aserio 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/
<github>
[hpx] hkaiser force-pushed find_symbols from a9e0248 to fc4482b: https://git.io/vbzvG
<zao>
Seems like I put /opt on the ZFS storage, so I don't have to dig out the system drive.
<heller>
woohoo. std::thread execution time is 32647ns and hpx::thread is 5885ns
<heller>
woohoo
<heller>
who's your daddy?
<zao>
Nice.
K-ballo has joined #ste||ar
hkaiser has joined #ste||ar
mcopik has quit [Ping timeout: 264 seconds]
<github>
[hpx] msimberg closed pull request #3040: Fix a mistake about exception handling in asynchronous execution of scan_partitioner. (master...tg_fix_scan_partitioner_async) https://git.io/vbOTk
<jbjnr>
lol - I'm running pycicle on my windows machine in the bash shell. That machine is always on and it is in charge of daint now.
hkaiser has quit [Quit: bye]
aserio has joined #ste||ar
akheir has joined #ste||ar
akheir has quit [Remote host closed the connection]
tjtn has joined #ste||ar
mbremer has joined #ste||ar
eschnett has joined #ste||ar
parsa has joined #ste||ar
akheir has joined #ste||ar
<github>
[hpx] aserio created fixing_3075 (+2 new commits): https://git.io/vbVrf
<github>
hpx/fixing_3075 109a3a2 Adrian Serio: Fixed checkpoint.cpp to make floats constant...
<github>
hpx/fixing_3075 5514818 Adrian Serio: Initializing split_gids_...
<akheir>
diehlpk_work: installed the libpng-devel
hkaiser has joined #ste||ar
<github>
[hpx] aserio opened pull request #3076: Fixing 3075: Part 1 (master...fixing_3075) https://git.io/vbVrP
daissgr1 has joined #ste||ar
EverYoung has joined #ste||ar
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
<hkaiser>
heller: you happy with #3072 now?
<hkaiser>
heller: also, what about #3061 now?
patg[[w]] has joined #ste||ar
<patg[[w]]>
aserio: see pm
<diehlpk_work>
terminate called after throwing an instance of 'hpx::detail::exception_with_info<hpx::exception>'
parsa has quit [Quit: Zzzzzzzzzzzz]
hkaiser has quit [Quit: bye]
<diehlpk_work>
Can me someone explain how this error could come from?
david_pfander has quit [Ping timeout: 272 seconds]
wash has quit [Ping timeout: 248 seconds]
aserio has quit [Ping timeout: 272 seconds]
Smasher has joined #ste||ar
wash has joined #ste||ar
hkaiser has joined #ste||ar
<heller>
hkaiser: still skeptical if it's really a better solution than what we had before ;)
<heller>
the code itself looks ok
<hkaiser>
heller: before we had to pay for one turnaround (possibly over the network) for each entry, now we pay one network operation per locality
<heller>
the broadcast didn't change
<hkaiser>
right, but now we don't have the callback invocations anymore
<heller>
right, but the return value
<hkaiser>
shrug
<heller>
the usecase before was just to print out all names, IIUC
<heller>
it's different now, of course
<hkaiser>
much cheaper to send one (possibly big map) than to invoke hundrets of callbacks over the wire
<heller>
the callbacks were invoked on service locality, not on the caller site
<heller>
or did I miss that?
<hkaiser>
no
<hkaiser>
depended on the function object
<heller>
but yes, I concur that if you need to collect all the data, it is cheaper that way
<K-ballo>
diehlpk_work: is that all the information you have about the error?
<diehlpk_work>
K-ballo, Investigating more
<hkaiser>
heller: good
<heller>
hkaiser: I'd still would like to ask you to consider if it's really the best approach for your usecase at hand
<hkaiser>
heller: the use case just tells us that we miss an introspection API for the symbol namespace
<hkaiser>
what I did so far in Phylanx is really just the tip of the iceberg
<heller>
my imagination is failing me here ;)
<hkaiser>
mine not, I have plenty of ideas - but yah, I'd agree to leave the old API in place in addition to th enew one
<heller>
I mean ... in most cases, the names you are interested are application specific (or at least the patterns are well known) and the actual set of names you are interested in, and you should be able to easily recompute those names (if they aren't stored already)
<heller>
so what you get with the new API, is the set of all names (with their bound GID) in the symbol namespace
<heller>
right?
<hkaiser>
yes
<heller>
so what it saves is the application specific recomputation of those names, correct?
<hkaiser>
recomputing might not be an easy option
<heller>
yes, I can see a usecase where you just want to connect a process to a running set and get going from there
mbremer has quit [Ping timeout: 260 seconds]
<hkaiser>
the compilation of the AST into the expression-tree is not a trivial step, so recomputing the names is not 'just' a tree traversal
<heller>
the expression tree contains the clients, right?
<hkaiser>
yes
<heller>
and the clients store their names ;)
<hkaiser>
I can't traverse the expression tree
<hkaiser>
well, without addiing a special API for this
<hkaiser>
also, the expression tree is distributed itself
<hkaiser>
so traversal on that level is not cheap per se either
<heller>
hmm
<heller>
might still be cheaper than doing a broadcast over all localities ;)
<heller>
or well, in that case, I guess it is almost equivalent
<heller>
if you can't travers the expression tree, how are the expressions evaluated?
<hkaiser>
the evaluation is done through the API each of the treenodes exposes - this API is not a generic traversal, rather a specialized 'calculate me'
<hkaiser>
heller: I'd like not to conflate the Phylanx use case with the missing symbol namespace API
<heller>
might be something worth considering ...
<heller>
oh, I am not, I think I approved the PR already, didn't I?
<hkaiser>
there might be better solutions for Phylanx, that doesn't mean that we wouldn't need an introspection API on the symbol namespace
<heller>
I didn't say that
<hkaiser>
heller: do you want me to add back the old API ?
<heller>
it's not used anyway, is it?
<heller>
I am impartial here
<hkaiser>
not currently, it was used for the command line option only - I changed that to use the new one
<heller>
hkaiser: the reason I want to discuss the phylanx use case is because this seems to have driven the desire to add this API. And I am just wondering if that's really what you want
<heller>
yeah
<heller>
that's fine
<hkaiser>
heller: I'm not 100% sure where all of this will go, what I know is that I need to discover the nodes of the expression tree dynamically at runtime, possibly for trees that were serialized to disk
<hkaiser>
and the AGAS symbolic namespace seems to provide the functionality I need without having to reinvent the wheel
<heller>
yeah
<heller>
that is what it is for indeed
<heller>
hkaiser: to me, that sounds more like introspection for the expression tree than symbolic namespace. That doesn't mean that the first can't be implemented with the latter
<hkaiser>
nod
<hkaiser>
the symbolic namspace allows to discover the tree even if there are multiple tree around and the root of the tree is inaccessible to the caller
<heller>
*nod*
<heller>
however, getting the root of the tree is probably a known name of some sort?
<hkaiser>
yah, sure
<hkaiser>
as I said, I don't have the full picture yet
<hkaiser>
just trying to stay on the safe side here
<heller>
sure
<heller>
speaking of pictures and nice things ... looked at buildbot recently?
<hkaiser>
:D
<hkaiser>
getting there
<heller>
almost back to normal...
<heller>
the scan timeouts worry me
<heller>
wasn't able to reproduce them here
K-ballo has quit [Quit: K-ballo]
K-ballo has joined #ste||ar
aserio has joined #ste||ar
akheir has quit [Remote host closed the connection]
<hkaiser>
aserio: could you give me the link for the call, pls ? :/
<hkaiser>
aserio: ^^?
<aserio>
hkaiser: woops
<aserio>
you should have it
<aserio>
I forgot to tell you that I sent it
<hkaiser>
I did have it for sure ;)
<hkaiser>
don't have it anymore, though
<aserio>
You have it now right?
<hkaiser>
no
<hkaiser>
sorry, must have deleted the email
<aserio>
Just now?
<hkaiser>
got it, tks
aserio has quit [Ping timeout: 255 seconds]
aserio has joined #ste||ar
<hkaiser>
aserio: no audio today?
parsa has joined #ste||ar
<patg[[w]]>
jmjss&of81
<patg[[w]]>
ugh
eschnett has quit [Quit: eschnett]
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
eschnett has joined #ste||ar
<aserio>
hkaiser: yt?
<hkaiser>
here
<aserio>
please see pm
patg[[w]] has quit [Quit: Leaving]
aserio has quit [Ping timeout: 240 seconds]
EverYoung has quit [Ping timeout: 272 seconds]
parsa has quit [Quit: Zzzzzzzzzzzz]
parsa has joined #ste||ar
mcopik has joined #ste||ar
aserio has joined #ste||ar
Smasher has quit [Remote host closed the connection]
parsa has quit [Quit: Zzzzzzzzzzzz]
tjtn has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]