compilation using fftw

[Luispain]

Hello everyone !

I am trying to make a fresh compilation of PBBFMM3D linking with a local installation of fftw-3.3.9 .

I have just cloned the PBBFMM3D sources today (28th april '21).

The only modification I have made to Makefile is the adjustment of variable FFTW_ROOT (which now points towards fftw-3.3.9).

Running make binary_file_mykernel produces the following error :

examples/binary_file_mykernel.cpp:64:14: error: cannot declare variable ‘Atree’ to be of abstract type ‘myKernel’
     myKernel Atree(L, tree_level, interpolation_order, eps, use_chebyshev);
              ^
examples/binary_file_mykernel.cpp:4:7: note:   because the following virtual functions are pure within ‘myKernel’:
 class myKernel: public H2_3D_Tree {
       ^
In file included from ./include/kernel_Types.hpp:7:0,
                 from ./include/compute.hpp:9,
                 from ./include/test.hpp:9,
                 from ./include/bbfmm3d.hpp:9,
                 from examples/binary_file_mykernel.cpp:2:
./include/H2_3D_Tree.hpp:94:20: note: 	virtual double H2_3D_Tree::EvaluateKernel(const vector3&, const vector3&)
     virtual double EvaluateKernel(const vector3& targetpos, const vector3& sourcepos) = 0;
                    ^
In file included from ./include/bbfmm3d.hpp:9:0,
                 from examples/binary_file_mykernel.cpp:2:
./include/test.hpp: In instantiation of ‘void DirectCalc3D(T*, const std::vector<_vector3>&, const std::vector<_vector3>&, std::vector<double>&, int, std::vector<double>&, int) [with T = myKernel]’:
examples/binary_file_mykernel.cpp:90:77:   required from here
./include/test.hpp:28:74: error: no matching function for call to ‘myKernel::EvaluateKernel(const _vector3&, const _vector3&)’
                 double val = FMMtree->EvaluateKernel(target[i], source[j]);
                                                                          ^
./include/test.hpp:28:74: note: candidate is:
examples/binary_file_mykernel.cpp:12:20: note: virtual double myKernel::EvaluateKernel(vector3&, vector3&)
     virtual double EvaluateKernel(vector3& targetpos, vector3& sourcepos) {
                    ^
examples/binary_file_mykernel.cpp:12:20: note:   no known conversion for argument 1 from ‘const _vector3’ to ‘vector3& {aka _vector3&}’
make: *** [examples/binary_file_mykernel.o] Error 1

I may be doing something wrong.

Any help would be greatly appreciated,

Thank you very much in advance !
Luis

[Charles-Chao-Chen]

Hello Luis,

Thanks for letting us know. It seems we accidentally pushed some developing codes to the main branch. We should be able to fix it quickly.

[Charles-Chao-Chen]

Hi Luis,

Can you pull and try it again?

[Luispain]

Hi Charles,

Thank you for your help !

Yes, it works now ! I git pulled last version and recompiled (using fftw 3.3.4, because fftw 3.3.9 gives a different error message, perhaps I messed up the installation of fftw 3.3.9).

I compiled binary_file_mykernel and binary_file_standard_kernel and run the executables with the following results :

for binary_file_mykernel :

Precompute files exist.
 Distribute points = 0.00113201 Interaction list = 0.00171781
 FMM distribute = 0.00296497
upward time = 0.00696611 interaction time = 0.012738 downward time = 0.00111008
 neartargetCompute = 0.00329685
Relative Error for first 5000 : 2.06614e-05
Pre-computation time: 0.00694799
FMM computing time:   0.0974002
Exact computing time: 0.299929

And for binary_file_standard_kernel :

Precompute files exist.
 Distribute points = 0.00115013 Interaction list = 0.001894
 FMM distribute = 0.00318193
upward time = 0.00530887 interaction time = 0.00829697 downward time = 0.00281096
 neartargetCompute = 0.00388002
Relative Error for first 5000 : 2.06454e-05
Pre-computation time: 0.012223
FMM computing time:   0.0740399
Exact computing time: 0.329635

I guess everything is running good now.
My next step will consist in installing a Python wrap. I see you have a boost-based solution. I will try that.

Thank you again and have a nice day !
Luis

[Charles-Chao-Chen]

Hi Luis,

Glad to know it is working now. The outputs look good. We have tested the code with fftw3 3.3.8. Please don't hesitate to contact us if you have any questions.