Hello
Running the example shows the following error on an NVIDIA V100 GPU. The CUDA compiler is 11.7, from NVHPC SDK. Do you have some suggestions ? Thanks.

# of arguments: 31
--lxyz : 100 100 350
--pxyz : 0.0 0.0 -175
--nxyz : 200 200 350
--spot_energy : 200.0 0.0
--spot_position : 0 0 0.5
--spot_size : 30.0 30.0
--histories : 100000
--phantom_path : ./water_phantom.raw
--output_prefix : ./
--gpu_id : 0
./
...
...
From phantom: (0, 0, 0): 0.000000
Initialization for geometry done 667.443308 ms
num spots 1
h0 0 h1 100000 28
Vertices total memory 32508 MB Free memory 31830 MB
Scorer offset total memory 32508 MB Free memory 31830 MB
After thread init total memory 32508 MB Free memory 31824 MB
blocks 197 threads 512
After run total memory 32508 MB Free memory 27530 MB
CUDA error (transport particle table): too many resources requested for launch

Hi! Thank you for your incredible work!

I found a bug in the general PBS treatment machine class (base/mqi_treatment_machine_pbs.hpp). I provided the machine parameter from an input file and I obtained random beam positions. In the PBS definition line 105 was missing the assignment of the treatment_machine_ion::SAD

The following is my version of the corrected code.

Best, Giorgio Cartechini

pbs(const std::string filename) : cm2mm(10.0), mm2cm(0.0) {
    this->load_beamdata(filename);
    if (geometry_spec_.SAD[0] == 0)
        treatment_machine_ion<T>::SAD_[0] = std::numeric_limits<T>::infinity();
    else 
        treatment_machine_ion<T>::SAD_[0] = geometry_spec_.SAD[0];
        
    if (geometry_spec_.SAD[1] == 0)
        treatment_machine_ion<T>::SAD_[1] = std::numeric_limits<T>::infinity();
    else
        treatment_machine_ion<T>::SAD_[1] = geometry_spec_.SAD[1];
}

Hi @leehoy!
First, thanks a lot for your work on this nice tool.

I have been working in modifying the build system so that it is compatible with Cuda 8 and NVIDIA Tesla C2070 cards, and cleaning up the CMake files. See https://github.com/ferdymercury/moquimc/

Feel free to pull the changes to your current version. Thanks!

Hello,

First of all, many thanks for your outstanding work on the GPU MC for the quick proton dose calculation. I would love to use it to our clinical beamline at the University of Kansas Medical Center and do some other further research on this codes. I do have a few questions, though:

1. I have completed the testing program for the water phantom. Now I want to use the codes for the CT dicom data and plan dicom data generated by our own system, both generated by RayStation. But there are obvious issues with the output results: firstly, the raw output cannot see my original CT image, and secondly, the output cannot see the dose. I would like to know how to use this code if I want to obtain the CT results with a dose as mentioned in your paper.
   Let me first explain my current approach: the main program I am using now is "tps" in the test file. The following is my input file, both CT and plan are exported from RayStation.
   
   By modifying the program, my “treatment machine” selected "pbs" and used the read file operation in the code to define the parameters of the "pbs" system. Below is my parameter file. Finally, there is the result graph I ran.
   
   
   I would like to know if there are any problems with my above approach and if there are any areas in the code that need to be modified or read from files?

2. Due to the ”1.“ can't achieve good results, I chose to use the test data from the "dicom-interface" project mentioned in your article (Abdomen-1. dicom, 2. dicom, 3. dicom, rtip demo. dicom). But i found that "block(aperture)" is used in this plan, but the "block" section displayed in this code is "todo". If I want this code to be used for plans with blocks, do I need to define the block section myself? Is the test data in the "dicom-interface" project mentioned above not applicable to this code?

3. Is the "range shift" in this code need to define parameters? As I saw in your paper that "The H&N plan has a range shift made of Lucite with density 1.191 g cm-3".

4. In the definition file of "treatment machine", do I need to provide parameters for each energy in order to fully run the program, as shown in the following figure? I am not sure about all the energy parameters of my system at present, and is this writing the reason why it is unable to output normal results. Now I just want to preliminarily test the feasibility. Do I need all the energy parameters here to obtain the results mentioned in your paper? Is it feasible to customize some parameters?
   

5. Based on your code, we would like to add a material at the output of the beam to simulate our research requirements. Can I use a method similar to "range shift" to define my type and use it in "beamline" to achieve my requirements?

I'm sorry that these are so many questions, but I would really love to use your GPU MC code to do some research. I really appreciate your work, and would be happy if you could support me.

Best,

Aoxiang