hkaiser 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/ | GSoC2018: https://wp.me/p4pxJf-k1
<parsa>
hkaiser: ping
<hkaiser>
parsa: here
<parsa>
i had an issue i couldn't quite describe… see #414
<hkaiser>
parsa: I don't see the problem, [1, 2] _is_ a vector
<parsa>
ah, paste problem… i meant [[1], [2]]
<parsa>
updated
<hkaiser>
k
<hkaiser>
strange indeed
<parsa>
the add1d2d ends up getting called in case of that code
<parsa>
hkaiser: none of those three functions ever return anything except false for this code… it never reaches that line
<hkaiser>
parsa: hmmm
<hkaiser>
but that's where the np.array() is converted to the blaze type
<hkaiser>
and it will happen only if one of those returns true
eschnett has joined #ste||ar
<parsa>
hkaiser: no joke… load(src, convert) is called twice with the updated python code before add starts, both times all three functions called in it return false… if that's not where the mapping happens i have no clue where it does, but add primitive does receive that numpy matrix as a vector
<github>
hpx/master 2415fd5 Mikael Simberg: Fix using pu-step and pu-offset
<github>
hpx/master a452255 Mikael Simberg: Merge pull request #3313 from msimberg/fix-pu-offset-step...
nikunj has quit [Ping timeout: 260 seconds]
nikunj has joined #ste||ar
verganz has joined #ste||ar
hkaiser has joined #ste||ar
nikunj has quit [Ping timeout: 250 seconds]
simbergm_ has joined #ste||ar
diehlpk has joined #ste||ar
nikunj has joined #ste||ar
parsa has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
diehlpk has quit [Ping timeout: 256 seconds]
<hkaiser>
heller: yt?
stmatengss has joined #ste||ar
jaafar has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
jaafar has quit [Read error: Connection reset by peer]
jaafar has joined #ste||ar
jaafar has quit [Read error: Connection reset by peer]
jaafar has joined #ste||ar
jaafar has quit [Ping timeout: 240 seconds]
parsa has joined #ste||ar
<hkaiser>
parsa: yt?
<parsa>
yas
<hkaiser>
hey
<hkaiser>
wrt #414
<hkaiser>
the issue is that for np.array([[1], [2]]) the dtype in Int32
<hkaiser>
that causes a bunch of conversion problems leading to what you see
<hkaiser>
dtype _is_ Int32
<parsa>
i don't know much about how the integration works, how do i approach this issue?
<parsa>
also does that mean if i use [[1.], [2.]] that'd work?
<hkaiser>
one possible solution might be replacing the std::int64_t in the primitive_argument_type with node_data<std::int64_t>
<hkaiser>
yes, if you'd use [[1.], [2.]] it should work
<hkaiser>
well, there is another bug in the python translation layer that might prevent even for this to work
<parsa>
and does this also mean right now all numpy arrays of int end up creating a additional copy of themselves as a node_data<double>?
<hkaiser>
yes
<parsa>
:))))
<hkaiser>
parsa: I think we should implement node_data<int64_t>
<hkaiser>
we need it anyways
<parsa>
or force int arrays to be converted to double?
<hkaiser>
yet another refactoring requiring a lot of work
<parsa>
mapped*
<hkaiser>
yah, I thoght about this but I don't know yet how to do that
<parsa>
rabbit holes everywhere
<hkaiser>
yes
<parsa>
hkaiser: we don't have an imminent need for it… let's just add to the ticket and leave it be for now. we've got other pressing issues
<hkaiser>
k
<hkaiser>
parsa: done
<parsa>
thank you
<parsa>
shouldn't it be np.dtype('d') instead of float?
<parsa>
we don't have node_data<float>
<hkaiser>
parsa: works either way
<parsa>
why does it work? it's a different data type?
<hkaiser>
float is propagated to double just fine
<hkaiser>
int32_t is not
<hkaiser>
don't ask, that's the way numpy behaves
<hkaiser>
but feel free to add that to the ticket
<parsa>
i mean, users send phylanx a different data type than what they receive
<hkaiser>
nod
<hkaiser>
nothing we can do about this at this point
verganz has quit [Quit: Leaving.]
stmatengss has quit [Quit: Leaving.]
<parsa>
hkaiser: i created #415 but is not finished. can you take a quick look at it? i think the code is unreasonably large for a simple operation but i don't know how to express it better
Vir- has quit [Ping timeout: 240 seconds]
Vir- has joined #ste||ar
Vir- has quit [Ping timeout: 265 seconds]
Vir has joined #ste||ar
simbergm_ has quit [Quit: WeeChat 2.1]
Vir has quit [Ping timeout: 240 seconds]
Vir has joined #ste||ar
Vir has quit [Ping timeout: 248 seconds]
Vir has joined #ste||ar
<nikunj>
first successful attempt at running hpx threads directly from compiler's main. It runs fine on my pc (ubuntu 18.04). Please test it on your pc (preferably linux for now) and let me know of the results :)