<Yorlik>
I am not so much concerned about the general case. What concerns me is high density, highly dynamic scenarios. Imagine a battlefiled with 2oo participating parties.
<Yorlik>
We want to push as far as we can on that.
<Yorlik>
One idea is to separate queries of grounded units versus flying uniots
<hkaiser>
sure
<Yorlik>
MNajority will always be grounded.
<Yorlik>
Also we might need to mix in category based search, like: Give me all players in region or all monsters in region etc ...
<hkaiser>
sure, that 4d tree supports that very well
<hkaiser>
I used it for a geo-information systems with millions of objects
<hkaiser>
20 years ago
<Yorlik>
There are very different types of search. E.G. Collision is involving all objects with collision, but sometimes it's not about collision but a type of object in an area.
<hkaiser>
anyways, r-trees or kd-trees are the way to go
<Yorlik>
I think I'll just use the Boost R-Tree for now and see when it needs optimizing / special code and just hack into it .
* Yorlik
ditches writing his own quadtree for now
<Yorlik>
Lazyness wins :)
jaafar has joined #ste||ar
hkaiser has quit [Quit: bye]
hkaiser has joined #ste||ar
<hkaiser>
hey nikunj, yt?
<nikunj>
hkaiser: here
<nikunj>
hkaiser: I was looking at the appendix parsa gave the other day
<nikunj>
it looks different than the ones SC wants
<hkaiser>
well, whatever it takes
<hkaiser>
nikunj
<nikunj>
alright then
<hkaiser>
brb
<nikunj>
ok
<nikunj>
hkaiser: brb
hkaiser has quit [Read error: Connection reset by peer]
hkaiser has joined #ste||ar
<nikunj>
hkaiser: yt?
<hkaiser>
nikunj: hey
<nikunj>
hkaiser: I updated the paper and added the appendix at the end
<nikunj>
for simplicity I provide the url to the hpxr_data repo for running the benchmarks and generating the graphs
<nikunj>
and I explain how to do so in the appendix
<hkaiser>
nikunj: thanks!
<hkaiser>
shouldn't you run that bash script on Cori and add its output?
<nikunj>
hkaiser: I will now update the hpxr_data repo and add the urls in the appendix
<hkaiser>
ahh cool!
<nikunj>
hkaiser: I have added the bash script in the hpxr_data along with the expected output
<hkaiser>
perfect
<nikunj>
expected output -> my output
<hkaiser>
lol
<nikunj>
I also add the graph scripts that I use to generate the graphs I use in paper
<nikunj>
this way we are completely transparent with our results
<hkaiser>
thanks for doing all of this!
<nikunj>
no problems ;)
<nikunj>
also about the porting to c++11 thingy, I have another problem
<hkaiser>
did you solve the sequence_index issue?
<nikunj>
using the pack thingy results in this "template argument for template type parameter must be a type'
<hkaiser>
I can look into the index_sequence
<nikunj>
alright
<nikunj>
let me push the current code then
<hkaiser>
can you do the (auto&&)[]{}?
<nikunj>
I have resolved that
<hkaiser>
perfect
<hkaiser>
I'll do that over the weekend
<nikunj>
I created a validate_ function and used compile time result type determination
<nikunj>
and provided that as validate_<result_type> as a function parameter
<hkaiser>
well decltype for the win ;-)
<nikunj>
I can't figure out how to resolve the capture intialization within the lambda as well
<nikunj>
c++11 does not allow to initialize captures within lambda
<hkaiser>
use HPX macros HPX_CAPTURE_MOVE() and HPX_CAPTURE_FORWARD()
<nikunj>
ohh great
<nikunj>
I'll do that
<hkaiser>
grep the hpx source code to see how its done
<nikunj>
sure
<nikunj>
great then, let me make the change and push it for you
<hkaiser>
thanks
<nikunj>
no, thank you! :)
<hkaiser>
nikunj: pls let me know once you have pushed your changes