GitLab

a template.
So this template has to be given to FOctree, FFmm and FKernels.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@114 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@113 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@112 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@111 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@110 2616d619-271b-44dc-8df4-d4a8f33a7222

and avoid computation when needed (especially for the M2L)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@109 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@108 2616d619-271b-44dc-8df4-d4a8f33a7222

(FTreeCordinate)
To gain time.
Now will test the efficiency.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@107 2616d619-271b-44dc-8df4-d4a8f33a7222

(a littre more efficient)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@106 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@105 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@104 2616d619-271b-44dc-8df4-d4a8f33a7222

This is a simple parameter given to the tree (FOctree) constructor.
(Instead of a template)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@103 2616d619-271b-44dc-8df4-d4a8f33a7222

(step 2/2)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@102 2616d619-271b-44dc-8df4-d4a8f33a7222

(step1/2)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@101 2616d619-271b-44dc-8df4-d4a8f33a7222

(durty way to do that)

In fmm algorithm thread this is an optimization to do P2P
with partitionning.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@100 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@99 2616d619-271b-44dc-8df4-d4a8f33a7222

But to test it now you have to use :
at the top of FList header.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@98 2616d619-271b-44dc-8df4-d4a8f33a7222

Now the leafs contains a list of real particles (and not pointers).
It make the P2P running faster (excepted with openmp, I'm going to fix it now).

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@97 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@96 2616d619-271b-44dc-8df4-d4a8f33a7222

In Extesions we previously return object instead of ref for F3DPosition
Now this is ref (it seems that the compiler already did this because the efficiency is unchanged).

The P2P time contains L2P.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@95 2616d619-271b-44dc-8df4-d4a8f33a7222

now the cells to compute are taken not from the address in the momery
but from the morton index.

Then to compute P2P between two cells we do something like :
if( cell1.index < cell2.index)
    compute(cell1,cell2)
else
    //it has already been computed or will be next time

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@94 2616d619-271b-44dc-8df4-d4a8f33a7222

- change the kernels to be mutual P2P by default

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@93 2616d619-271b-44dc-8df4-d4a8f33a7222

(number of cells, average of child per cells, etc.)
May evolve day after day.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@92 2616d619-271b-44dc-8df4-d4a8f33a7222

(Repartition etc.)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@91 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@90 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@89 2616d619-271b-44dc-8df4-d4a8f33a7222

be used to compute.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@88 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@87 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@86 2616d619-271b-44dc-8df4-d4a8f33a7222

clean/refactor tests

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@85 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@84 2616d619-271b-44dc-8df4-d4a8f33a7222

+
Remove a bug in the mpi version.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@83 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@82 2616d619-271b-44dc-8df4-d4a8f33a7222

Now it works well the results are equal when comparing mpi & no mpi vers.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@81 2616d619-271b-44dc-8df4-d4a8f33a7222

because the width was not correct.
Also change the testMortonIndex.cpp to make
the octree default height value to 6

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@80 2616d619-271b-44dc-8df4-d4a8f33a7222

- to print box properties from files to user
- to because of a mistake in the file reading (width was * 2!
so the program used 4 when width was 2)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@79 2616d619-271b-44dc-8df4-d4a8f33a7222

Valgrind says there is not memory lack (excepted because of openmp)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@78 2616d619-271b-44dc-8df4-d4a8f33a7222

But here is a new tools :
testMortonIndex.cpp
and
FHLoader.hpp
and
Data/testH.txt

This is loading a file and printing all the morton index of the boxes containing particles.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@77 2616d619-271b-44dc-8df4-d4a8f33a7222

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@76 2616d619-271b-44dc-8df4-d4a8f33a7222

But the program failled to dealloc a memory in hybrid mode.
Because of MPI an array is used, and when it is used with openmp a dealloc problem exists.
But this problem does not exist on platfrim...
Need to check again.

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@75 2616d619-271b-44dc-8df4-d4a8f33a7222