13:48 <diehlpk_work> jbjnr, yet?

13:48 <jbjnr> yet what?

13:48 <diehlpk_work> Do you have time to shorten the related work?

13:49 <diehlpk_work> I will send an reminder to the committee today and ask them for their response

13:50 <jbjnr> I can look at it tomorrow or tonight maybe

13:50 <jbjnr> the related work needs to be long - they asked for more, not less, we have to shorten everything else

13:54 <diehlpk_work> I meant by shorten to rephrase our German structured sentence to shorter Engish sentence

13:54 <diehlpk_work> *English

14:02 <jbjnr> ok

16:30 <zao> https://stjepang.github.io/2019/01/29/lock-free-rust-crossbeam-in-2019.html

16:30 <zao> I wonder if any of these juicy things can be ported to HPX

16:44 <hkaiser> FWIW, we do have channels

17:29 <zao> Aye. Just interesting to see a bunch of fundamental data types that Rust-land uses to build libraries like rayon and tokio on.

20:10 <jpenuchot> hkaiser: I really think there will be no way to implement Blaze CUDA without modifying Blaze itself... It was clearly not designed with anything else than SMP in mind as alternate way to assign expressions. I've been exploring overloading smpAssign() functions, assign() functions and making my own cudaAssign() functions but in all three cases some modifications *will* have to be made and we must

20:10 <jpenuchot> convince Klaus to do them. I'm thinking of minor modifications though such as adapting the EnableIf conditions for the SMP backend and making the UseAssign_v trait public.