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/
eschnett has joined #ste||ar
eschnett has quit [Quit: eschnett]
K-ballo has quit [Quit: K-ballo]
bita has joined #ste||ar
<bita>
hkaiser, should extract_boolean_value of -2 be true?
<hkaiser>
bita: I think so, everything != 0 is true
<hkaiser>
except for None/nil which is false
<bita>
I believe currently it produces 254
<hkaiser>
well, that's true too ;-)
<bita>
it doesn't work correctly for negative integers
<hkaiser>
interesting
<bita>
oh, Okay
<bita>
I thought it should have mapped everything to 0 or 1
<hkaiser>
we internally represent bools as uint8 values, those however should get converted to np.bool_
<bita>
thanks
<hkaiser>
any time
hkaiser has quit [Quit: bye]
eschnett has joined #ste||ar
bita has quit [Quit: Leaving]
eschnett has quit [Quit: eschnett]
nikunj has quit [Ping timeout: 264 seconds]
nikunj has joined #ste||ar
nikunj has quit [Ping timeout: 240 seconds]
nikunj has joined #ste||ar
nikunj has quit [Quit: Leaving]
david_pfander has joined #ste||ar
hkaiser has joined #ste||ar
david_pfander has quit [Ping timeout: 250 seconds]
K-ballo has joined #ste||ar
david_pfander has joined #ste||ar
david_pfander has quit [Ping timeout: 268 seconds]
nikunj has joined #ste||ar
eschnett has joined #ste||ar
<K-ballo>
what's the difference between HPX_CONFIG_DEFINITIONS and HPX_CONFIG_COND_DEFINITIONS?
<hkaiser>
K-ballo: seee the generated config defines.hpp
<hkaiser>
COND_DEFINITIONS wrap the PP constant into a #ifndef/#endif to avoid duplicate definitions
<K-ballo>
I see
<K-ballo>
btw I just noticed a bunch of stale definitions in there
<jbjnr>
sorry hkaiser went away and missed your reply. I see the problem with the paper. We start by talking about v1309 and the importance of modelling it, then if we don't have the light curve, the results we show are pointless.
<jbjnr>
but ...
<jbjnr>
we could conclude along the lines of - this project was done as a test project on the cscs computer with limited compute hours and demonstrates that the problem that would take N weeks on machine X at LSU could be achieved in M days/hours on daint at cscs under a full production project etc etc
<jbjnr>
anyway, just thinking aloud
<jbjnr>
I'll post that to the list
<jbjnr>
where;'s diehlpk_work ? I want a new graph! I'm busting to see it
<hkaiser>
jbjnr: I agree, we should put things the way as you outlined
<hkaiser>
jbjnr: Patrick will be back a bit later
<jbjnr>
I added the 4096 point to the plot myself using my best guess of what numbers he used for the calculation and it looks about right
<jbjnr>
I'm happy
<hkaiser>
jbjnr: yes, looks very good
<hkaiser>
could you commit what you have, pls?
<jbjnr>
no need, I only added one point with speedup 900 to the 4096 LF plot and 2.88 speedup compared to MPI, the text part I'm thinking about and will write tomorrow
<jbjnr>
diehlpk_work: will update the plots with the correct figures as he knows the exact calcualtion
<jbjnr>
I'm proofreading at the moment the other parts so that I know what's being said elsewhere
<jbjnr>
I didn't read the paper at all until this evening
<hkaiser>
jbjnr: a factor of 2.88 sounds great
diehlpk_mobile has joined #ste||ar
<hkaiser>
hey diehlpk_mobile
<hkaiser>
diehlpk_mobile: jbnr has new data
<jbjnr>
hkaiser: I'm going to have to go to my superiors and tell them that these results with LF are too interesting not to warrant a paper of their own and submit to IPDPS later this year (oct). Then I can test the LF code on our linear algebra as well as octotiger and get results for other applications.
<jbjnr>
... or something
<hkaiser>
sounds great
<jbjnr>
anyway. See ya tomorrow
<hkaiser>
jbjnr: we should have the current results in this paper however
<jbjnr>
for sure, but since it'll be rejected, I can reuse them later
<hkaiser>
lol
<hkaiser>
ok
<hkaiser>
sounds like a plan
<jbjnr>
Got to plan ahead
<jbjnr>
^^
<hkaiser>
sure, your pessimism breaks all records today ;-)
<jbjnr>
bah. I've done better. I'm in a good mood. It's not often you speed a code up by 2.88 on 4096 nodes!
<jbjnr>
just shows how shit our mpi stuff is I suppose. let's be honest :)
<jbjnr>
I hope kevin ran LF on cori ....
<jbjnr>
right. I'm off to bed.
<diehlpk_mobile>
jbjnr, the plot should be updated by tomorrow morning
<diehlpk_mobile>
the paper will be accepted, the results are not too bad
<diehlpk_mobile>
we just need a good story
<diehlpk_mobile>
hkaiser, see pm
diehlpk_mobile has quit [Quit: Yaaic - Yet another Android IRC client - http://www.yaaic.org]