K-ballo changed the topic of #ste||ar to: STE||AR: Systems Technology, Emergent Parallelism, and Algorithm Research | stellar.cct.lsu.edu | HPX: A cure for performance impaired parallel applications | github.com/STEllAR-GROUP/hpx | Buildbot: http://rostam.cct.lsu.edu/ | Log: http://irclog.cct.lsu.edu/
hkaiser has joined #ste||ar
hkaiser has quit [Quit: bye]
<bering[m]>
Are there user-defined literals for std::size_t in HPX?
<Deepak1411[m]>
Hey, there I was trying to build hpx on my windows but, when you go for installing boost x64 it throws access denied . Can someone please help me out. I have given the images regarding issue
<srinivasyadav227>
idk if this helps, but can you try with cmd as admin?
<Deepak1411[m]>
I tried it in cmd only
<srinivasyadav227>
Deepak1411: no i mean run cmd as "Run as administrator"
<Deepak1411[m]>
srinivasyadav227: yes installed it
<srinivasyadav227>
Deepak1411: ok, sorry, i am not sure what might be the actual program, hkasier might be right person to help you with this :), anyways assuming yours is 64bit windows machine and have latest Visual studio, you may try installing vcpkg cleanly again from the above link , and try again, as this was working for me.
<Deepak1411[m]>
ok fine, I'll try once again and yes using a 64bit machine
<Deepak1411[m]>
thanks for responding 🙂
<srinivasyadav227>
:)
klaus[m] has quit [Ping timeout: 245 seconds]
gdaiss[m] has quit [Ping timeout: 245 seconds]
bering[m] has quit [Ping timeout: 245 seconds]
Girlwithandroid[ has quit [Ping timeout: 245 seconds]
rachitt_shah[m] has quit [Ping timeout: 245 seconds]
klaus[m] has joined #ste||ar
gdaiss[m] has joined #ste||ar
bering[m] has joined #ste||ar
rachitt_shah[m] has joined #ste||ar
Girlwithandroid[ has joined #ste||ar
<zao>
Deepak1411[m]: You may be running into antivirus or other protection like SmartScreen preventing access to things. Also make sure that the Windows installation is 64-bit (easiest seen in the Win+Break About screen).
nanmiao has quit [Quit: Connection closed]
hkaiser has joined #ste||ar
<gonidelis[m]>
hkaiser: do you have 20 minutes within the day?
<hkaiser>
gonidelis[m]: I could meet at 10am CDT
<gonidelis[m]>
cool
<gonidelis[m]>
thanks!
<Deepak1411[m]>
zao: Yes finally it is done, it was the antivirus protection stoping it , 🙂
<zao>
Sweet.
chuanqiu has joined #ste||ar
<diehlpk_work>
ms[m], gdaiss[m] Could you please add some content to 7.5. Kokkos vs CUDA and add the HIPS results there?
chuanqiu has quit [Quit: Connection closed]
chuanqiu has joined #ste||ar
<hkaiser>
gonidelis[m]: I'm available now
<gonidelis[m]>
hkaiser: i am having lunch, would you be available in 15?
<diehlpk_work>
ms[m], Please add your affiliation to the paper
<hkaiser>
gonidelis[m]: sure, np
<gnikunj[m]>
hkaiser: sent you the email. Let me know if you want me to add/delete anything.
<gonidelis[m]>
hkaiser: link
chuanqiu has quit [Quit: Connection closed]
<bering[m]>
There is a ~300% performance regression for the callback mode of the CUDA executor, using its performance test.
<bering[m]>
The regression is introduced by commit
<bering[m]>
b5a7913aad (Fix check in unregister_thread and add register bool, 2020-10-14)
nanmiao has joined #ste||ar
<hkaiser>
bering[m]: could you please create a ticket for this?
<bering[m]>
Alright.
<rori>
Thanks for noticing, this is on me, when I did the hip support, will have a closer look!
<ms[m]>
bering: that's specifically that commit, correct? the immediately preceding commit works perfectly fine (and performs better) for you?
<bering[m]>
Yes.
<ms[m]>
also thanks for finding the commit!
<ms[m]>
all right, we'll see what we can do
<ms[m]>
how critical is the callback version for you? do events work well for you?
<bering[m]>
It is not critical at all. So, no worries. I use it only for comparison and can simply use the results from the good version.
<ms[m]>
diehlpk_work: will do
<bering[m]>
I was not quite on current master while testing, though. So, I will double-check on master and create a ticket then. Probably it went away since then.
<bering[m]>
The event mode does not seem to be affected.
<ms[m]>
bering: most likely it won't have been fixed on master, unless there's some spooky action at a distance... the changes from that commit are still on latest master
<gonidelis[m]>
hkaiser: without optimizations all threads are active 99% for a cilk app. with the optimizations all threads are active 50%...
chuanqiu has joined #ste||ar
chuanqiu has quit [Quit: Connection closed]
<hkaiser>
gonidelis[m]: ok, so at least mutliple cores are activated ;-)
<gonidelis[m]>
Could it be that clang++ calculates the loop on compile time ??
<hkaiser>
look at the assembly, that will tell you
<hkaiser>
gonidelis[m]: or do a relwithdebinfo build (i.e. with optimization and debug symbols) and use a debugger
<zao>
Deepak1411[m]: There's way more information in the Output pane.
<zao>
Output -> Build, that is.
<hkaiser>
Deepak1411[m]: what version of VC do you use?
<hkaiser>
VStudio, that is
<zao>
hkaiser: For reference, I just did an x64 RelWithDebInfo build of hpx master examples+tools with VS 16.10.0, Boost 1.76.0, hwloc 2.4.1; deps from upstream binaries.
<zao>
(successfully, I might add)
<gonidelis[m]>
zao: now try doing one on Cilk ;p
<zao>
Blissfully unaware what Cilk is :D
<gonidelis[m]>
the best parallel runtime on the world
<Deepak1411[m]>
<hkaiser "Deepak1411: what version of VC d"> VS 2019
<gnikunj[m]>
hkaiser: am I on the wrong zoom, or is there no meeting today?
<zao>
Deepak1411[m]: What particular minor version, there's a lot of them :)
<zao>
(you can see it in the installer or the About window)
<Deepak1411[m]>
zao: Microsoft Visual Studio Community 2019
<Deepak1411[m]>
Version 16.9.5
<gonidelis[m]>
* the best parallel runtime in the world
<zao>
I'm on 16.10.0, not sure how far back we are supposed to support.
<zao>
hkaiser: ^
<hkaiser>
that's fairly recent
<hkaiser>
Deepak1411[m]: we'll need more information to be able to help
<zao>
Deepak1411[m]: I recommend taking the text from the Output pane's "Build Order" listing to get the actual warnings, errors and notes; and put it in a GitHub Gist and link that here.
<zao>
The errors listing is a bit too short-worded and skips most of the context needed to understand an error.