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/
diehlpk has joined #ste||ar
hkaiser has quit [Quit: bye]
diehlpk has quit [Ping timeout: 240 seconds]
hkaiser has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
hkaiser has quit [Quit: bye]
daissgr has quit [Ping timeout: 256 seconds]
verganz has quit [Ping timeout: 260 seconds]
nanashi55 has quit [Ping timeout: 260 seconds]
nanashi55 has joined #ste||ar
<github>
[hpx] sithhell created fix_circleci_workspace (+1 new commit): https://git.io/vx7lG
<github>
hpx/fix_circleci_workspace d0b2fcf Thomas Heller: Don't store the inspect results in the workspace
david_pfander has joined #ste||ar
<github>
[hpx] sithhell pushed 1 new commit to fix_circleci_workspace: https://git.io/vx747
<github>
hpx/fix_circleci_workspace 865e7de Thomas Heller: Properly storing test results for circle ci
<github>
[hpx] sithhell pushed 1 new commit to fix_circleci_workspace: https://git.io/vx7RN
<github>
hpx/fix_circleci_workspace 5b54a42 Thomas Heller: Using ninja instead of make for CircleCI
<heller>
simbergm: morning
<heller>
simbergm: I think the failure was not because install depended on all the tests, more because inspect and core wrote to the workspace in parallel
<simbergm>
heller: good morning
<heller>
currently trying to resolve it as well
<heller>
and have the install and deploy step to run always
<simbergm>
very possible, I'm very much just guessing what could be wrong
<simbergm>
would be very nice if we can keep the dependencies and not have problems with this workspace thing
<jbjnr_>
simbergm: heller I'm in trouble. I rebased my executor branch on to the lazy_init and remove_wait_or_add_new, but if they are not good and it is actually slower, then I need to go back to master. Probalem is, that produced a ton of conflicts that I resolved. took a whole day and I do not have the old branches around :( - what are the chances that we won't use those features?
<heller>
jbjnr_: for me, the chances to not use those features goes to zero
<heller>
simbergm: the docker image is only generated on master
<jbjnr_>
but simbergm tells me that removing the staged tasks makes things slower
<simbergm>
jbjnr_: don't know about the chances but if it actually is faster on cholesky it's not bad, and it will only get better once stack reuse is there (again)
<jbjnr_>
I cannot say that it is faster on cholesky. Flipping between those branches is too much work to test easily
<simbergm>
but it's only on future_overhead, it might still be faster with other improvements
<simbergm>
+ it's much cleaner and easier to understand
<jbjnr_>
and it's probabyl only a minor part of many of the tweaks and optimiations we're making
<simbergm>
ok
<simbergm>
so even if it's not faster in the end, I think it's worth it for the simplicity
<simbergm>
as long as it's not significantly slower of course...
<jbjnr_>
you were the one who measure it
<jbjnr_>
^d
<simbergm>
yes, but that's only one benchmark
<heller>
yes, it's one part of many
<heller>
and it makes the code soo much cleaner
<github>
[hpx] sithhell force-pushed fix_circleci_workspace from 31a6a08 to c29e36d: https://git.io/vx7g9
<github>
hpx/fix_circleci_workspace 89901ab Thomas Heller: Properly storing test results for circle ci
<github>
hpx/fix_circleci_workspace d4158e5 Thomas Heller: Using ninja instead of make for CircleCI
<github>
hpx/fix_circleci_workspace c29e36d Thomas Heller: Adding ldconfig to the install step to update ld's cache
<simbergm>
heller: you like ninja, no? ;)
<heller>
simbergm: ninja is awesome
<heller>
simbergm: also, look at the timings of the header tests
<heller>
simbergm: significant improvement
<simbergm>
huh, because of ninja?
<heller>
yes
<heller>
ninja is much quicker than make to compute the dependencies
<heller>
when doing a `ctest -R tests.headers`, we invoke make/ninja for each single test
<simbergm>
hmm, nice
<simbergm>
I changed to make because the header tests were OOMing or something similar with ninja before
<heller>
oh, ok
<heller>
we'll see
<heller>
simply including a header should certainly not take more than 2GB of RAM
<heller>
we have to fix this ;)
<simbergm>
they seemed to pass now, so all good
<simbergm>
I know, it seemed strange but I checked it a few times with a clean build, with ninja clang was killed and with make no problems (but slow)
<heller>
how about we fix the actual cause instead of the symptoms ;)?
<github>
[hpx] msimberg closed pull request #3275: WIP: Make CircleCI install step only depend on examples (master...msimberg-patch-2) https://git.io/vx9Db
<github>
[hpx] sithhell deleted msimberg-patch-2 at b5e10b2: https://git.io/vx7MR
K-ballo has joined #ste||ar
daissgr has quit [Ping timeout: 268 seconds]
hkaiser has joined #ste||ar
nikunj has joined #ste||ar
nikunj_ has joined #ste||ar
nikunj has quit [Ping timeout: 260 seconds]
<jbjnr_>
hkaiser: or anyone - I wanted to add the serialization paper to the publications page, but I do not know how to upload a pdf to the /pub/ folder. Other publications have a pdf link ...
nikunj_ has quit [Ping timeout: 260 seconds]
nikunj has joined #ste||ar
<hkaiser>
jbjnr_: send it to Adrian, he'll manages those pages