pmcx.cpp File Reference

Python interface using Pybind11 for MCX. More...

#include <pybind11/pybind11.h>
#include <pybind11/numpy.h>
#include <iostream>
#include <string>
#include "mcx_utils.h"
#include "mcx_core.h"
#include "mcx_const.h"
#include "mcx_shapes.h"
#include <pybind11/iostream.h>

Include dependency graph for pmcx.cpp:

Macros
#define	PYBIND11_DETAILED_ERROR_MESSAGES
#define	RAND_WORD_LEN   4
#define	GET_SCALAR_FIELD(src_pydict, dst_mcx_config, property, py_type)
#define	GET_VEC3_FIELD(src, dst, prop, type)
#define	GET_VEC4_FIELD(src, dst, prop, type)
#define	GET_VEC34_FIELD(src, dst, prop, type)

Functions
	pybind11::PYBIND11_RUNTIME_EXCEPTION (runtime_error, PyExc_RuntimeError)
void	parseVolume (const py::dict &user_cfg, Config &mcx_config)
void	parse_config (const py::dict &user_cfg, Config &mcx_config)
void	cleanup_configs (MCXGPUInfo *&gpu_info, MCXConfig &mcx_config)
py::dict	pmcx_interface (const py::dict &user_cfg)
int	mcx_throw_exception (const int id, const char *msg, const char *filename, const int linenum)
	Error reporting function in PMCX, equivalent to mcx_error in binary mode. More...
void	print_mcx_usage ()
void	mcx_python_flush ()
	Force matlab refresh the command window to print all buffered messages.
py::dict	pmcx_interface_wargs (py::args args, const py::kwargs &kwargs)
py::str	print_version ()
py::list	get_GPU_info ()
	PYBIND11_MODULE (_pmcx, m)

Variables
float *	det_ps = nullptr
int	dim_det_ps [2] = {0, 0}
	buffer to receive data from cfg.detphotons field
int	seed_byte = 0
	dimensions of the cfg.detphotons array

Detailed Description

Python interface using Pybind11 for MCX.

Macro Definition Documentation

GET_SCALAR_FIELD

#define GET_SCALAR_FIELD	(		src_pydict,
			dst_mcx_config,
			property,
			py_type
	)

Value:

if ((src_pydict).contains(#property))\
    {try {(dst_mcx_config).property = py_type((src_pydict)[#property]);\
            std::cout << #property << ": " << (float) (dst_mcx_config).property << std::endl;} \
        catch (const std::runtime_error &err)\
        {throw py::type_error(std::string("Failed to assign MCX property " + std::string(#property) + ". Reason: " + err.what()));}\
    }

Macro to find and extract a scalar property from a source Python dictionary configuration and assign it in a destination MCX Config. The scalar is cast to the python type before assignment.

GET_VEC34_FIELD

#define GET_VEC34_FIELD	(		src,
			dst,
			prop,
			type
	)

Value:

if (src.contains(#prop)) {try {auto list = py::list(src[#prop]);\
            dst.prop = {list[0].cast<type>(), list[1].cast<type>(), list[2].cast<type>(), list.size() == 4 ? list[3].cast<type>() : dst.prop.w}; \
            std::cout << #prop << ": [" << dst.prop.x << ", " << dst.prop.y << ", " << dst.prop.z << ", " << dst.prop.w << "]\n";} \
        catch (const std::runtime_error &err ) {throw py::type_error(std::string("Failed to assign MCX property " + std::string(#prop) + ". Reason: " + err.what()));}                                                                 \
    }

GET_VEC3_FIELD

#define GET_VEC3_FIELD	(		src,
			dst,
			prop,
			type
	)

Value:

if (src.contains(#prop)) {try {auto list = py::list(src[#prop]);\
            dst.prop = {list[0].cast<type>(), list[1].cast<type>(), list[2].cast<type>()};\
            std::cout << #prop << ": [" << dst.prop.x << ", " << dst.prop.y << ", " << dst.prop.z << "]\n";} \
        catch (const std::runtime_error &err ) {throw py::type_error(std::string("Failed to assign MCX property " + std::string(#prop) + ". Reason: " + err.what()));}}

GET_VEC4_FIELD

#define GET_VEC4_FIELD	(		src,
			dst,
			prop,
			type
	)

Value:

if (src.contains(#prop)) {try {auto list = py::list(src[#prop]);\
            dst.prop = {list[0].cast<type>(), list[1].cast<type>(), list[2].cast<type>(), list[3].cast<type>()}; \
            std::cout << #prop << ": [" << dst.prop.x << ", " << dst.prop.y << ", " << dst.prop.z << ", " << dst.prop.w << "]\n";} \
        catch (const std::runtime_error &err ) {throw py::type_error(std::string("Failed to assign MCX property " + std::string(#prop) + ". Reason: " + err.what()));}}

RAND_WORD_LEN

#define RAND_WORD_LEN   4

number of Words per RNG state

Function Documentation

cleanup_configs()

void cleanup_configs ( MCXGPUInfo *&  gpu_info, MCXConfig &  mcx_config  )

inline

Function that's called to cleanup any memory/configs allocated by PMCX. It is used in both normal and exceptional termination of the application

Parameters

gpu_info	reference to an array of MCXGPUInfo data structure
mcx_config	reference to MCXConfig data structure

Here is the call graph for this function:

get_GPU_info()

py::list get_GPU_info

(

)

mcxconfig: structure to store all simulation parameters

gpuinfo: structure to store GPU information

Here is the call graph for this function:

mcx_throw_exception()

int mcx_throw_exception	(	const int	id,
		const char *	msg,
		const char *	filename,
		const int	linenum
	)

Error reporting function in PMCX, equivalent to mcx_error in binary mode.

Parameters

[in]	id	a single integer for the types of the error
[in]	msg	the error message string
[in]	filename	the unit file name where this error is raised
[in]	linenum	the line number in the file where this error is raised

parseVolume()

void parseVolume	(	const py::dict &	user_cfg,
		Config &	mcx_config
	)

Determines the type of volume passed to the interface and decides how to copy it to MCXConfig.

Parameters

user_cfg
mcx_config

float to half conversion https://stackoverflow.com/questions/3026441/float32-to-float16/5587983#5587983 https://gamedev.stackexchange.com/a/17410 (for denorms)

pmcx_interface()

py::dict pmcx_interface

(

const py::dict &

user_cfg

)

gpuInfo: structure to store GPU information

activeDev: count of total active GPUs to be used

The next step, we identify gpu number and query all GPU info

Validate all input fields, and warn incompatible inputs

One must define the domain and properties

Initialize all buffers necessary to store the output variables

Start multiple threads, one thread to run portion of the simulation on one CUDA GPU, all in parallel

Enclose all simulation calls inside a try/catch construct for exception handling

Call the main simulation host function to start the simulation

If error is detected, gracefully terminate the mex and return back to Python

return the final optical properties for polarized MCX simulation

Clear up simulation data structures by calling the destructors

print_version()

py::str print_version

(

)

mcxconfig: structure to store all simulation parameters

Here is the call graph for this function: