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
tufei_ has joined #ste||ar
tufei has quit [Ping timeout: 255 seconds]
hkaiser has quit [Quit: Bye!]
prakhar has joined #ste||ar
Yorlik__ has joined #ste||ar
Yorlik_ has quit [Ping timeout: 252 seconds]
prakhar has quit [Quit: Client closed]
HHN93 has joined #ste||ar
prakhar has joined #ste||ar
<HHN93> would vectorization be enabled if I have compiled hpx in debug mode?
<HHN93> debug mode should turn off all optimisations, right?
<satacker[m]> In `-O0` there aren't any
HHN9310 has joined #ste||ar
HHN9310 has quit [Client Quit]
HHN936 has joined #ste||ar
<satacker[m]> RelWithDebInfo has optimized code with debug symbols
<HHN936> ok will try it out, thank you
HHN936 has quit [Client Quit]
<satacker[m]> You can just print the flags from inside your cmakelists
HHN93 has quit [Ping timeout: 260 seconds]
prakhar has quit [Ping timeout: 260 seconds]
prakhar has joined #ste||ar
<prakhar> When following the HPX quick start guide, I used ```vcpkg install hpx``` to install hpx and then followed the hello world example but I was unable to build it, so I used the GitHub repo instead, and had a few dependency issues like asio and some others, after fixing those, and running sudo make install in build, I was able to generate
<prakhar> hpxconfig.cmake and hence include it in the example code. I just wanted to point this out, since I couldn't find these while following the quick start guide.
K-ballo1 has joined #ste||ar
K-ballo has quit [Ping timeout: 268 seconds]
K-ballo1 is now known as K-ballo
prakhar has quit [Quit: Client closed]
prakhar has joined #ste||ar
prakhar has quit [Quit: Client closed]
hkaiser has joined #ste||ar
<zao> Note that normally RelWithDebInfo gets you -O2 and symbols while Release gets you -O3 and no symbols.
<zao> As things like the various amounts of -ftree-vectorize vary between those optimization levels, you may see a significant difference in vectorization between O2 and O3, something that also has changed lately with newer GCC versions as they enable slightly more in O2.
<Yorlik__> Isn't HPX always compiled with O2 at max?
Yorlik__ is now known as Yorlik
<hkaiser> Yorlik: not with Release mode (gcc/clang)
<Yorlik> IC. Thanks for the clarification.
<hkaiser> well, we don't do anything special, it's all using cmake defaults
<Yorlik> I guess I confused something I remembered from RelWDebInfo with Release.
Guest19 has joined #ste||ar
Guest87 has joined #ste||ar
Guest87 has quit [Client Quit]
Guest19 has quit [Ping timeout: 260 seconds]
K-ballo1 has joined #ste||ar
diehlpk_work has joined #ste||ar
K-ballo has quit [Ping timeout: 276 seconds]
K-ballo1 is now known as K-ballo
K-ballo has quit [Remote host closed the connection]
K-ballo has joined #ste||ar
HHN93 has joined #ste||ar
<HHN93> for benchmarking is it better if I build HPX again in a separate location in release mode or should I rather use an container for it?
<HHN93> even if I build it again in release mode I would have to deal with making sure CMAKE is linking with the correct HPX binaries, any idea how I can do it?
HHN93 has quit [Client Quit]
diehlpk_work has quit [Ping timeout: 246 seconds]
<sarkar_t[m]> <hkaiser> "sarkar_t: I think you could..." <- Hi hkaiser I found the following parallel algorithms that already support executors that are based on s/r:... (full message at <https://libera.ems.host/_matrix/media/v3/download/libera.chat/07828f175388599f08b184e375763cbda522ac20>)
<sarkar_t[m]> Please let me know if there is any discrepancy.
<sarkar_t[m]> Also, can you suggest what can I do now? Should I try to add s/r based executor support for any parallel algorithm? If so, do you have any suggestions on which one to pickup?
<sarkar_t[m]> Or do you suggest I work on some other task that can help me more?
<hkaiser> sarkar_t[m]: yes, try to find out what's necessary for that and what it means
<hkaiser> that list is probably too short - although it's the list of algorithms we have tests for