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
K-ballo has quit [Quit: K-ballo]
hkaiser has quit [Quit: bye]
<jedi18[m]>
gonidelis: No min_max_result wasn't implemented but I've added it now
bita has quit [Ping timeout: 264 seconds]
surya69 has joined #ste||ar
<surya69>
Hello Is this a right place to discuss about HPX
<heller1>
Hi, yes
<surya69>
Hi I choosed the project implementing iterative solvers for gsoc 2021
<surya69>
Can i get some microtask and bugs to know more about te codebase
<surya69>
the*
surya69 has quit [Quit: Connection closed]
<jedi18[m]>
Will I have to add an algorithm_result_helper for min_max_result?
<jedi18[m]>
In dispatch.hpp
<zao>
Heh, patient person, gave the channel almost nine minutes before bailing :D
<gonidelis[m]>
zao: lol
<gonidelis[m]>
jedi18: did you work it out?
<jedi18[m]>
Yeah it did
<gonidelis[m]>
so you added an overload on dispatch?
<jedi18[m]>
Yep
<gonidelis[m]>
how did you figure it out? that's impressive
<jedi18[m]>
Er.. the other implementations? Still haven't finished testing so I'm not sure if it's right xD
<gonidelis[m]>
jedi18: great work! bravo
<jedi18[m]>
Thanks but it still could be wrong :D
<gonidelis[m]>
let's see
sauravjoshi23 has joined #ste||ar
sauravjoshi23 has quit [Client Quit]
K-ballo has joined #ste||ar
surya69 has joined #ste||ar
<surya69>
Hey All,
<gonidelis[m]>
hey
<surya69>
I choosed the project Implement Iterative Solvers which looks appealing for me because of the scope of problem and I am proficient in C++ .Is this the right place to discuss About this?If Yes can anybody give me some micro tasks or small bugs to perform so that I will get better understanding of the codebase.
<gonidelis[m]>
have you compiled hpx on your machine?
<gonidelis[m]>
have you seen our introductory guide?
<surya69>
Yes I Downloaded dependencies
<surya69>
but got few errors while compiling
<gonidelis[m]>
did you installed it ? just downloading the dependencies won't do the job
<surya69>
Yes installed
<gonidelis[m]>
try to search on our issues for you errors
<gonidelis[m]>
you just said you had a couple of errors
<surya69>
Yes I have...I will try to compile
<K-ballo>
which system are you on? which compiler are you using?
<surya69>
Linux ubuntu
<surya69>
g++ compiler
<gonidelis[m]>
what version of g++
<gonidelis[m]>
?
<surya69>
Ide is VScode and sometimes code::blocks
<surya69>
Sorrry my compiler in ubuntu is gcc
<surya69>
version 9.3.0
<sestro[m]>
Hi, I have a rather specific question regarding the pkg-config files generated HPX. What is the reasoning behind using the full library paths (e.g., `$PREFIX/lib/libhpx_parallelism.so`) instead of the typical `-L` and `-lhpx_parallelism` combination?
<gonidelis[m]>
surya69: post your eror in a gist maybe ;)
hkaiser has joined #ste||ar
tarzeau_ is now known as tarzeau
surya69 has quit [Ping timeout: 240 seconds]
<K-ballo>
sestro[m]: I don't know the answer but I'm curious what prompts you to ask
<srinivasyadav227>
namespace v1 is experimental right?
<sestro[m]>
K-ballo: Our build system (unfortunately not CMake...) has problems with the ordering of the linker flags in case the linker is using `--as-needed`.
<hkaiser>
rori: yt?
<hkaiser>
ms[m]: I would need some guidance for #5238 - I'd appreciate it if you had some time to discuss
<hkaiser>
I meant #5221
<srinivasyadav227>
hkaiser: is #5235 working on your local?
<hkaiser>
srinivasyadav227: I can compile things, yes
nanmiao has joined #ste||ar
<srinivasyadav227>
should we pass any build flags? im still getting build errors
<hkaiser>
srinivasyadav227: ok, what errors do you see?
<hkaiser>
I have to admit that I did not compile everything :/
<hkaiser>
srinivasyadav227: let me correct what I said
<srinivasyadav227>
ok
<hkaiser>
seq/par/par_unseq are in hpx::execution
nanmiao51 has joined #ste||ar
nanmiao51 has quit [Client Quit]
<hkaiser>
datapar and friends should go into hpx::experimental::execution
nanmiao59 has joined #ste||ar
nanmiao59 has quit [Client Quit]
<hkaiser>
srinivasyadav227: the v1 is an inline namespace so shouldn't be relevant most of the time
<srinivasyadav227>
you mean they should be in hpx::experimental::execution or they are already present?
nanmiao has quit [Ping timeout: 240 seconds]
<hkaiser>
srinivasyadav227: they should eventually go there
<srinivasyadav227>
oh ok..I will change them
<hkaiser>
we can move them around with out compatibility work as nobody is using the code currently anyways
<srinivasyadav227>
yea..I will move them, if works I will push the changes soon!
<hkaiser>
srinivasyadav227: let's make it work as is first
<hkaiser>
then we can move things around in the context of the larger refactoring
<hkaiser>
srinivasyadav227: have you looked into the std::experimental::simd types in gcc's standard library yet?
<ms[m]>
hkaiser: here, what's up?
<hkaiser>
hey Mikael
<hkaiser>
#5221
<hkaiser>
the hip builder is failing now, apparently because some code is #ifdef'ed for device code
<hkaiser>
ms[m]: what are the rules to wrap code in HPX_COMPUTE_DEVICE_CODE?
<srinivasyadav227>
hkaiser: ok, I will first try making it work first
<hkaiser>
ms[m]: it did compile before, all I have done is to move the files around (I think)
<ms[m]>
hkaiser: almost everything distributed is wrapped in that, I don't know what the exceptions are though
<hkaiser>
ok
<ms[m]>
ok, I can have a look
<hkaiser>
couldn't we just disable the modules that are distributed eventually?
<srinivasyadav227>
hkaiser: I didn't read completely but yes I read part of it, I got my end semester exams now(would finish by march 17) , so now just focusing on existing datapar build fixing
<ms[m]>
not easily with the current setup at least
<hkaiser>
ms[m]: no, please don't waste your time, I just would like to understand the intent
<hkaiser>
ms[m]: once the headers are moved to modules this should work, shouldn't it?
<hkaiser>
srinivasyadav227: sounds good
<ms[m]>
eventually we should still be able to use hipcc only for certain files, but as long as the usage is setting CMAKE_CXX_COMPILER to hipcc we can't do that
<ms[m]>
headers don't really make a difference
<hkaiser>
nod
<hkaiser>
I'll add some more #ifdef's then
<ms[m]>
👍️
<hkaiser>
thanks
<ms[m]>
I'd like to revisit using hipcc only locally at some point, it may just need some time from the hip toolchain (if it's not already possible)
<hkaiser>
sure
<hkaiser>
no pressure, just trying to understand the issues
<ms[m]>
or ideally we'd get our host/device annotations in check, that would be the correct thing... but that's one of those things that could take a week, or a year :P
<hkaiser>
yah
<hkaiser>
ms[m]: so was/is the src directory excluded by default from being compiled by hipcc?
<ms[m]>
hkaiser: no, literally everything is compiled by hipcc
<hkaiser>
ok, then I don't understand why things compiled with hip before ...
<hkaiser>
ms[m]: what is produced by the build system on the hip side? a functioning (local only) runtime system?
<hkaiser>
brb
hkaiser has quit [Read error: Connection reset by peer]
hkaiser has joined #ste||ar
<gonidelis[m]>
hkaiser: see pm please :)
<ms[m]>
hkaiser: just the stuff that's annotated for the device, which means some utilities, executors and things like that
<ms[m]>
the runtime (local or distributed) itself doesn't actually get compiled for the device
nanmiao has joined #ste||ar
<ms[m]>
hkaiser: do you know if vc 1.3.3 would do for now (it's what's in the ubuntu repos, and it's from 2017), or would it need a newer one? I guess you just got the latest from vcpkg?
<hkaiser>
ms[m]: i don't know
<hkaiser>
ms[m]: so I can remove the whole runtime class from hipcc?
<hkaiser>
gonidelis[m]: see pm, pls
<ms[m]>
hkaiser: probably yes, if you mean from device code, but it should most likely not be necessary