<hkaiser>
well, not sure if that would be a good idea to make it constructable from any unrelated type...
<hkaiser>
bita: where would you stop?
<bita>
the thing is right now that we are using physl code and pass the arguments to functions using define, some of those arguments are size_t (can be represented as int64_t) but the constructor is explicit and ...
<bita>
I don't have a problem with that and I will save those variables using integers :)
wate123__ has quit [Remote host closed the connection]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 272 seconds]
<hkaiser>
bita: I understand
<hkaiser>
but really where should we stop? size_t? int? long? short? char?
<hkaiser>
unsigned int?
<hkaiser>
and what the other constructors that take blaze data types, should those be converting from any integral type?
<hkaiser>
*what about*
<bita>
after looking at the primitive_argument_type.hpp I know that of course I can save it as int64_t
<bita>
thank you
<weilewei>
hmmm...sounds like c++ template code to c interface, just an irrelevant comment
<hkaiser>
bita: it's a matter of taste almost, there is no strict rule
<hkaiser>
rule of thumb: interfaces must be a restricted as possible while providing all necessary functionaly
<bita>
that's true
<hkaiser>
bita: the primitive_argument_type has a gazillion of constructors already, I'd say we try not to add more
<bita>
yes, I totaaly agree
<hkaiser>
:D
hkaiser has quit [Quit: bye]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Remote host closed the connection]
wate123_Jun has joined #ste||ar
bita has quit [Quit: Leaving]
wate123_Jun has quit [Ping timeout: 246 seconds]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 252 seconds]
weilewei has quit [Remote host closed the connection]
kale_ has quit [Remote host closed the connection]
heller1 has joined #ste||ar
wate123_Jun has quit [Ping timeout: 240 seconds]
parsa[m] has joined #ste||ar
parsa[m] is now known as Guest20891
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 256 seconds]
Hashmi has joined #ste||ar
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 240 seconds]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 240 seconds]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 252 seconds]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 252 seconds]
Hashmi has quit [Quit: Connection closed for inactivity]
mcopik has joined #ste||ar
mcopik has quit [Client Quit]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 252 seconds]
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 252 seconds]
hkaiser has joined #ste||ar
Pranavug has joined #ste||ar
Pranavug has quit [Client Quit]
wate123_Jun has joined #ste||ar
gonidelis has joined #ste||ar
wate123_Jun has quit [Ping timeout: 252 seconds]
nikunj97 has joined #ste||ar
<simbergm>
kokkos meeting in 23 minutes, and hpx meeting in 53 minutes
<simbergm>
hkaiser: you good with hosting again?
<hkaiser>
simbergm: yes, absolutely - sorry forgot to schedule the meeting, will do right away
<simbergm>
ah, no worries, I didn't expect you to :P
<hkaiser>
sent
<simbergm>
thanks
<simbergm>
nikunj97: what do you mean by git and 1.4.1 being different? the 1.4.1 release is tagged on git
<simbergm>
nikunj97: you might also want to try the thread_pool_executor on master
<simbergm>
yeah, it is... almost
<nikunj97>
mdiers_, It was with git. I thought HPX git was same as 1.4.1. Just recently I realized that 1.4.1 had cherry picked commits so is different from git
<nikunj97>
I should re-build HPX in this case
<nikunj97>
I thought master had 1.4.1 and some more commits
<nikunj97>
simbergm, will do, thanks!
<nikunj97>
git as in master
<nikunj97>
but as you told the other day that master is different this time bcoz of cherry picked commits in 1.4.1
<hkaiser>
nikunj97: you should use the stable tag if anytihng
<nikunj97>
hkaiser, I had 1.4.1 and master
<hkaiser>
stable is the same as the last commit on master that successfully passed the ci
<nikunj97>
aah, alright
<simbergm>
hkaiser: do you want to join the kokkos meeting this week?
<simbergm>
or others?
<simbergm>
otherwise we'll go ahead
wate123_Jun has joined #ste||ar
wate123_Jun has quit [Ping timeout: 240 seconds]
rtohid has joined #ste||ar
<simbergm>
pmc meeting everyone!
<simbergm>
let me know if you need a meeting number
<rori>
sorry I have some wifi problems today, I may be often disconnected..
<hkaiser>
jbjnr: ^^
<hkaiser>
rori: can you hear us?
<rori>
yep !
<rori>
But I 'm disconnected sometimes
<hkaiser>
ok
diehlpk_work has joined #ste||ar
shahrzad has joined #ste||ar
<rori>
yeah as john's said in the meeting, my invite was in my spam folder ^^, but I've redirected it now :)
nan11 has joined #ste||ar
<hkaiser>
diehlpk_work: yt?
<diehlpk_work>
yes
<hkaiser>
see pm, pls
shahrzad has quit [Ping timeout: 252 seconds]
weilewei has joined #ste||ar
<nikunj97>
heller1, it is pretty difficult to extend my existing code to add block based stencil :/
wate123_Jun has quit []
wate123_Jun has joined #ste||ar
karame_ has joined #ste||ar
<heller1>
He ;)
<hkaiser>
diehlpk_work: se pm, pls
<diehlpk_work>
There is none
shahrzad has joined #ste||ar
shahrzad has quit [Ping timeout: 252 seconds]
gonidelis has quit [Ping timeout: 240 seconds]
ibalampanis has joined #ste||ar
<ibalampanis>
Hello to everyone! Have a nice day!
<hkaiser>
hey ibalampanis
<ibalampanis>
simbergm, hkaiser: I 'd like to ask you if the coord meeting is stable every Thursday. I checked late my mail and I couldn't join to meeting today.
<hkaiser>
ibalampanis: it's every other Thursday
<ibalampanis>
Same time?
<hkaiser>
yah
<ibalampanis>
So I make it iteratively in my calendar! Thanks
<hkaiser>
ibalampanis: attending is certainly not a requirement, the discussions are very specific to current HPX problems, but feel free to join
<ibalampanis>
Yes, I have it in my mind. I want it just for experience!
<simbergm>
ibalampanis: if you're interesting in following that more often you might want to join the hpx-devel mailing list as well (you'll also find a calendar invitation for the coordination meetings in the archives)
<simbergm>
if details change it'll be announced there
<ibalampanis>
Yes, I have already joined. This mail I checked late..
<ibalampanis>
Thanks
<simbergm>
👍️
<simbergm>
hkaiser: btw, you said hpx-users/devel@stellar-group.org work now right? if yes, I'll just update the links on the homepage
<diehlpk_work>
weilewei, yet?
<weilewei>
diehlpk_work yes
<hkaiser>
simbergm: yes, please
<hkaiser>
simbergm: I updated a bunch of places already, but might have missed some
<simbergm>
hkaiser: yep, I wanted to add hpx-devel and matrix to the support page
<simbergm>
wordpress is giving me a hard time!
<hkaiser>
it does that at times
<hkaiser>
let me know if I can help
<simbergm>
thanks
<simbergm>
I think I got it
<simbergm>
somehow it stripped all the newlines from the markdown and just left 'n's
<heller1>
So btree for fast lookup, space filling curve for decomposition and skip list for storage?
<heller1>
Yeah, I understand
<Yorlik>
Yes
<Yorlik>
And tie it to physics data
<Yorlik>
So it's all nicely together
<Yorlik>
Like mass, velocity and such
<Yorlik>
I'm thinking about using tiles as buckets, of maybe 8*8 or 16*16 virtual meters
<Yorlik>
And above having that structure
<heller1>
What's the typical interaction range (in virtual meters)?
<Yorlik>
For combat it's usually small - melee range
<Yorlik>
But we'll also have firearms
<Yorlik>
Magic is kinda in between
<Yorlik>
That's why I though about 8-16 meters
<Yorlik>
And making 1/1024 meters our unit
<Yorlik>
So ~1mm precision
<Yorlik>
maybe a bit finer if needed
<heller1>
And the structure above is within those tiles or globally?
<Yorlik>
Using a 32 bit integer for coordinates that would allow for a damn huge world already
<Yorlik>
It would be globally
<Yorlik>
I am thinki9ng about simply using vector search within a bucket
<Yorlik>
Maybe 8x8 is ok
<Yorlik>
Crowded situations are always the tough part
<heller1>
That's about 64k elements per bucket
<Yorlik>
If you assume 1 sqm per person
<Yorlik>
OK - we might have bugs and flies and stuff too
<Yorlik>
But a 1sqm would be 64 in a bucket when its super crowded
<heller1>
And the projectiles
<Yorlik>
projectiles would not use physocs
<Yorlik>
They are too fast - you need prediction and a hit checvk
<Yorlik>
bullets just an LOS check
<Yorlik>
Arrows an animation or simple effect with a short delay
<heller1>
I see
<heller1>
Sounds reasonable then
<Yorlik>
We need to cheat our way around - after all we don't want correct physics, but entertaining physics.
<Yorlik>
So - it just needs to be believable enough. Not moire
<heller1>
Is your distributed simulation running?
<Yorlik>
Not yet distributed
<Yorlik>
I'm on the scripting in the moment
<heller1>
Aw, too bad
<Yorlik>
I can create objects from blueprints, attach scripts and react to events already
<Yorlik>
Also objects now get correctly finalized on destruction
<Yorlik>
Clustering is the next milestone
<heller1>
Jump on the hype train and make an epidemic game as the first mvp ;)
<Yorlik>
Current milestone is the local simulation
<Yorlik>
lol
<Yorlik>
I was fantasiting that, but we need MUCH more time.
<Yorlik>
Diseases will be a thing though
<Yorlik>
Also a body simulation with injuries an afflictions
<Yorlik>
However - still on the basics, and now spatial indexing is knocking at trhe door
<Yorlik>
I need it for clustering too, since we are load balancing using tiles.
<Yorlik>
Thats why all this Hilbert stuff
<hkaiser>
Yorlik: space filling curves will not give you too much, I believe
<Yorlik>
I need some good sorting fro the R-tree
<hkaiser>
your buckets are small enough to use brute-force search for neighborhood analysis
<Yorlik>
And hat I like about the SFC is the locality preservation
<hkaiser>
in between buckets things like kd-trees are usually perfect
<Yorlik>
How good is a KD tree with dynamic data?
<hkaiser>
quite nice as you split nodes on demand only
<Yorlik>
So I can do lazy calculations?
<heller1>
Something like plague inc doesn't look too complicated
<Yorlik>
Plague is fun
<hkaiser>
kd-trees are not evenly split, normally, so you can get away with imbalances nicely
<Yorlik>
played it long ago.
<Yorlik>
What I'm scared off is too many chacks too oftenb
<Yorlik>
When an object moves I need to check if I have to reindex it
<Yorlik>
That check should be super cheap
<Yorlik>
Actually I was connsidering a simple hierarchical AABB tree with fixed sizes
<hkaiser>
I'd rather keep track of an imbalance measure and completely rebuild the tree once in a while
<Yorlik>
Since the buckets are fixed you can get your bucket index by just >> your coords
rtohid has joined #ste||ar
<heller1>
Well, that can still remain as is
<Yorlik>
The poblem is moving objects leave their buckets
<hkaiser>
Yorlik: works as well, if you assume that one object can't be in more than one bucket at a time
<heller1>
Just cluster your buckets onto the leaves of the tree
<Yorlik>
I did a writeup of what searches i usually need
<heller1>
And have a mapping of bucket to leave node
<Yorlik>
And range and area based searches are mostly smallish
<hkaiser>
can objects be in more than one bucket?
<Yorlik>
A large dragon migth be
<Yorlik>
But I'd chach neighbouring buckets as well.
<hkaiser>
see, that might require you touse something like a kd-tree
<hkaiser>
and objects being not in a leaf-node but further up in the hierarchy
<Yorlik>
Or I make different trees for different types of objects
<Yorlik>
E.g. we want to have flight
<Yorlik>
But every object marked grounded counls stay in a simple 2D tree
<Yorlik>
statics, like large indestructible rocks are a different case
<Yorlik>
Houises might get destroyed but are static
<Yorlik>
Not sure if partitioning the problem like this would really be beneficial
<Yorlik>
Because I might have to do sveral queries in some cases, e.g. LOS or bullet hit checks
<Yorlik>
But "Give me all possible enemies around" would never check houses or rocks
<Yorlik>
So - what I'm currently doing is snoopiong around but at the same time trying to better define my problem.
<Yorlik>
So I need something that is optimal for the sum of questions I have: Pathfinding, Ranged search, Area search, LOS checks/Hit Checks, Collision and Physics interaction, Viewing cone checks, fast insert/delete/move object
<Yorlik>
Uh - and we want aerial vehicles, like balloons and stuff.
wate123_Jun has quit [Remote host closed the connection]
<heller1>
BTW, do you have some preliminary art work you can share?
wate123_Jun has joined #ste||ar
<Yorlik>
Not really. I'm working with an artist, but she still has a ton to learn.
<Yorlik>
Asset acquisition might become an issue at some point, but there is stuff around that is good and dirt cheap also - but not really original then.
<Yorlik>
We think of a mildly stylized style.
<Yorlik>
So - not like a comic, but not photorealistic either
<heller1>
a bit like diablo?
wate123_Jun has quit [Ping timeout: 260 seconds]
<Yorlik>
Somewhat, yes. We also want Action RPG elemnts in our Combat
<Yorlik>
It will not be round based - all real time.
Abhishek09 has joined #ste||ar
<heller1>
ms: I don't like the local runtime PR at all, I still think it strips away too much
<weilewei>
heller1 what is this software you are using for chat?
<heller1>
not good
<heller1>
riot.im
<weilewei>
can it connect to irc?
<heller1>
it should, yes
<simbergm>
heller: are you on #ste||ar or #hpx?
<simbergm>
also, thanks for going through the pr! I'll write up a proper response in the morning
<heller1>
ms[m]: thanks. I am on both
<simbergm>
but you're writing from only one...
<heller1>
yes
<heller1>
right now, I am writing from #hpx
<heller1>
tiago.fg: I think it is a bug, still there on master
Abhishek09 has quit [Remote host closed the connection]
<heller1>
tiago.fg: on the emitter, add --hpx:expect-connecting-localities, on the worker and collector add --hpx:expect-connecting-localities --hpx:worker
<diehlpk_work>
I wanted to delete a folder and got device busy and coudl nto delete the folder
<weilewei>
I don't have this though, and just tested to delete a folder, it works fine
<weilewei>
maybe too many people using headnode
<weilewei>
Summit hangs sometimes
<diehlpk_work>
Ok, I logged out and in again and it worked
<weilewei>
Summit has 6 different login nodes, if I remember correctly
<diehlpk_work>
Yes, right now I had to compile all again, since I was not able to get spectrum-mpi working with my own gcc build
<weilewei>
ok, hope that will work
<weilewei>
do you have any problem with Summit's own gcc?
<diehlpk_work>
Not yet
<diehlpk_work>
I just had to change my build scripts a bit
<weilewei>
ok, let me know then
<diehlpk_work>
Are you using jemalloc ro tcmalloc?
<weilewei>
tcmalloc
<weilewei>
can you cd /gpfs/alpine/proj-shared/cph102/
<weilewei>
do you have access to this folder?
<diehlpk_work>
Let me check, I just wanted to know what you sue
<diehlpk_work>
*use
wate123_Jun has quit [Remote host closed the connection]
<diehlpk_work>
weilewei, Can you share a job script with me?
<Yorlik>
Any suggestions for a good lock-free list (may be singly linked)?
<weilewei>
diehlpk_work see DM
<weilewei>
my job scripts, but you might need to change project number
<diehlpk_work>
hkaiser, simbergm Any known issues with HPX's Cmake for finding the MPI compiler?
<diehlpk_work>
I have to run cmake twice to find the MPI compiler on Summit
<weilewei>
what I did on Summit, is export CC=mpicc, export CXX=mpicxx, not sure if this helps
<weilewei>
Feel free to module load hpx on Summit as well. Maybe that helps as well. Just a suggestion
<diehlpk_work>
Did the same
wate123_Jun has joined #ste||ar
bita_ has joined #ste||ar
bita_ has quit [Client Quit]
bita_ has joined #ste||ar
bita has quit [Ping timeout: 252 seconds]
<hkaiser>
Yorlik: we have one in HPX
<hkaiser>
Yorlik: you sure you need a lock-free one?
<Yorlik>
hkaiser: the locator system will have contention.
<Yorlik>
Because there will be space based interaction between objects in tasks on different threads
<Yorlik>
Two objects could change bucket at the samne time, one queries one changes, etc ...
<zao>
weilewei: -DCMAKE_C_COMPILER and -DCMAKE_CXX_COMPILER are the ways you pass compilers in if you don't have the environment set. The cool kids use "toolchains" to specify a platform's compilers and behaviour, more used for cross-compilers: https://github.com/stellar-group/hpx/cmake/toolchains
<diehlpk_work>
Anyone has issues with GitHub? It is so slow for me?
<zao>
Gone for me.
<zao>
(I made the above URL by hand :D)
<diehlpk_work>
weilewei, which boost version are you using?
<Yorlik>
hkaiser: I'm aware that avoiding contention must be the first thought.
<hkaiser>
rtohid: yes, they sent an email around that they would down today
<rtohid>
hkaiser thanks. Do you know when it is expected to be back up?
<hkaiser>
later tonight
<weilewei>
zao thanks for sharing
<bita_>
Is github server working fine for you? nan11 and I can only open every other page
<hkaiser>
github is down, currently
<hkaiser>
give them some time, it will recover soon enough
<bita_>
uhum, sure
<weilewei>
yea, no work can be done :-)
<diehlpk_work>
github is back
<bita_>
yeay
<zao>
> Update - We have rolled out a fix and are monitoring recovery.
<zao>
> Apr 2, 21:06 UTC
<diehlpk_work>
hkaiser, What is the minim boost version for hpx 1.4?
<weilewei>
1.68.0 works for me for hpx 1.4.1 though
<hkaiser>
diehlpk_work: I think 1.61
<hkaiser>
suggested is anything >= 1.67, IIRC
rtohid has quit [Ping timeout: 240 seconds]
rtohid has joined #ste||ar
<diehlpk_work>
Lol, I got Cannot allocate memory from gcc
shahrzad has joined #ste||ar
<diehlpk_work>
Ok, weilewei I got only a memory error, I shoudl read the error log more carefully
<weilewei>
diehlpk_work memory error when compiling things?
<diehlpk_work>
Anyway it seems that hpx 1.4 using boost 1.70
<diehlpk_work>
Yes, gcc ran out of memory
<weilewei>
ok, as long as works for you, then it is fine
<diehlpk_work>
On the login node you can use only 16 cores on 1 CPU and restricted memory
<diehlpk_work>
Make -j did not worked and I had to run on less cores
<weilewei>
really? I tried make -j10 not a problem for me
<bita_>
diehlpk_work, the schedule for Fall is open, I think your course code is Math4997. I don't see if it is cross-listed and as what code it is cross-listed
<diehlpk_work>
Math4997-1
<diehlpk_work>
Jerry never came back to me
<diehlpk_work>
You might ask for permission to take the course
<bita_>
I like to send him an email. can I copy you, hkaiser, dr. ram and nan11, too?
<diehlpk_work>
Yes
<nan11>
sure
<hkaiser>
bita_: it's not cross-listed, but we got explicit acknowledgement that you could use the course for credits
<weilewei>
does this math course count towards CSC degree?
<diehlpk_work>
hkaiser, But this was from CS or?
<diehlpk_work>
weilewei, Yes, CS students can take my course.
<diehlpk_work>
I had one student
<hkaiser>
ece should work as well, if not we need to poke Ram
<bita_>
hkaiser, can we get credit as a graduate (7000) course?
<hkaiser>
ask Jerry
<diehlpk_work>
I had Math graduate students
<bita_>
sure, so I will send an email
<diehlpk_work>
An the course is approved from grad school
<diehlpk_work>
hkaiser, hpx build on Summit
<diehlpk_work>
Next is octotiger
<diehlpk_work>
yeah, octotiger complied on summit
<weilewei>
diehlpk_work I think I can take your course as well
<weilewei>
yea!
rtohid has quit [Ping timeout: 240 seconds]
<weilewei>
diehlpk_work but, can your course count as 7000+ level in CS?
<weilewei>
hkaiser is it possible to have CSC 7700 special topic with you, like a reading class, on Fall2020?
shahrzad has quit [Ping timeout: 240 seconds]
<zao>
In my minds eye, I see "storytime with Uncle Kaiser" with a comfy leather chair and nice dimmed lightning.
<hkaiser>
lol
<hkaiser>
weilewei: sure, we'll read Stepanov's 'Elements of Programming'
<weilewei>
hkaiser I would like to have that this fall, as well as 3 credit hours of CSC9000
<weilewei>
nice!
<hkaiser>
weilewei: and since everything will be still happening online, this is a good topic
<bita_>
hkaiser, is it Okay if I ask which example of HPX uses the number of localities as a parameter in its cmake?
<hkaiser>
sec
<weilewei>
oh, this fall seems like still stay-at-home statu
<weilewei>
status
<hkaiser>
weilewei: highly likely
<hkaiser>
bita_: can't find it right now - give me a bit of time
<wate123_Jun>
....
<weilewei>
ok, that's great, stay at home for a year...
<bita_>
of course, thanks
<weilewei>
I will send an email to you and Symone then for requesting courses
<hkaiser>
weilewei: it's probably early toregister for the fall, isn't it?
<weilewei>
hkaiser right, I will do it when Fall comes closer then
<hkaiser>
weilewei: btw, I was serious about the topic
<weilewei>
hkaiser yes, I understood this
<weilewei>
Elements of Programming is very difficult to read