hkaiser changed the topic of #ste||ar to: STE||AR: Systems Technology, Emergent Parallelism, and Algorithm Research | stellar-group.org | HPX: A cure for performance impaired parallel applications | github.com/STEllAR-GROUP/hpx | This channel is logged: irclog.cct.lsu.edu
diehlpk_work_ has quit [Ping timeout: 240 seconds]
jehelset has quit [Ping timeout: 240 seconds]
nanmiao has joined #ste||ar
diehlpk_work_ has joined #ste||ar
<diehlpk_work_> hkaiser, gnikunj[m] gonidelis[m] Open source projects can apply to be mentoring organizations from February 7- 21 at 1800 UTC.
diehlpk_work_ has quit [Remote host closed the connection]
<hkaiser> ok, that gives us a bit of time
<gnikunj[m]> We need to figure out more projects meanwhile
<hkaiser> indeed
<gonidelis[m]> gnikunj: hkaiser Patrick Diehl those two guys just ruined the whole thing. they completed every given project we got and now we need to figure knew stuff. i knew we shouldn't take them. thanks for that! srinivasyadav227 jedi18
<hkaiser> lol
<gnikunj[m]> Hahaha
hkaiser has quit [Quit: Bye!]
jehelset has joined #ste||ar
diehlpk has joined #ste||ar
<srinivasyadav227> gonidelis: 😂😂
diehlpk has quit [Quit: Leaving.]
jehelset has quit [Ping timeout: 240 seconds]
jehelset has joined #ste||ar
jehelset has quit [Ping timeout: 240 seconds]
jehelset has joined #ste||ar
hkaiser has joined #ste||ar
<jedi18[m]> hkaiser: gnikunj will you be having the sender receiver discussion in person or through a video meet?
<gnikunj[m]> jedi18: I’m planning to do a meet tomorrow. Feel free to join. Heads up - it will mostly be based on some of the doubts I have to implement them and decisions taken towards it.
<jedi18[m]> Yep I'd like to join if it isn't too late for me. Sure I'll just listen in, will probably help in understanding it
jehelset has quit [Ping timeout: 240 seconds]
<hkaiser> gnikunj[m], jedi18[m]: we could do that right in the morning at 9am, later in the day I will be tied up
<gnikunj[m]> hkaiser: I’m free at 9. Let’s do it then!
diehlpk has joined #ste||ar
diehlpk has quit [Quit: Leaving.]
diehlpk has joined #ste||ar
diehlpk_work has joined #ste||ar
<jedi18[m]> hkaiser: sure!
diehlpk has quit [Quit: Leaving.]
jehelset has joined #ste||ar
diehlpk has joined #ste||ar
diehlpk has quit [Client Quit]
<pedro_barbosa[m]> is it possible to add the struct float4 to HPXCL?
<hkaiser> pedro_barbosa[m]: I don't see a reason why that shouldn't be possible?
<pedro_barbosa[m]> I meant to implement it directly in HPXCL so that I wouldn't need to use
<pedro_barbosa[m]> ```
<pedro_barbosa[m]> #include "vector_types.h"
<pedro_barbosa[m]> ```
<pedro_barbosa[m]> s///, s///
<pedro_barbosa[m]> I'm having some performance issues and I don't think this would solve it but I'm not sure what else could be causing it
<hkaiser> pedro_barbosa[m]: well, you know HPXCL by now much better than anybody here - you should tell us ;-)
<pedro_barbosa[m]> I wish I knew, I'm going to try a few more things to see if I can solve it
<diehlpk_work> pedro_barbosa[m], Do you mean that here
<pedro_barbosa[m]> Not quite, I have an example in pure CUDA where the kernel takes 20s-22s to run, when I do it with HPXCL, which has an array instead of a float4 it takes 30s-32s
<diehlpk_work> Can you show me the exmaple?
<pedro_barbosa[m]> this is the HPXCL version
<pedro_barbosa[m]> the cuda version you can go to /usr/local/cuda-11.2/samples/5_Simulations/nbody
<pedro_barbosa[m]> assuming that's where you have CUDA
<diehlpk_work> pedro_barbosa[m], You can add flaot 4 to HPXCL
<diehlpk_work> Just wrap the cuda commands into the action as you did for p2p
<diehlpk_work> Have you measured the kernel runtime?
<diehlpk_work> Or is the time above the total time?
<pedro_barbosa[m]> I've measured the kernel itself with nvprof
<pedro_barbosa[m]> This is the HPXCL execution
<pedro_barbosa[m]> This is the CUDA
<pedro_barbosa[m]> <diehlpk_work> "Just wrap the cuda commands into..." <- I'll try to do it with the #include "vector_types.h" and if it doesn't get better I'll try to implement it into HPXCL
jehelset has quit [Ping timeout: 240 seconds]