anisotropy of bodies at the end of the simulation (when nbRoutine > 1)

thomaspeugeot / tkv

10 000

MIT License

0 stars 0 forks source link

anisotropy of bodies at the end of the simulation (when nbRoutine > 1) #1

Closed thomaspeugeot closed 8 years ago

thomaspeugeot commented 8 years ago

Sequence:

load conf-fra-00000.bods
run the simulation till bodies are stabilized

Problem:

an anisotropy appears at the end
it seems short range repulsion is overtaken with a local repulsion field

see

https://github.com/thomaspeugeot/tkv/barnes-hut/Anisotopry%20issue.pptx

thomaspeugeot commented 8 years ago

there is a problem with the hypothesis of long range field anisotropy that creates a body spacing anisotropy because all the rest of the square is evenly spaced. Why would an anisotropic field appears ?

heavy work :

render the repulsion field around the anisotropy
render repulsion vectors

light work :

make short repulsion stronger (does not work)

thomaspeugeot commented 8 years ago

attempt with 1/r3 repulsion vector intensity makes no changes. This goes along that the the field anisotropy must be rendered

thomaspeugeot commented 8 years ago

Note: the issue is not dependant on the siee (8000 or 200000 bodies Note: the issue is not dependant on the theta of the barnes hut (try to lower theta to 0 at the end has no effect) A programming issue ?

thomaspeugeot commented 8 years ago

the problem is solved when NbRoutine == 1. So the problem is concurrency provoked

thomaspeugeot commented 8 years ago

solved : commit log !

    startIndex := (i*sliceLen)/nbRoutine

endIndex := ((i+1)*sliceLen)/nbRoutine -1
endIndex := ((i+1)*sliceLen)/nbRoutine