diehlpk has joined #ste||ar
K-ballo has quit [Quit: K-ballo]
diehlpk has quit [Ping timeout: 250 seconds]
rori has joined #ste||ar
hkaiser has joined #ste||ar
K-ballo has joined #ste||ar
hkaiser has quit [Quit: bye]
hkaiser has joined #ste||ar
rori has quit [Quit: bye]
<
hkaiser>
simbergm: the all_to_all problem is solved now
<
hkaiser>
I left the fix on the all_reduce branch (#3950)
<
simbergm>
hkaiser: that's fine, thanks!
<
simbergm>
hkaiser: hmm, why does that fix it?
<
K-ballo>
sequencing?
<
K-ballo>
the id being moved before the async call is evaluated?
<
hkaiser>
simbergm: explicitly define the sequence of operations wrt the move(id)
<
simbergm>
ok, I wouldn't have thought the move would happen before
<
simbergm>
learned something
<
K-ballo>
it does since 17
<
K-ballo>
wait no, it happens
*after*
<
K-ballo>
wouldn't happen before, all good
<
hkaiser>
simbergm: it's gcc6 doing the wrong thing
<
hkaiser>
we were seeing the problem for that compiler only
<
simbergm>
oh right, forgot that it was only there
<
simbergm>
darn gcc, it's nothing but trouble
<
simbergm>
getting better though it seems
<
simbergm>
you are goldmines
hkaiser has quit [Quit: bye]
hkaiser has joined #ste||ar
jaafar has quit [Quit: Konversation terminated!]
jaafar has joined #ste||ar
<
diehlpk_work>
Modern C++ and CUDA by wash[m]