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/
grostig has quit [Ping timeout: 240 seconds]
jaafar has quit [Quit: Konversation terminated!]
hkaiser has quit [Quit: bye]
parsa is now known as parsa_
jbjnr has joined #ste||ar
jbjnr has quit [Ping timeout: 250 seconds]
hkaiser has joined #ste||ar
daissgr has joined #ste||ar
<K-ballo>
we have cyclic order dependent includes between dataflow and parallel executor
<zao>
?
<K-ballo>
I'm thinking maybe we knew? there's some funny business with #include out of header guards in dataflow.hpp
<hkaiser>
K-ballo: yes, I'm aware of this
<hkaiser>
I tried to break that dependency before, but was not successful
<hkaiser>
dataflow needs to be changed to use some lower level interfaces that are used by the executor as well avoiding to pull in dataflow for the implementation of the executor customization points
<hkaiser>
K-ballo: so should we create specializations for parallel_policy_executor<[sync|deferred]>?
<hkaiser>
or is it not worth the trouble?
jbjnr has joined #ste||ar
<K-ballo>
hkaiser: no, I think we should just have an is_syn_policy predicatec
<K-ballo>
it is or may be a runtime value after all
daissgr has quit [Ping timeout: 240 seconds]
jbjnr has quit [Ping timeout: 240 seconds]
jaafar has joined #ste||ar
david_pfander has joined #ste||ar
<hkaiser>
K-ballo: you're on the roll!
jbjnr has joined #ste||ar
david_pfander has quit [Ping timeout: 268 seconds]
<K-ballo>
I'm almost done here.. maybe another week or two
<K-ballo>
hkaiser: will you do any work on parallel executor? if not I'll add the sync check to skip the hierarchical spawning
<hkaiser>
K-ballo: pls go ahead
<K-ballo>
hkaiser: does not help, I knew it and forgot.. goes through the fallback then_execute which just spawns a new thread regardless
<K-ballo>
that's why I said my changes cause the test to reveal the wrong thing is being done by deferred, because it was not due to my changes
<K-ballo>
I'll push what I have and you can take a look at it from there
<hkaiser>
K-ballo: ok
<hkaiser>
we need to implement then_execute and bulk_then_execute
jbjnr has quit [Ping timeout: 240 seconds]
<K-ballo>
or make the fallback one respect policies
<K-ballo>
else the fallback is not going to be of much use