mlp6 / fem Goto Github PK

instead of having to always re-run Field II when using a new mesh, we should have a function that will take an existing 3D Field II intensity output and 3D map it onto a new grid

original attempt at a test failed b/c the py.test.tmdir fixture created a dynamic path written to the PVD file, so the filecmp.cmp() operation would fail with the validation data. Need to change the code to not write the entire path to the PSD file, but instead simply the basename of the VTR file.

  File "/dscrhome/mlp6/pkgs/fem/post/create_res_sim_mat.py", line 120, in extract_arfi_data
    arfidata[j, i, t - 1] = disp_scale * zdisp[nodeid]
IndexError: index 25897041 is out of bounds for axis 0 with size 25897041

Seems to happen all the way back to versions 6.5.x. This error is happening only on a HUGE mesh, so not sure what is going on.

Generating the disp.dat file from nodout can take a while, and it would be nice to see progressive output status to know how many timesteps have been completed.

Branch mlp6-thermal has two files that have hard-coded adaptations for face heating simulations: field2dynaFaceHeating.m and makeFaceHeating.m. These could either be incorporated into the existing field2dyna.m and makeLoadsTemps.m or simplified to operate as standalone components in a face heating simulation workflow for later combination with the other thermal simulations
The hard-coded transducer dimensions should be revised to read from the probe configuration file.

The current extract_dt() function in create_res_sim only looks for the DATABASE_NODOUT keywork field to find the time step increment, but some models only have DATABASE_BINARY_D3PLOT defined. Would be good to have this be more robust. Either look for both and choose one by default, have the user specify the string to look for, or maybe allow the user to just specify it. ??

from @ndanieley

"impulseResponse":
{
        "f0": 4000000.0,
        "phase": 0.0,
        "bw": 55.0,
        "wavetype": "gaussian"
},

I don't use any of those variables in my code, and as far as I can tell, you
don't use them in the matlab code. am I missing something?

Sam discovered that the axial lens correction is also incorrectly being applied in the lateral and elevation dimensions, creating an incorrect offset for the solved fields.

The CLI syntax for this example needs to be update to reflect the explicit choice of PML (--pml) or NONREFLECT (--nonreflect) since the latest version does not have a default.

last time step can sometimes be incompletely written; either need to check node count and use only if full, or just ignore the last timestep

https://github.com/mlp6/fem/blob/master/docs/thermal.rst

The two variables being passed into format are incorrectly put into a tuple, which creates an IndexError:

These inputs should not be in a tuple!

Attempt to accelerate the post-processing by reading each timestep from nodout in parallel instead of serially.

examples/ are not being pulled down since they are not importable parts of the package, but then the example files are not available locally for someone to easily copy and use. Would be worth pulling these down for a normal end user to have accessible.

The method currently assumes there is a single magnitude at a single, unique node definition in the point loads file; that is not true when there are multiple load components at a node. I think this will currently overwrite an image plane location with whatever the last component magnitude is in the point loads file.

utilize the C version of Field II Pro to run on systems w/o Matlab licenses (but don't drop Matlab support)

getting a strange class error when trying to use linear.m in place of proprietary probe definitions

been a while since i wrote a git issue so apologies if this isn't the correct formatting

1. even though there is a nargin catch in field2dyna it still errors if it does not have the correct number of inputs (there are catches for nargin 7,8, and 11 but 9 and 10 are missing). I think i can fix this by changing the nargin catch for threads to 8 not 7, and for lownslow to 9 not 8. (assuming those without pro will not be running multithreaded or messing with lownslow.
2. in dynaField two lines cause matlab to crash if you do not have field_ii_pro (which students in my class do not):
   set_field('threads', threads);
   disp(sprintf('PARALLEL THREADS: %d', threads));
   If you comment these out it runs fine. but non-pro field doesn't have set_field i think

let me know if you want me to submit my version with these changes? n

scale_disp_dat.cpp doesn't actually use the scaleFactor to scale the displacement data (uses a hard-coded value of 0.1)

In running a mesh that is half-symmetry, and not centered about lat = 0, the following error occurred when trying to create the res_sim.mat file from disp.dat:

In [3]: create_res_sim_mat(DYNADECK)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-3-a01d3b524844> in <module>()
----> 1 create_res_sim_mat(DYNADECK)

~/fem/post/create_res_sim_mat.py in run(dynadeck, disp_comp, disp_scale, ressim, nodedyn, dispout, legacynodes)
     40     [snic, axes] = fem_mesh.SortNodeIDs(node_id_coords)
     41
---> 42     image_plane = extract_image_plane(snic, axes, ele_pos=0.0)
     43
     44     header = read_header(dispout)

~/fem/post/create_res_sim_mat.py in extract_image_plane(snic, axes, ele_pos)
    196
    197     ele0 = np.min(np.where(axes[0] >= ele_pos))
--> 198     image_plane = np.squeeze(snic[ele0, :, :]).astype(int)
    199
    200     return image_plane

ValueError: Can't cast from structure to non-structure, except if the structure only has a single field.

create_res_sim currently extracts all timesteps... would be nice to limit this if a user just needs some or a sparse set for memory concerns, etc. might be a feature to enable in combinations with issue #83

Getting the following compilation error:

$ make
gcc -I . -I /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library checkOnAxis.o checkUniform.o correctAxialLens.o dynaField.o dynaWrite.o field2dyna.o gaussPulse.o readMpn.o main.c cJSON.c /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a -lpthread -lm -O
main.c:8:1: warning: return type defaults to ‘int’ [-Wimplicit-int]
 main()
 ^~~~
/usr/bin/x86_64-linux-gnu-ld: field2dyna.o: relocation R_X86_64_32 against `.rodata' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: gaussPulse.o: relocation R_X86_64_32 against `.rodata' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: readMpn.o: relocation R_X86_64_32 against `.rodata' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(field_II_main.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(transducers.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(calculations_par.o): relocation R_X86_64_32 against symbol `calc_scat_call' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(general.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(thread_lib.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(calculations.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(triangles.o): relocation R_X86_64_32 against `.rodata.str1.8' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: /home/mlp6/Documents/MATLAB/Field_II_Pro/c_library/lib_Field_II.a(lines.o): relocation R_X86_64_32 against `.rodata.str1.1' can not be used when making a PIE object; recompile with -fPIC
/usr/bin/x86_64-linux-gnu-ld: final link failed: Nonrepresentable section on output
collect2: error: ld returned 1 exit status
Makefile:5: recipe for target 'main' failed
make: *** [main] Error 1

@ndanieley did you hit any errors like this?

associated with PR #51

make: *** No rule to make target create_disp_dat_wrap.c, needed by _create_disp_dat.so. Stop.

Need to find a way to automatically build the SWIG module... maybe setup.py?

Time indices for Start/Stop Index are hard-coded to +/- 400 ns, causing errors for very low frequencies probes. Need to more elegantly code these limits.

Some files (vf105) in the ultrasound/probes repository use FIELD_PARAMS.focus_m instead of FIELD_PARAMS.focus. Change one to be in concordance with the other?

currently this method is called multiple times when different components of displacement have to be extracted; would be nice to just pass a vector of the components to extract

we don't need the full dynamic range of double-precision numbers, which are eating up a lot of extra space; will convert to single-precision floats with truncates range

@ndanieley isn't sure what this should be returning and why it appears to be returning nothing (see the fieldC branch code review in progress)

create_disp_dat.py can create the binary data file and the VTK-formatted displacement file from a raw nodout file, but disp.dat files cannot be directly converted to VTK format.

Need to add the ability to offset an excitation from the "center" (x = 0 and y = 0) of a probe.

formatExpImpResp.c:5:28: fatal error: gsl/gsl_interp.h: No such file or directory
 #include <gsl/gsl_interp.h>

This error was on durmstrang.

Utilize a try/except structure to attempt to import the SWIG create_disp_dat, and default to python implementation if it fails.

Memory usage in my tests is high, even when running single-threaded, which usually is not a memory hog. This memory usage also seems to grow, so this isn't something static that is preallocated.

Need to profile this is more detail.

legacy mess of code; use common module that can then write loads or temps

Including the *DATABASE_TPRINT card in the .dyn file that controls the simulation results in temperatures at each node being output for the timestep specified. A program should be written to parse this output and export it into a .mat file. The program should have the option of selecting the entire node set temperature data for exporting, or just exporting the data associated with a desired face of the simulated volume.