, any advised change on how to fix it?

08:34 <HHN93> is there any macro to deal with UNUSED variables?

09:42 <K-ballo> how can policy be unused? waasn't it there before already

12:50 <HHN93> any help on why there are 2 perf test reports per commit? and why they are often different

12:53 <zao> HHN93: They're from two different clusters, Rostam (at LSU?) and Piz Daint (at CSCS in Switzerland).

12:54 <zao> As such, they may have a bit different performance characteristics from CPUs, interconnects, etc.

12:54 <HHN93> ohkay, thank you

12:57 <hkaiser> HHN93: please ignore the report from rostam, that one is known to be broken

12:58 <hkaiser> the other one (coming from daint) is more reliable, but can be ignored by you as well as your changes didn't affect performance for sure

12:58 <HHN93> yeah, I was puzzled by the rostam report

12:59 <hkaiser> we're currently working on fixing that

13:00 <HHN93> any brief explanation on what the problem is? it seems to consistently flag PRs as negatively performing?

13:01 <hkaiser> HHN93: two things: a) we store a perf baseline in our repository that is not correct, and b) the rostam results have a large variation (up to 20%), currently

13:03 <HHN93> ok got it.

13:08 <hkaiser> HHN93: is your PR ready now?

13:08 <HHN93> par_unseq?

13:09 <hkaiser> #6200

13:09 <HHN93> requested changes have been pushed

13:10 <hkaiser> it complains about cmake-formatting being incorrect: https://circleci-tasks-prod.s3.us-east-1.amazonaws.com/forks/storage/artifacts/7eaf89a1-5802-4ada-8ae5-685e6d373f40/585531300/fc9ba0e1-e98f-4412-8549-0e49ba7a1331/0/hpx/artifacts/modified_cmake_format_files.txt?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=ASIAQVFQINEOHAKFAG6O%2F20230318%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20230318T130853Z&X-Amz-Expires=60&X-Amz-Sec

13:10 <hkaiser> urity-Token=IQoJb3JpZ2luX2VjEH0aCXVzLWVhc3QtMSJGMEQCIBjiePMaOi%2B7DvHkrUYU5Vna56NfB0KQ40fjV9YMozrJAiAKhIGfuLUFXeDPceW1aM3HQnvTkbv8Mr0J8t5HkVHgbCqrAghGEAMaDDA0NTQ2NjgwNjU1NiIMJAQpDyG2gQtC6%2F3CKogCTu3YQiEoiIgOd4ffkYab9XRMDRikc%2BE8qBSjREitVrJxfrhs7WqkdKxiDkIs4Z2%2FDzHVbE2zxCgZS4ZfK8sqkzQjT6FmMjsDmDsmCR2e%2F4S%2B97x7JWOud%2BD%2BE7rqsTjB2n3Hzvb93RcEGlZ4dTxPf99bkwRteqMPaVDfQ2NQj3N6TpEuUKoL7sndxY2C3y6Y8Cxsmg5YBVsl3pDcdx81lfEM3ywOJtX0

13:10 <hkaiser> s5wbeWen9Eg9X6HgCzE8rwYEV1cGuoH4JjPuLnvKd64mcEG7V1%2FH3tKv4J50kk7%2B%2FPF0HXT4W8lkgN4ILqtdQn1u%2FOrQNEsqzJgmaAr0S%2BlELX2629t5OUoC%2F3NaeB86MKzr1qAGOp4BtQQMJyRN8t4KfZwGrjJNIX6sxfJVgsUj%2F1t%2F5n2xdRjwSCioFA92kBWuuZF6GZMicxC19iCJdD7PwW2eNZd8%2Fz9goBLIvSoIFDXKUhiPx%2FPb9%2FWA2XvOON39xmMKuGno5O3JPyMB97kSV9vnqX9%2FSDGR%2Blrg89t52BvLIPQLnbMfnR13hl046k1TOsnjXLaXwiISATEv8w1zyWIGjUQ%3D&X-Amz-SignedHeaders=host&x-id=GetObject&X-Amz-Signa

13:10 <hkaiser> ture=5d3853a7382ec0140bc7abcb6bd9e525ba372af51a2bd66533aba9cec22fc696

13:10 <hkaiser> uggh

13:10 <hkaiser> https://gist.github.com/hkaiser/c5ad5dcb1aa0d00cfa78cff9622b329e

13:11 <HHN93> is there any script to fix the formatting?

13:11 <hkaiser> cmake-format

13:11 <hkaiser> https://cmake-format.readthedocs.io/en/latest/cmake-format.html

13:19 <HHN93> btw I am also facing an issue with clang format

13:19 <HHN93>             // clang-format off

13:19 <HHN93>             HPX_IVDEP HPX_UNROLL HPX_VECTORIZE for (std::size_t i = 0;

13:19 <HHN93>                                                     i != count; ++i)

13:19 <HHN93>             {

13:19 <HHN93>                 sift_down(first, comp, proj, len, start - i);

13:19 <HHN93>             }

13:19 <HHN93>             // clang-format on

13:19 <HHN93> formats into this

13:24 <HHN93> pushed #6199 #6200 @hk

13:24 <HHN93> hkaiser

13:24 <hkaiser> HHN93: is it because of the `// clang-format off` comment?

13:25 <HHN93> honestly, no idea. I just reverted and did the same thing again and it worked

14:01 <HHN93> How do we do performance analysis, I have mostly just used chrono to check the time taken to run a function. Are there any better methods?

14:34 <Neeraj> Hii hkaiser can you tell me that ,What are the most important performance requirements for the GID associative container in our specific use case?

14:35 <Neeraj> and give me the location where the current hash map and binary tree implementation is performing in practice

14:38 <Neeraj> You gave me those link and material I went through them and get familiar with AGAS and parallel programing now I am focusing on project

14:40 <Neeraj> and give me some simple bud/issue those can make me get involve with HPX

14:43 <hkaiser> Neeraj: that's all good questions

14:45 <Neeraj> I am also implemented a little bit N-array tree in that every Node contain ,component and component GIDs

14:46 <hkaiser> I think part of your project could be creating an artificial benchmark

14:46 <hkaiser> in the phd I linked, Parsa was using both, an artificial benchmark and a real application, IIRC

14:49 <Neeraj> yeah in this project basically made HPX runtime system more faster , am I right

14:51 <hkaiser> certain functionalities, yes

14:51 <Neeraj> like searching GIDs and insertion also

14:52 <hkaiser> yes, lookup of global objects - find out where an object is located

14:52 <Neeraj> I made a advantages and disadvantages of those data structures whose relevant to the project

14:53 <hkaiser> Neeraj: btw, we do have an occasional test failure related to this - I'd appreciate it if you were interested in trying to fix that

14:54 <Neeraj> yeah gave me instruction I'll trying my best to solve that bug

14:57 <Neeraj> Sure, I would be happy to work on fixing the test failure. Can you provide more details on what the issue is and how I can get started with fixing it?

15:00 <Neeraj> I am waiting here from early morning( INDIA) I think that time You all were sleeping

15:00 <hkaiser> Neeraj: I'll find the link and post it here

15:01 <hkaiser> might take a moment or two...

15:02 <Neeraj> Ok

15:16 <Neeraj4> sorry hkaiser there is a network problem

15:16 <Neeraj4> can you share again that link

15:19 <Neeraj15> now send so I secure that link

15:19 <Neeraj15> hkaiser

15:19 <Neeraj15> are you here

15:20 <zao> Note that there's channel logs in the topic if you're offline.

15:23 <Neeraj15> from where I can get those chat

15:27 <Neeraj15> hkaiser

15:30 <zao> (there has been no links shared yet)

15:31 <Neeraj15> hkaiser He still didn't share the link

15:49 <HHN93> Neeraj15 https://irclog.cct.lsu.edu/ste~b~~b~ar/2023-03-18

15:49 <HHN93> you can check channel logs here

15:55 <Neeraj15> ok HHN93

16:29 <Neeraj> hkaiser are you here ?

16:35 <Neeraj88> ok good night hkaiser share the link and some instruction also with that I can get more about it ,I'll get that link by link HHN93 provided

17:20 <pansysk75[m]> hkaiser: Would a distributed algorithm using a partitioned vector be sufficient to add as a perf-test report? Or do you perhaps have something else in mind?

17:27 <hkaiser> pansysk75[m]: that would be a good starting point

21:30 <HHN93> hkaiser any advice on how performance analysis is to be done?