DetectorChamber vtk handler

Reviewed-on: #1
added CUDA for images and raytracing, added python bindings via pybind11, removed LTK, added documentation

.clangd

@@ -0,0 +1,52 @@
+CompileFlags:
+  CompilationDatabase: build
+  Add: 
+    - -I/home/rigoni/devel/cmt/ulib/src
+    - -isystem/home/share/micromamba/envs/mutom/include
+    - -isystem/home/share/micromamba/envs/mutom/include/eigen3
+    - -isystem/home/share/micromamba/envs/mutom/targets/x86_64-linux/include
+    - -isystem/home/share/micromamba/envs/mutom/lib/gcc/x86_64-conda-linux-gnu/14.3.0/include/c++
+    - -isystem/isystem/home/share/micromamba/envs/mutom/lib/gcc/x86_64-conda-linux-gnu/14.3.0/include/c++/x86_64-conda-linux-gnu
+    - -isystem/home/share/micromamba/envs/mutom/x86_64-conda-linux-gnu/sysroot/usr/include
+    - "--gcc-toolchain=/home/share/micromamba/envs/mutom"
+    - -D_ULIB_DETAIL_SIGNAL_EMIT
+    - -DUSE_CUDA
+    - -std=c++17
+    - "-D__host__="
+    - "-D__device__="
+    - "-D__global__="
+    - "-D__constant__="
+    - "-D__shared__="
+    - "-D__align__(x)="
+    - "-D__forceinline__=inline"
+    - "-D__launch_bounds__(x)="
+
+Diagnostics:
+  UnusedIncludes: None
+  MissingIncludes: None
+
+---
+If:
+  PathExclude: [/home/rigoni/devel/cmt/ulib/src/.*]
+Diagnostics:
+  Suppress: ["*"]
+
+---
+If:
+  PathMatch: [.*\.cu, .*/src/Math/testing/VoxRaytracerTest.cpp, .*/src/Math/VoxRaytracer.cpp, .*/src/Math/VoxImage.cpp]
+CompileFlags:
+  Add: 
+    - "-x"
+    - "cuda"
+    - "--cuda-path=/home/share/micromamba/envs/mutom"
+    - "--cuda-gpu-arch=sm_61"
+    - "--gcc-toolchain=/home/share/micromamba/envs/mutom"
+    - "-L/home/share/micromamba/envs/mutom/lib"
+    - "-lcudart"
+    - "-lcuda"
+    - "-U__host__"
+    - "-U__device__"
+    - "-U__global__"
+    - "-U__constant__"
+    - "-U__shared__"
+    - "-U__forceinline__"

.gitea/workflows/publish-docs.yaml

@@ -0,0 +1,42 @@
+name: MkDocs Subpath Deploy
+
+on:
+  workflow_dispatch: # trigger manuale
+  push:
+    branches:
+      - main  # Trigger on main branch
+
+jobs:
+  build-and-deploy:
+    runs-on: mildpub # Runner that can access to SSH_YFINPUB_HOST
+    
+    steps:
+      - name: Checkout del codice
+        uses: actions/checkout@v4
+
+      - name: Configura Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.x'
+
+      - name: Installa dipendenze
+        run: |
+          python -m pip install --upgrade pip
+          pip install mkdocs-material 
+          pip install -r docs/docker/requirements.txt
+
+      - name: Build del sito
+        run: mkdocs build
+
+      - name: Deploy via SSH (SCP)
+        uses: https://github.com/appleboy/scp-action@master
+        with:
+          host: ${{ vars.SSH_YFINPUB_HOST }}
+          username: ${{ vars.SSH_YFINPUB_USER }}
+          key: ${{ secrets.MILD_PUB }}
+          port: 22
+          source: "site/*"
+          # Il percorso sul server deve corrispondere alla tua sottopagina
+          target: "/var/www/docs/cmt/uLib/"
+          strip_components: 1 # Rimuove la cartella "site/" e mette solo il contenuto
+          rm: true            # Pulisce la cartella prima di copiare (opzionale, stile Vercel)

.gitignore

@@ -1,3 +1,15 @@
 CMakeFiles/
 build/
 .cache/
+build_warnings*.log
+final_build.log
+cmake_configure.log
+compile_commands.json
+
+dist/
+build_python/
+src/Python/uLib/*.so*
+src/Python/uLib/*.pyd
+src/Python/uLib/*.pyc
+src/Python/uLib/__pycache__
+src/Python/uLib/.nfs*

.vscode/settings.json

@@ -1,8 +1,32 @@
 {
     "clangd.fallbackFlags": [
-        "-I${workspaceFolder}/src",
-        "-I/home/share/micromamba/envs/mutom/include",
-        "-I/home/rigoni/.conan2/p/eigen5481853932f72/p/include/eigen3"
+        "-I/home/rigoni/devel/cmt/ulib/src",
+        "-isystem/home/share/micromamba/envs/mutom/include",
+        "-isystem/home/share/micromamba/envs/mutom/include/eigen3",
+        "-isystem/home/share/micromamba/envs/mutom/targets/x86_64-linux/include",
+        "-isystem/home/share/micromamba/envs/mutom/lib/gcc/x86_64-conda-linux-gnu/14.3.0/include/c++",
+        "-isystem/home/share/micromamba/envs/mutom/lib/gcc/x86_64-conda-linux-gnu/14.3.0/include/c++/x86_64-conda-linux-gnu",
+        "-isystem/home/share/micromamba/envs/mutom/x86_64-conda-linux-gnu/sysroot/usr/include",
+        "--gcc-toolchain=/home/share/micromamba/envs/mutom",
+        "-D__host__=",
+        "-D__device__=",
+        "-D__global__=",
+        "-D__constant__=",
+        "-D__shared__=",
+        "-DUSE_CUDA",
+        "-D__CUDACC__"
     ],
-    "clangd.semanticHighlighting.enable": true
+    "clangd.semanticHighlighting.enable": true,
+    "clangd.arguments": [
+        "--compile-commands-dir=build",
+        "--query-driver=/home/share/micromamba/envs/mutom/bin/g++,/home/share/micromamba/envs/mutom/bin/gcc,/home/share/micromamba/envs/mutom/bin/nvcc",
+        "--suppress-system-warnings",
+        "--all-scopes-completion",
+        "--completion-style=detailed",
+        "--header-insertion=never",
+        "-j=4",
+        "--pch-storage=memory",
+        "--background-index",
+        "--log=verbose"
+    ]
 }

CMake/uLibTargetMacros.cmake

@@ -91,7 +91,9 @@ macro(uLib_add_tests name)
 
     # custom target to compile all tests
     add_custom_target(all-${name}-tests)
-    add_dependencies(all-${name}-tests ${TESTS})
+    if(TESTS)
+        add_dependencies(all-${name}-tests ${TESTS})
+    endif()
 endmacro(uLib_add_tests name)
 
 

CMakeLists.txt

@@ -4,11 +4,31 @@
 ################################################################################
 
 cmake_minimum_required (VERSION 3.26)
+if(POLICY CMP0167)
+    cmake_policy(SET CMP0167 NEW)
+endif()
+
 
 ## -------------------------------------------------------------------------- ##
 
 project(uLib)
 
+# CUDA Toolkit seems to be missing locally. Toggle ON if nvcc is made available.
+option(USE_CUDA "Enable CUDA support" ON)
+if(USE_CUDA)
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -allow-unsupported-compiler")
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} --expt-relaxed-constexpr")
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Wno-deprecated-gpu-targets")
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcudafe \"--diag_suppress=20012\"")
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcudafe \"--diag_suppress=20014\"")
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcudafe \"--diag_suppress=20015\"")
+    find_package(CUDAToolkit REQUIRED)
+    enable_language(CUDA)
+    set(CMAKE_CUDA_ARCHITECTURES 61)
+    include_directories(${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES})
+    add_compile_definitions(USE_CUDA)
+endif()
+
 # The version number.
 set(PROJECT_VERSION_MAJOR 0)
 set(PROJECT_VERSION_MINOR 6)
@@ -86,6 +106,7 @@ set(Boost_USE_STATIC_LIBS OFF)
 set(Boost_USE_MULTITHREADED ON)
 set(Boost_USE_STATIC_RUNTIME OFF)
 message(STATUS "CMAKE_PREFIX_PATH is ${CMAKE_PREFIX_PATH}")
+
 find_package(Boost 1.45.0 COMPONENTS program_options serialization unit_test_framework REQUIRED)
 include_directories(${Boost_INCLUDE_DIRS})
 
@@ -99,6 +120,7 @@ include(${ROOT_USE_FILE})
 find_package(VTK REQUIRED)
 # include(${VTK_USE_FILE})
 
+find_package(pybind11 REQUIRED)
 
 
 option(CENTOS_SUPPORT "VTK definitions for CentOS" OFF)
@@ -124,7 +146,8 @@ else()
                                 RenderingFreeType
                                 RenderingGL2PSOpenGL2
                                 RenderingOpenGL2
-                                RenderingVolumeOpenGL2)
+                                RenderingVolumeOpenGL2
+                                 IOGeometry)
 endif()
 
 set(CMAKE_REQUIRED_INCLUDES CMAKE_REQUIRED_INCLUDES math.h)
@@ -182,8 +205,8 @@ add_subdirectory(${SRC_DIR}/Core)
 include_directories(${SRC_DIR}/Math)
 add_subdirectory(${SRC_DIR}/Math)
 
-include_directories(${SRC_DIR}/Detectors)
-add_subdirectory(${SRC_DIR}/Detectors)
+include_directories(${SRC_DIR}/HEP)
+add_subdirectory(${SRC_DIR}/HEP)
 
 include_directories(${SRC_DIR}/Root)
 add_subdirectory(${SRC_DIR}/Root)
@@ -191,6 +214,8 @@ add_subdirectory(${SRC_DIR}/Root)
 include_directories(${SRC_DIR}/Vtk)
 add_subdirectory(${SRC_DIR}/Vtk)
 
+add_subdirectory(${SRC_DIR}/Python)
+
 #add_subdirectory("${SRC_DIR}/utils/make_recipe")
 
 ## Documentation and packages

CMakePresets.json

@@ -0,0 +1,16 @@
+{
+    "version": 8,
+    "configurePresets": [
+        {
+            "name": "andrea",
+            "displayName": "Custom configure preset",
+            "description": "Sets Ninja generator, build and install directory",
+            "generator": "Ninja",
+            "binaryDir": "${sourceDir}/out/build/${presetName}",
+            "cacheVariables": {
+                "CMAKE_BUILD_TYPE": "Debug",
+                "CMAKE_INSTALL_PREFIX": "${sourceDir}/out/install/${presetName}"
+            }
+        }
+    ]
+}

README.md

@@ -61,3 +61,10 @@ cmake --preset conan-release
 ```bash
 cmake --build build -j10
 ```
+
+### Make python package
+
+```bash
+micromamba run -n mutom env USE_CUDA=ON poetry install
+```
+

build_python.py

@@ -0,0 +1,54 @@
+import os
+import subprocess
+import sys
+import shutil
+
+def build(setup_kwargs):
+    """
+    Build the C++ extension using CMake.
+    This function is called by poetry-core during the build process.
+    The binary is placed directly inside the uLib directory in src/Python.
+    """
+    # Root of the whole project where this build_extension.py is located
+    project_root = os.path.abspath(os.path.dirname(__file__))
+    
+    # Where the extension should go
+    package_dir = os.path.join(project_root, "src/Python/uLib")
+    
+    # Ensure package directory exists
+    os.makedirs(package_dir, exist_ok=True)
+    
+    # Temporary build directory
+    build_temp = os.path.join(project_root, "build_python")
+    os.makedirs(build_temp, exist_ok=True)
+
+    print(f"--- Running CMake build in {build_temp} ---")
+    print(f"Project root: {project_root}")
+    print(f"Target binary dir: {package_dir}")
+
+    # Determine if CUDA should be enabled
+    use_cuda = os.environ.get("USE_CUDA", "OFF").upper()
+    if use_cuda in ["ON", "1", "TRUE", "YES"]:
+        use_cuda = "ON"
+    else:
+        use_cuda = "OFF"
+    
+    # CMake configuration
+    cmake_args = [
+        f"-DCMAKE_LIBRARY_OUTPUT_DIRECTORY={package_dir}",
+        f"-DPYTHON_EXECUTABLE={sys.executable}",
+        "-DCMAKE_BUILD_TYPE=Release",
+        f"-DUSE_CUDA={use_cuda}", 
+        "-G", "Unix Makefiles",
+    ]
+
+    # Use micromamba to ensure Boost and VTK are found during the build
+    subprocess.check_call(["cmake", project_root] + cmake_args, cwd=build_temp)
+    subprocess.check_call(["cmake", "--build", ".", "--parallel", "--target", "uLib_python"], cwd=build_temp)
+    
+    # Ensure the package is found by poetry during the wheel creation process.
+    # Return setup_kwargs for poetry-core.
+    return setup_kwargs
+
+if __name__ == "__main__":
+    build({})

conanfile.txt

@@ -1,6 +1,7 @@
 [requires]
 eigen/3.4.0
 boost/1.83.0
+pybind11/3.0.2
 
 [generators]
 CMakeDeps

docs/assets/css/extensions/tabbed.css

@@ -0,0 +1,41 @@
+.tabbed-set {
+  display: flex;
+  position: relative;
+  flex-wrap: wrap;
+}
+
+.tabbed-set .highlight {
+  background: #ddd;
+}
+
+.tabbed-set .tabbed-content {
+  display: none;
+  order: 99;
+  width: 100%;
+}
+
+.tabbed-set label {
+  width: auto;
+  margin: 0 0.5em;
+  padding: 0.25em;
+  font-size: 120%;
+  cursor: pointer;
+  color: #ffffff !important;
+}
+
+.tabbed-set input {
+  position: absolute;
+  opacity: 0;
+}
+
+.tabbed-set input:nth-child(n+1) {
+  color: #333333;
+}
+
+.tabbed-set input:nth-child(n+1):checked + label {
+    color: cyan !important;
+}
+
+.tabbed-set input:nth-child(n+1):checked + label + .tabbed-content {
+    display: block;
+}

docs/assets/css/extra.css

@@ -0,0 +1,17 @@
+@import "extensions/tabbed.css";
+
+.md-grid {
+    max-width: 100%;
+}
+
+.md-main__inner {
+    margin-top: 0;
+    padding-top: 0;
+}
+
+.md-sidebar--secondary {
+    right: 1.5rem;
+    top: 4.8rem;
+    transform: none;
+    width: 18rem;
+}

docs/docker/Dockerfile

@@ -0,0 +1,30 @@
+# Stage 1: Build the static site using MkDocs
+FROM python:3.9-slim-buster as builder
+
+# Set the working directory
+WORKDIR /app
+
+# Copy the requirements file
+COPY requirements.txt .
+
+# Install the Python dependencies
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the rest of the application files
+COPY ../.. .
+
+# Build the MkDocs site
+RUN mkdocs build
+
+
+# Stage 2: Serve the static files with Nginx
+FROM nginx:alpine
+
+# Copy the built site from the builder stage
+COPY --from=builder /app/site /usr/share/nginx/html
+
+# Expose port 80 for the web server
+EXPOSE 80
+
+# Command to run Nginx in the foreground
+CMD ["nginx", "-g", "daemon off;"]

docs/docker/Dockerfile.dev

@@ -0,0 +1,14 @@
+# Dockerfile for development with live-reloading
+FROM python:3.9-slim-buster
+
+WORKDIR /app
+
+# Copy and install dependencies
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Expose the port MkDocs serve will run on
+EXPOSE 8000
+
+# Command to run the development server
+CMD ["mkdocs", "serve", "--dev-addr", "0.0.0.0:8000"]

docs/docker/docker-compose.yml

@@ -0,0 +1,13 @@
+version: '3.8'
+
+services:
+  mkdocs:
+    build:
+      context: .
+      dockerfile: Dockerfile.dev
+    ports:
+      - "8000:8000"
+    volumes:
+      - ../..:/app
+    environment:
+      - GIT_DISCOVERY_ACROSS_FILESYSTEM=1 

docs/docker/requirements.txt

@@ -0,0 +1,17 @@
+# ------------------------------------------------------------------
+# MkDocs runtime dependencies for the docs Docker image
+# ------------------------------------------------------------------
+
+# Core: theme (provides mkdocs itself as a transitive dep)
+mkdocs-material==9.7.1
+
+# pymdownx.* extensions used in mkdocs.yml:
+#   arithmatex, highlight, superfences, tabbed, details, blocks.caption
+# (also a hard dep of mkdocs-material, pinned here for reproducibility)
+pymdown-extensions>=10.0
+
+# Markdown math rendering support (arithmatex generic mode)
+# JS side is loaded via CDN (polyfill.io + MathJax), no extra Python pkg needed
+
+# Optional: PDF export plugin (exporter: block, currently commented out in mkdocs.yml)
+mkdocs-exporter

docs/docker/runtime.txt

@@ -0,0 +1 @@
+3.7

docs/index.md

@@ -0,0 +1,63 @@
+# uLib
+
+[![DOI](https://zenodo.org/badge/36926725.svg)](https://zenodo.org/badge/latestdoi/36926725)
+
+**uLib** is the base toolkit library for the **CMT (Cosmic Muon Tomography)** project, developed at the University of Padova and INFN Sezione di Padova, Italy.
+
+It provides:
+
+- **Core** – object model, timers, configuration, UUID utilities.
+- **Math** – linear algebra (Eigen3), structured grids, voxel images, ray-tracing, image filters.
+- **Python bindings** – full pybind11 interface for scripting and analysis workflows.
+- Optional **CUDA** acceleration for voxel filtering (transparent RAM ↔ VRAM management).
+
+---
+
+## Quick Start
+
+=== "Users (pip / poetry)"
+    ```bash
+    # Activate your conda/micromamba environment first
+    micromamba activate mutom
+
+    poetry install            # CPU build
+    USE_CUDA=ON poetry install # GPU build
+    ```
+
+=== "Developers (CMake)"
+    ```bash
+    conan install . --output-folder=build --build=missing
+    cmake --preset conan-release
+    cmake --build build --target uLib_python -j$(nproc)
+
+    export PYTHONPATH="$(pwd)/build/src/Python:$(pwd)/src/Python"
+    ```
+
+Then in Python:
+
+```python
+import uLib
+
+# Core
+timer = uLib.Core.Timer()
+timer.Start()
+
+# Math
+grid = uLib.Math.StructuredGrid([10, 10, 10])
+grid.SetSpacing([1.0, 1.0, 1.0])
+
+img = uLib.Math.VoxImage([10, 10, 10])
+img.SetValue(0, 3.14)
+print(img.GetValue(0))
+```
+
+---
+
+## Documentation Sections
+
+| Section | Description |
+|---|---|
+| [Python – Installation](python/installation.md) | Environment setup, user install, developer build |
+| [Python – API Usage](python/usage.md) | Full API reference with examples |
+| [Python – Developer Guide](python/developer_guide.md) | Adding bindings, running tests, build details |
+| [C++ Build – Usage & CUDA](usage/usage.md) | CMake build, CUDA configuration |

docs/python/developer_guide.md

@@ -0,0 +1,179 @@
+# Developer Guide – Python Bindings
+
+This guide is aimed at contributors who want to extend or modify the Python bindings for `uLib`.
+
+---
+
+## Repository Layout
+
+```
+ulib/
+├── src/
+│   └── Python/
+│       ├── module.cpp               # pybind11 module entry point
+│       ├── core_bindings.cpp        # uLib::Core bindings
+│       ├── math_bindings.cpp        # uLib::Math bindings
+│       ├── math_filters_bindings.cpp# VoxImageFilter bindings
+│       ├── CMakeLists.txt           # builds uLib_python shared lib
+│       ├── testing/                 # Python unit tests
+│       │   ├── pybind_test.py
+│       │   ├── core_pybind_test.py
+│       │   ├── math_pybind_test.py
+│       │   └── math_filters_test.py
+│       └── uLib/                    # Python package (uLib_python.so lands here)
+│           └── __init__.py
+├── build_python.py                  # poetry build hook (calls CMake)
+├── pyproject.toml                   # poetry metadata
+└── condaenv.yml                     # conda/micromamba environment
+```
+
+---
+
+## Adding a New Binding
+
+All bindings live in the four source files listed above. The module entry point `module.cpp` calls `init_core()`, `init_math()`, and `init_math_filters()` in order.
+
+### 1. Pick (or create) the right binding file
+
+| C++ header location | Binding file |
+|---|---|
+| `src/Core/` | `core_bindings.cpp` |
+| `src/Math/` (geometry, grids, VoxImage) | `math_bindings.cpp` |
+| `src/Math/VoxImageFilter*.hpp` | `math_filters_bindings.cpp` |
+
+### 2. Add the `#include` directive
+
+```cpp
+// math_bindings.cpp
+#include "Math/MyNewClass.h"
+```
+
+### 3. Write the pybind11 binding inside the appropriate `init_*` function
+
+```cpp
+void init_math(py::module_ &m) {
+    // ... existing bindings ...
+
+    py::class_<MyNewClass>(m, "MyNewClass")
+        .def(py::init<>())
+        .def("MyMethod",  &MyNewClass::MyMethod)
+        .def("AnotherMethod", &MyNewClass::AnotherMethod,
+             py::arg("x"), py::arg("y") = 0.0f);
+}
+```
+
+### 4. Rebuild only the Python target
+
+```bash
+cmake --build build --target uLib_python -j$(nproc)
+```
+
+### 5. Write a Python test
+
+Add a new test class to the relevant test file (or create a new one under `src/Python/testing/`):
+
+```python
+# src/Python/testing/math_pybind_test.py
+class TestMyNewClass(unittest.TestCase):
+    def test_basic(self):
+        obj = uLib.Math.MyNewClass()
+        result = obj.MyMethod()
+        self.assertAlmostEqual(result, expected_value)
+```
+
+Register the test in `src/Python/CMakeLists.txt` if you add a new file:
+
+```cmake
+add_test(NAME pybind_my_new
+         COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/testing/my_new_test.py)
+set_tests_properties(pybind_my_new PROPERTIES
+    ENVIRONMENT "PYTHONPATH=$<TARGET_FILE_DIR:uLib_python>:${PROJECT_SOURCE_DIR}/src/Python")
+```
+
+---
+
+## Build System Details
+
+### CMakeLists.txt (`src/Python/`)
+
+`pybind11_add_module` compiles the shared library `uLib_python` and links it against the C++ static/shared libraries `uLibCore` and `uLibMath`. The install target copies the `.so` into the standard library directory.
+
+```cmake
+pybind11_add_module(uLib_python
+    module.cpp core_bindings.cpp math_bindings.cpp math_filters_bindings.cpp)
+
+target_link_libraries(uLib_python PRIVATE uLibCore uLibMath)
+```
+
+### poetry / build_python.py
+
+`pyproject.toml` declares `build_python.py` as the custom build hook. When `poetry install` or `poetry build` is invoked it:
+
+1. Calls `cmake <root> -DCMAKE_LIBRARY_OUTPUT_DIRECTORY=<pkg_dir> ...` in `build_python/`.
+2. Builds only the `uLib_python` target.
+3. The resulting `.so` is placed inside `src/Python/uLib/` so it is picked up by Poetry as a package data file.
+
+The `USE_CUDA` environment variable gates CUDA support at build time:
+
+```bash
+USE_CUDA=ON poetry install   # with CUDA
+USE_CUDA=OFF poetry install  # CPU only (default)
+```
+
+---
+
+## Running All Tests
+
+```bash
+# From the repository root, with PYTHONPATH set:
+export PYTHONPATH="$(pwd)/build/src/Python:$(pwd)/src/Python"
+
+python -m pytest src/Python/testing/ -v
+```
+
+Or through CMake's test runner (after building the full project):
+
+```bash
+cd build
+ctest --output-on-failure -R pybind
+```
+
+Expected output (all passing):
+
+```
+    Start 1: pybind_general
+1/4 Test #1: pybind_general .............   Passed
+    Start 2: pybind_core
+2/4 Test #2: pybind_core ................   Passed
+    Start 3: pybind_math
+3/4 Test #3: pybind_math ................   Passed
+    Start 4: pybind_math_filters
+4/4 Test #4: pybind_math_filters ........   Passed
+```
+
+---
+
+## Memory Management Notes
+
+`uLib::Vector<T>` has explicit GPU memory management. When wrapping methods that return references to internal data, use `py::return_value_policy::reference_internal` to avoid dangling references:
+
+```cpp
+.def("Data", &VoxImage<Voxel>::Data,
+     py::return_value_policy::reference_internal)
+```
+
+For objects held by `std::unique_ptr` without Python-side deletion, use `py::nodelete`:
+
+```cpp
+py::class_<Abstract::VoxImageFilter,
+           std::unique_ptr<Abstract::VoxImageFilter, py::nodelete>>(m, "AbstractVoxImageFilter")
+```
+
+---
+
+## Useful References
+
+- [pybind11 documentation](https://pybind11.readthedocs.io)
+- [pybind11 – STL containers](https://pybind11.readthedocs.io/en/stable/advanced/cast/stl.html)
+- [pybind11 – Eigen integration](https://pybind11.readthedocs.io/en/stable/advanced/cast/eigen.html)
+- [CMake – pybind11 integration](https://pybind11.readthedocs.io/en/stable/compiling.html)

docs/python/installation.md

@@ -0,0 +1,146 @@
+# Python Installation
+
+The `uLib` Python package exposes the Core and Math C++ libraries via [pybind11](https://pybind11.readthedocs.io) bindings. There are two ways to install it: as an **end user** (pre-built wheel / pip) or as a **developer** (editable build from source).
+
+---
+
+## Prerequisites
+
+`uLib` depends on native C++ libraries that must be compiled. Ensure the following are available in your environment before installing:
+
+| Dependency | Minimum version | Notes |
+|---|---|---|
+| Python | 3.9 | |
+| CMake | 3.12 | |
+| pybind11 | 2.6.0 | |
+| Conan | 2.x | for Eigen3 / Boost |
+| micromamba / conda | any | recommended – provides ROOT, VTK |
+
+### Creating the `mutom` Conda/Micromamba Environment
+
+A ready-to-use environment definition is provided as `condaenv.yml` at the repository root.
+
+=== "Micromamba"
+    ```bash
+    micromamba env create -f condaenv.yml
+    micromamba activate mutom
+    ```
+
+=== "Conda"
+    ```bash
+    conda env create -f condaenv.yml
+    conda activate mutom
+    ```
+
+The environment installs CMake, Conan, ROOT, VTK, and the compiler toolchain.
+
+> **CUDA (optional)**  
+> If you want GPU-accelerated voxel filtering, you also need NVCC inside the environment:
+> ```bash
+> micromamba install cuda-nvcc -c conda-forge
+> ```
+
+---
+
+## User Installation (wheel / pip)
+
+Once the native dependencies are present in your environment, install the package with Poetry or pip:
+
+```bash
+# Activate your environment first
+micromamba activate mutom
+
+# Build and install (CUDA disabled by default)
+poetry install
+
+# Build and install with CUDA support
+USE_CUDA=ON poetry install
+```
+
+After installation the module is importable from anywhere in the environment:
+
+```python
+import uLib
+print(dir(uLib.Core))
+print(dir(uLib.Math))
+```
+
+---
+
+## Developer Installation (editable / in-tree build)
+
+For development you typically want to skip the packaging layer and work directly against the CMake build tree.
+
+### Step 1 – Install Conan dependencies
+
+```bash
+conan profile detect          # first time only
+conan install . --output-folder=build --build=missing
+```
+
+### Step 2 – Configure and build
+
+```bash
+# Standard release build
+cmake --preset conan-release
+
+# …or manually
+cmake -B build \
+      -DCMAKE_TOOLCHAIN_FILE=build/conan_toolchain.cmake \
+      -DCMAKE_BUILD_TYPE=Release \
+      -DUSE_CUDA=OFF          # set to ON when a GPU is available
+
+cmake --build build --target uLib_python -j$(nproc)
+```
+
+The shared library (`uLib_python*.so`) is written to `build/src/Python/`.
+
+### Step 3 – Make the module importable
+
+Point `PYTHONPATH` at the build output **and** the Python source directory (the latter carries the `uLib/__init__.py` that stitches sub-modules together):
+
+```bash
+export PYTHONPATH="$(pwd)/build/src/Python:$(pwd)/src/Python:$PYTHONPATH"
+python -c "import uLib; print(uLib.__version__)"
+```
+
+Or, for a one-shot check:
+
+```bash
+PYTHONPATH="build/src/Python:src/Python" python src/Python/testing/pybind_test.py
+```
+
+### Step 4 – Run the tests
+
+CMake registers the Python tests alongside the C++ ones; use `ctest` from the build directory:
+
+```bash
+cd build
+ctest --output-on-failure -R pybind
+```
+
+Individual test scripts can also be run directly once `PYTHONPATH` is set:
+
+```bash
+python src/Python/testing/core_pybind_test.py
+python src/Python/testing/math_pybind_test.py
+python src/Python/testing/math_filters_test.py
+```
+
+---
+
+## Verifying the Installation
+
+```python
+import uLib
+
+# Core module
+obj    = uLib.Core.Object()
+timer  = uLib.Core.Timer()
+timer.Start()
+elapsed = timer.StopWatch()   # float, seconds
+
+# Math module
+v3 = uLib.Math.Vector3f([1.0, 0.0, 0.0])
+print(v3[0])  # 1.0
+```

docs/python/usage.md

@@ -0,0 +1,373 @@
+# Python API Usage
+
+The `uLib` Python package is split into two sub-modules mirroring the C++ library:
+
+| Sub-module | Contents |
+|---|---|
+| `uLib.Core` | Low-level utilities: `Object`, `Timer`, `Options`, `TypeRegister` |
+| `uLib.Math` | Geometry, grids, voxel images, ray-tracing, image filters |
+
+```python
+import uLib
+# Sub-modules are accessible as attributes
+uLib.Core   # core utilities
+uLib.Math   # mathematical structures
+```
+
+---
+
+## uLib.Core
+
+### Object
+
+Base class for uLib objects; exposed to Python for type-hierarchy purposes.
+
+```python
+obj = uLib.Core.Object()
+copy = obj.DeepCopy()
+```
+
+### Timer
+
+Precision wall-clock timer.
+
+```python
+import time
+timer = uLib.Core.Timer()
+timer.Start()
+time.sleep(0.5)
+elapsed = timer.StopWatch()   # returns elapsed seconds as float
+print(f"Elapsed: {elapsed:.3f} s")
+```
+
+### Options
+
+Wraps Boost.ProgramOptions for INI-style configuration files.
+
+```python
+opt = uLib.Core.Options("My Program")
+opt.parse_config_file("config.ini")   # load settings
+n = opt.count("my_key")               # check if key exists
+opt.save_config_file("out.ini")
+```
+
+---
+
+## uLib.Math – Linear Algebra
+
+The math module exposes Eigen3 vectors and matrices as well-typed Python objects with NumPy interoperability.
+
+### Fixed-size Vectors
+
+```python
+import numpy as np
+import uLib
+
+# Construct from list
+v3 = uLib.Math.Vector3f([1.0, 2.0, 3.0])
+print(v3[0], v3[1], v3[2])   # 1.0 2.0 3.0
+
+# Construct from NumPy array
+arr = np.array([4.0, 5.0, 6.0], dtype=np.float32)
+v3b = uLib.Math.Vector3f(arr)
+
+# Zero-initialise
+v4d = uLib.Math.Vector4d()    # all zeros
+
+# Available types
+# Vector1f / 2f / 3f / 4f  (float32)
+# Vector1d / 2d / 3d / 4d  (float64)
+# Vector1i / 2i / 3i / 4i  (int32)
+```
+
+### Fixed-size Matrices
+
+```python
+# 2-by-2 float matrix
+m2f = uLib.Math.Matrix2f()
+m2f[0, 0] = 1;  m2f[0, 1] = 2
+m2f[1, 0] = 3;  m2f[1, 1] = 4
+
+# From list (row-major)
+m4f = uLib.Math.Matrix4f([1,0,0,0, 0,1,0,0, 0,0,1,0, 0,0,0,1])
+
+# From NumPy (2-D array)
+mat = np.eye(3, dtype=np.float32)
+m3f = uLib.Math.Matrix3f(mat)
+
+# Dynamic matrices
+mXf = uLib.Math.MatrixXf(4, 4)   # 4×4 float, zeros
+```
+
+### Homogeneous Types
+
+```python
+# HPoint3f – a 3-D point in homogeneous coordinates (w = 1)
+p = uLib.Math.HPoint3f(1.0, 2.0, 3.0)
+
+# HVector3f – a free vector (w = 0)
+v = uLib.Math.HVector3f(0.0, 1.0, 0.0)
+
+# HLine3f – a parametric ray
+line = uLib.Math.HLine3f()
+line.origin    = uLib.Math.HPoint3f(0, 0, 0)
+line.direction = uLib.Math.HVector3f(0, 0, 1)
+```
+
+---
+
+## uLib.Math – Transforms and Geometry
+
+### AffineTransform
+
+A rigid-body / affine transform stored as a 4×4 matrix.
+
+```python
+tf = uLib.Math.AffineTransform()
+
+tf.SetPosition([1.0, 0.0, 0.0])   # translate
+tf.Translate([0.0, 1.0, 0.0])     # cumulative translate
+tf.Scale([2.0, 2.0, 2.0])         # uniform scale
+tf.Rotate(uLib.Math.Vector3f([0, 0, 3.14159]))  # Euler angles (rad)
+
+mat4 = tf.GetWorldMatrix()   # 4×4 matrix
+pos  = tf.GetPosition()      # Vector3f
+```
+
+### Geometry
+
+Inherits `AffineTransform`; converts points between world and local frames.
+
+```python
+geo = uLib.Math.Geometry()
+geo.SetPosition([1.0, 1.0, 1.0])
+
+world_pt = uLib.Math.Vector4f([2.0, 3.0, 2.0, 1.0])
+local_pt = geo.GetLocalPoint(world_pt)
+back     = geo.GetWorldPoint(local_pt)
+# back ≈ [2, 3, 2, 1]
+```
+
+### ContainerBox
+
+An axis-aligned bounding box with an associated transform.
+
+```python
+box = uLib.Math.ContainerBox()
+box.SetOrigin([-1.0, -1.0, -1.0])
+box.SetSize([2.0, 2.0, 2.0])
+print(box.GetSize())    # [2, 2, 2]
+```
+
+---
+
+## uLib.Math – Structured Grids
+
+### StructuredGrid (3-D)
+
+A 3-D voxel grid (origin, spacing, and integer dimensions).
+
+```python
+dims = uLib.Math.Vector3i([10, 10, 10])
+grid = uLib.Math.StructuredGrid(dims)
+grid.SetSpacing([1.0, 1.0, 1.0])
+grid.SetOrigin([0.0, 0.0, 0.0])
+
+print(grid.GetSpacing())          # [1, 1, 1]
+print(grid.IsInsideBounds([5, 5, 5, 1]))  # True
+idx = grid.Find([2.5, 2.5, 2.5]) # returns grid cell index
+```
+
+### Structured2DGrid / Structured4DGrid
+
+```python
+grid2d = uLib.Math.Structured2DGrid()
+grid2d.SetDims([100, 100])
+grid2d.SetPhysicalSpace([0, 0], [1, 1])
+print(grid2d.GetSpacing())
+```
+
+---
+
+## uLib.Math – VoxImage
+
+`VoxImage` is a 3-D voxel volume where each cell stores a `Voxel` ( `.Value` + `.Count`).
+
+```python
+dims = uLib.Math.Vector3i([20, 20, 20])
+img  = uLib.Math.VoxImage(dims)
+img.SetSpacing([0.5, 0.5, 0.5])
+
+# Access by linear index
+img.SetValue(0, 42.0)
+print(img.GetValue(0))    # 42.0
+
+# Access by 3-D index
+img.SetValue(uLib.Math.Vector3i([1, 1, 1]), 7.5)
+print(img.GetValue(uLib.Math.Vector3i([1, 1, 1])))  # 7.5
+
+# Clipping / masking helpers
+cropped = img.clipImage(uLib.Math.Vector3i([2, 2, 2]),
+                        uLib.Math.Vector3i([18, 18, 18]))
+masked  = img.maskImage(0.0, 100.0, 0.0)   # mask outside [0, 100]
+
+# I/O
+img.ExportToVti("output.vti")
+img.ImportFromVti("output.vti")
+```
+
+### Voxel (element type)
+
+```python
+vox = uLib.Math.Voxel()
+vox.Value = 1.5
+vox.Count = 3
+
+data = img.Data()        # returns the underlying Vector_Voxel
+vox0 = data[0]
+print(vox0.Value, vox0.Count)
+```
+
+---
+
+## uLib.Math – VoxRaytracer
+
+Performs ray-tracing through a `StructuredGrid` and returns per-voxel chord lengths.
+
+```python
+import numpy as np
+import uLib
+
+grid = uLib.Math.StructuredGrid([10, 10, 10])
+grid.SetSpacing([1.0, 1.0, 1.0])
+grid.SetOrigin([0.0, 0.0, 0.0])
+
+rt = uLib.Math.VoxRaytracer(grid)
+
+# Trace a ray between two homogeneous points (x, y, z, w=1)
+p1 = np.array([0.5, 0.5, -1.0, 1.0], dtype=np.float32)
+p2 = np.array([0.5, 0.5, 11.0, 1.0], dtype=np.float32)
+result = rt.TraceBetweenPoints(p1, p2)
+
+print("Voxels crossed:", result.Count())
+print("Total length  :", result.TotalLength())
+
+elements = result.Data()
+for i in range(result.Count()):
+    print(f"  vox_id={elements[i].vox_id}  L={elements[i].L:.4f}")
+```
+
+---
+
+## uLib.Math – Image Filters
+
+All filters share the same interface: construct with a kernel size, attach a `VoxImage`, optionally set parameters, then call `.Run()`.
+
+```python
+import uLib
+
+dims = uLib.Math.Vector3i([10, 10, 10])
+img  = uLib.Math.VoxImage(dims)
+for i in range(10**3):
+    img.SetValue(i, float(i))
+
+kernel_dims = uLib.Math.Vector3i([3, 3, 3])
+```
+
+### Linear (Gaussian / Box) Filter
+
+```python
+filt = uLib.Math.VoxFilterAlgorithmLinear(kernel_dims)
+filt.SetImage(img)
+filt.SetKernelNumericXZY([1.0] * 27)   # uniform box kernel, length = product of dims
+filt.Run()
+```
+
+### ABTrim Filter
+
+Applies alpha-beta trimming to remove outliers before averaging.
+
+```python
+filt = uLib.Math.VoxFilterAlgorithmAbtrim(kernel_dims)
+filt.SetImage(img)
+filt.SetKernelNumericXZY([1.0] * 27)
+filt.SetABTrim(2, 2)    # trim 2 low and 2 high values
+filt.Run()
+```
+
+### Bilateral Filter
+
+Edge-preserving smoothing controlled by a spatial sigma (from the kernel shape) and an intensity sigma.
+
+```python
+filt = uLib.Math.VoxFilterAlgorithmBilateral(kernel_dims)
+filt.SetImage(img)
+filt.SetKernelNumericXZY([1.0] * 27)
+filt.SetIntensitySigma(0.3)
+filt.Run()
+```
+
+### Threshold Filter
+
+Zeros voxels below a threshold.
+
+```python
+filt = uLib.Math.VoxFilterAlgorithmThreshold(kernel_dims)
+filt.SetImage(img)
+filt.SetKernelNumericXZY([1.0] * 27)
+filt.SetThreshold(0.5)
+filt.Run()
+```
+
+### Median Filter
+
+```python
+filt = uLib.Math.VoxFilterAlgorithmMedian(kernel_dims)
+filt.SetImage(img)
+filt.SetKernelNumericXZY([1.0] * 27)
+filt.Run()
+```
+
+---
+
+## uLib.Math – Accumulators
+
+Accumulators collect scalar samples and return a summary statistic.
+
+```python
+# Arithmetic mean
+acc = uLib.Math.Accumulator_Mean_f()
+acc(10.0)
+acc(20.0)
+mean = acc()       # 15.0
+
+# Alpha-beta trimmed mean
+acc2 = uLib.Math.Accumulator_ABTrim_f()
+acc2.SetABTrim(0.1, 0.1)   # trim bottom 10 % and top 10 %
+acc2 += 1.0
+acc2 += 9999.0             # outlier
+acc2 += 5.0
+result = acc2()            # trimmed mean ≈ 3.0
+```
+
+---
+
+## Dynamic Vectors (`uLib.Math.Vector_*`)
+
+Typed dynamic arrays backed by `uLib::Vector<T>` with optional CUDA memory management.
+
+```python
+# Integer vector
+vi = uLib.Math.Vector_i()
+vi.append(1); vi.append(2); vi.append(3)
+print(len(vi), vi[0])
+
+# Float vector with CUDA management
+vf = uLib.Math.Vector_f()
+vf.append(1.5)
+vf.MoveToVRAM()   # copy to GPU (no-op when CUDA is absent)
+vf.MoveToRAM()    # copy back to CPU
+
+# Other types: Vector_ui, Vector_l, Vector_ul, Vector_d
+# Compound element types: Vector_Vector3f, Vector_Vector4f, Vector_Voxel …
+```

docs/usage/usage.md

@@ -0,0 +1,60 @@
+# Usage and Installation Guide
+
+## Requirements
+
+### Compiling with CUDA Support
+
+The library supports running VoxImage filtering operations directly on CUDA cores via transparent RAM/VRAM memory transfers.
+
+By default, the `CMakeLists.txt` build system sets `USE_CUDA=ON` and will attempt to locate `nvcc` and the NVIDIA CUDA Toolkit. If the toolkit is missing, `CMake` will fail unless you explicitly configure the project with `-DUSE_CUDA=OFF`.
+
+### 1. Installing CUDA Environment via Micromamba
+
+If you are developing inside an isolated Conda/Micromamba environment (e.g., `mutom`), you can inject the CUDA compilers directly into your environment rather than relying on global system dependencies:
+
+```bash
+# Add the conda-forge channel if not already available
+micromamba config append channels conda-forge
+
+# Install nvcc and the necessary CUDA toolkit components
+micromamba install cuda-nvcc
+```
+
+Verify your installation:
+
+```bash
+nvcc --version
+```
+
+### 2. Building the Project
+
+Configure and compile the project using standard CMake flows:
+
+```bash
+mkdir -p build && cd build
+
+# Configure CMake
+# (Optional) Explicitly toggle CUDA: cmake -DUSE_CUDA=ON ..
+cmake ..
+
+# Compile the project and tests
+make -j $(nproc)
+```
+
+### 3. Validating CUDA Support
+
+You can verify that the CUDA kernels are launching correctly and allocating device memory through `DataAllocator` by running the mathematical unit tests.
+
+```bash
+# From the build directory
+./src/Math/testing/VoxImageFilterTest
+
+# Output should show:
+# "Data correctly stayed in VRAM after CUDA execution!"
+```
+
+## How It Works Under The Hood
+
+The `DataAllocator<T>` container automatically wraps memory allocations to transparently map to CPU RAM, or GPU VRAM. Standard iteration automatically pulls data backwards using implicit `MoveToRAM()` calls. 
+
+Filters using `#ifdef USE_CUDA` explicitly dictate `<buffer>.MoveToVRAM()` allocating directly on device bounds seamlessly. Fallbacks to Host compute iterations handle themselves automatically. Chaining specific filters together safely chains continuous VRAM operations avoiding costly Host copies in between iterations.

mkdocs.yml

@@ -0,0 +1,561 @@
+# site_name: uLib Documentation
+# site_description: CMT Cosmic Muon Tomography – uLib toolkit
+# site_author: Andrea Rigoni Garola
+# repo_url: https://github.com/cmt/ulib
+# docs_dir: docs
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | This is the main file used by MkDocs to build the pages.                                               |
+# | It contains a lot of information and settings for you to read and use.                                 |
+# | Comments may contain "Read More" URLs to more in-depth documentation about the option for you to read. |
+# |                                                                                                        |
+# | You can check out https://www.mkdocs.org/user-guide/configuration/ for a more detailed explanation of  |
+# | all the options MkDocs offers by default.                                                              |
+# |                                                                                                        |
+# +------------------------------------------------- NOTE -------------------------------------------------+
+# |                                                                                                        |
+# | Some of the options listed here are only available through the usage of Material for MkDocs.           |
+# | Those options will usually have a link to the docs of this Theme and also mention "Material" as name.  |
+# | The actual name of the theme is "Material for MkDocs" and "Material" is used for simplicity reasons.   |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | Main Page Settings for MkDocs.                                                                         |
+# | Those settings are site name, site description, Site author and also Site URL (Canonical URL)          |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#site_name                                           |
+# | - https://www.mkdocs.org/user-guide/configuration/#site_description                                    |
+# | - https://www.mkdocs.org/user-guide/configuration/#site_author                                         |
+# | - https://www.mkdocs.org/user-guide/configuration/#site_url                                            |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+site_name: OpenCMT uLib Documentation
+site_url: https://docs.mildstone.org/uLib/ # <--- project subfolder
+use_directory_urls: true
+site_description: 'Documentation for OpenCMT uLib'
+site_author: 'Andrea Rigoni Garola'
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | This setting allows you to define your own Copyright notice.                                           |
+# | The text is treated as HTML code so you can use things like <a> tags or &copy; to display the          |
+# | Copyright icon.                                                                                        |
+# |                                                                                                        |
+# | Where or IF the Copyright is displayed depends on the theme you use.                                   |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#copyright                                           |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+copyright: |
+  &copy; Author
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | The base folder to use.                                                                                |
+# | Any Markdown files you put into this folder will be turned into a static HTML page once you build or   |
+# | publish your page.                                                                                     |
+# |                                                                                                        |
+# | It is also used as the base directory for other settings like the "extra_css" or "extra_javascript"    |
+# | option.                                                                                                |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+docs_dir: docs/
+
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | These options allow to define a Repository to link to.                                                 |
+# | The result will, depending on the theme, be a link somewhere shown on the page that links to the       |
+# | Repository with the specified repo_name.                                                               |
+# |                                                                                                        |
+# | This will also enable a "edit" button on the page itself that allows the direct editing of the page.   |
+# | You can disable this by setting "edit_uri" to an empty String.                                         |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#repo_name                                           |
+# | - https://www.mkdocs.org/user-guide/configuration/#repo_url                                            |
+# | - https://www.mkdocs.org/user-guide/configuration/#edit_uri                                            |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+repo_name: OpenCMT/uLib
+repo_url: https://gitea.mildstone.org/OpenCMT/uLib.git
+#edit_uri: tree/master/docs # Uncomment to define a different URI/URL for the "edit" option
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | The "nav" option is where you define the navigation to show in MkDocs.                                 |
+# |                                                                                                        |
+# | Depending on the theme you use will the resulting Navigation look different.                           |
+# |                                                                                                        |
+# | You can set different types of navigations. Either just the path, the path with a separate title or    |
+# | an external URL.                                                                                       |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#documentation-layout                                |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+
+nav:
+  - Home: index.md
+  - Python:
+    - Installation: python/installation.md
+    - API Usage: python/usage.md
+    - Developer Guide: python/developer_guide.md
+  - C++ Build:
+    - Usage & CUDA: usage/usage.md
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | The "theme" section allows you to define what theme to use.                                            |
+# | It is also used for theme-specific options, but also for advanced stuff such as theme-extensions, if   |
+# | the theme actually supports it.                                                                        |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#theme                                               |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+theme:
+
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "name" option is where you define the theme to use.                                              |
+  # |                                                                                                      |
+  # | Note that not all themes are included by default and will require you to install them first.         |
+  # | The Material theme is one of them. See the "Read More" link for instructions on how to install it.   |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/getting-started/                                       |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  name: 'material'
+
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The Material theme allows "theme-extsnions", meaning that you can override parts of it by either     |
+  # | overriding a particular file, or only parts (blocks) of it.                                          |
+  # |                                                                                                      |
+  # | If you want to override parts of Material, uncomment the "custom_dir" option below and set the       |
+  # | folder (relative to the mkdocs.yml file) where your theme extensions will be located at.             |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://www.mkdocs.org/user-guide/configuration/#custom_dir                                        |
+  # | - https://squidfunk.github.io/mkdocs-material/customization/#extending-the-theme                     |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #custom_dir: 'theme'
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "favicon" option allows you to set your own image/icon to use in the browser-tab.                |
+  # |                                                                                                      |
+  # | Pretty much all image types are supported, but it's recommended to use a PNG, SVG or ICO image for   |
+  # | the favicon.                                                                                         |
+  # |                                                                                                      |
+  # | The directory is relative to the "docs_dir".                                                         |
+  # |                                                                                                      |
+  # | Example: Having a favicon.png in docs/assets/images will result in the "favicon" setting showing     |
+  # |          'assets/images/favicon.png'                                                                 |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-logo-and-icons/#favicon             |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #favicon: 'assets/images/favicon.png'
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "palette" section is a Material option and allows you to define specific style options such as   |
+  # | Color-Scheme, and primary and secondary Color.                                                       |
+  # |                                                                                                      |
+  # | You can also define multiple palettes that can have different Color Schemses and primary and/or      |
+  # | secondary Colors.                                                                                    |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-colors/                             |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-colors/#color-palette-toggle        |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  palette:
+  
+    # Palette toggle for light mode
+    - media: "(prefers-color-scheme: light)"
+      scheme: default
+      primary: 'indigo'
+      accent: 'indigo'
+      toggle:
+        icon: material/brightness-7
+        name: Switch to dark mode
+
+    # Palette toggle for dark mode
+    - media: "(prefers-color-scheme: dark)"
+      scheme: slate
+      primary: 'indigo'
+      accent: 'indigo'
+      toggle:
+        icon: material/brightness-4
+        name: Switch to light mode
+    
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | With the "font" option can you set a different font to use.                                          |
+  # |                                                                                                      |
+  # | Material supports all Google fonts, but you can also define your own ones if you choose so.          |
+  # |                                                                                                      |
+  # | The "text" option is used for the regular font while "code" is used for code blocks, inline code and |
+  # | similar.                                                                                             |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-fonts/                              |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #font:
+  #  text: 'Roboto'
+  #  code: 'Roboto Mono'
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | Material suppors more than 40 different languages which you can set using the "language" option      |
+  # | below.                                                                                               |
+  # |                                                                                                      |
+  # | The default language is "en" (English).                                                              |
+  # |                                                                                                      |
+  # | You can also enable/set a "selector" to allow switching between languages.                           |
+  # | See the "alternate" option in the "extra" section below for more information on this topic.          |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-language/                           |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #language: 'en'
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "direction" option is commonly used together with the "language" option.                         |
+  # |                                                                                                      |
+  # | It allows you to change the text direction from the default left-to-right (ltr) to right-to-left     |
+  # | (rtl) which is used in certain languages.                                                            |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-language/#directionality            |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #direction: 'ltr'
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "features" option allows you to enable specific features of Material, by adding them to the      |
+  # | list.                                                                                                |
+  # |                                                                                                      |
+  # | Features are in the format <category>.<name>. As an example, the feature to enable tabs is called    |
+  # | navigation.tabs.                                                                                     |
+  # |                                                                                                      |
+  # | The list below contains all known features of Material.                                              |
+  # |                                                                                                      |
+  # | Features marked with a * are currently Insiders-only. (Last update: 11th December 2021)              |
+  # | https://squidfunk.github.io/mkdocs-material/insiders/                                                |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/setting-up-navigation/                           |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  features:
+  # Announce
+  #
+  #- announce.dismiss # Adds a "X" button to dismiss a news banner/mark it as read.*
+  
+  # Header
+  #
+  #- header.autohide # Hide header when user scrolls past a specific point.
+  
+  # Navigation:
+  #
+  #- navigation.expand      # Expand all collapsable sections.
+  #- navigation.instant     # Instant loading pages.
+  #- navigation.indexes     # Attach pages directly to Sections. Incompatible with "toc.integrate"
+  #- navigation.sections    # Render top sections as groups.
+  - navigation.tabs         # Render top sections as tabs at the top.
+  #- navigation.tabs.sticky # Tabs won't disappear when scrolling down. Requires "navigation.tabs".
+  #- navigation.top         # Adds a "Back to top" that is shown when scrolling up.
+  #- navigation.tracking    # Updates the url with highlighted section anchor.
+  
+  # Search
+  #
+  #- search.highlight # Search will highlight the searched word(s) on the page.*
+  #- search.share     # Adds an option to share a search query link.*
+  #- search.suggest   # Search will suggest the likeliest completion for a word.*
+  
+  # Table of Contents
+  #
+  #- toc.integrate          # Include the TOC sections in the left navugation.
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "icon" section allows you to define a icon to use for the logo and/or repository.                |
+  # |                                                                                                      |
+  # | To use already available icons will you need to set the right path for it, depending on which you    |
+  # | want to use.                                                                                         |
+  # |                                                                                                      |
+  # | Available icons:                                                                                     |
+  # | - FontAwesome                                                                                        |
+  # |   - Brands:  fontawesome/brands/...  (https://fontawesome.com/icons?d=gallery&p=2&s=brands&m=free)   |
+  # |   - Regular: fontawesome/regular/... (https://fontawesome.com/icons?d=gallery&p=2&s=regular&m=free)  |
+  # |   - Solid:   fontawesome/solid/...   (https://fontawesome.com/icons?d=gallery&p=2&s=solid&m=free)    |
+  # |                                                                                                      |
+  # | - Material Design Icons: material/... (https://materialdesignicons.com/)                             |
+  # |                                                                                                      |
+  # | - Octicons: octicons/... (https://primer.style/octicons/)                                            |
+  # |                                                                                                      |
+  # | You can also define your own Image for the logo. To do that, remove the "logo" option from "icon"    |
+  # | instead add a "logo" option on the same level as the "icon" one, where you then set the path         |
+  # | (relative to the "docs_dir") to the icon to use. Supported are all images types, including SVG.      |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-logo-and-icons/#logo                |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  icon:
+    logo: 'material/library'
+    repo: 'material/library'
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "admonition" option allows you to set a different icon for each admonition type.                 |
+  # |                                                                                                      |
+  # | This is currently a Insiders-only feature. (Last update: 7th October 2021)                           |
+  # | https://squidfunk.github.io/mkdocs-material/insiders/                                                |
+  # |                                                                                                      |
+  # | Supported are all bundled icons:                                                                     |
+  # | - FontAwesome                                                                                        |
+  # |   - Brands:  fontawesome/brands/...  (https://fontawesome.com/icons?d=gallery&p=2&s=brands&m=free)   |
+  # |   - Regular: fontawesome/regular/... (https://fontawesome.com/icons?d=gallery&p=2&s=regular&m=free)  |
+  # |   - Solid:   fontawesome/solid/...   (https://fontawesome.com/icons?d=gallery&p=2&s=solid&m=free)    |
+  # |                                                                                                      |
+  # | - Material Design Icons: material/... (https://materialdesignicons.com/)                             |
+  # |                                                                                                      |
+  # | - Octicons: octicons/... (https://primer.style/octicons/)                                            |
+  # |                                                                                                      |
+  # | You can also create and use your own icons. See the documentation for more information.              |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/reference/admonitions/#changing-the-icons              |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #admonition:
+  #    note: 'octicons/tag-16'
+  #    abstract: 'octicons/checklist-16'
+  #    info: 'octicons/info-16'
+  #    tip: 'octicons/squirrel-16'
+  #    success: 'octicons/check-16'
+  #    question: 'octicons/question-16'
+  #    warning: 'octicons/alert-16'
+  #    failure: 'octicons/x-circle-16'
+  #    danger: 'octicons/zap-16'
+  #    bug: 'octicons/bug-16'
+  #    example: 'octicons/beaker-16'
+  #    quote: 'octicons/quote-16'
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | With the "extra_css" option can you add your own (S)CSS files to enhance the documentation.            |
+# |                                                                                                        |
+# | The path to the file is relative to the "docs_dir".                                                    |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#extra_css                                           |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+extra_css:
+  - assets/css/extra.css
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | Similar to the "extra_css" option does the "extra_javascript" option allow you to set custom JS files  |
+# | to add extra featurues.                                                                                |
+# |                                                                                                        |
+# | The path to the file is relative to the "docs_dir".                                                    |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#extra_javascript                                    |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+extra_javascript:
+  - https://polyfill.io/v3/polyfill.min.js?features=es6
+  - https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | The "extra" section contains pretty much anything you want, as long as it is a valid key-value pair.   |
+# |                                                                                                        |
+# | Material uses this section for different custom settings that wouldn't fit in the theme section.       |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#extra                                               |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+extra:
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The social section allows you to set a list of entries which would be displayed in the footer of the |
+  # | page.                                                                                                |
+  # |                                                                                                      |
+  # | Each entry has the exact same options:                                                               |
+  # | - icon: Path to the SVG icon to use. See "icon" section for available icon sets.                     |
+  # | - link: URL to which the icon should link.                                                           |
+  # | - name: Optional Name that would be displayed as title on hover.                                     |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/setting-up-the-footer/#social-links              |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  social:
+    - icon: 'fontawesome/brands/github'
+      link: 'https://github.com/Andre601/mkdocs-template'
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | Allows you to hide the "Made with Material for MkDocs" text in the footer of the pages by setting    |
+  # | this to "true".                                                                                      |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/setting-up-the-footer/#generator-notice          |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #generator: true
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "manifest" option allows you to define a .manifest file to use.                                  |
+  # |                                                                                                      |
+  # | A .manifest file makes the doc act like a web-application and tells it how to behave when installed. |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/reference/meta-tags/#adding-a-web-app-manifest         |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #manifest: manifest.webmanifest
+  
+  # +------------------------------------------------------------------------------------------------------+
+  # |                                                                                                      |
+  # | The "alternate" option can be used to create a selector to switch languages.                         |
+  # |                                                                                                      |
+  # | Using this requires you to create a specific, more complicated MkDocs setup.                         |
+  # |                                                                                                      |
+  # | A Setup Guide for multi-language docs can be found here:                                             |
+  # | https://github.com/squidfunk/mkdocs-material/discussions/2346                                        |
+  # |                                                                                                      |
+  # | Read More:                                                                                           |
+  # | - https://squidfunk.github.io/mkdocs-material/setup/changing-the-language/#site-language-selector    |
+  # |                                                                                                      |
+  # +------------------------------------------------------------------------------------------------------+
+  #alternate:
+
+# +--------------------------------------------------------------------------------------------------------+
+# |                                                                                                        |
+# | MkDocs allows the usage of Markdown extensions which can do various things.                            |
+# |                                                                                                        |
+# | Material includes the pymdownx extension which provides a lot of useful features to use.               |
+# |                                                                                                        |
+# | Note that some extensions may use specific settings that you need to set.                              |
+# | Please check out the official documentation of PyMdownx for more information:                          |
+# | https://facelessuser.github.io/pymdown-extensions/                                                     |
+# |                                                                                                        |
+# | Material already provides required CSS and JS values for the PyMdownX Extensions, which means you do   |
+# | not need to set them up yourself.                                                                      |
+# |                                                                                                        |
+# | Read More:                                                                                             |
+# | - https://www.mkdocs.org/user-guide/configuration/#markdown_extensions                                 |
+# |                                                                                                        |
+# +--------------------------------------------------------------------------------------------------------+
+markdown_extensions:
+  - markdown.extensions.admonition:
+  - markdown.extensions.codehilite:
+      guess_lang: false
+  - markdown.extensions.toc:
+      permalink: true
+  - pymdownx.arithmatex:
+      generic: true
+  - attr_list
+  - md_in_html
+  - pymdownx.blocks.caption
+  - admonition
+  - pymdownx.highlight:
+      anchor_linenums: true
+  - pymdownx.superfences
+  - pymdownx.tabbed:
+      alternate_style: true
+  - pymdownx.details
+  - attr_list
+  - tables
+
+  #- pymdownx.b64:
+  #- pymdownx.betterem:
+  #- pymdownx.caret:
+  #- pymdownx.critic:
+  #- pymdownx.details:
+  #- pymdownx.emoji:
+  #- pymdownx.escapeall:
+  #- pymdownx.extra:
+  #- pymdownx.extrarawhtml:
+  #- pymdownx.highlight:
+  #- pymdownx.inlinehilite:
+  #- pymdownx.keys:
+  #- pymdownx.magiclink:
+  #- pymdownx.mark:
+  #- pymdownx.pathconverter:
+  #- pymdownx.progressbar:
+  #- pymdownx.smartsymbols:
+  #- pymdownx.snippets:
+  #- pymdownx.striphtml:
+  #- pymdownx.superfences:
+  #- pymdownx.tabbed:
+  #- pymdownx.tasklist:
+  #- pymdownx.tilde:
+  # - exporter:
+  #     formats:
+  #       pdf:
+  #         enabled: !ENV [MKDOCS_EXPORTER_PDF, true]
+  #         concurrency: 8
+  #         stylesheets:
+  #           - resources/stylesheets/pdf.scss
+  #         covers:
+  #           front: resources/templates/covers/front.html.j2
+  #           back: resources/templates/covers/back.html.j2
+  #         aggregator:
+  #           enabled: true
+  #           output: .well-known/site.pdf
+  #           covers: all
+
+# theme:
+#   name: material
+#   palette:
+#     - scheme: default
+#       primary: indigo
+#       accent: blue
+#       toggle:
+#         icon: material/brightness-7
+#         name: Switch to dark mode
+#     - scheme: slate
+#       primary: indigo
+#       accent: blue
+#       toggle:
+#         icon: material/brightness-4
+#         name: Switch to light mode
+#   features:
+#     - navigation.tabs
+#     - navigation.sections
+#     - navigation.top
+#     - content.code.copy
+#     - content.tabs.link
+
+# plugins:
+#   - search
+
+# markdown_extensions:
+
+

poetry.lock

@@ -0,0 +1,7 @@
+# This file is automatically @generated by Poetry 2.3.1 and should not be changed by hand.
+package = []
+
+[metadata]
+lock-version = "2.1"
+python-versions = ">=3.9"
+content-hash = "db9b4c08b159b17b239e26c67ead7c37b82d9f9eb06550245ae3134c095f98f7"

pyproject.toml

@@ -0,0 +1,15 @@
+[tool.poetry]
+name = "uLib"
+version = "0.6.0"
+description = "CMT Cosmic Muon Tomography project uLib python bindings"
+authors = ["Andrea Rigoni Garola <andrea.rigoni@pd.infn.it>"]
+readme = "README.md"
+packages = [{ include = "uLib", from = "src/Python" }]
+build = "build_python.py"
+
+[tool.poetry.dependencies]
+python = ">=3.9"
+
+[build-system]
+requires = ["poetry-core>=2.0.0", "pybind11>=2.6.0", "cmake>=3.12"]
+build-backend = "poetry.core.masonry.api"

src/Core/CMakeLists.txt

@@ -1,9 +1,38 @@
 
-set(HEADERS Archives.h Array.h Collection.h Debug.h Export.h Function.h Macros.h Mpl.h Object.h Options.h Serializable.h Signal.h Singleton.h SmartPointer.h StaticInterface.h StringReader.h Types.h Uuid.h Vector.h)
+set(HEADERS 
+    Archives.h 
+    Array.h 
+    Collection.h 
+    DataAllocator.h 
+    Debug.h 
+    Export.h 
+    Function.h 
+    Macros.h 
+    Mpl.h 
+    Object.h 
+    Options.h 
+    Serializable.h 
+    Signal.h 
+    Singleton.h 
+    SmartPointer.h 
+    StaticInterface.h 
+    StringReader.h 
+    Types.h 
+    Uuid.h 
+    Vector.h
+)
 
-set(SOURCES Archives.cpp Debug.cpp Object.cpp Options.cpp Serializable.cpp Signal.cpp Uuid.cpp)
+set(SOURCES 
+    Archives.cpp 
+    Debug.cpp 
+    Object.cpp 
+    Options.cpp 
+    Serializable.cpp 
+    Signal.cpp 
+    Uuid.cpp
+)
 
-set(LIBRARIES Boost::program_options)
+set(LIBRARIES Boost::program_options Boost::serialization)
 
 set(libname ${PACKAGE_LIBPREFIX}Core)
 set(ULIB_SHARED_LIBRARIES ${ULIB_SHARED_LIBRARIES} ${libname} PARENT_SCOPE)
@@ -13,10 +42,14 @@ add_library(${libname} SHARED ${SOURCES})
 set_target_properties(${libname} PROPERTIES
                       VERSION ${PROJECT_VERSION}
                       SOVERSION ${PROJECT_SOVERSION})
+if(USE_CUDA)
+  set(LIBRARIES ${LIBRARIES} CUDA::cudart)
+endif()
+
 target_link_libraries(${libname} ${LIBRARIES})
 
 install(TARGETS ${libname}
-        EXPORT "${PROJECT_NAME}Targets"
+        EXPORT "uLibTargets"
         RUNTIME DESTINATION ${INSTALL_BIN_DIR} COMPONENT bin
         LIBRARY DESTINATION ${INSTALL_LIB_DIR} COMPONENT lib)
 

src/Core/DataAllocator.h

@@ -0,0 +1,260 @@
+/*//////////////////////////////////////////////////////////////////////////////
+// CMT Cosmic Muon Tomography project //////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////
+
+  Copyright (c) 2014, Universita' degli Studi di Padova, INFN sez. di Padova
+  All rights reserved
+
+  Authors: Andrea Rigoni Garola < andrea.rigoni@pd.infn.it >
+
+  ------------------------------------------------------------------
+  This library is free software;  you  can  redistribute  it  and/or
+  modify it  under the  terms  of  the  GNU  Lesser  General  Public
+  License as published  by  the  Free  Software  Foundation;  either
+  version 3.0 of the License, or (at your option) any later version.
+
+  This library is  distributed in  the hope that it will  be useful,
+  but  WITHOUT ANY WARRANTY;  without  even  the implied warranty of
+  MERCHANTABILITY  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of  the GNU Lesser General  Public
+  License along with this library.
+
+//////////////////////////////////////////////////////////////////////////////*/
+
+#ifndef U_MATH_DATAALLOCATOR_H
+#define U_MATH_DATAALLOCATOR_H
+
+#include <algorithm>
+#include <cstring>
+#include <iostream>
+#include <stdexcept>
+#include <vector>
+
+#ifdef USE_CUDA
+#include <cuda_runtime.h>
+#include <thrust/device_vector.h>
+#endif
+
+namespace uLib {
+
+enum class MemoryDevice { RAM, VRAM };
+
+template <typename T> class DataAllocator {
+public:
+  DataAllocator(size_t size = 0, bool owns_objects = true)
+      : m_Size(size), m_RamData(nullptr), m_VramData(nullptr),
+        m_Device(MemoryDevice::RAM), m_OwnsObjects(owns_objects) {
+    if (m_Size > 0) {
+      if (m_OwnsObjects)
+        m_RamData = new T[m_Size]();
+      else
+        m_RamData = static_cast<T *>(::operator new(m_Size * sizeof(T)));
+    }
+  }
+
+  DataAllocator(const DataAllocator<T> &other)
+      : m_Size(other.m_Size), m_RamData(nullptr), m_VramData(nullptr),
+        m_Device(other.m_Device), m_OwnsObjects(other.m_OwnsObjects) {
+    if (m_Size > 0) {
+      if (other.m_RamData) {
+        if (m_OwnsObjects)
+          m_RamData = new T[m_Size];
+        else
+          m_RamData = static_cast<T *>(::operator new(m_Size * sizeof(T)));
+        std::memcpy(m_RamData, other.m_RamData, m_Size * sizeof(T));
+      }
+#ifdef USE_CUDA
+      if (other.m_VramData) {
+        cudaMalloc((void **)&m_VramData, m_Size * sizeof(T));
+        cudaMemcpy(m_VramData, other.m_VramData, m_Size * sizeof(T),
+                   cudaMemcpyDeviceToDevice);
+      }
+#endif
+    }
+  }
+
+  ~DataAllocator() {
+    if (m_RamData) {
+      if (m_OwnsObjects)
+        delete[] m_RamData;
+      else
+        ::operator delete(m_RamData);
+    }
+#ifdef USE_CUDA
+    if (m_VramData) {
+      cudaFree(m_VramData);
+    }
+#endif
+  }
+
+  DataAllocator &operator=(const DataAllocator &other) {
+    if (this != &other) {
+      m_OwnsObjects = other.m_OwnsObjects;
+      resize(other.m_Size);
+      m_Device = other.m_Device;
+      if (other.m_RamData) {
+        if (!m_RamData) {
+          if (m_OwnsObjects)
+            m_RamData = new T[m_Size];
+          else
+            m_RamData = static_cast<T *>(::operator new(m_Size * sizeof(T)));
+        }
+        std::memcpy(m_RamData, other.m_RamData, m_Size * sizeof(T));
+      }
+#ifdef USE_CUDA
+      if (other.m_VramData) {
+        if (!m_VramData)
+          cudaMalloc((void **)&m_VramData, m_Size * sizeof(T));
+        cudaMemcpy(m_VramData, other.m_VramData, m_Size * sizeof(T),
+                   cudaMemcpyDeviceToDevice);
+      }
+#endif
+    }
+    return *this;
+  }
+
+  void MoveToRAM() {
+    if (m_Device == MemoryDevice::RAM)
+      return;
+    if (!m_RamData && m_Size > 0) {
+      if (m_OwnsObjects)
+        m_RamData = new T[m_Size]();
+      else
+        m_RamData = static_cast<T *>(::operator new(m_Size * sizeof(T)));
+    }
+#ifdef USE_CUDA
+    if (m_VramData && m_Size > 0) {
+      cudaMemcpy(m_RamData, m_VramData, m_Size * sizeof(T),
+                 cudaMemcpyDeviceToHost);
+    }
+#endif
+    m_Device = MemoryDevice::RAM;
+  }
+
+  void MoveToVRAM() {
+    if (m_Device == MemoryDevice::VRAM)
+      return;
+#ifdef USE_CUDA
+    if (!m_VramData && m_Size > 0) {
+      cudaMalloc((void **)&m_VramData, m_Size * sizeof(T));
+    }
+    if (m_RamData && m_Size > 0) {
+      cudaMemcpy(m_VramData, m_RamData, m_Size * sizeof(T),
+                 cudaMemcpyHostToDevice);
+    }
+#endif
+    m_Device = MemoryDevice::VRAM;
+  }
+
+  void resize(size_t size) {
+    if (m_Size == size)
+      return;
+
+    T *newRam = nullptr;
+    T *newVram = nullptr;
+
+    if (size > 0) {
+      if (m_OwnsObjects)
+        newRam = new T[size]();
+      else
+        newRam = static_cast<T *>(::operator new(size * sizeof(T)));
+
+      if (m_RamData) {
+        std::memcpy(newRam, m_RamData, std::min(m_Size, size) * sizeof(T));
+      }
+
+#ifdef USE_CUDA
+      cudaMalloc((void **)&newVram, size * sizeof(T));
+      if (m_VramData) {
+        cudaMemcpy(newVram, m_VramData, std::min(m_Size, size) * sizeof(T),
+                   cudaMemcpyDeviceToDevice);
+      }
+#endif
+    }
+
+    if (m_RamData) {
+      if (m_OwnsObjects)
+        delete[] m_RamData;
+      else
+        ::operator delete(m_RamData);
+    }
+#ifdef USE_CUDA
+    if (m_VramData)
+      cudaFree(m_VramData);
+#endif
+
+    m_Size = size;
+    m_RamData = newRam;
+    m_VramData = newVram;
+  }
+
+  size_t size() const { return m_Size; }
+
+  T &at(size_t index) {
+    MoveToRAM();
+    if (index >= m_Size)
+      throw std::out_of_range("Index out of range");
+    return m_RamData[index];
+  }
+
+  const T &at(size_t index) const {
+    const_cast<DataAllocator *>(this)->MoveToRAM();
+    if (index >= m_Size)
+      throw std::out_of_range("Index out of range");
+    return m_RamData[index];
+  }
+
+  T &operator[](size_t index) {
+    MoveToRAM();
+    return m_RamData[index];
+  }
+
+  const T &operator[](size_t index) const {
+    const_cast<DataAllocator *>(this)->MoveToRAM();
+    return m_RamData[index];
+  }
+
+  T *data() { return (m_Device == MemoryDevice::RAM) ? m_RamData : m_VramData; }
+  const T *data() const {
+    return (m_Device == MemoryDevice::RAM) ? m_RamData : m_VramData;
+  }
+
+  T *GetRAMData() { return m_RamData; }
+  const T *GetRAMData() const { return m_RamData; }
+
+  T *GetVRAMData() { return m_VramData; }
+  const T *GetVRAMData() const { return m_VramData; }
+
+  MemoryDevice GetDevice() const { return m_Device; }
+
+  // Iterator support for RAM operations
+  T *begin() {
+    MoveToRAM();
+    return m_RamData;
+  }
+  T *end() {
+    MoveToRAM();
+    return m_RamData + m_Size;
+  }
+  const T *begin() const {
+    const_cast<DataAllocator *>(this)->MoveToRAM();
+    return m_RamData;
+  }
+  const T *end() const {
+    const_cast<DataAllocator *>(this)->MoveToRAM();
+    return m_RamData + m_Size;
+  }
+
+private:
+  size_t m_Size;
+  T *m_RamData;
+  T *m_VramData;
+  MemoryDevice m_Device;
+  bool m_OwnsObjects;
+};
+
+} // namespace uLib
+
+#endif // U_MATH_DATAALLOCATOR_H

src/Core/Object.h

@@ -116,16 +116,15 @@ public:
   connect(typename FunctionPointer<Func1>::Object *sender, Func1 sigf,
           typename FunctionPointer<Func2>::Object *receiver, Func2 slof) {
     SignalBase *sigb = sender->findOrAddSignal(sigf);
-    typedef boost::signals2::signal<
-        typename FunctionPointer<Func2>::SignalSignature>
-        SigT;
-    ConnectSignal(sigb, slof, receiver);
+    ConnectSignal<typename FunctionPointer<Func1>::SignalSignature>(sigb, slof,
+                                                                    receiver);
     return true;
   }
 
   template <typename FuncT>
   static inline bool connect(SignalBase *sigb, FuncT slof, Object *receiver) {
-    ConnectSignal(sigb, slof, receiver);
+    ConnectSignal<typename FunctionPointer<FuncT>::SignalSignature>(sigb, slof,
+                                                                    receiver);
     return true;
   }
 

src/Core/Signal.h

@@ -50,8 +50,11 @@ using namespace boost::placeholders;
 #define SIGNAL(a) BOOST_STRINGIZE(a)
 
 #define _ULIB_DETAIL_SIGNAL_EMIT(_name, ...)                                   \
-  static BOOST_AUTO(sig, this->findOrAddSignal(&_name));                       \
-  sig->operator()(__VA_ARGS__);
+  do {                                                                         \
+    BOOST_AUTO(sig, this->findOrAddSignal(&_name));                            \
+    if (sig)                                                                   \
+      sig->operator()(__VA_ARGS__);                                            \
+  } while (0)
 
 /**
  * Utility macro to implement signal emission implementa una delle seguenti:
@@ -84,66 +87,61 @@ template <typename T> struct Signal {
 
 namespace detail {
 
-template <typename FuncT, int arity> struct ConnectSignal {};
+template <typename FuncT, typename SigSignature, int arity>
+struct ConnectSignal {};
 
-template <typename FuncT> struct ConnectSignal<FuncT, 0> {
+template <typename FuncT, typename SigSignature>
+struct ConnectSignal<FuncT, SigSignature, 0> {
   static void connect(SignalBase *sigb, FuncT slof,
                       typename FunctionPointer<FuncT>::Object *receiver) {
-    typedef
-        typename Signal<typename FunctionPointer<FuncT>::SignalSignature>::type
-            SigT;
+    typedef typename Signal<SigSignature>::type SigT;
     reinterpret_cast<SigT *>(sigb)->connect(slof);
   }
 };
 
-template <typename FuncT> struct ConnectSignal<FuncT, 1> {
+template <typename FuncT, typename SigSignature>
+struct ConnectSignal<FuncT, SigSignature, 1> {
   static void connect(SignalBase *sigb, FuncT slof,
                       typename FunctionPointer<FuncT>::Object *receiver) {
-    typedef
-        typename Signal<typename FunctionPointer<FuncT>::SignalSignature>::type
-            SigT;
+    typedef typename Signal<SigSignature>::type SigT;
     reinterpret_cast<SigT *>(sigb)->connect(boost::bind(slof, receiver));
   }
 };
 
-template <typename FuncT> struct ConnectSignal<FuncT, 2> {
+template <typename FuncT, typename SigSignature>
+struct ConnectSignal<FuncT, SigSignature, 2> {
   static void connect(SignalBase *sigb, FuncT slof,
                       typename FunctionPointer<FuncT>::Object *receiver) {
-    typedef
-        typename Signal<typename FunctionPointer<FuncT>::SignalSignature>::type
-            SigT;
+    typedef typename Signal<SigSignature>::type SigT;
     reinterpret_cast<SigT *>(sigb)->connect(boost::bind(slof, receiver, _1));
   }
 };
 
-template <typename FuncT> struct ConnectSignal<FuncT, 3> {
+template <typename FuncT, typename SigSignature>
+struct ConnectSignal<FuncT, SigSignature, 3> {
   static void connect(SignalBase *sigb, FuncT slof,
                       typename FunctionPointer<FuncT>::Object *receiver) {
-    typedef
-        typename Signal<typename FunctionPointer<FuncT>::SignalSignature>::type
-            SigT;
+    typedef typename Signal<SigSignature>::type SigT;
     reinterpret_cast<SigT *>(sigb)->connect(
         boost::bind(slof, receiver, _1, _2));
   }
 };
 
-template <typename FuncT> struct ConnectSignal<FuncT, 4> {
+template <typename FuncT, typename SigSignature>
+struct ConnectSignal<FuncT, SigSignature, 4> {
   static void connect(SignalBase *sigb, FuncT slof,
                       typename FunctionPointer<FuncT>::Object *receiver) {
-    typedef
-        typename Signal<typename FunctionPointer<FuncT>::SignalSignature>::type
-            SigT;
+    typedef typename Signal<SigSignature>::type SigT;
     reinterpret_cast<SigT *>(sigb)->connect(
         boost::bind(slof, receiver, _1, _2, _3));
   }
 };
 
-template <typename FuncT> struct ConnectSignal<FuncT, 5> {
+template <typename FuncT, typename SigSignature>
+struct ConnectSignal<FuncT, SigSignature, 5> {
   static void connect(SignalBase *sigb, FuncT slof,
                       typename FunctionPointer<FuncT>::Object *receiver) {
-    typedef
-        typename Signal<typename FunctionPointer<FuncT>::SignalSignature>::type
-            SigT;
+    typedef typename Signal<SigSignature>::type SigT;
     reinterpret_cast<SigT *>(sigb)->connect(
         boost::bind(slof, receiver, _1, _2, _3, _4));
   }
@@ -152,15 +150,16 @@ template <typename FuncT> struct ConnectSignal<FuncT, 5> {
 } // namespace detail
 
 template <typename FuncT> SignalBase *NewSignal(FuncT f) {
-  // seems to work wow !
-  return new Signal<void()>::type;
+  return new
+      typename Signal<typename FunctionPointer<FuncT>::SignalSignature>::type;
 }
 
-template <typename FuncT>
+template <typename SigSignature, typename FuncT>
 void ConnectSignal(SignalBase *sigb, FuncT slof,
                    typename FunctionPointer<FuncT>::Object *receiver) {
-  detail::ConnectSignal<FuncT, FunctionPointer<FuncT>::arity>::connect(
-      sigb, slof, receiver);
+  detail::ConnectSignal<FuncT, SigSignature,
+                        FunctionPointer<FuncT>::arity>::connect(sigb, slof,
+                                                                receiver);
 }
 
 } // namespace uLib

src/Core/Uuid.h

@@ -23,69 +23,49 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef U_CORE_UUID_H
 #define U_CORE_UUID_H
 
 #include <iostream>
 #include <vector>
 
-#include <boost/uuid/uuid.hpp>
 #include <boost/uuid/name_generator.hpp>
 #include <boost/uuid/random_generator.hpp>
+#include <boost/uuid/uuid.hpp>
 #include <boost/uuid/uuid_io.hpp>
 
 #include "Core/Mpl.h"
 #include "Core/Object.h"
 
-
 namespace uLib {
 
-
-
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 // Object Registration //
 
-
 typedef boost::uuids::uuid uuid_t;
 
 extern uuid_t uLib_dns_uuid;
 
-template < typename T >
-class type_id : public boost::uuids::uuid {
+template <typename T> class type_id : public boost::uuids::uuid {
 public:
-    type_id() :
-        m_size(sizeof(T)),
-        uuid(boost::uuids::name_generator(uLib_dns_uuid)(typeid(T).name()))
-    {
-        std::cout << "Request for register new type\n" <<
-                     "name: " << typeid(T).name() << "\n" <<
-                     "uuid: " << to_string(*this) << "\n";
-    }
+  type_id()
+      : m_size(sizeof(T)),
+        uuid(boost::uuids::name_generator(uLib_dns_uuid)(typeid(T).name())) {
+    std::cout << "Request for register new type\n"
+              << "name: " << typeid(T).name() << "\n"
+              << "uuid: " << to_string(*this) << "\n";
+  }
 
-    explicit type_id(boost::uuids::uuid const& u)
-            : boost::uuids::uuid(u) {}
+  explicit type_id(boost::uuids::uuid const &u) : boost::uuids::uuid(u) {}
 
-    operator boost::uuids::uuid() {
-        return static_cast<boost::uuids::uuid&>(*this);
-    }
-
-    operator boost::uuids::uuid() const {
-        return static_cast<boost::uuids::uuid const&>(*this);
-    }
-
-    unsigned int size() const { return m_size; }
+  unsigned int size() const { return m_size; }
 
 private:
-    unsigned int m_size;
+  unsigned int m_size;
 };
 
-
-
-
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
@@ -94,70 +74,57 @@ private:
 namespace detail {
 
 class TypeRegister {
-    typedef boost::uuids::name_generator IDGen_t;
+  typedef boost::uuids::name_generator IDGen_t;
+
 public:
-    struct RegisterEntry {
-        uuid_t id;
-        int size;
-    };
+  struct RegisterEntry {
+    uuid_t id;
+    int size;
+  };
 
-    TypeRegister(uuid_t const &dns) :
-        gen(dns) {}
-
-
-    template< typename T >
-    RegisterEntry * AddType(T *t = NULL) {
-        RegisterEntry en = { gen(typeid(T).name()), sizeof(T) };
-        for(int i=0; i < m_registry.size(); ++i)
-            if(en.id == m_registry[i].id) return &(m_registry[i]);
-        m_registry.push_back(en);
-        return &m_registry.back();
-    }
-
-    void PrintSelf(std::ostream &o) {
-        std::cout << "RegisterController: \n";
-        for (int i=0; i<m_registry.size(); ++i)
-            o << "type [" << i << "]: "
-              << to_string(m_registry[i].id) << " "
-              << m_registry[i].size << "\n";
-        o << "\n";
-    }
+  TypeRegister(uuid_t const &dns) : gen(dns) {}
 
+  template <typename T> RegisterEntry *AddType(T *t = NULL) {
+    RegisterEntry en = {gen(typeid(T).name()), sizeof(T)};
+    for (int i = 0; i < m_registry.size(); ++i)
+      if (en.id == m_registry[i].id)
+        return &(m_registry[i]);
+    m_registry.push_back(en);
+    return &m_registry.back();
+  }
 
+  void PrintSelf(std::ostream &o) {
+    std::cout << "RegisterController: \n";
+    for (int i = 0; i < m_registry.size(); ++i)
+      o << "type [" << i << "]: " << to_string(m_registry[i].id) << " "
+        << m_registry[i].size << "\n";
+    o << "\n";
+  }
 
 private:
-    IDGen_t gen;
-    std::vector<RegisterEntry> m_registry;
-
+  IDGen_t gen;
+  std::vector<RegisterEntry> m_registry;
 };
 
-} // detail
-
-
-
+} // namespace detail
 
 class TypeRegister : public detail::TypeRegister {
 public:
-    typedef detail::TypeRegister                BaseClass;
-    typedef detail::TypeRegister::RegisterEntry Entry;
+  typedef detail::TypeRegister BaseClass;
+  typedef detail::TypeRegister::RegisterEntry Entry;
 
-    static TypeRegister* Controller();
+  static TypeRegister *Controller();
 
 private:
-    TypeRegister();                  // Blocks constructor
-    static TypeRegister *s_Instance; // Singleton instance
+  TypeRegister();                  // Blocks constructor
+  static TypeRegister *s_Instance; // Singleton instance
 };
 
-
-
-
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 // OBJECT REGISTER //
 
-
-
-} // uLib
+} // namespace uLib
 
 #endif // UUID_H

src/Core/Vector.h

@@ -23,156 +23,433 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef U_CORE_VECTOR_H
 #define U_CORE_VECTOR_H
 
-#include <vector>
 #include <iostream>
+#include <map>
+#include <mutex>
+#include <vector>
+
+#include <Core/DataAllocator.h>
 
-#include <Core/StaticInterface.h>
-#include <Core/SmartPointer.h>
 #include <Core/CommaInitializer.h>
+#include <Core/SmartPointer.h>
+#include <Core/StaticInterface.h>
 
+#ifdef USE_CUDA
+#include <thrust/device_ptr.h>
+#include <thrust/device_vector.h>
+#endif
 
 namespace uLib {
 
-// Vector Implemetation ... wraps std::vector
-template <typename T>
-class Vector : public std::vector<T, std::allocator<T> >
-{    
-    typedef std::vector< T,std::allocator<T> > BaseClass;
-    typedef std::allocator<T> Allocator;
-public:       
-    typedef T TypeData;
-    typedef __gnu_cxx::__normal_iterator<T*, BaseClass >       Iterator;
-    typedef __gnu_cxx::__normal_iterator<const T*, BaseClass>  ConstIterator;
-
-    typedef CommaInitializer< Vector<T> , T > VectorCommaInit;
-
-    Vector(unsigned int size) : BaseClass(size) {}
-    Vector(unsigned int size, T &value) : BaseClass(size,value) {}
-    Vector() : BaseClass(0) {}
-
-    inline VectorCommaInit operator <<(T scalar) {
-        return VectorCommaInit(this, scalar);
-    }
-
-    inline void PrintSelf(std::ostream &o);
-
-    void remove_element(unsigned int index) {
-        std::swap(this->at(index),this->back());
-        this->pop_back();
-    }
-
-    void remove_element(T &t) {
-      std::swap(t, this->back());
-      this->pop_back();
-    }   
-
-};
-
-template<typename T>
-void Vector<T>::PrintSelf(std::ostream &o)
-{
-    o << " *** uLib Vector *** \n";
-    o << " n. of items = " << this->size() << "\n";
-    for(int i=0; i< this->size(); ++i)
-        o << (T)this->at(i) << " ";
-    o << "\n";
-}
-
-template <typename T>
-std::ostream & operator << (std::ostream &o, const Vector<T> &v) {
-    for(int i=0; i< v.size(); ++i)
-        o << (T)v.at(i) << " ";
-    o << "\n";
-    return o;
-}
-
-template <typename T>
-std::ofstream & operator << (std::ofstream &o, const Vector<T> &v) {
-    for(int i=0; i< v.size(); ++i)
-        o << (T)v.at(i) << " ";
-    return o;
-}
-
-
-template < typename T >
-std::istream & operator >> (std::istream &is, Vector<T> &v) {
-    T value;
-    while( is >> value ) {
-        if(is.fail()) v.push_back(0);
-        else v.push_back( value );
-    }
-    return is;
-}
-
-
-
-// Smart pointer Vector Implementation //
-
-
-template <typename T>
-class SmartVector : public SmartPointer< Vector<T> > {
-    typedef SmartPointer< Vector<T> > Base;
+// MetaAllocator Implementation ...
+template <typename T> class MetaAllocator {
 public:
+  using value_type = T;
+  using pointer = T *;
+  using const_pointer = const T *;
+  using reference = T &;
+  using const_reference = const T &;
+  using size_type = std::size_t;
+  using difference_type = std::ptrdiff_t;
 
-  SmartVector() : Base(new Vector<T>()) { }
-  SmartVector( const SmartVector &copy) : Base(copy) { }
-  SmartVector(unsigned int size) : Base(new Vector<T>((int)size)) { }
+  template <class U> struct rebind {
+    using other = MetaAllocator<U>;
+  };
 
-  virtual ~SmartVector() {}
+  MetaAllocator() noexcept = default;
 
-  T& operator[](int p) {
-      return Base::get()->at(p);
+  template <class U>
+  constexpr MetaAllocator(const MetaAllocator<U> &) noexcept {}
+
+  T *allocate(std::size_t n) {
+    if (n == 0)
+      return nullptr;
+
+    DataAllocator<T> *da = new DataAllocator<T>(n, false);
+    T *ptr = da->GetRAMData();
+
+    std::lock_guard<std::mutex> lock(GetMutex());
+    GetAllocationMap()[ptr] = da;
+    return ptr;
   }
 
-  void swap_elements(unsigned int first, unsigned int second) {
-      std::swap(Base::get()->at(first),Base::get()->at(second));
+  void deallocate(T *p, std::size_t /*n*/) noexcept {
+    if (!p)
+      return;
+
+    std::lock_guard<std::mutex> lock(GetMutex());
+    auto &map = GetAllocationMap();
+    auto it = map.find(p);
+    if (it != map.end()) {
+      delete it->second;
+      map.erase(it);
+    }
+  }
+
+  static DataAllocator<T> *GetDataAllocator(T *p) {
+    if (!p)
+      return nullptr;
+    std::lock_guard<std::mutex> lock(GetMutex());
+    auto &map = GetAllocationMap();
+    auto it = map.find(p);
+    if (it != map.end()) {
+      return it->second;
+    }
+    return nullptr;
+  }
+
+private:
+  static std::map<T *, DataAllocator<T> *> &GetAllocationMap() {
+    static std::map<T *, DataAllocator<T> *> allocMap;
+    return allocMap;
+  }
+
+  static std::mutex &GetMutex() {
+    static std::mutex mtx;
+    return mtx;
+  }
+};
+
+template <class T, class U>
+bool operator==(const MetaAllocator<T> &, const MetaAllocator<U> &) {
+  return true;
+}
+template <class T, class U>
+bool operator!=(const MetaAllocator<T> &, const MetaAllocator<U> &) {
+  return false;
+}
+
+// Vector Implemetation ... wraps std::vector
+template <typename T> class Vector : public std::vector<T, MetaAllocator<T>> {
+  typedef std::vector<T, MetaAllocator<T>> BaseClass;
+  typedef MetaAllocator<T> Allocator;
+
+public:
+  typedef T TypeData;
+  typedef __gnu_cxx::__normal_iterator<T *, BaseClass> Iterator;
+  typedef __gnu_cxx::__normal_iterator<const T *, BaseClass> ConstIterator;
+
+  typedef CommaInitializer<Vector<T>, T> VectorCommaInit;
+  typedef typename BaseClass::iterator iterator;
+  typedef typename BaseClass::const_iterator const_iterator;
+  typedef typename BaseClass::size_type size_type;
+  typedef typename BaseClass::reference reference;
+
+  Vector(unsigned int size) : BaseClass(size) {}
+  Vector(unsigned int size, T &value) : BaseClass(size, value) {}
+  Vector() : BaseClass(0) {}
+  Vector(std::initializer_list<T> init) : BaseClass(init) {}
+
+  inline VectorCommaInit operator<<(T scalar) {
+    return VectorCommaInit(this, scalar);
+  }
+
+  void MoveToVRAM() {
+    if (auto alloc = MetaAllocator<T>::GetDataAllocator(BaseClass::data())) {
+      alloc->MoveToVRAM();
+    }
+  }
+
+  void MoveToRAM() {
+    if (auto alloc = MetaAllocator<T>::GetDataAllocator(BaseClass::data())) {
+      alloc->MoveToRAM();
+    }
+  }
+
+  T *GetVRAMData() {
+    if (auto alloc = MetaAllocator<T>::GetDataAllocator(BaseClass::data())) {
+      return alloc->GetVRAMData();
+    }
+    return nullptr;
+  }
+
+  const T *GetVRAMData() const {
+    if (auto alloc = MetaAllocator<T>::GetDataAllocator(
+            const_cast<T *>(BaseClass::data()))) {
+      return alloc->GetVRAMData();
+    }
+    return nullptr;
+  }
+
+#ifdef USE_CUDA
+  /// Returns a thrust::device_ptr to the VRAM data (valid after MoveToVRAM()).
+  /// thrust::device_ptr<T> is itself a random-access iterator compatible with
+  /// all thrust algorithms (thrust::transform, thrust::sort,
+  /// thrust::for_each…).
+  thrust::device_ptr<T> DeviceData() {
+    if (auto alloc = MetaAllocator<T>::GetDataAllocator(BaseClass::data())) {
+      return thrust::device_pointer_cast(alloc->GetVRAMData());
+    }
+    return thrust::device_ptr<T>(nullptr);
+  }
+
+  thrust::device_ptr<const T> DeviceData() const {
+    if (auto alloc = MetaAllocator<T>::GetDataAllocator(
+            const_cast<T *>(BaseClass::data()))) {
+      return thrust::device_pointer_cast(
+          static_cast<const T *>(alloc->GetVRAMData()));
+    }
+    return thrust::device_ptr<const T>(nullptr);
+  }
+
+  /// Device-side begin iterator (valid after MoveToVRAM()).
+  thrust::device_ptr<T> DeviceBegin() { return DeviceData(); }
+
+  /// Device-side end iterator (valid after MoveToVRAM()).
+  thrust::device_ptr<T> DeviceEnd() {
+    return DeviceData() + static_cast<std::ptrdiff_t>(BaseClass::size());
+  }
+
+  thrust::device_ptr<const T> DeviceBegin() const { return DeviceData(); }
+
+  thrust::device_ptr<const T> DeviceEnd() const {
+    return DeviceData() + static_cast<std::ptrdiff_t>(BaseClass::size());
+  }
+#endif // USE_CUDA
+
+  inline void PrintSelf(std::ostream &o);
+
+  // Overrides for auto-sync //
+  T &operator[](size_t i) {
+    this->MoveToRAM();
+    return BaseClass::operator[](i);
+  }
+  const T &operator[](size_t i) const {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::operator[](i);
+  }
+  T &at(size_t i) {
+    this->MoveToRAM();
+    return BaseClass::at(i);
+  }
+  const T &at(size_t i) const {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::at(i);
+  }
+  T &front() {
+    this->MoveToRAM();
+    return BaseClass::front();
+  }
+  const T &front() const {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::front();
+  }
+  T &back() {
+    this->MoveToRAM();
+    return BaseClass::back();
+  }
+  const T &back() const {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::back();
+  }
+  T *data() noexcept {
+    this->MoveToRAM();
+    return BaseClass::data();
+  }
+  const T *data() const noexcept {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::data();
+  }
+  Iterator begin() noexcept {
+    this->MoveToRAM();
+    return BaseClass::begin();
+  }
+  ConstIterator begin() const noexcept {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::begin();
+  }
+  Iterator end() noexcept {
+    this->MoveToRAM();
+    return BaseClass::end();
+  }
+  ConstIterator end() const noexcept {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::end();
+  }
+  auto rbegin() noexcept {
+    this->MoveToRAM();
+    return BaseClass::rbegin();
+  }
+  auto rbegin() const noexcept {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::rbegin();
+  }
+  auto rend() noexcept {
+    this->MoveToRAM();
+    return BaseClass::rend();
+  }
+  auto rend() const noexcept {
+    const_cast<Vector *>(this)->MoveToRAM();
+    return BaseClass::rend();
+  }
+
+  void push_back(const T &x) {
+    this->MoveToRAM();
+    BaseClass::push_back(x);
+  }
+  void push_back(T &&x) {
+    this->MoveToRAM();
+    BaseClass::push_back(std::move(x));
+  }
+  template <typename... Args> reference emplace_back(Args &&...args) {
+    this->MoveToRAM();
+    return BaseClass::emplace_back(std::forward<Args>(args)...);
+  }
+  void pop_back() {
+    this->MoveToRAM();
+    BaseClass::pop_back();
+  }
+
+  template <typename... Args>
+  iterator emplace(const_iterator pos, Args &&...args) {
+    this->MoveToRAM();
+    return BaseClass::emplace(pos, std::forward<Args>(args)...);
+  }
+  iterator insert(const_iterator pos, const T &x) {
+    this->MoveToRAM();
+    return BaseClass::insert(pos, x);
+  }
+  iterator insert(const_iterator pos, T &&x) {
+    this->MoveToRAM();
+    return BaseClass::insert(pos, std::move(x));
+  }
+  template <typename InputIt>
+  iterator insert(const_iterator pos, InputIt first, InputIt last) {
+    this->MoveToRAM();
+    return BaseClass::insert(pos, first, last);
+  }
+  iterator erase(const_iterator pos) {
+    this->MoveToRAM();
+    return BaseClass::erase(pos);
+  }
+  iterator erase(const_iterator first, const_iterator last) {
+    this->MoveToRAM();
+    return BaseClass::erase(first, last);
+  }
+
+  void resize(size_t n) {
+    this->MoveToRAM();
+    BaseClass::resize(n);
+  }
+  void resize(size_t n, const T &x) {
+    this->MoveToRAM();
+    BaseClass::resize(n, x);
+  }
+  void reserve(size_t n) {
+    this->MoveToRAM();
+    BaseClass::reserve(n);
+  }
+  void clear() noexcept {
+    this->MoveToRAM();
+    BaseClass::clear();
+  }
+  template <typename InputIt> void assign(InputIt first, InputIt last) {
+    this->MoveToRAM();
+    BaseClass::assign(first, last);
+  }
+  void assign(size_type count, const T &value) {
+    this->MoveToRAM();
+    BaseClass::assign(count, value);
   }
 
   void remove_element(unsigned int index) {
-      std::swap(Base::get()->at(index),Base::get()->back());
-      Base::get()->pop_back();
+    this->MoveToRAM();
+    std::swap(this->at(index), this->back());
+    this->pop_back();
+  }
+
+  void remove_element(T &t) {
+    this->MoveToRAM();
+    std::swap(t, this->back());
+    this->pop_back();
+  }
+};
+
+template <typename T> void Vector<T>::PrintSelf(std::ostream &o) {
+  o << " *** uLib Vector *** \n";
+  o << " n. of items = " << this->size() << "\n";
+  for (int i = 0; i < this->size(); ++i)
+    o << (T)this->at(i) << " ";
+  o << "\n";
+}
+
+template <typename T>
+std::ostream &operator<<(std::ostream &o, const Vector<T> &v) {
+  for (int i = 0; i < v.size(); ++i)
+    o << (T)v.at(i) << " ";
+  o << "\n";
+  return o;
+}
+
+template <typename T>
+std::ofstream &operator<<(std::ofstream &o, const Vector<T> &v) {
+  for (int i = 0; i < v.size(); ++i)
+    o << (T)v.at(i) << " ";
+  return o;
+}
+
+template <typename T> std::istream &operator>>(std::istream &is, Vector<T> &v) {
+  T value;
+  while (is >> value) {
+    if (is.fail())
+      v.push_back(0);
+    else
+      v.push_back(value);
+  }
+  return is;
+}
+
+// Smart pointer Vector Implementation //
+
+template <typename T> class SmartVector : public SmartPointer<Vector<T>> {
+  typedef SmartPointer<Vector<T>> Base;
+
+public:
+  SmartVector() : Base(new Vector<T>()) {}
+  SmartVector(const SmartVector &copy) : Base(copy) {}
+  SmartVector(unsigned int size) : Base(new Vector<T>((int)size)) {}
+
+  virtual ~SmartVector() {}
+
+  T &operator[](int p) { return Base::get()->at(p); }
+
+  void swap_elements(unsigned int first, unsigned int second) {
+    std::swap(Base::get()->at(first), Base::get()->at(second));
+  }
+
+  void remove_element(unsigned int index) {
+    std::swap(Base::get()->at(index), Base::get()->back());
+    Base::get()->pop_back();
   }
 
   void remove_element(T &t) {
     std::swap(t, Base::get()->back());
     Base::get()->pop_back();
-    }
+  }
 };
 
-
-
-
 // ------ Utils ------------------------------------------------------------- //
 
-
-
 // RIFARE con iteratore !
 template <typename _Tp, class _CmpT>
-inline const unsigned long
-VectorSplice(const _Tp &_it, const _Tp &_end, const float value, _CmpT _comp)
-{
-    _Tp it = _it;
-    _Tp end = _end-1;
-    for(it; it != end;)
-    {
-        if (_comp(*it,value))       ++it;
-        else if(_comp(*end,value))  std::swap(*it,*end--);
-        else                        --end;
-    }
-    return it - _it;
+inline unsigned long VectorSplice(const _Tp &_it, const _Tp &_end,
+                                  const float value, _CmpT _comp) {
+  _Tp it = _it;
+  _Tp end = _end - 1;
+  for (; it != end;) {
+    if (_comp(*it, value))
+      ++it;
+    else if (_comp(*end, value))
+      std::swap(*it, *end--);
+    else
+      --end;
+  }
+  return it - _it;
 }
 
-
-
-} // uLib
-
-
-
-
+} // namespace uLib
 
 #endif // VECTOR_H

src/Core/testing/CMakeLists.txt

@@ -8,7 +8,7 @@ set( TESTS
         ObjectCopyTest
         StaticInterfaceTest
         CommaInitTest
-        DebugTTreeDumpTest
+        # DebugTTreeDumpTest
         BoostTest
         BoostAccumulatorTest
         PropertiesTest
@@ -19,6 +19,8 @@ set( TESTS
         UuidTest
         TypeIntrospectionTraversal
         OptionsTest
+        PingPongTest
+        VectorMetaAllocatorTest
 )
 
 set(LIBRARIES
@@ -29,3 +31,8 @@ set(LIBRARIES
        ${ROOT_LIBRARIES}
 )
 uLib_add_tests(Core)
+
+if(USE_CUDA)
+    set_source_files_properties(VectorMetaAllocatorTest.cpp PROPERTIES LANGUAGE CUDA)
+endif()
+

src/Core/testing/PingPongTest.cpp

@@ -0,0 +1,52 @@
+#include "Core/Object.h"
+#include "Core/Signal.h"
+#include "testing-prototype.h"
+#include <iostream>
+
+using namespace uLib;
+
+class Ping : public Object {
+public:
+signals:
+  void PingSignal(int count);
+public slots:
+  void OnPong(int count) {
+    std::cout << "Ping received Pong " << count << std::endl;
+    if (count > 0)
+      ULIB_SIGNAL_EMIT(Ping::PingSignal, count - 1);
+  }
+};
+
+void Ping::PingSignal(int count) { ULIB_SIGNAL_EMIT(Ping::PingSignal, count); }
+
+class Pong : public Object {
+public:
+signals:
+  void PongSignal(int count);
+public slots:
+  void OnPing(int count) {
+    std::cout << "Pong received Ping " << count << std::endl;
+    if (count > 0)
+      ULIB_SIGNAL_EMIT(Pong::PongSignal, count - 1);
+  }
+};
+
+void Pong::PongSignal(int count) { ULIB_SIGNAL_EMIT(Pong::PongSignal, count); }
+
+int main() {
+  BEGIN_TESTING(PingPong);
+  Ping ping;
+  Pong pong;
+
+  std::cout << "Connecting ping to pong" << std::endl;
+  Object::connect(&ping, &Ping::PingSignal, &pong, &Pong::OnPing);
+
+  std::cout << "Connecting pong to ping" << std::endl;
+  Object::connect(&pong, &Pong::PongSignal, &ping, &Ping::OnPong);
+
+  std::cout << "Emitting PingSignal(5)" << std::endl;
+  ping.PingSignal(5);
+
+  END_TESTING;
+  return 0;
+}

src/Core/testing/SignalTest.cpp

@@ -23,93 +23,63 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #include <iostream>
 #include <typeinfo>
 
-
-#include "testing-prototype.h"
-#include "Core/Types.h"
 #include "Core/Object.h"
 #include "Core/Signal.h"
-
+#include "Core/Types.h"
+#include "testing-prototype.h"
 
 using namespace uLib;
 
-
-
-
 class Ob1 : public Object {
 public:
-
-
 signals:
-    void V0();
-
-    int V1(int a);
+  void V0();
 
+  void V1(int a);
 };
 
-
-
-
 // should be done by moc //
-void Ob1::V0() {
-    ULIB_SIGNAL_EMIT(Ob1::V0);
-}
-
-int Ob1::V1(int a) {
-    ULIB_SIGNAL_EMIT(Ob1::V1,a);
-}
-
-
-
+void Ob1::V0() { ULIB_SIGNAL_EMIT(Ob1::V0); }
 
+void Ob1::V1(int a) { ULIB_SIGNAL_EMIT(Ob1::V1, a); }
 
 class Ob2 : public Object {
 
 public slots:
-    void PrintV0() {
-        std::cout << "Ob2 prints V0\n" << std::flush;
-    }
+  void PrintV0() { std::cout << "Ob2 prints V0\n" << std::flush; }
 };
 
 class Ob3 : public Object {
 
 public slots:
-    void PrintV0() {
-        std::cout << "Ob3 prints V0\n" << std::flush;
-    }
+  void PrintV0() { std::cout << "Ob3 prints V0\n" << std::flush; }
 
-    void PrintNumber(int n) {
-        std::cout << "Ob3 is printing number: " << n << "\n";
-    }
+  void PrintNumber(int n) {
+    std::cout << "Ob3 is printing number: " << n << "\n";
+  }
 };
 
-
-
-
 int main() {
-    BEGIN_TESTING(Signals);
+  BEGIN_TESTING(Signals);
 
-    Ob1 ob1;
-    Ob2 ob2;
-    Ob3 ob3;
+  Ob1 ob1;
+  Ob2 ob2;
+  Ob3 ob3;
 
-    Object::connect(&ob1,&Ob1::V0,&ob2,&Ob2::PrintV0);
-    Object::connect(&ob1,&Ob1::V0,&ob3,&Ob3::PrintV0);
-    Object::connect(&ob1,&Ob1::V1,&ob3,&Ob3::PrintNumber);
+  Object::connect(&ob1, &Ob1::V0, &ob2, &Ob2::PrintV0);
+  Object::connect(&ob1, &Ob1::V0, &ob3, &Ob3::PrintV0);
+  Object::connect(&ob1, &Ob1::V1, &ob3, &Ob3::PrintNumber);
 
-    //  not working yet
-    //    Object::connect(&ob1,SIGNAL(V0(),&ob2,SLOT(PrintV0())
+  //  not working yet
+  //    Object::connect(&ob1,SIGNAL(V0(),&ob2,SLOT(PrintV0())
 
-    ob1.PrintSelf(std::cout);
+  ob1.PrintSelf(std::cout);
 
-    emit ob1.V0();
-    emit ob1.V1(5552368);
+  emit ob1.V0();
+  emit ob1.V1(5552368);
 
-    END_TESTING;
+  END_TESTING;
 }
-
-

src/Core/testing/VectorMetaAllocatorTest.cpp

@@ -0,0 +1,97 @@
+/*//////////////////////////////////////////////////////////////////////////////
+// CMT Cosmic Muon Tomography project //////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////
+
+  Copyright (c) 2014, Universita' degli Studi di Padova, INFN sez. di Padova
+  All rights reserved
+
+  Authors: Andrea Rigoni Garola < andrea.rigoni@pd.infn.it >
+
+//////////////////////////////////////////////////////////////////////////////*/
+
+#include "testing-prototype.h"
+#include <Core/Vector.h>
+
+#ifdef USE_CUDA
+#include <thrust/device_ptr.h>
+#include <thrust/transform.h>
+
+struct DoubleFunctor {
+  __host__ __device__ int operator()(int x) const { return x * 2; }
+};
+#endif
+
+int main() {
+  BEGIN_TESTING(VectorMetaAllocator);
+
+  uLib::Vector<int> v;
+  std::cout << "Pushing elements...\n";
+  v << 1, 2, 3, 4, 5;
+
+  std::cout << "Initial RAM contents: ";
+  for (size_t i = 0; i < v.size(); ++i) {
+    std::cout << v[i] << " ";
+    if (v[i] != (int)(i + 1)) {
+      std::cout << "\nError: Value mismatch at index " << i << "\n";
+      exit(1);
+    }
+  }
+  std::cout << "\n";
+
+#ifdef USE_CUDA
+  std::cout << "Moving to VRAM...\n";
+  v.MoveToVRAM();
+
+  int *vram_ptr = v.GetVRAMData();
+  if (vram_ptr) {
+    std::cout << "Successfully obtained VRAM pointer: " << vram_ptr << "\n";
+  } else {
+    std::cout << "Error: Failed to obtain VRAM pointer!\n";
+    exit(1);
+  }
+
+  // Verify DeviceData() matches GetVRAMData()
+  {
+    thrust::device_ptr<int> dev_ptr = v.DeviceData();
+    if (dev_ptr.get() != vram_ptr) {
+      std::cout << "Error: DeviceData() does not match GetVRAMData()!\n";
+      exit(1);
+    }
+    std::cout << "DeviceData() matches GetVRAMData(). OK\n";
+  }
+
+  // Use thrust::transform via DeviceBegin()/DeviceEnd() to double all elements
+  // on device
+  std::cout << "Doubling elements on device via thrust::transform...\n";
+  thrust::transform(v.DeviceBegin(), v.DeviceEnd(), v.DeviceBegin(),
+                    DoubleFunctor{});
+
+  std::cout << "Moving back to RAM...\n";
+  v.MoveToRAM();
+
+  std::cout << "RAM contents after VRAM trip + thrust transform: ";
+  for (size_t i = 0; i < v.size(); ++i) {
+    std::cout << v[i] << " ";
+    if (v[i] != (int)((i + 1) * 2)) {
+      std::cout << "\nError: Data corrupted after RAM->VRAM->thrust->RAM trip "
+                   "at index "
+                << i << "\n";
+      exit(1);
+    }
+  }
+  std::cout << "\n";
+#else
+  std::cout << "USE_CUDA not defined, skipping VRAM tests.\n";
+#endif
+
+  std::cout << "Scaling vector...\n";
+  for (size_t i = 0; i < v.size(); ++i)
+    v[i] *= 10;
+
+  std::cout << "Final contents: ";
+  for (size_t i = 0; i < v.size(); ++i)
+    std::cout << v[i] << " ";
+  std::cout << "\n";
+
+  END_TESTING;
+}

src/Core/testing/VectorTest.cpp

@@ -23,62 +23,47 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
-#include <Core/Vector.h>
 #include "testing-prototype.h"
+#include <Core/Vector.h>
 
 #include <algorithm>
 
-template < typename T >
-struct __Cmp {
-    bool operator()(const T &data, const float value) {
-        return data <= value;
-    }
+template <typename T> struct __Cmp {
+  bool operator()(const T &data, const float value) { return data <= value; }
 };
 
+template <typename _Tp, typename _CmpT>
+inline const unsigned long VectorSplice(const _Tp &_it, const _Tp &_end,
+                                        const float value, _CmpT _comp) {
 
-
-
-template<typename _Tp, typename _CmpT>
-inline const unsigned long
-VectorSplice(const _Tp &_it, const _Tp &_end, const float value, _CmpT _comp)
-  {
-
-    _Tp it = _it;
-    _Tp end = _end-1;
-    for(it; it != end; )
-    {
-        if ( _comp(*it, value) ) it++;
-        else if( _comp(*end, value) )
-        {
-            std::swap(*it,*end--);
-        }
-        else --end;
-    }
-    return it - _it;
+  _Tp it = _it;
+  _Tp end = _end - 1;
+  for (it; it != end;) {
+    if (_comp(*it, value))
+      it++;
+    else if (_comp(*end, value)) {
+      std::swap(*it, *end--);
+    } else
+      --end;
+  }
+  return it - _it;
 }
 
+int main() {
+  BEGIN_TESTING(Vector);
 
-int main()
-{
-    BEGIN_TESTING(Vector);
+  uLib::Vector<float> v;
+  v << 5, 4, 3, 2, 6, 1, 2, 3, 65, 7, 32, 23, 4, 3, 45, 4, 34, 3, 4, 4, 3, 3, 4,
+      2, 2, 3;
 
-    uLib::Vector<float> v;
-    v << 5,4,3,2,6,1,2,3,65,7,32,23,4,3,45,4,34,3,4,4,3,3,4,2,2,3;
+  int id = ::VectorSplice(v.begin(), v.end(), 3, __Cmp<float>());
 
+  std::cout << "id: " << id << "\n";
+  std::cout << "vector: ";
+  for (uLib::Vector<float>::Iterator it = v.begin(); it != v.end(); it++)
+    std::cout << *it << " ";
+  std::cout << std::endl;
+  //    std::sort(v.begin(),v.end(),LT<float>());
 
-
-    int id = VectorSplice(v.begin(),v.end(),3,__Cmp<float>());
-
-    std::cout << "id: " << id << "\n";
-    std::cout << "vector: ";
-    for(uLib::Vector<float>::Iterator it = v.begin(); it!=v.end(); it++)
-        std::cout << *it <<" ";
-    std::cout << std::endl;
-//    std::sort(v.begin(),v.end(),LT<float>());
-
-
-    END_TESTING;
+  END_TESTING;
 }
-

src/Detectors/CMakeLists.txt

@@ -1,12 +0,0 @@
-set(HEADERS MuonScatter.h MuonError.h MuonEvent.h)
-
-set(ULIB_SELECTED_MODULES ${ULIB_SELECTED_MODULES} Detectors PARENT_SCOPE)
-
-
-install(FILES ${HEADERS}
-        DESTINATION ${INSTALL_INC_DIR}/Detectors)
-
-if(BUILD_TESTING)
-  include(uLibTargetMacros)
-  add_subdirectory(testing)
-endif()

src/HEP/CMakeLists.txt

@@ -0,0 +1,12 @@
+
+################################################################################
+##### HEP - High Energy Physics modules ########################################
+################################################################################
+
+include_directories(${SRC_DIR}/HEP)
+
+add_subdirectory(Detectors)
+add_subdirectory(Geant)
+
+set(ULIB_SHARED_LIBRARIES ${ULIB_SHARED_LIBRARIES} PARENT_SCOPE)
+set(ULIB_SELECTED_MODULES ${ULIB_SELECTED_MODULES} PARENT_SCOPE)

src/HEP/Detectors/CMakeLists.txt

@@ -0,0 +1,34 @@
+set(HEADERS
+    ChamberHitEvent.h
+    DetectorChamber.h
+    ExperimentFitEvent.h
+    HierarchicalEncoding.h
+    Hit.h
+    HitMC.h
+    LinearFit.h
+    MuonError.h
+    MuonEvent.h
+    MuonScatter.h
+)
+
+set(libname ${PACKAGE_LIBPREFIX}Detectors)
+set(ULIB_SHARED_LIBRARIES ${ULIB_SHARED_LIBRARIES} ${libname} PARENT_SCOPE)
+set(ULIB_SELECTED_MODULES ${ULIB_SELECTED_MODULES} Detectors PARENT_SCOPE)
+
+## Headers-only INTERFACE library
+add_library(${libname} INTERFACE)
+target_include_directories(${libname} INTERFACE
+    $<BUILD_INTERFACE:${SRC_DIR}>
+    $<INSTALL_INTERFACE:${INSTALL_INC_DIR}>
+)
+
+install(TARGETS ${libname}
+        EXPORT "uLibTargets")
+
+install(FILES ${HEADERS}
+        DESTINATION ${INSTALL_INC_DIR}/HEP/Detectors)
+
+if(BUILD_TESTING)
+  include(uLibTargetMacros)
+  add_subdirectory(testing)
+endif()

src/HEP/Geant/CMakeLists.txt

@@ -0,0 +1,52 @@
+
+################################################################################
+##### HEP/Geant - Geant4 integration library ###################################
+################################################################################
+
+find_package(Geant4 QUIET)
+
+if(NOT Geant4_FOUND)
+    message(STATUS "Geant4 not found - skipping mutomGeant library")
+    return()
+endif()
+
+message(STATUS "Geant4 found: ${Geant4_VERSION}")
+include(${Geant4_USE_FILE})
+
+set(HEADERS
+    GeantEvent.h
+    Matter.h
+    Scene.h
+    Solid.h
+)
+
+set(SOURCES
+    Scene.cpp
+    Solid.cpp
+)
+
+set(libname ${PACKAGE_LIBPREFIX}Geant)
+set(ULIB_SHARED_LIBRARIES ${ULIB_SHARED_LIBRARIES} ${libname} PARENT_SCOPE)
+set(ULIB_SELECTED_MODULES ${ULIB_SELECTED_MODULES} Geant PARENT_SCOPE)
+
+add_library(${libname} SHARED ${SOURCES})
+set_target_properties(${libname} PROPERTIES
+                      VERSION ${PROJECT_VERSION}
+                      SOVERSION ${PROJECT_SOVERSION})
+
+target_include_directories(${libname} PRIVATE ${Geant4_INCLUDE_DIRS})
+
+target_link_libraries(${libname}
+    ${PACKAGE_LIBPREFIX}Core
+    ${PACKAGE_LIBPREFIX}Math
+    ${PACKAGE_LIBPREFIX}Detectors
+    ${Geant4_LIBRARIES}
+)
+
+install(TARGETS ${libname}
+        EXPORT "uLibTargets"
+        RUNTIME DESTINATION ${INSTALL_BIN_DIR} COMPONENT bin
+        LIBRARY DESTINATION ${INSTALL_LIB_DIR} COMPONENT lib)
+
+install(FILES ${HEADERS}
+        DESTINATION ${INSTALL_INC_DIR}/HEP/Geant)

src/HEP/Geant/GeantEvent.h

@@ -23,8 +23,6 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef U_GEANTEVENT_H
 #define U_GEANTEVENT_H
 
@@ -38,32 +36,25 @@ namespace uLib {
 
 class GeantEventData {
 public:
-    uLibGetMacro      (EventID, Id_t    )
-    uLibGetMacro      (Momentum,Scalarf )
-    uLibConstRefMacro (GenPos,  Vector3f)
-    uLibConstRefMacro (GenDir,  Vector3f)
-    uLibConstRefMacro (ChEvents,Vector<ChamberHitEventData>)
+uLibGetMacro(EventID, Id_t) uLibGetMacro(Momentum, Scalarf)
+    uLibConstRefMacro(GenPos, Vector3f) uLibConstRefMacro(GenDir, Vector3f)
+        uLibConstRefMacro(ChEvents, Vector<ChamberHitEventData>)
 
-private:
-        friend class GeantEvent;
-    Id_t     m_EventID;
-    Scalarf  m_Momentum;
-    Vector3f m_GenPos;
-    Vector3f m_GenDir;
-    Vector<ChamberHitEventData> m_ChEvents;
+            private : friend class GeantEvent;
+  Id_t m_EventID;
+  Scalarf m_Momentum;
+  Vector3f m_GenPos;
+  Vector3f m_GenDir;
+  Vector<ChamberHitEventData> m_ChEvents;
 };
 
 class GeantEvent {
 public:
-    uLibSetMacro (EventID, Id_t    )
-    uLibSetMacro (Momentum,Scalarf )
-    uLibRefMacro (GenPos,  Vector3f)
-    uLibRefMacro (GenDir,  Vector3f)
-    uLibRefMacro (ChEvents,Vector<ChamberHitEventData>)
+  uLibSetMacro(EventID, Id_t) uLibSetMacro(Momentum, Scalarf)
+      uLibRefMacro(GenPos, Vector3f) uLibRefMacro(GenDir, Vector3f)
+          uLibRefMacro(ChEvents, Vector<ChamberHitEventData>)
 };
 
-
-}
-
+} // namespace uLib
 
 #endif // GEANTEVENT_H

src/Math/CMakeLists.txt

@@ -43,12 +43,18 @@ set(ULIB_SELECTED_MODULES ${ULIB_SELECTED_MODULES} Math PARENT_SCOPE)
 add_library(${libname} SHARED ${SOURCES})
 set_target_properties(${libname} PROPERTIES
                       VERSION ${PROJECT_VERSION}
-                      SOVERSION ${PROJECT_SOVERSION})
+                      SOVERSION ${PROJECT_SOVERSION}
+                      CXX_STANDARD 17
+                      CUDA_STANDARD 17)
 target_link_libraries(${libname} ${LIBRARIES})
 
+if(USE_CUDA)
+    set_source_files_properties(VoxRaytracer.cpp VoxImage.cpp PROPERTIES LANGUAGE CUDA)
+endif()
+
 
 install(TARGETS ${libname}
-        EXPORT "${PROJECT_NAME}Targets"
+        EXPORT "uLibTargets"
         RUNTIME DESTINATION ${INSTALL_BIN_DIR} COMPONENT bin
         LIBRARY DESTINATION ${INSTALL_LIB_DIR} COMPONENT lib)
 

src/Math/ContainerBox.h

@@ -29,63 +29,147 @@
 #define U_CONTAINERBOX_H
 
 #include "Geometry.h"
+#include "Math/Dense.h"
+#include "Math/Transform.h"
+#include <utility>
 
 
 namespace uLib {
 
 
 
+/**
+ * @brief Represents an oriented bounding box (OBB) within a hierarchical transformation system.
+ *
+ * ContainerBox inherits from AffineTransform, which defines its parent coordinate system.
+ * It contains an internal local transformation (m_LocalT) that defines the box's
+ * specific origin and size relative to its own coordinate system.
+ */
 class ContainerBox : public AffineTransform {
-public:
-    ContainerBox() : m_LocalT(this) {}
 
+    typedef AffineTransform BaseClass;
+
+public:
+    /**
+     * @brief Default constructor.
+     * Initializes the local transformation with this instance as its parent.
+     */
+    ContainerBox() : 
+        m_LocalT(this)  // BaseClass is Parent of m_LocalTransform
+    {}
+
+    /**
+     * @brief Copy constructor.
+     * @param copy The ContainerBox instance to copy from.
+     */
     ContainerBox(const ContainerBox &copy) :
-        m_LocalT(this),
+        m_LocalT(this), // BaseClass is Parent of m_LocalTransform
         AffineTransform(copy)
     {
-        // FIX for performance //
         this->SetOrigin(copy.GetOrigin());
         this->SetSize(copy.GetSize());
     }
 
+    /**
+     * @brief Sets the box origin relative to its coordinate system.
+     * @param v The origin position vector.
+     */
     inline void SetOrigin(const Vector3f &v) { m_LocalT.SetPosition(v); }
 
+    /**
+     * @brief Gets the box origin relative to its coordinate system.
+     * @return The origin position vector.
+     */
     inline Vector3f GetOrigin() const { return m_LocalT.GetPosition(); }
 
+    /**
+     * @brief Sets the size of the box.
+     * Re-initializes the local transformation and applies the new scale.
+     * @param v The size vector (width, height, depth).
+     */
     void SetSize(const Vector3f &v) {
         Vector3f pos = this->GetOrigin();
-        m_LocalT = AffineTransform(this);
+        m_LocalT = AffineTransform(this); // regenerate local transform
         m_LocalT.Scale(v);
         m_LocalT.SetPosition(pos);
     }
 
+    /**
+     * @brief Gets the current size (scale) of the box.
+     * @return The size vector.
+     */
     inline Vector3f GetSize() const { return m_LocalT.GetScale(); }
 
-    // FIX... //
+    /**
+     * @brief Swaps two local axes of the box.
+     * @param first Index of the first axis (0=X, 1=Y, 2=Z).
+     * @param second Index of the second axis (0=X, 1=Y, 2=Z).
+     */
     inline void FlipLocalAxes(int first, int second)
     { m_LocalT.FlipAxes(first,second); }
 
+    /**
+     * @brief Returns the world transformation matrix of the box's volume.
+     * @return A 4x4 transformation matrix.
+     */
     Matrix4f GetWorldMatrix() const { return m_LocalT.GetWorldMatrix(); }
 
+    /**
+     * @brief Returns the local transformation matrix of the box's volume.
+     * @return A 4x4 transformation matrix.
+     */
+    Matrix4f GetLocalMatrix() const { return m_LocalT.GetMatrix(); }
+
+    /**
+     * @brief Transforms a point from box-local space to world space.
+     * @param v The local point (4D homogeneous vector).
+     * @return The transformed point in world space.
+     */
     inline Vector4f GetWorldPoint(const Vector4f &v) const {
         return m_LocalT.GetWorldMatrix() * v;
     }
 
+    /**
+     * @brief Transforms a point from box-local space coordinates to world space.
+     * @param x X coordinate in local space.
+     * @param y Y coordinate in local space.
+     * @param z Z coordinate in local space.
+     * @return The transformed point in world space.
+     */
     inline Vector4f GetWorldPoint(const float x, const float y, const float z) {
         return this->GetWorldPoint(Vector4f(x,y,z,1));
     }
 
+    /**
+     * @brief Transforms a point from world space to box-local space.
+     * @param v The world point (4D homogeneous vector).
+     * @return The transformed point in box-local space.
+     */
     inline Vector4f GetLocalPoint(const Vector4f &v) const {
         return m_LocalT.GetWorldMatrix().inverse() * v;
     }
 
+    /**
+     * @brief Transforms a point from world space coordinates to box-local space.
+     * @param x X coordinate in world space.
+     * @param y Y coordinate in world space.
+     * @param z Z coordinate in world space.
+     * @return The transformed point in box-local space.
+     */
     inline Vector4f GetLocalPoint(const float x, const float y, const float z) {
          return this->GetLocalPoint(Vector4f(x,y,z,1));
     }
 
+    /** Translate using transformation chain */
+    using BaseClass::Translate;
 
-protected:
+    /** Rotate using transformation chain */
+    using BaseClass::Rotate;
 
+    /** Scale using transformation chain */
+    using BaseClass::Scale;
+
+    
 private:
     AffineTransform  m_LocalT;
 };

src/Math/Dense.h

@@ -47,6 +47,7 @@
 #ifndef ULIB_DENSEMATRIX_H
 #define ULIB_DENSEMATRIX_H
 
+// #include <Eigen/src/Core/Matrix.h>
 #include <stdlib.h>
 
 #include <Eigen/Dense>
@@ -114,6 +115,21 @@ typedef unsigned long Scalarul;
 typedef float Scalarf;
 typedef double Scalard;
 
+typedef Eigen::Matrix<int, 1, 1> Vector1i;
+typedef Eigen::Vector2i Vector2i;
+typedef Eigen::Vector3i Vector3i;
+typedef Eigen::Vector4i Vector4i;
+
+typedef Eigen::Matrix<float, 1, 1> Vector1f;
+typedef Eigen::Vector2f Vector2f;
+typedef Eigen::Vector3f Vector3f;
+typedef Eigen::Vector4f Vector4f;
+
+typedef Eigen::Matrix<double, 1, 1> Vector1d;
+typedef Eigen::Vector2d Vector2d;
+typedef Eigen::Vector3d Vector3d;
+typedef Eigen::Vector4d Vector4d;
+
 typedef Eigen::Matrix<int, 1, 1> Matrix1i;
 typedef Eigen::Matrix2i Matrix2i;
 typedef Eigen::Matrix3i Matrix3i;
@@ -124,15 +140,15 @@ typedef Eigen::Matrix2f Matrix2f;
 typedef Eigen::Matrix3f Matrix3f;
 typedef Eigen::Matrix4f Matrix4f;
 
-typedef Eigen::Matrix<int, 1, 1> Vector1i;
-typedef Eigen::Vector2i Vector2i;
-typedef Eigen::Vector3i Vector3i;
-typedef Eigen::Vector4i Vector4i;
+typedef Eigen::Matrix<double, 1, 1> Matrix1d;
+typedef Eigen::Matrix2d Matrix2d;
+typedef Eigen::Matrix3d Matrix3d;
+typedef Eigen::Matrix4d Matrix4d;
+
+typedef Eigen::MatrixXi MatrixXi;
+typedef Eigen::MatrixXf MatrixXf;
+typedef Eigen::MatrixXd MatrixXd;
 
-typedef Eigen::Matrix<float, 1, 1> Vector1f;
-typedef Eigen::Vector2f Vector2f;
-typedef Eigen::Vector3f Vector3f;
-typedef Eigen::Vector4f Vector4f;
 
 ////////////////////////////////////////////////////////////////////////////////
 // Vector String interaction ///////////////////////////////////////////////////
@@ -175,7 +191,7 @@ std::string VectorxT_ToString(const Eigen::Matrix<T, size, 1> &vec) {
 // }
 
 template <typename T, int size>
-void operator>>(std::string &str, Eigen::Matrix<T, size, 1> &vec) {
+void operator >> (std::string &str, Eigen::Matrix<T, size, 1> &vec) {
   VectorxT_StringTo(vec, str);
 }
 
@@ -188,6 +204,9 @@ public:
   typedef Eigen::Matrix<Scalarf, 4, 1> BaseClass;
 
   _HPoint3f() : BaseClass(0, 0, 0, p) {}
+  _HPoint3f(int rows, int cols) : BaseClass() {
+    this->operator()(3) = p;
+  }
   _HPoint3f(float x, float y, float z) : BaseClass(x, y, z, p) {}
   _HPoint3f(Vector3f &in) : BaseClass(in.homogeneous()) {
     this->operator()(3) = p;

src/Math/Geometry.h

@@ -36,7 +36,7 @@ namespace uLib {
 class Geometry : public AffineTransform {
 public:
 
-    inline Vector4f GetWorldPoint(const Vector4f &v) const {
+    inline Vector4f GetWorldPoint(const Vector4f v) const {
         return this->GetWorldMatrix() * v;
     }
 
@@ -44,7 +44,7 @@ public:
         return this->GetWorldPoint(Vector4f(x,y,z,1));
     }
 
-    inline Vector4f GetLocalPoint(const Vector4f &v) const {
+    inline Vector4f GetLocalPoint(const Vector4f v) const {
         return this->GetWorldMatrix().inverse() * v;
     }
 

src/Math/Transform.h

@@ -84,8 +84,8 @@ public:
 
     inline void SetParent(AffineTransform *name) { this->m_Parent = name; }
 
-    inline void SetMatrix (Matrix4f &mat) { m_T.matrix() = mat; }
-    inline Matrix4f& GetMatrix () { return m_T.matrix(); }
+    inline void SetMatrix (Matrix4f mat) { m_T.matrix() = mat; }
+    inline Matrix4f GetMatrix() const { return m_T.matrix(); }
 
     Matrix4f GetWorldMatrix() const
     {
@@ -93,22 +93,22 @@ public:
         else return  m_Parent->GetWorldMatrix() * m_T.matrix(); // T = B * A //
     }
 
-    inline void SetPosition(const Vector3f &v) { this->m_T.translation() = v; }
+    inline void SetPosition(const Vector3f v) { this->m_T.translation() = v; }
 
     inline Vector3f GetPosition() const { return this->m_T.translation(); }
 
-    inline void SetRotation(const Matrix3f &m) { this->m_T.linear() = m; }
+    inline void SetRotation(const Matrix3f m) { this->m_T.linear() = m; }
 
     inline Matrix3f GetRotation() const { return this->m_T.rotation(); }
 
-    inline void Translate(const Vector3f &v) { this->m_T.translate(v); }
+    inline void Translate(const Vector3f v) { this->m_T.translate(v); }
 
-    inline void Scale(const Vector3f &v) { this->m_T.scale(v); }
+    inline void Scale(const Vector3f v) { this->m_T.scale(v); }
 
     inline Vector3f GetScale() const { return this->m_T.linear() * Vector3f(1,1,1); } // FIXXXXXXX
 
 
-    inline void Rotate(const Matrix3f &m) { this->m_T.rotate(m); }
+    inline void Rotate(const Matrix3f m) { this->m_T.rotate(m); }
 
     inline void Rotate(const float angle, Vector3f axis)
     {
@@ -122,12 +122,12 @@ public:
         Rotate(angle,euler_axis);
     }
 
-    inline void PreRotate(const Matrix3f &m) { this->m_T.prerotate(m); }
+    inline void PreRotate(const Matrix3f m) { this->m_T.prerotate(m); }
 
-    inline void QuaternionRotate(const Vector4f &q)
+    inline void QuaternionRotate(const Vector4f q)
     { this->m_T.rotate(Eigen::Quaternion<float>(q)); }
 
-    inline void EulerYZYRotate(const Vector3f &e) {
+    inline void EulerYZYRotate(const Vector3f e) {
         Matrix3f mat;
         mat =     Eigen::AngleAxisf(e.x(), Vector3f::UnitY())
                 * Eigen::AngleAxisf(e.y(), Vector3f::UnitZ())

src/Math/VoxImage.h

@@ -23,8 +23,6 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef U_MATH_VOXIMAGE_H
 #define U_MATH_VOXIMAGE_H
 
@@ -36,6 +34,8 @@
 #include <stdlib.h>
 #include <vector>
 
+#include <Core/DataAllocator.h>
+
 namespace uLib {
 
 ////////////////////////////////////////////////////////////////////////////////
@@ -46,36 +46,36 @@ namespace Abstract {
 
 class VoxImage : public uLib::StructuredGrid {
 public:
-    typedef uLib::StructuredGrid BaseClass;
+  typedef uLib::StructuredGrid BaseClass;
 
-    virtual float GetValue(const Vector3i &id) const = 0;
-    virtual float GetValue(const int id) const = 0;
-    virtual void SetValue(const Vector3i &id, float value) = 0;
-    virtual void SetValue(const int id, float value) = 0;
+  virtual float GetValue(const Vector3i &id) const = 0;
+  virtual float GetValue(const int id) const = 0;
+  virtual void SetValue(const Vector3i &id, float value) = 0;
+  virtual void SetValue(const int id, float value) = 0;
 
-    virtual void SetDims(const Vector3i &size) = 0;
+  virtual void SetDims(const Vector3i &size) = 0;
 
-    void ExportToVtk(const char *file, bool density_type = 0);
+  void ExportToVtk(const char *file, bool density_type = 0);
 
-    // use this function to export to VTK binary format
-    void ExportToVti (const char *file, bool density_type = 0, bool compressed = 0);
-    
-    // this function has been deprecated in favor of ExportToVti
-    // but it is kept for backward compatibility and because it 
-    // does not depend on vtk library
-    void ExportToVtkXml(const char *file, bool density_type = 0);
+  // use this function to export to VTK binary format
+  void ExportToVti(const char *file, bool density_type = 0,
+                   bool compressed = 0);
 
-    int ImportFromVtk(const char *file, bool density_type = 0);
+  // this function has been deprecated in favor of ExportToVti
+  // but it is kept for backward compatibility and because it
+  // does not depend on vtk library
+  void ExportToVtkXml(const char *file, bool density_type = 0);
 
-    int ImportFromVti(const char *file, bool density_type = 0);
+  int ImportFromVtk(const char *file, bool density_type = 0);
 
+  int ImportFromVti(const char *file, bool density_type = 0);
+
+  virtual ~VoxImage() {}
 protected:
-
-    virtual ~VoxImage() {}
-    VoxImage(const Vector3i &size) : BaseClass(size) {}
+  VoxImage(const Vector3i &size) : BaseClass(size) {}
 };
 
-}
+} // namespace Abstract
 
 ////////////////////////////////////////////////////////////////////////////////
 //  VOXEL   ////////////////////////////////////////////////////////////////////
@@ -83,421 +83,416 @@ protected:
 
 namespace Interface {
 struct Voxel {
-    template<class Self> void check_structural() {
-        uLibCheckMember(Self,Value, Scalarf);
-    }
+  template <class Self> void check_structural() {
+    uLibCheckMember(Self, Value, Scalarf);
+  }
 };
-}
+} // namespace Interface
 
 struct Voxel {
-    Scalarf      Value;
+  Scalarf Value = 0.0f;
+  Scalari Count = 0;
 };
 
-
 ////////////////////////////////////////////////////////////////////////////////
 //  VOX IMAGE  /////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 
-
-template< typename T >
-class VoxImage :  public Abstract::VoxImage {
+template <typename T> class VoxImage : public Abstract::VoxImage {
 public:
-    typedef Abstract::VoxImage BaseClass;
+  typedef Abstract::VoxImage BaseClass;
 
-    VoxImage();
+  VoxImage();
 
-    VoxImage(const Vector3i &size);
+  VoxImage(const Vector3i &size);
 
-    VoxImage(const VoxImage<T> &copy) :
-        BaseClass(copy)
-    {
-        this->m_Data = copy.m_Data;
+  VoxImage(const VoxImage<T> &copy) : BaseClass(copy) {
+    this->m_Data = copy.m_Data;
+  }
+
+  inline DataAllocator<T> &Data() { return this->m_Data; }
+  inline const DataAllocator<T> &ConstData() const { return m_Data; }
+
+  inline const T &At(int i) const { return m_Data.at(i); }
+  inline const T &At(const Vector3i &id) const { return m_Data.at(Map(id)); }
+  inline T &operator[](unsigned int i) { return m_Data[i]; }
+  inline T &operator[](const Vector3i &id) { return m_Data[Map(id)]; }
+
+  // this implements Abstract interface //
+  inline Scalarf GetValue(const Vector3i &id) const {
+    return this->At(id).Value;
+  }
+  inline Scalarf GetValue(const int id) const { return this->At(id).Value; }
+
+  inline void SetValue(const Vector3i &id, Scalarf value) {
+    this->operator[](id).Value = value;
+  }
+  inline void SetValue(const int id, float value) {
+    this->operator[](id).Value = value;
+  }
+
+  inline void SetDims(const Vector3i &size) {
+    this->m_Data.resize(size.prod());
+    StructuredGrid::SetDims(size);
+  }
+
+  inline VoxImage<T> clipImage(const Vector3i begin, const Vector3i end) const;
+  inline VoxImage<T> clipImage(const HPoint3f begin, const HPoint3f end) const;
+  inline VoxImage<T> clipImage(const float density) const;
+  inline VoxImage<T> clipImage(const float densityMin,
+                               const float densityMax) const;
+
+  inline VoxImage<T> maskImage(const HPoint3f begin, const HPoint3f end,
+                               float value) const;
+  inline VoxImage<T> maskImage(const float threshold, float belowValue = 0,
+                               float aboveValue = 0) const;
+  inline VoxImage<T> fixVoxels(const float threshold, float tolerance) const;
+  inline VoxImage<T> fixVoxels(const float threshold, float tolerance,
+                               const HPoint3f begin, const HPoint3f end) const;
+  inline VoxImage<T> fixVoxelsAroundPlane(const float threshold,
+                                          float tolerance, const HPoint3f begin,
+                                          const HPoint3f end,
+                                          bool aboveAir) const;
+  inline VoxImage<T> fixVoxels(const HPoint3f begin, const HPoint3f end) const;
+  inline VoxImage<T> Abs() const;
+
+  inline void SelectScalarfComponent(T &element, Scalarf *scalar);
+
+  inline void InitVoxels(T t);
+
+  // MATH OPERATORS //
+  inline void operator*=(Scalarf scalar) {
+    for (unsigned int i = 0; i < m_Data.size(); ++i)
+      m_Data[i].Value *= scalar;
+  }
+  inline void operator+=(Scalarf scalar) {
+    for (unsigned int i = 0; i < m_Data.size(); ++i)
+      m_Data[i].Value += scalar;
+  }
+  inline void operator/=(Scalarf scalar) {
+    for (unsigned int i = 0; i < m_Data.size(); ++i)
+      m_Data[i].Value /= scalar;
+  }
+  inline void operator-=(Scalarf scalar) {
+    for (unsigned int i = 0; i < m_Data.size(); ++i)
+      m_Data[i].Value -= scalar;
+  }
+
+  // MATH VoxImage Operators //
+  template <typename S> void operator+=(VoxImage<S> &sibling) {
+    if (this->GetDims() != sibling.GetDims()) {
+      // printf("Warning when adding VoxImages: I'm NOT doing it!\n");
+      return;
+    } // WARNING! You must Warn the user!
+    for (unsigned int i = 0; i < m_Data.size(); ++i) {
+      m_Data[i].Value += sibling.At(i).Value;
     }
+  }
 
-    inline std::vector<T> & Data() { return this->m_Data; }
-    inline const std::vector<T>& ConstData() const { return m_Data; }
-
-    inline const T& At(int i)              const { return m_Data.at(i); }
-    inline const T& At(const Vector3i &id) const { return m_Data.at(Map(id)); }
-    inline T& operator[](unsigned int i)     { return m_Data[i]; }
-    inline T& operator[](const Vector3i &id) { return m_Data[Map(id)]; }
-
-    // this implements Abstract interface //
-    inline Scalarf GetValue(const Vector3i &id) const {
-        return this->At(id).Value;
-    }
-    inline Scalarf GetValue(const int id) const {
-        return this->At(id).Value;
+  template <typename S> void operator-=(VoxImage<S> &sibling) {
+    if (this->GetDims() != sibling.GetDims()) {
+      // printf("Warning when subtracting VoxImages: I'm NOT doing it!\n");
+      return;
+    } // WARNING! You must Warn the user!
+    for (unsigned int i = 0; i < m_Data.size(); ++i) {
+      m_Data[i].Value -= sibling.At(i).Value;
     }
+  }
 
-    inline void SetValue(const Vector3i &id, Scalarf value) {
-        this->operator [](id).Value = value;
-    }
-    inline void SetValue(const int id, float value) {
-        this->operator [](id).Value = value;
+  template <typename S> void operator*=(VoxImage<S> &sibling) {
+    if (this->GetDims() != sibling.GetDims()) {
+      // printf("Warning when multiplying VoxImages: I'm NOT doing it!\n");
+      return;
+    } // WARNING! You must Warn the user!
+    for (unsigned int i = 0; i < m_Data.size(); ++i) {
+      m_Data[i].Value *= sibling.At(i).Value;
     }
+  }
 
-    inline void SetDims(const Vector3i &size) {
-        this->m_Data.resize(size.prod());
-        BaseClass::BaseClass::SetDims(size); // FIX horrible coding style !
-    }
-
-    inline VoxImage<T> clipImage(const Vector3i begin, const Vector3i end) const;
-    inline VoxImage<T> clipImage(const HPoint3f begin, const HPoint3f end) const;
-    inline VoxImage<T> clipImage(const float density) const;
-    inline VoxImage<T> clipImage(const float densityMin, const float densityMax) const;
-
-    inline VoxImage<T> maskImage(const HPoint3f begin, const HPoint3f end, float value) const;
-    inline VoxImage<T> maskImage(const float threshold, float belowValue=0, float aboveValue=0) const;
-    inline VoxImage<T> fixVoxels(const float threshold, float tolerance) const;
-    inline VoxImage<T> fixVoxels(const float threshold, float tolerance, const HPoint3f begin, const HPoint3f end) const;
-    inline VoxImage<T> fixVoxelsAroundPlane(const float threshold, float tolerance, const HPoint3f begin, const HPoint3f end, bool aboveAir) const;
-    inline VoxImage<T> fixVoxels(const HPoint3f begin, const HPoint3f end) const;
-    inline VoxImage<T> Abs() const;
-
-    inline void SelectScalarfComponent(T &element, Scalarf *scalar);
-
-    inline void InitVoxels(T t);
-
-    // MATH OPERATORS //
-    inline void operator *=(Scalarf scalar) {
-        for(unsigned int i = 0; i < m_Data.size(); ++i)
-            m_Data[i].Value *= scalar;
-    }
-    inline void operator +=(Scalarf scalar) {
-        for(unsigned int i = 0; i < m_Data.size(); ++i)
-            m_Data[i].Value += scalar;
-    }
-    inline void operator /=(Scalarf scalar) {
-        for(unsigned int i = 0; i < m_Data.size(); ++i)
-            m_Data[i].Value /= scalar;
-    }
-    inline void operator -=(Scalarf scalar) {
-        for(unsigned int i = 0; i < m_Data.size(); ++i)
-            m_Data[i].Value -= scalar;
-    }
-
-    // MATH VoxImage Operators //
-    template <typename S>
-    void operator +=(VoxImage<S> &sibling) {
-        if (this->GetDims() != sibling.GetDims()) {
-            //printf("Warning when adding VoxImages: I'm NOT doing it!\n");
-            return;
-        }// WARNING! You must Warn the user!
-        for(unsigned int i = 0; i < m_Data.size(); ++i) {
-            m_Data[i].Value += sibling.At(i).Value;
-        }
-    }
-
-    template <typename S>
-    void operator -=(VoxImage<S> &sibling) {
-        if (this->GetDims() != sibling.GetDims()) {
-            //printf("Warning when subtracting VoxImages: I'm NOT doing it!\n");
-            return;
-        }// WARNING! You must Warn the user!
-        for(unsigned int i = 0; i < m_Data.size(); ++i) {
-            m_Data[i].Value -= sibling.At(i).Value;
-        }
-    }
-
-    template <typename S>
-    void operator *=(VoxImage<S> &sibling) {
-        if (this->GetDims() != sibling.GetDims()) {
-            //printf("Warning when multiplying VoxImages: I'm NOT doing it!\n");
-            return;
-        }// WARNING! You must Warn the user!
-        for(unsigned int i = 0; i < m_Data.size(); ++i) {
-            m_Data[i].Value *= sibling.At(i).Value;
-        }
-    }
-
-    template <typename S>
-    void operator /=(VoxImage<S> &sibling) {
-        if (this->GetDims() != sibling.GetDims()) {
-            //printf("Warning when dividing VoxImages: I'm NOT doing it!\n");
-            return;
-        }// WARNING! You must Warn the user!
-        for(unsigned int i = 0; i < m_Data.size(); ++i) {
-            m_Data[i].Value /= sibling.At(i).Value;
-        }
+  template <typename S> void operator/=(VoxImage<S> &sibling) {
+    if (this->GetDims() != sibling.GetDims()) {
+      // printf("Warning when dividing VoxImages: I'm NOT doing it!\n");
+      return;
+    } // WARNING! You must Warn the user!
+    for (unsigned int i = 0; i < m_Data.size(); ++i) {
+      m_Data[i].Value /= sibling.At(i).Value;
     }
+  }
 
 private:
-    std::vector<T> m_Data;
+  DataAllocator<T> m_Data;
 };
 
-
-template<typename T>
-VoxImage<T>::VoxImage() :
-    m_Data(0),
-    BaseClass(Vector3i(0,0,0))
-{ Interface::IsA <T,Interface::Voxel>(); /* structural check for T */ }
-
-template<typename T>
-VoxImage<T>::VoxImage(const Vector3i &size) :
-    m_Data(size.prod()),
-    BaseClass(size)
-{ Interface::IsA <T,Interface::Voxel>(); /* structural check for T */ }
-
-
 template <typename T>
-VoxImage<T> VoxImage<T>::clipImage(const Vector3i begin, const Vector3i end) const
-{
-        Vector3i dim = (end-begin)+Vector3i(1,1,1);
-        VoxImage<T> out(*this);
-        out.SetDims(dim);
-        out.SetPosition(this->GetPosition() + this->GetSpacing().cwiseProduct(begin.cast<float>()) );
-
-        for(uint x = 0; x<dim(0); ++x )
-            for(uint y = 0; y<dim(1); ++y )
-                for(uint z = 0; z<dim(2); ++z )
-                {
-                    Vector3i id = Vector3i(x,y,z);
-                    out[id] = this->At(begin + id);
-                }
-        return out;
+VoxImage<T>::VoxImage() : m_Data(0), BaseClass(Vector3i(0, 0, 0)) {
+  Interface::IsA<T, Interface::Voxel>(); /* structural check for T */
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::clipImage(const HPoint3f begin, const HPoint3f end) const
-{
-    Vector3i v1 = this->Find(begin);
-    Vector3i v2 = this->Find(end);
-    return this->clipImage(v1,v2);
+VoxImage<T>::VoxImage(const Vector3i &size)
+    : m_Data(size.prod()), BaseClass(size) {
+  Interface::IsA<T, Interface::Voxel>(); /* structural check for T */
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::clipImage(const float density) const
-{
-    Vector3i v1 = this->GetDims();
-    Vector3i v2 = Vector3i(0,0,0);
-    for(uint i=0; i< this->m_Data.size(); ++i) {
-        if( this->GetValue(i) >= density ) {
-            Vector3i id = this->UnMap(i);
-            v1 = v1.array().min(id.array());
-            v2 = v2.array().max(id.array());
-        }
+VoxImage<T> VoxImage<T>::clipImage(const Vector3i begin,
+                                   const Vector3i end) const {
+  Vector3i dim = (end - begin) + Vector3i(1, 1, 1);
+  VoxImage<T> out(*this);
+  out.SetDims(dim);
+  out.SetPosition(this->GetPosition() +
+                  this->GetSpacing().cwiseProduct(begin.cast<float>()));
+
+  for (uint x = 0; x < dim(0); ++x)
+    for (uint y = 0; y < dim(1); ++y)
+      for (uint z = 0; z < dim(2); ++z) {
+        Vector3i id = Vector3i(x, y, z);
+        out[id] = this->At(begin + id);
+      }
+  return out;
+}
+
+template <typename T>
+VoxImage<T> VoxImage<T>::clipImage(const HPoint3f begin,
+                                   const HPoint3f end) const {
+  Vector3i v1 = this->Find(begin);
+  Vector3i v2 = this->Find(end);
+  return this->clipImage(v1, v2);
+}
+
+template <typename T>
+VoxImage<T> VoxImage<T>::clipImage(const float density) const {
+  Vector3i v1 = this->GetDims();
+  Vector3i v2 = Vector3i(0, 0, 0);
+  for (uint i = 0; i < this->m_Data.size(); ++i) {
+    if (this->GetValue(i) >= density) {
+      Vector3i id = this->UnMap(i);
+      v1 = v1.array().min(id.array());
+      v2 = v2.array().max(id.array());
     }
-    return this->clipImage(v1,v2);
+  }
+  return this->clipImage(v1, v2);
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::clipImage(const float densityMin, const float densityMax) const
-{
-    Vector3i v1 = this->GetDims();
-    Vector3i v2 = Vector3i(0,0,0);
-    for(uint i=0; i< this->m_Data.size(); ++i) {
-        if( this->GetValue(i) >= densityMin && this->GetValue(i) <= densityMax) {
-            Vector3i id = this->UnMap(i);
-            v1 = v1.array().min(id.array());
-            v2 = v2.array().max(id.array());
-        }
+VoxImage<T> VoxImage<T>::clipImage(const float densityMin,
+                                   const float densityMax) const {
+  Vector3i v1 = this->GetDims();
+  Vector3i v2 = Vector3i(0, 0, 0);
+  for (uint i = 0; i < this->m_Data.size(); ++i) {
+    if (this->GetValue(i) >= densityMin && this->GetValue(i) <= densityMax) {
+      Vector3i id = this->UnMap(i);
+      v1 = v1.array().min(id.array());
+      v2 = v2.array().max(id.array());
     }
-    return this->clipImage(v1,v2);
+  }
+  return this->clipImage(v1, v2);
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::maskImage(const HPoint3f begin, const HPoint3f end, float value) const
-{
-    VoxImage<T> out(*this);
-    out.SetDims(this->GetDims());
-    out.SetPosition(this->GetPosition());
+VoxImage<T> VoxImage<T>::maskImage(const HPoint3f begin, const HPoint3f end,
+                                   float value) const {
+  VoxImage<T> out(*this);
+  out.SetDims(this->GetDims());
+  out.SetPosition(this->GetPosition());
 
-    Vector3i voxB = this->Find(begin);
-    Vector3i voxE = this->Find(end);
+  Vector3i voxB = this->Find(begin);
+  Vector3i voxE = this->Find(end);
 
-    Vector3i ID;
+  Vector3i ID;
 
-    for(int ix=voxB(0); ix<voxE(0); ix++)
-        for(int iy=voxB(1); iy<voxE(1); iy++)
-            for(int iz=voxB(2); iz<voxE(2); iz++){
-                ID << ix,iy,iz;
-                out.SetValue(ID,value*1.E-6);
-            }
+  for (int ix = voxB(0); ix < voxE(0); ix++)
+    for (int iy = voxB(1); iy < voxE(1); iy++)
+      for (int iz = voxB(2); iz < voxE(2); iz++) {
+        ID << ix, iy, iz;
+        out.SetValue(ID, value * 1.E-6);
+      }
 
-    return out;
+  return out;
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::maskImage(const float threshold, float belowValue, float aboveValue) const
-{
-    std::cout << "VoxImage: maskImage, fixing voxels under threshold " << threshold;
-    if(belowValue)
-        std::cout << " at value " << belowValue;
-    else
-        std::cout << " at -value";
-    std::cout << ", voxels above threshold at value ";
-    if(aboveValue)
-        std::cout << aboveValue;
-    else
-        std::cout << "found";
+VoxImage<T> VoxImage<T>::maskImage(const float threshold, float belowValue,
+                                   float aboveValue) const {
+  std::cout << "VoxImage: maskImage, fixing voxels under threshold "
+            << threshold;
+  if (belowValue)
+    std::cout << " at value " << belowValue;
+  else
+    std::cout << " at -value";
+  std::cout << ", voxels above threshold at value ";
+  if (aboveValue)
+    std::cout << aboveValue;
+  else
+    std::cout << "found";
 
+  VoxImage<T> out(*this);
+  out.SetDims(this->GetDims());
+  out.SetPosition(this->GetPosition());
 
-    VoxImage<T> out(*this);
-    out.SetDims(this->GetDims());
-    out.SetPosition(this->GetPosition());
-
-    for(uint i=0; i< this->m_Data.size(); ++i) {
-        // skip negative voxels: they are already frozen
-        if( this->GetValue(i) >= 0 ){
-            // voxels under threshold
-            if( this->GetValue(i) <= threshold*1.E-6 ){
-                if(belowValue){
-                    //                std::cout << "vox " << i << ", " << this->GetValue(i);
-                    //                std::cout << " ----> set to " << -1.*belowValue*1.E-6 << std::endl;
-                    out.SetValue(i,-1.*belowValue*1.E-6);}
-                else
-                    out.SetValue(i,-1.*this->GetValue(i));
-            }
-            // voxels over threshold
-            else{
-                if(aboveValue)
-                    out.SetValue(i,aboveValue*1.E-6);
-                else
-                    out.SetValue(i,this->GetValue(i));
-            }
-        }
+  for (uint i = 0; i < this->m_Data.size(); ++i) {
+    // skip negative voxels: they are already frozen
+    if (this->GetValue(i) >= 0) {
+      // voxels under threshold
+      if (this->GetValue(i) <= threshold * 1.E-6) {
+        if (belowValue) {
+          //                std::cout << "vox " << i << ", " <<
+          //                this->GetValue(i); std::cout << " ----> set to " <<
+          //                -1.*belowValue*1.E-6 << std::endl;
+          out.SetValue(i, -1. * belowValue * 1.E-6);
+        } else
+          out.SetValue(i, -1. * this->GetValue(i));
+      }
+      // voxels over threshold
+      else {
+        if (aboveValue)
+          out.SetValue(i, aboveValue * 1.E-6);
+        else
+          out.SetValue(i, this->GetValue(i));
+      }
     }
-    return out;
+  }
+  return out;
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::fixVoxels(const float threshold, float tolerance) const
-{
-    std::cout << "VoxImage: fixing voxels with value " << threshold << std::endl;
+VoxImage<T> VoxImage<T>::fixVoxels(const float threshold,
+                                   float tolerance) const {
+  std::cout << "VoxImage: fixing voxels with value " << threshold << std::endl;
 
-    VoxImage<T> out(*this);
-    out.SetDims(this->GetDims());
-    out.SetPosition(this->GetPosition());
+  VoxImage<T> out(*this);
+  out.SetDims(this->GetDims());
+  out.SetPosition(this->GetPosition());
 
-    for(uint i=0; i< this->m_Data.size(); ++i) {
+  for (uint i = 0; i < this->m_Data.size(); ++i) {
+    // voxels around threshold
+    if (fabs(this->GetValue(i) - threshold * 1.E-6) < tolerance * 1.E-6) {
+      //            std::cout << "vox " << i << ", " << this->GetValue(i);
+      //            std::cout << " ----> set to " << -1.*this->GetValue(i) <<
+      //            std::endl;
+      out.SetValue(i, -1. * this->GetValue(i));
+    }
+  }
+  return out;
+}
+
+template <typename T> VoxImage<T> VoxImage<T>::Abs() const {
+  std::cout << "VoxImage: set abs voxels value " << std::endl;
+
+  VoxImage<T> out(*this);
+  out.SetDims(this->GetDims());
+  out.SetPosition(this->GetPosition());
+
+  for (uint i = 0; i < this->m_Data.size(); ++i)
+    out.SetValue(i, fabs(this->GetValue(i)));
+
+  return out;
+}
+
+template <typename T>
+VoxImage<T> VoxImage<T>::fixVoxels(const float threshold, float tolerance,
+                                   const HPoint3f begin,
+                                   const HPoint3f end) const {
+  VoxImage<T> out(*this);
+  out.SetDims(this->GetDims());
+  out.SetPosition(this->GetPosition());
+
+  Vector3i voxB = this->Find(begin);
+  Vector3i voxE = this->Find(end);
+
+  Vector3i ID;
+
+  for (int ix = voxB(0); ix < voxE(0); ix++)
+    for (int iy = voxB(1); iy < voxE(1); iy++)
+      for (int iz = voxB(2); iz < voxE(2); iz++) {
+        ID << ix, iy, iz;
         // voxels around threshold
-        if( fabs(this->GetValue(i) - threshold*1.E-6) < tolerance* 1.E-6 ){
-//            std::cout << "vox " << i << ", " << this->GetValue(i);
-//            std::cout << " ----> set to " << -1.*this->GetValue(i) << std::endl;
-                out.SetValue(i,-1.*this->GetValue(i));
+        if (fabs(this->GetValue(ID) - threshold * 1.E-6) < tolerance * 1.E-6) {
+          out.SetValue(ID, -1. * this->GetValue(ID));
         }
-    }
-    return out;
+      }
+
+  return out;
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::Abs() const
-{
-    std::cout << "VoxImage: set abs voxels value " << std::endl;
+VoxImage<T> VoxImage<T>::fixVoxels(const HPoint3f begin,
+                                   const HPoint3f end) const {
+  VoxImage<T> out(*this);
+  out.SetDims(this->GetDims());
+  out.SetPosition(this->GetPosition());
 
-    VoxImage<T> out(*this);
-    out.SetDims(this->GetDims());
-    out.SetPosition(this->GetPosition());
+  Vector3i voxB = this->Find(begin);
+  Vector3i voxE = this->Find(end);
 
-    for(uint i=0; i< this->m_Data.size(); ++i)
-        out.SetValue(i,fabs(this->GetValue(i)));
+  Vector3i ID;
 
-    return out;
+  for (int ix = voxB(0); ix < voxE(0); ix++)
+    for (int iy = voxB(1); iy < voxE(1); iy++)
+      for (int iz = voxB(2); iz < voxE(2); iz++) {
+        ID << ix, iy, iz;
+        // voxels around threshold
+        out.SetValue(ID, -1. * this->GetValue(ID));
+      }
+  return out;
 }
 
 template <typename T>
-VoxImage<T> VoxImage<T>::fixVoxels( const float threshold, float tolerance, const HPoint3f begin, const HPoint3f end) const
-{
-    VoxImage<T> out(*this);
-    out.SetDims(this->GetDims());
-    out.SetPosition(this->GetPosition());
+VoxImage<T> VoxImage<T>::fixVoxelsAroundPlane(const float threshold,
+                                              float tolerance, const HPoint3f B,
+                                              const HPoint3f E,
+                                              bool aboveAir) const {
+  VoxImage<T> out(*this);
+  Vector3i dim = this->GetDims();
+  out.SetDims(dim);
+  out.SetPosition(this->GetPosition());
 
-    Vector3i voxB = this->Find(begin);
-    Vector3i voxE = this->Find(end);
+  HPoint3f Bcoll = this->GetPosition().homogeneous();
 
-    Vector3i ID;
+  Vector3i ID;
+  for (int ix = 0; ix < dim(0); ix++)
+    for (int iy = 0; iy < dim(1); iy++)
+      for (int iz = 0; iz < dim(2); iz++) {
+        ID << ix, iy, iz;
 
-    for(int ix=voxB(0); ix<voxE(0); ix++)
-        for(int iy=voxB(1); iy<voxE(1); iy++)
-            for(int iz=voxB(2); iz<voxE(2); iz++){
-                ID << ix,iy,iz;
-                // voxels around threshold
-                if( fabs(this->GetValue(ID) - threshold*1.E-6) < tolerance*1.E-6 ){
-                    out.SetValue(ID,-1.*this->GetValue(ID));
-                }
-            }
+        // B, E voxel position
+        Vector3i iv(ix, iy, iz);
+        Vector3f v =
+            Vector3f(iv.cast<float>().cwiseProduct(this->GetSpacing()));
+        HPoint3f Bvox = Bcoll + HPoint3f(v);
+        HPoint3f Evox = Bvox + this->GetSpacing().homogeneous();
+        HPoint3f V = Bvox + 0.5 * (this->GetSpacing().homogeneous());
 
-    return out;
+        // if distance point (x0,y0) from line by points (x1,y1) and (x2,y2) is
+        // less than tolerance
+        float x1 = B[1];
+        float y1 = B[2];
+        float x2 = E[1];
+        float y2 = E[2];
+        float x0 = V[1];
+        float y0 = V[2];
+        float dist = fabs((x2 - x1) * (y1 - y0) - ((x1 - x0) * (y2 - y1))) /
+                     sqrt((x2 - x1) * (x2 - x1) + ((y2 - y1) * (y2 - y1)));
+        float distSign = (x2 - x1) * (y1 - y0) - ((x1 - x0) * (y2 - y1));
+
+        // set voxel air value
+        if (dist < tolerance) {
+          // std::cout << "voxel " << iv << ", line " << dist << ", tolerance "
+          // << tolerance << std::endl;
+          out.SetValue(ID, threshold * 1.E-6);
+        } else
+          out.SetValue(ID, this->GetValue(ID));
+
+        if ((distSign > 0 && aboveAir) || (distSign < 0 && !aboveAir))
+          out.SetValue(ID, threshold * 1.E-6);
+      }
+  return out;
 }
 
-template <typename T>
-VoxImage<T> VoxImage<T>::fixVoxels(const HPoint3f begin, const HPoint3f end) const
-{
-    VoxImage<T> out(*this);
-    out.SetDims(this->GetDims());
-    out.SetPosition(this->GetPosition());
-
-    Vector3i voxB = this->Find(begin);
-    Vector3i voxE = this->Find(end);
-
-    Vector3i ID;
-
-    for(int ix=voxB(0); ix<voxE(0); ix++)
-        for(int iy=voxB(1); iy<voxE(1); iy++)
-            for(int iz=voxB(2); iz<voxE(2); iz++){
-                ID << ix,iy,iz;
-                // voxels around threshold
-                out.SetValue(ID,-1.*this->GetValue(ID));
-            }
-    return out;
+template <typename T> void VoxImage<T>::InitVoxels(T t) {
+  std::fill(m_Data.begin(), m_Data.end(), t); // warning... stl function //
 }
 
-
-template <typename T>
-VoxImage<T> VoxImage<T>::fixVoxelsAroundPlane( const float threshold, float tolerance, const HPoint3f B, const HPoint3f E, bool aboveAir) const
-{
-    VoxImage<T> out(*this);
-    Vector3i dim = this->GetDims();
-    out.SetDims(dim);
-    out.SetPosition(this->GetPosition());
-
-    HPoint3f  Bcoll = this->GetPosition().homogeneous();
-
-    Vector3i ID;
-    for(int ix=0; ix<dim(0); ix++)
-        for(int iy=0; iy<dim(1); iy++)
-            for(int iz=0; iz<dim(2); iz++){
-                ID << ix,iy,iz;
-
-                // B, E voxel position
-                Vector3i iv(ix,iy,iz);
-                Vector3f v = Vector3f(iv.cast<float>().cwiseProduct(this->GetSpacing()));
-                HPoint3f Bvox = Bcoll + HPoint3f(v);
-                HPoint3f Evox = Bvox + this->GetSpacing().homogeneous();
-                HPoint3f V = Bvox + 0.5*(this->GetSpacing().homogeneous());
-
-                // if distance point (x0,y0) from line by points (x1,y1) and (x2,y2) is less than tolerance
-                float x1 = B[1];
-                float y1 = B[2];
-                float x2 = E[1];
-                float y2 = E[2];
-                float x0 = V[1];
-                float y0 = V[2];
-                float dist = fabs( (x2-x1)*(y1-y0) - ((x1-x0)*(y2-y1))) / sqrt( (x2-x1)*(x2-x1)+((y2-y1)*(y2-y1)));
-                float distSign = (x2-x1)*(y1-y0) - ((x1-x0)*(y2-y1));
-
-                // set voxel air value
-                if(dist < tolerance){
-                    //std::cout << "voxel " << iv << ", line " << dist << ", tolerance " << tolerance << std::endl;
-                    out.SetValue(ID,threshold*1.E-6);
-                }
-                else
-                    out.SetValue(ID,this->GetValue(ID));
-
-                if((distSign>0 && aboveAir) || (distSign<0 && !aboveAir)  )
-                    out.SetValue(ID,threshold*1.E-6);
-            }
-    return out;
-}
-
-
-template<typename T>
-void VoxImage<T>::InitVoxels(T t)
-{
-    std::fill( m_Data.begin(), m_Data.end(), t ); // warning... stl function //
-}
-
-}
+} // namespace uLib
 
 #endif // VOXIMAGE_H

src/Math/VoxImageFilter.h

@@ -23,8 +23,6 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTER_H
 #define VOXIMAGEFILTER_H
 
@@ -33,96 +31,83 @@
 
 #include "Math/VoxImage.h"
 
-
 namespace uLib {
 
-
 namespace Interface {
 struct VoxImageFilterShape {
-    template <class Self> void check_structural() {
-        uLibCheckFunction(Self,operator(),float,float);
-        uLibCheckFunction(Self,operator(),float,const Vector3f&);
-    }
+  template <class Self> void check_structural() {
+    uLibCheckFunction(Self, operator(), float, float);
+    uLibCheckFunction(Self, operator(), float, const Vector3f &);
+  }
 };
-}
-
-template < typename VoxelT > class Kernel;
+} // namespace Interface
 
+template <typename VoxelT> class Kernel;
 
 namespace Abstract {
 class VoxImageFilter {
 public:
-    virtual void Run() = 0;
+  virtual void Run() = 0;
 
-    virtual void SetImage(Abstract::VoxImage *image) = 0;
+  virtual void SetImage(Abstract::VoxImage *image) = 0;
 
 protected:
-    virtual ~VoxImageFilter() {}
+  virtual ~VoxImageFilter() {}
 };
-}
+} // namespace Abstract
 
-
-template < typename VoxelT, typename AlgorithmT >
-class VoxImageFilter : public Abstract::VoxImageFilter
-{
+template <typename VoxelT, typename AlgorithmT>
+class VoxImageFilter : public Abstract::VoxImageFilter {
 
 public:
-    VoxImageFilter(const Vector3i &size);
+  VoxImageFilter(const Vector3i &size);
 
-    void Run();
+  void Run();
 
-    void SetKernelNumericXZY(const std::vector<float> &numeric);
+  void SetKernelNumericXZY(const std::vector<float> &numeric);
 
-    void SetKernelSpherical(float (*shape)(float));
+  void SetKernelSpherical(float (*shape)(float));
 
-    template < class ShapeT >
-    void SetKernelSpherical( ShapeT shape );
+  template <class ShapeT> void SetKernelSpherical(ShapeT shape);
 
-    void SetKernelWeightFunction(float (*shape)(const Vector3f &));
+  void SetKernelWeightFunction(float (*shape)(const Vector3f &));
 
-    template < class ShapeT >
-    void SetKernelWeightFunction( ShapeT shape );
+  template <class ShapeT> void SetKernelWeightFunction(ShapeT shape);
 
-    inline Kernel<VoxelT> GetKernelData() const { return this->m_KernelData; }
+  inline const Kernel<VoxelT> &GetKernelData() const {
+    return this->m_KernelData;
+  }
+  inline Kernel<VoxelT> &GetKernelData() { return this->m_KernelData; }
 
-    inline VoxImage<VoxelT>* GetImage() const { return this->m_Image; }
+  inline VoxImage<VoxelT> *GetImage() const { return this->m_Image; }
 
-    void SetImage(Abstract::VoxImage *image);
+  void SetImage(Abstract::VoxImage *image);
 
 protected:
+  float Convolve(const VoxImage<VoxelT> &buffer, int index); // remove //
 
-    float Convolve(const VoxImage<VoxelT> &buffer, int index); // remove //
+  void SetKernelOffset();
 
-    void SetKernelOffset();
+  float Distance2(const Vector3i &v);
 
-    float Distance2(const Vector3i &v);
-
-    // protected members for algorithm access //
-    Kernel<VoxelT>    m_KernelData;
-    VoxImage<VoxelT> *m_Image;
+  // protected members for algorithm access //
+  Kernel<VoxelT> m_KernelData;
+  VoxImage<VoxelT> *m_Image;
 
 private:
-    AlgorithmT       *t_Algoritm;
-
+  AlgorithmT *t_Algoritm;
 };
 
-
-
-
-
-
-}
-
+} // namespace uLib
 
 #endif // VOXIMAGEFILTER_H
 
 #include "VoxImageFilter.hpp"
 
-#include "VoxImageFilterLinear.hpp"
-#include "VoxImageFilterThreshold.hpp"
-#include "VoxImageFilterMedian.hpp"
+#include "VoxImageFilter2ndStat.hpp"
 #include "VoxImageFilterABTrim.hpp"
 #include "VoxImageFilterBilateral.hpp"
-#include "VoxImageFilter2ndStat.hpp"
 #include "VoxImageFilterCustom.hpp"
-
+#include "VoxImageFilterLinear.hpp"
+#include "VoxImageFilterMedian.hpp"
+#include "VoxImageFilterThreshold.hpp"

src/Math/VoxImageFilter.hpp

@@ -23,280 +23,238 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTER_HPP
 #define VOXIMAGEFILTER_HPP
 
-#include <Math/Dense.h>
 #include "Math/StructuredData.h"
 #include "Math/VoxImage.h"
 #include "VoxImageFilter.h"
+#include <Math/Dense.h>
 
 namespace uLib {
 
-
 // KERNEL //////////////////////////////////////////////////////////////////////
 
-template < typename T >
-class Kernel : public StructuredData {
-    typedef StructuredData BaseClass;
+template <typename T> class Kernel : public StructuredData {
+  typedef StructuredData BaseClass;
+
 public:
-    Kernel(const Vector3i &size);
+  Kernel(const Vector3i &size);
 
-    inline T& operator[](const Vector3i &id) { return m_Data[Map(id)]; }
-    inline T& operator[](const int &id) { return m_Data[id]; }
-    inline int GetCenterData() const;
+  inline T &operator[](const Vector3i &id) { return m_Data[Map(id)]; }
+  inline T &operator[](const int &id) { return m_Data[id]; }
+  inline int GetCenterData() const;
 
-    inline std::vector<T> & Data() { return this->m_Data; }
+  inline DataAllocator<T> &Data() { return this->m_Data; }
 
-    inline const std::vector<T>& ConstData() const { return this->m_Data; }
+  inline const DataAllocator<T> &ConstData() const { return this->m_Data; }
 
-    void PrintSelf(std::ostream &o) const;
+  void PrintSelf(std::ostream &o) const;
 
 private:
-    std::vector<T> m_Data;
+  DataAllocator<T> m_Data;
 };
 
-template < typename T >
-Kernel<T>::Kernel(const Vector3i &size) :
-    BaseClass(size),
-    m_Data(size.prod())
-{
-    Interface::IsA<T,Interface::Voxel>();
+template <typename T>
+Kernel<T>::Kernel(const Vector3i &size) : BaseClass(size), m_Data(size.prod()) {
+  Interface::IsA<T, Interface::Voxel>();
 }
 
-template < typename T >
-inline int Kernel<T>::GetCenterData() const
-{
-    static int center = Map(this->GetDims() / 2);
-    return center;
+template <typename T> inline int Kernel<T>::GetCenterData() const {
+  static int center = Map(this->GetDims() / 2);
+  return center;
 }
 
-template < typename T >
-void Kernel<T>::PrintSelf(std::ostream &o) const
-{
-    o << " Filter Kernel Dump [XZ_Y]: \n";
-    Vector3i index;
-    o << "\n Value: \n\n"
-      << "------------------------------------------------- \n";
-    for (int y = 0 ; y < this->GetDims()(1); ++y ) {
-        o << "[y=" << y << "]\n";
-        for (int z = 0 ; z < this->GetDims()(2); ++z ) {
-            for (int x = 0 ; x < this->GetDims()(0); ++x ) {
-                index << x,y,z;
-                o << m_Data[Map(index)].Value << " ";
-            } o << "\n";
-        } o << " --------------------------------------------------- \n";
+template <typename T> void Kernel<T>::PrintSelf(std::ostream &o) const {
+  o << " Filter Kernel Dump [XZ_Y]: \n";
+  Vector3i index;
+  o << "\n Value: \n\n"
+    << "------------------------------------------------- \n";
+  for (int y = 0; y < this->GetDims()(1); ++y) {
+    o << "[y=" << y << "]\n";
+    for (int z = 0; z < this->GetDims()(2); ++z) {
+      for (int x = 0; x < this->GetDims()(0); ++x) {
+        index << x, y, z;
+        o << m_Data[Map(index)].Value << " ";
+      }
+      o << "\n";
     }
-    o << "\n Offset: \n"
-      << "------------------------------------------------- \n";
-    for (int y = 0 ; y < this->GetDims()(1); ++y ) {
-        o << "[y=" << y << "]\n";
-        for (int z = 0 ; z < this->GetDims()(2); ++z ) {
-            for (int x = 0 ; x < this->GetDims()(0); ++x ) {
-                index << x,y,z;
-                o << m_Data[Map(index)].Count << " ";
-            } o << "\n";
-        } o << " --------------------------------------------------- \n";
+    o << " --------------------------------------------------- \n";
+  }
+  o << "\n Offset: \n"
+    << "------------------------------------------------- \n";
+  for (int y = 0; y < this->GetDims()(1); ++y) {
+    o << "[y=" << y << "]\n";
+    for (int z = 0; z < this->GetDims()(2); ++z) {
+      for (int x = 0; x < this->GetDims()(0); ++x) {
+        index << x, y, z;
+        o << m_Data[Map(index)].Count << " ";
+      }
+      o << "\n";
     }
+    o << " --------------------------------------------------- \n";
+  }
 }
 
 ////////////////////////////////////////////////////////////////////////////////
 
-
-
-
-
-
-
-
-#define _TPL_  template < typename VoxelT , typename AlgorithmT >
-#define _TPLT_ VoxelT,AlgorithmT
-
-
+#define _TPL_ template <typename VoxelT, typename AlgorithmT>
+#define _TPLT_ VoxelT, AlgorithmT
 
 _TPL_
-VoxImageFilter<_TPLT_>::VoxImageFilter(const Vector3i &size) :
-    m_KernelData(size),
-    t_Algoritm(static_cast<AlgorithmT *>(this))
-{
+VoxImageFilter<_TPLT_>::VoxImageFilter(const Vector3i &size)
+    : m_KernelData(size), t_Algoritm(static_cast<AlgorithmT *>(this)) {}
 
+_TPL_
+void VoxImageFilter<_TPLT_>::Run() {
+  VoxImage<VoxelT> buffer = *m_Image;
+#pragma omp parallel for
+  for (int i = 0; i < m_Image->Data().size(); ++i)
+    m_Image->operator[](i).Value = this->t_Algoritm->Evaluate(buffer, i);
+#pragma omp barrier
 }
 
 _TPL_
-void VoxImageFilter<_TPLT_>::Run()
-{
-    VoxImage<VoxelT> buffer = *m_Image;
-    #pragma omp parallel for
-    for(int i=0 ; i < m_Image->Data().size() ; ++i)
-        m_Image->operator [](i).Value = this->t_Algoritm->Evaluate(buffer,i);
-    #pragma omp barrier
-}
-
-_TPL_
-void VoxImageFilter<_TPLT_>::SetKernelOffset()
-{
-    Vector3i id(0,0,0);
-    for( int z=0 ; z < m_KernelData.GetDims()(2); ++z ) {
-        for( int x=0 ; x < m_KernelData.GetDims()(0); ++x ) {
-            for( int y=0 ; y < m_KernelData.GetDims()(1); ++y ) {
-                id << x,y,z;
-                m_KernelData[id].Count = id.transpose() * m_Image->GetIncrements();
-            }
-        }
+void VoxImageFilter<_TPLT_>::SetKernelOffset() {
+  Vector3i id(0, 0, 0);
+  for (int z = 0; z < m_KernelData.GetDims()(2); ++z) {
+    for (int x = 0; x < m_KernelData.GetDims()(0); ++x) {
+      for (int y = 0; y < m_KernelData.GetDims()(1); ++y) {
+        id << x, y, z;
+        m_KernelData[id].Count = id.transpose() * m_Image->GetIncrements();
+      }
     }
+  }
 }
 
 _TPL_
-float VoxImageFilter<_TPLT_>::Distance2(const Vector3i &v)
-{
-    Vector3i tmp = v;
-    const Vector3i &dim = this->m_KernelData.GetDims();
-    Vector3i center = dim / 2;
-    tmp = tmp - center;
-    center = center.cwiseProduct(center);
-    tmp    = tmp.cwiseProduct(tmp);
-    return (float)(tmp.sum()) / (float)( center.sum() + 0.25 *
-                                         (3 - (dim(0) % 2) - (dim(1) % 2) - (dim(2) % 2)));
+float VoxImageFilter<_TPLT_>::Distance2(const Vector3i &v) {
+  Vector3i tmp = v;
+  const Vector3i &dim = this->m_KernelData.GetDims();
+  Vector3i center = dim / 2;
+  tmp = tmp - center;
+  center = center.cwiseProduct(center);
+  tmp = tmp.cwiseProduct(tmp);
+  return (float)(tmp.sum()) /
+         (float)(center.sum() +
+                 0.25 * (3 - (dim(0) % 2) - (dim(1) % 2) - (dim(2) % 2)));
 }
 
-
 _TPL_
-void VoxImageFilter<_TPLT_>::SetKernelNumericXZY(const std::vector<float> &numeric)
-{
-    // set data order //
-    StructuredData::Order order = m_KernelData.GetDataOrder();
-    //m_KernelData.SetDataOrder(StructuredData::XZY);
-    Vector3i id;
-    int index = 0;
-    for( int y=0 ; y < m_KernelData.GetDims()(1); ++y ) {
-        for( int z=0 ; z < m_KernelData.GetDims()(2); ++z ) {
-            for( int x=0 ; x < m_KernelData.GetDims()(0); ++x ) {
-                id << x,y,z;
-                m_KernelData[id].Value = numeric[index++];
-            }
-        }
+void VoxImageFilter<_TPLT_>::SetKernelNumericXZY(
+    const std::vector<float> &numeric) {
+  // set data order //
+  StructuredData::Order order = m_KernelData.GetDataOrder();
+  // m_KernelData.SetDataOrder(StructuredData::XZY);
+  Vector3i id;
+  int index = 0;
+  for (int y = 0; y < m_KernelData.GetDims()(1); ++y) {
+    for (int z = 0; z < m_KernelData.GetDims()(2); ++z) {
+      for (int x = 0; x < m_KernelData.GetDims()(0); ++x) {
+        id << x, y, z;
+        m_KernelData[id].Value = numeric[index++];
+      }
     }
-    //m_KernelData.SetDataOrder(order);
+  }
+  // m_KernelData.SetDataOrder(order);
 }
 
 _TPL_
-void VoxImageFilter<_TPLT_>::SetKernelSpherical(float(* shape)(float))
-{
-    Vector3i id;
-    for( int y=0 ; y < m_KernelData.GetDims()(1); ++y ) {
-        for( int z=0 ; z < m_KernelData.GetDims()(2); ++z ) {
-            for( int x=0 ; x < m_KernelData.GetDims()(0); ++x ) {
-                id << x,y,z;
-                m_KernelData[id].Value = shape(this->Distance2(id));
-            }
-        }
+void VoxImageFilter<_TPLT_>::SetKernelSpherical(float (*shape)(float)) {
+  Vector3i id;
+  for (int y = 0; y < m_KernelData.GetDims()(1); ++y) {
+    for (int z = 0; z < m_KernelData.GetDims()(2); ++z) {
+      for (int x = 0; x < m_KernelData.GetDims()(0); ++x) {
+        id << x, y, z;
+        m_KernelData[id].Value = shape(this->Distance2(id));
+      }
     }
+  }
 }
 
-
 _TPL_ template <class ShapeT>
-void VoxImageFilter<_TPLT_>::SetKernelSpherical(ShapeT shape)
-{
-    Interface::IsA<ShapeT,Interface::VoxImageFilterShape>();
-    Vector3i id;
-    for( int y=0 ; y < m_KernelData.GetDims()(1); ++y ) {
-        for( int z=0 ; z < m_KernelData.GetDims()(2); ++z ) {
-            for( int x=0 ; x < m_KernelData.GetDims()(0); ++x ) {
-                id << x,y,z;
-                m_KernelData[id].Value = shape(this->Distance2(id));
-            }
-        }
+void VoxImageFilter<_TPLT_>::SetKernelSpherical(ShapeT shape) {
+  Interface::IsA<ShapeT, Interface::VoxImageFilterShape>();
+  Vector3i id;
+  for (int y = 0; y < m_KernelData.GetDims()(1); ++y) {
+    for (int z = 0; z < m_KernelData.GetDims()(2); ++z) {
+      for (int x = 0; x < m_KernelData.GetDims()(0); ++x) {
+        id << x, y, z;
+        m_KernelData[id].Value = shape(this->Distance2(id));
+      }
     }
+  }
 }
 
 _TPL_
-void VoxImageFilter<_TPLT_>::SetKernelWeightFunction(float (*shape)(const Vector3f &))
-{
-    const Vector3i &dim = m_KernelData.GetDims();
-    Vector3i id;
-    Vector3f pt;
-    for( int y=0 ; y < dim(1); ++y ) {
-        for( int z=0 ; z < dim(2); ++z ) {
-            for( int x=0 ; x < dim(0); ++x ) {
-                // get voxels centroid coords from kernel center //
-                id << x,y,z;
-                pt << id(0) - dim(0)/2 + 0.5 * !(dim(0) % 2),
-                      id(1) - dim(1)/2 + 0.5 * !(dim(1) % 2),
-                      id(2) - dim(2)/2 + 0.5 * !(dim(2) % 2);
-                // compute function using given shape //
-                m_KernelData[id].Value = shape(pt);
-            }
-        }
+void VoxImageFilter<_TPLT_>::SetKernelWeightFunction(
+    float (*shape)(const Vector3f &)) {
+  const Vector3i &dim = m_KernelData.GetDims();
+  Vector3i id;
+  Vector3f pt;
+  for (int y = 0; y < dim(1); ++y) {
+    for (int z = 0; z < dim(2); ++z) {
+      for (int x = 0; x < dim(0); ++x) {
+        // get voxels centroid coords from kernel center //
+        id << x, y, z;
+        pt << id(0) - dim(0) / 2 + 0.5 * !(dim(0) % 2),
+            id(1) - dim(1) / 2 + 0.5 * !(dim(1) % 2),
+            id(2) - dim(2) / 2 + 0.5 * !(dim(2) % 2);
+        // compute function using given shape //
+        m_KernelData[id].Value = shape(pt);
+      }
     }
+  }
 }
 
-_TPL_ template < class ShapeT >
-void VoxImageFilter<_TPLT_>::SetKernelWeightFunction(ShapeT shape)
-{
-    Interface::IsA<ShapeT,Interface::VoxImageFilterShape>();
-    const Vector3i &dim = m_KernelData.GetDims();
-    Vector3i id;
-    Vector3f pt;
-    for( int y=0 ; y < dim(1); ++y ) {
-        for( int z=0 ; z < dim(2); ++z ) {
-            for( int x=0 ; x < dim(0); ++x ) {
-                // get voxels centroid coords from kernel center //
-                id << x,y,z;
-                pt << id(0) - dim(0)/2 + 0.5 * !(dim(0) % 2),
-                      id(1) - dim(1)/2 + 0.5 * !(dim(1) % 2),
-                      id(2) - dim(2)/2 + 0.5 * !(dim(2) % 2);
-                // compute function using given shape //
-                m_KernelData[id].Value = shape(pt);
-            }
-        }
+_TPL_ template <class ShapeT>
+void VoxImageFilter<_TPLT_>::SetKernelWeightFunction(ShapeT shape) {
+  Interface::IsA<ShapeT, Interface::VoxImageFilterShape>();
+  const Vector3i &dim = m_KernelData.GetDims();
+  Vector3i id;
+  Vector3f pt;
+  for (int y = 0; y < dim(1); ++y) {
+    for (int z = 0; z < dim(2); ++z) {
+      for (int x = 0; x < dim(0); ++x) {
+        // get voxels centroid coords from kernel center //
+        id << x, y, z;
+        pt << id(0) - dim(0) / 2 + 0.5 * !(dim(0) % 2),
+            id(1) - dim(1) / 2 + 0.5 * !(dim(1) % 2),
+            id(2) - dim(2) / 2 + 0.5 * !(dim(2) % 2);
+        // compute function using given shape //
+        m_KernelData[id].Value = shape(pt);
+      }
     }
+  }
 }
 
-
-
-
 _TPL_
-void VoxImageFilter<_TPLT_>::SetImage(Abstract::VoxImage *image)
-{
-    this->m_Image = reinterpret_cast<VoxImage<VoxelT> *> (image);
-    this->SetKernelOffset();
+void VoxImageFilter<_TPLT_>::SetImage(Abstract::VoxImage *image) {
+  this->m_Image = reinterpret_cast<VoxImage<VoxelT> *>(image);
+  this->SetKernelOffset();
 }
 
-
 _TPL_
-float VoxImageFilter<_TPLT_>::Convolve(const VoxImage<VoxelT> &buffer, int index)
-{
-    const std::vector<VoxelT> &vbuf = buffer.ConstData();
-    const std::vector<VoxelT> &vker = m_KernelData.ConstData();
-    int vox_size = vbuf.size();
-    int ker_size = vker.size();
-    int pos;
-    float conv = 0, ksum = 0;
-    for (int ik = 0; ik < ker_size; ++ik) {
-        pos = index + vker[ik].Count - vker[m_KernelData.GetCenterData()].Count;
-        pos = (pos + vox_size) % vox_size;
-        conv += vbuf[pos].Value * vker[ik].Value;
-        ksum += vker[ik].Value;
-    }
-    return conv / ksum;
+float VoxImageFilter<_TPLT_>::Convolve(const VoxImage<VoxelT> &buffer,
+                                       int index) {
+  const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+  const DataAllocator<VoxelT> &vker = m_KernelData.ConstData();
+  int vox_size = vbuf.size();
+  int ker_size = vker.size();
+  int pos;
+  float conv = 0, ksum = 0;
+  for (int ik = 0; ik < ker_size; ++ik) {
+    pos = index + vker[ik].Count - vker[m_KernelData.GetCenterData()].Count;
+    pos = (pos + vox_size) % vox_size;
+    conv += vbuf[pos].Value * vker[ik].Value;
+    ksum += vker[ik].Value;
+  }
+  return conv / ksum;
 }
 
-
-
 #undef _TPLT_
 #undef _TPL_
 
-
-
-
-
-
-
-}
-
-
-
+} // namespace uLib
 
 #endif // VOXIMAGEFILTER_HPP

src/Math/VoxImageFilter2ndStat.hpp

@@ -23,14 +23,12 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTER2NDSTAT_HPP
 #define VOXIMAGEFILTER2NDSTAT_HPP
 
-#include <Math/Dense.h>
 #include "Math/VoxImage.h"
 #include "VoxImageFilter.h"
+#include <Math/Dense.h>
 
 ////////////////////////////////////////////////////////////////////////////////
 /////  VOXIMAGE FILTER ABTRIM  /////////////////////////////////////////////////
@@ -39,45 +37,42 @@
 namespace uLib {
 
 template <typename VoxelT>
-class VoxFilterAlgorithm2ndStat :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithm2ndStat<VoxelT> > {
+class VoxFilterAlgorithm2ndStat
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithm2ndStat<VoxelT>> {
 
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithm2ndStat<VoxelT> > BaseClass;
-    VoxFilterAlgorithm2ndStat(const Vector3i &size) :
-        BaseClass(size)
-    { }
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithm2ndStat<VoxelT>> BaseClass;
+  VoxFilterAlgorithm2ndStat(const Vector3i &size) : BaseClass(size) {}
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        const std::vector<VoxelT> &vbuf = buffer.ConstData();
-        const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-        int vox_size = vbuf.size();
-        int ker_size = vker.size();
-        int pos;
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+    const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+    int vox_size = vbuf.size();
+    int ker_size = vker.size();
+    int pos;
 
-        // mean //
-        float conv = 0, ksum = 0;
-        for (int ik = 0; ik < ker_size; ++ik) {
-            pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + vox_size) % vox_size;
-            conv += vbuf[pos].Value * vker[ik].Value;
-            ksum += vker[ik].Value;
-        }
-        float mean = conv / ksum;
-
-        // rms //
-        conv = 0;
-        for (int ik = 0; ik < ker_size; ++ik) {
-            pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + vox_size) % vox_size;
-            conv += pow((vbuf[pos].Value * vker[ik].Value) - mean , 2);
-        }
-        return conv / (vker.size() - 1) ;
+    // mean //
+    float conv = 0, ksum = 0;
+    for (int ik = 0; ik < ker_size; ++ik) {
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + vox_size) % vox_size;
+      conv += vbuf[pos].Value * vker[ik].Value;
+      ksum += vker[ik].Value;
     }
+    float mean = conv / ksum;
 
-
+    // rms //
+    conv = 0;
+    for (int ik = 0; ik < ker_size; ++ik) {
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + vox_size) % vox_size;
+      conv += pow((vbuf[pos].Value * vker[ik].Value) - mean, 2);
+    }
+    return conv / (vker.size() - 1);
+  }
 };
 
-}
+} // namespace uLib
 #endif // VOXIMAGEFILTER2NDSTAT_HPP

src/Math/VoxImageFilterABTrim.hpp

@@ -23,14 +23,12 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTERABTRIM_HPP
 #define VOXIMAGEFILTERABTRIM_HPP
 
-#include <Math/Dense.h>
 #include "Math/VoxImage.h"
 #include "VoxImageFilter.h"
+#include <Math/Dense.h>
 
 ////////////////////////////////////////////////////////////////////////////////
 /////  VOXIMAGE FILTER ABTRIM  /////////////////////////////////////////////////
@@ -38,142 +36,257 @@
 
 namespace uLib {
 
+#if defined(USE_CUDA) && defined(__CUDACC__)
 template <typename VoxelT>
-class VoxFilterAlgorithmAbtrim :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithmAbtrim<VoxelT> > {
+__global__ void ABTrimFilterKernel(const VoxelT *in, VoxelT *out,
+                                   const VoxelT *kernel, int vox_size,
+                                   int ker_size, int center_count, int mAtrim,
+                                   int mBtrim) {
+  int index = blockIdx.x * blockDim.x + threadIdx.x;
+  if (index < vox_size) {
+    // Allocate space for sorting
+    extern __shared__ char shared_mem[];
+    VoxelT *mfh =
+        (VoxelT *)&shared_mem[threadIdx.x * ker_size * sizeof(VoxelT)];
 
-    struct KernelSortAscending
-    {
-        bool operator()(const VoxelT& e1, const VoxelT& e2)
-        { return e1.Value < e2.Value; }
-    };
+    for (int i = 0; i < ker_size; ++i) {
+      mfh[i].Count = i;
+    }
+
+    for (int ik = 0; ik < ker_size; ik++) {
+      int pos = index + kernel[ik].Count - center_count;
+      if (pos < 0) {
+        pos += vox_size * ((-pos / vox_size) + 1);
+      }
+      pos = pos % vox_size;
+      mfh[ik].Value = in[pos].Value;
+    }
+
+    // Simple bubble sort for small arrays
+    for (int i = 0; i < ker_size - 1; i++) {
+      for (int j = 0; j < ker_size - i - 1; j++) {
+        if (mfh[j].Value > mfh[j + 1].Value) {
+          VoxelT temp = mfh[j];
+          mfh[j] = mfh[j + 1];
+          mfh[j + 1] = temp;
+        }
+      }
+    }
+
+    float ker_sum = 0;
+    float fconv = 0;
+    for (int ik = 0; ik < mAtrim; ik++) {
+      ker_sum += kernel[mfh[ik].Count].Value;
+    }
+    for (int ik = mAtrim; ik < ker_size - mBtrim; ik++) {
+      fconv += mfh[ik].Value * kernel[mfh[ik].Count].Value;
+      ker_sum += kernel[mfh[ik].Count].Value;
+    }
+    for (int ik = ker_size - mBtrim; ik < ker_size; ik++) {
+      ker_sum += kernel[mfh[ik].Count].Value;
+    }
+
+    out[index].Value = fconv / ker_sum;
+  }
+}
+#endif
+
+template <typename VoxelT>
+class VoxFilterAlgorithmAbtrim
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithmAbtrim<VoxelT>> {
+
+  struct KernelSortAscending {
+    bool operator()(const VoxelT &e1, const VoxelT &e2) {
+      return e1.Value < e2.Value;
+    }
+  };
 
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmAbtrim<VoxelT> > BaseClass;
-    VoxFilterAlgorithmAbtrim(const Vector3i &size) :
-        BaseClass(size)
-    {
-        mAtrim = 0;
-        mBtrim = 0;
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmAbtrim<VoxelT>> BaseClass;
+  VoxFilterAlgorithmAbtrim(const Vector3i &size) : BaseClass(size) {
+    mAtrim = 0;
+    mBtrim = 0;
+  }
+
+#if defined(USE_CUDA) && defined(__CUDACC__)
+  void Run() {
+    if (this->m_Image->Data().GetDevice() == MemoryDevice::VRAM ||
+        this->m_KernelData.Data().GetDevice() == MemoryDevice::VRAM) {
+
+      this->m_Image->Data().MoveToVRAM();
+      this->m_KernelData.Data().MoveToVRAM();
+
+      VoxImage<VoxelT> buffer = *(this->m_Image);
+      buffer.Data().MoveToVRAM();
+
+      int vox_size = buffer.Data().size();
+      int ker_size = this->m_KernelData.Data().size();
+
+      VoxelT *d_img_out = this->m_Image->Data().GetVRAMData();
+      const VoxelT *d_img_in = buffer.Data().GetVRAMData();
+      const VoxelT *d_kernel = this->m_KernelData.Data().GetVRAMData();
+      int center_count =
+          this->m_KernelData[this->m_KernelData.GetCenterData()].Count;
+
+      int threadsPerBlock = 256;
+      int blocksPerGrid = (vox_size + threadsPerBlock - 1) / threadsPerBlock;
+      size_t shared_mem_size = threadsPerBlock * ker_size * sizeof(VoxelT);
+
+      ABTrimFilterKernel<<<blocksPerGrid, threadsPerBlock, shared_mem_size>>>(
+          d_img_in, d_img_out, d_kernel, vox_size, ker_size, center_count,
+          mAtrim, mBtrim);
+      cudaDeviceSynchronize();
+    } else {
+      BaseClass::Run();
+    }
+  }
+#endif
+
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+    const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+    int vox_size = vbuf.size();
+    int ker_size = vker.size();
+    int pos;
+
+    std::vector<VoxelT> mfh(ker_size);
+    for (int i = 0; i < ker_size; ++i)
+      mfh[i].Count = i; // index key for ordering function
+    for (int ik = 0; ik < ker_size; ik++) {
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + vox_size) % vox_size;
+      mfh[ik].Value = vbuf[pos].Value;
     }
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        const std::vector<VoxelT> &vbuf = buffer.ConstData();
-        const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-        int vox_size = vbuf.size();
-        int ker_size = vker.size();
-        int pos;
-
-        std::vector<VoxelT> mfh(ker_size);
-        for (int i = 0; i < ker_size; ++i)
-            mfh[i].Count = i; //index key for ordering function
-        for (int ik = 0; ik < ker_size; ik++) {
-            pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + vox_size) % vox_size;
-            mfh[ik].Value = vbuf[pos].Value;
-        }
-
-        std::sort(mfh.begin(), mfh.end(), KernelSortAscending());
-        float ker_sum = 0;
-        float fconv = 0;
-        for (int ik = 0; ik < mAtrim; ik++)
-            ker_sum += vker[ mfh[ik].Count ].Value;
-        for (int ik = mAtrim; ik < ker_size - mBtrim; ik++) {
-            fconv += mfh[ik].Value * vker[ mfh[ik].Count ].Value; // convloution //
-            ker_sum += vker[ mfh[ik].Count ].Value;
-        }
-        for (int ik = ker_size - mBtrim; ik < ker_size; ik++)
-            ker_sum += vker[ mfh[ik].Count ].Value;
-
-        return fconv / ker_sum;
+    std::sort(mfh.begin(), mfh.end(), KernelSortAscending());
+    float ker_sum = 0;
+    float fconv = 0;
+    for (int ik = 0; ik < mAtrim; ik++)
+      ker_sum += vker[mfh[ik].Count].Value;
+    for (int ik = mAtrim; ik < ker_size - mBtrim; ik++) {
+      fconv += mfh[ik].Value * vker[mfh[ik].Count].Value; // convloution //
+      ker_sum += vker[mfh[ik].Count].Value;
     }
+    for (int ik = ker_size - mBtrim; ik < ker_size; ik++)
+      ker_sum += vker[mfh[ik].Count].Value;
 
-    inline void SetABTrim(int a, int b) { mAtrim = a; mBtrim = b; }
+    return fconv / ker_sum;
+  }
+
+  inline void SetABTrim(int a, int b) {
+    mAtrim = a;
+    mBtrim = b;
+  }
 
 private:
-    int mAtrim;
-    int mBtrim;
+  int mAtrim;
+  int mBtrim;
 };
 
-
-
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 ////////////////////////////////////////////////////////////////////////////////
 // Roberspierre Filter //
 
-
-
 template <typename VoxelT>
-class VoxFilterAlgorithmSPR :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithmSPR<VoxelT> > {
+class VoxFilterAlgorithmSPR
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithmSPR<VoxelT>> {
 
-    struct KernelSortAscending
-    {
-        bool operator()(const VoxelT& e1, const VoxelT& e2)
-        { return e1.Value < e2.Value; }
-    };
+  struct KernelSortAscending {
+    bool operator()(const VoxelT &e1, const VoxelT &e2) {
+      return e1.Value < e2.Value;
+    }
+  };
 
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmSPR<VoxelT> > BaseClass;
-    VoxFilterAlgorithmSPR(const Vector3i &size) :
-        BaseClass(size)
-    {
-        mAtrim = 0;
-        mBtrim = 0;
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmSPR<VoxelT>> BaseClass;
+  VoxFilterAlgorithmSPR(const Vector3i &size) : BaseClass(size) {
+    mAtrim = 0;
+    mBtrim = 0;
+  }
+
+#if defined(USE_CUDA) && defined(__CUDACC__)
+  void Run() {
+    if (this->m_Image->Data().GetDevice() == MemoryDevice::VRAM ||
+        this->m_KernelData.Data().GetDevice() == MemoryDevice::VRAM) {
+
+      this->m_Image->Data().MoveToVRAM();
+      this->m_KernelData.Data().MoveToVRAM();
+
+      VoxImage<VoxelT> buffer = *(this->m_Image);
+      buffer.Data().MoveToVRAM();
+
+      int vox_size = buffer.Data().size();
+      int ker_size = this->m_KernelData.Data().size();
+
+      VoxelT *d_img_out = this->m_Image->Data().GetVRAMData();
+      const VoxelT *d_img_in = buffer.Data().GetVRAMData();
+      const VoxelT *d_kernel = this->m_KernelData.Data().GetVRAMData();
+      int center_count =
+          this->m_KernelData[this->m_KernelData.GetCenterData()].Count;
+
+      int threadsPerBlock = 256;
+      int blocksPerGrid = (vox_size + threadsPerBlock - 1) / threadsPerBlock;
+      size_t shared_mem_size = threadsPerBlock * ker_size * sizeof(VoxelT);
+
+      ABTrimFilterKernel<<<blocksPerGrid, threadsPerBlock, shared_mem_size>>>(
+          d_img_in, d_img_out, d_kernel, vox_size, ker_size, center_count,
+          mAtrim, mBtrim);
+      cudaDeviceSynchronize();
+    } else {
+      BaseClass::Run();
+    }
+  }
+#endif
+
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+    const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+    int vox_size = vbuf.size();
+    int ker_size = vker.size();
+    int pos;
+
+    std::vector<VoxelT> mfh(ker_size);
+    for (int i = 0; i < ker_size; ++i)
+      mfh[i].Count = i; // index key for ordering function
+    for (int ik = 0; ik < ker_size; ik++) {
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + vox_size) % vox_size;
+      mfh[ik].Value = vbuf[pos].Value;
     }
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        const std::vector<VoxelT> &vbuf = buffer.ConstData();
-        const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-        int vox_size = vbuf.size();
-        int ker_size = vker.size();
-        int pos;
+    std::sort(mfh.begin(), mfh.end(), KernelSortAscending());
+    float spr = vbuf[index].Value;
+    if ((mAtrim > 0 && spr <= mfh[mAtrim - 1].Value) ||
+        (mBtrim > 0 && spr >= mfh[ker_size - mBtrim].Value)) {
+      float ker_sum = 0;
+      float fconv = 0;
+      for (int ik = 0; ik < mAtrim; ik++)
+        ker_sum += vker[mfh[ik].Count].Value;
+      for (int ik = mAtrim; ik < ker_size - mBtrim; ik++) {
+        fconv += mfh[ik].Value * vker[mfh[ik].Count].Value;
+        ker_sum += vker[mfh[ik].Count].Value;
+      }
+      for (int ik = ker_size - mBtrim; ik < ker_size; ik++)
+        ker_sum += vker[mfh[ik].Count].Value;
 
-        std::vector<VoxelT> mfh(ker_size);
-        for (int i = 0; i < ker_size; ++i)
-            mfh[i].Count = i; //index key for ordering function
-        for (int ik = 0; ik < ker_size; ik++) {
-            pos = index + vker[ik].Count -
-                    vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + vox_size) % vox_size;
-            mfh[ik].Value = vbuf[pos].Value;
-        }
+      return fconv / ker_sum;
+    } else
+      return spr;
+  }
 
-        std::sort(mfh.begin(), mfh.end(), KernelSortAscending());
-        float spr = vbuf[index].Value;
-        if( (mAtrim > 0 && spr <= mfh[mAtrim-1].Value) ||
-                (mBtrim > 0 && spr >= mfh[ker_size - mBtrim].Value) )
-        {
-            float ker_sum = 0;
-            float fconv = 0;
-            for (int ik = 0; ik < mAtrim; ik++)
-                ker_sum += vker[ mfh[ik].Count ].Value;
-            for (int ik = mAtrim; ik < ker_size - mBtrim; ik++) {
-                fconv += mfh[ik].Value * vker[ mfh[ik].Count ].Value;
-                ker_sum += vker[ mfh[ik].Count ].Value;
-            }
-            for (int ik = ker_size - mBtrim; ik < ker_size; ik++)
-                ker_sum += vker[ mfh[ik].Count ].Value;
-
-            return fconv / ker_sum;
-        }
-        else
-            return spr;
-    }
-
-    inline void SetABTrim(int a, int b) { mAtrim = a; mBtrim = b; }
+  inline void SetABTrim(int a, int b) {
+    mAtrim = a;
+    mBtrim = b;
+  }
 
 private:
-    int mAtrim;
-    int mBtrim;
+  int mAtrim;
+  int mBtrim;
 };
 
-
-
-
-}
+} // namespace uLib
 
 #endif // VOXIMAGEFILTERABTRIM_HPP

src/Math/VoxImageFilterBilateral.hpp

@@ -23,14 +23,12 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTERBILATERAL_HPP
 #define VOXIMAGEFILTERBILATERAL_HPP
 
-#include <Math/Dense.h>
 #include "Math/VoxImage.h"
 #include "VoxImageFilter.h"
+#include <Math/Dense.h>
 
 ////////////////////////////////////////////////////////////////////////////////
 /////  VOXIMAGE FILTER LINEAR  /////////////////////////////////////////////////
@@ -38,115 +36,119 @@
 
 namespace uLib {
 
-
 template <typename VoxelT>
-class VoxFilterAlgorithmBilateral :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateral<VoxelT> > {
+class VoxFilterAlgorithmBilateral
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateral<VoxelT>> {
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateral<VoxelT> > BaseClass;
-    VoxFilterAlgorithmBilateral(const Vector3i &size) : BaseClass(size) {
-        m_sigma = 1;
-    }
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateral<VoxelT>> BaseClass;
+  VoxFilterAlgorithmBilateral(const Vector3i &size) : BaseClass(size) {
+    m_sigma = 1;
+  }
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        const std::vector<VoxelT> &vbuf = buffer.ConstData();
-        const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-        int vox_size = vbuf.size();
-        int ker_size = vker.size();
-        int pos;
-        float conv = 0, ksum = 0;
-        float gamma_smooth;
-        for (int ik = 0; ik < ker_size; ++ik) {
-            //            if (ik==this->m_KernelData.GetCenterData()) continue;
-            pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + vox_size) % vox_size;
-            gamma_smooth = compute_gauss( fabs(vbuf[index].Value - vbuf[pos].Value) * 1.E6 );
-            conv += vbuf[pos].Value * vker[ik].Value * gamma_smooth;
-            ksum += vker[ik].Value * gamma_smooth;
-        }
-        return conv / ksum;
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+    const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+    int vox_size = vbuf.size();
+    int ker_size = vker.size();
+    int pos;
+    float conv = 0, ksum = 0;
+    float gamma_smooth;
+    for (int ik = 0; ik < ker_size; ++ik) {
+      //            if (ik==this->m_KernelData.GetCenterData()) continue;
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + vox_size) % vox_size;
+      gamma_smooth =
+          compute_gauss(fabs(vbuf[index].Value - vbuf[pos].Value) * 1.E6);
+      conv += vbuf[pos].Value * vker[ik].Value * gamma_smooth;
+      ksum += vker[ik].Value * gamma_smooth;
     }
+    return conv / ksum;
+  }
 
-    inline void SetIntensitySigma(const float s) { m_sigma = s; }
+  inline void SetIntensitySigma(const float s) { m_sigma = s; }
 
 private:
-    inline float compute_gauss(const float x) {
-        return 1/(sqrt(2*M_PI)* m_sigma) * exp(-0.5*(x*x)/(m_sigma*m_sigma));
-    }
+  inline float compute_gauss(const float x) {
+    return 1 / (sqrt(2 * M_PI) * m_sigma) *
+           exp(-0.5 * (x * x) / (m_sigma * m_sigma));
+  }
 
-    Scalarf m_sigma;
+  Scalarf m_sigma;
 };
 
-
 template <typename VoxelT>
-class VoxFilterAlgorithmBilateralTrim :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateralTrim<VoxelT> > {
+class VoxFilterAlgorithmBilateralTrim
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateralTrim<VoxelT>> {
 
-    typedef std::pair<float,float> FPair;
+  typedef std::pair<float, float> FPair;
 
-    struct KernelSortAscending
-    {
-        bool operator()(const FPair& e1, const FPair& e2)
-        { return e1.second < e2.second; }
-    };
+  struct KernelSortAscending {
+    bool operator()(const FPair &e1, const FPair &e2) {
+      return e1.second < e2.second;
+    }
+  };
 
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateralTrim<VoxelT> > BaseClass;
-    VoxFilterAlgorithmBilateralTrim(const Vector3i &size) : BaseClass(size) {
-        m_sigma = 1;
-        mAtrim = 0;
-        mBtrim = 0;
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmBilateralTrim<VoxelT>>
+      BaseClass;
+  VoxFilterAlgorithmBilateralTrim(const Vector3i &size) : BaseClass(size) {
+    m_sigma = 1;
+    mAtrim = 0;
+    mBtrim = 0;
+  }
+
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+    const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+    int img_size = vbuf.size();
+    int ker_size = vker.size();
+    int pos;
+
+    std::vector<FPair> mfh(ker_size);
+    for (int i = 0; i < ker_size; ++i)
+      mfh[i].first = vker[i].Value; // kernel value in first
+    for (int ik = 0; ik < ker_size; ik++) {
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + img_size) % img_size;
+      mfh[ik].second = vbuf[pos].Value; // image value in second
     }
+    std::sort(mfh.begin(), mfh.end(), KernelSortAscending());
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        const std::vector<VoxelT> &vbuf = buffer.ConstData();
-        const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-        int img_size = vbuf.size();
-        int ker_size = vker.size();
-        int pos;
-
-
-
-        std::vector<FPair> mfh(ker_size);
-        for (int i = 0; i < ker_size; ++i)
-            mfh[i].first = vker[i].Value; // kernel value in first
-        for (int ik = 0; ik < ker_size; ik++) {
-            pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + img_size) % img_size;
-            mfh[ik].second = vbuf[pos].Value; // image value in second
-        }
-        std::sort(mfh.begin(), mfh.end(), KernelSortAscending());
-
-        float conv = 0, ksum = 0;
-        float gamma_smooth;
-        //        for (int ik = 0; ik < mAtrim; ik++)
-        //            ksum += mfh[ik].first;
-        for (int ik = mAtrim; ik < ker_size - mBtrim; ik++) {
-            gamma_smooth = compute_gauss( fabs(vbuf[index].Value - mfh[ik].second) * 1.E6 );
-            conv += mfh[ik].first * mfh[ik].second * gamma_smooth;
-            ksum += mfh[ik].first * gamma_smooth;
-        }
-        //        for (int ik = ker_size - mBtrim; ik < ker_size; ik++)
-        //            ksum += mfh[ik].first;
-
-        return conv / ksum;
+    float conv = 0, ksum = 0;
+    float gamma_smooth;
+    //        for (int ik = 0; ik < mAtrim; ik++)
+    //            ksum += mfh[ik].first;
+    for (int ik = mAtrim; ik < ker_size - mBtrim; ik++) {
+      gamma_smooth =
+          compute_gauss(fabs(vbuf[index].Value - mfh[ik].second) * 1.E6);
+      conv += mfh[ik].first * mfh[ik].second * gamma_smooth;
+      ksum += mfh[ik].first * gamma_smooth;
     }
+    //        for (int ik = ker_size - mBtrim; ik < ker_size; ik++)
+    //            ksum += mfh[ik].first;
 
-    inline void SetIntensitySigma(const float s) { m_sigma = s; }
-    inline void SetABTrim(int a, int b) { mAtrim = a; mBtrim = b; }
+    return conv / ksum;
+  }
+
+  inline void SetIntensitySigma(const float s) { m_sigma = s; }
+  inline void SetABTrim(int a, int b) {
+    mAtrim = a;
+    mBtrim = b;
+  }
 
 private:
-    inline float compute_gauss(const float x) {
-        return 1/(sqrt(2*M_PI)* m_sigma) * exp(-0.5*(x*x)/(m_sigma*m_sigma));
-    }
+  inline float compute_gauss(const float x) {
+    return 1 / (sqrt(2 * M_PI) * m_sigma) *
+           exp(-0.5 * (x * x) / (m_sigma * m_sigma));
+  }
 
-    Scalarf m_sigma;
-    int mAtrim;
-    int mBtrim;
+  Scalarf m_sigma;
+  int mAtrim;
+  int mBtrim;
 };
 
-}
+} // namespace uLib
 
 #endif // VOXIMAGEFILTERBILATERAL_HPP

src/Math/VoxImageFilterCustom.hpp

@@ -23,14 +23,12 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTERCUSTOM_HPP
 #define VOXIMAGEFILTERCUSTOM_HPP
 
-#include <Math/Dense.h>
 #include "Math/VoxImage.h"
 #include "VoxImageFilter.h"
+#include <Math/Dense.h>
 
 #define likely(expr) __builtin_expect(!!(expr), 1)
 
@@ -41,50 +39,50 @@
 namespace uLib {
 
 template <typename VoxelT>
-class VoxFilterAlgorithmCustom :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithmCustom<VoxelT> > {
+class VoxFilterAlgorithmCustom
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithmCustom<VoxelT>> {
 
+  typedef float (*FunctionPt)(const std::vector<Scalarf> &);
 
-    typedef float (* FunctionPt)(const std::vector<Scalarf> &);
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmCustom<VoxelT> > BaseClass;
-    VoxFilterAlgorithmCustom(const Vector3i &size) :
-        BaseClass(size), m_CustomEvaluate(NULL)
-    {}
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmCustom<VoxelT>> BaseClass;
+  VoxFilterAlgorithmCustom(const Vector3i &size)
+      : BaseClass(size), m_CustomEvaluate(NULL) {}
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        if(likely(m_CustomEvaluate)) {
-            const std::vector<VoxelT> &vbuf = buffer.ConstData();
-            const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-            int vox_size = vbuf.size();
-            int ker_size = vker.size();
-            int pos;
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    if (likely(m_CustomEvaluate)) {
+      const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+      const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+      int vox_size = vbuf.size();
+      int ker_size = vker.size();
+      int pos;
 
-            float ker_sum = 0;
-            std::vector<Scalarf> mfh(ker_size);
-            for (int ik = 0; ik < ker_size; ik++) {
-                pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-                pos = (pos + vox_size) % vox_size;
-                mfh[ik] = vbuf[pos].Value * vker[ik].Value;
-                ker_sum += vker[ik].Value;
-            }
-
-            return this->m_CustomEvaluate(mfh);
-        }
-        else
-            std::cerr << "Custom evaluate function is NULL \n" <<
-                         "No operation performed by filter.\n";
+      float ker_sum = 0;
+      std::vector<Scalarf> mfh(ker_size);
+      for (int ik = 0; ik < ker_size; ik++) {
+        pos = index + vker[ik].Count -
+              vker[this->m_KernelData.GetCenterData()].Count;
+        pos = (pos + vox_size) % vox_size;
+        mfh[ik] = vbuf[pos].Value * vker[ik].Value;
+        ker_sum += vker[ik].Value;
+      }
 
+      return this->m_CustomEvaluate(mfh);
+    } else {
+      std::cerr << "Custom evaluate function is NULL \n"
+                << "No operation performed by filter.\n";
+      return 0;
     }
+  }
 
-    inline void SetCustomEvaluate(FunctionPt funPt) { this->m_CustomEvaluate = funPt; }
+  inline void SetCustomEvaluate(FunctionPt funPt) {
+    this->m_CustomEvaluate = funPt;
+  }
 
 private:
-    FunctionPt m_CustomEvaluate;
+  FunctionPt m_CustomEvaluate;
 };
 
-}
-
+} // namespace uLib
 
 #endif // VOXIMAGEFILTERCUSTOM_HPP

src/Math/VoxImageFilterLinear.hpp

@@ -23,14 +23,12 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTERLINEAR_HPP
 #define VOXIMAGEFILTERLINEAR_HPP
 
-#include <Math/Dense.h>
 #include "Math/VoxImage.h"
 #include "VoxImageFilter.h"
+#include <Math/Dense.h>
 
 ////////////////////////////////////////////////////////////////////////////////
 /////  VOXIMAGE FILTER LINEAR  /////////////////////////////////////////////////
@@ -38,32 +36,86 @@
 
 namespace uLib {
 
+#if defined(USE_CUDA) && defined(__CUDACC__)
+template <typename VoxelT>
+__global__ void LinearFilterKernel(const VoxelT *in, VoxelT *out,
+                                   const VoxelT *kernel, int vox_size,
+                                   int ker_size, int center_count) {
+  int index = blockIdx.x * blockDim.x + threadIdx.x;
+  if (index < vox_size) {
+    float conv = 0;
+    float ksum = 0;
+    for (int ik = 0; ik < ker_size; ++ik) {
+      int pos = index + kernel[ik].Count - center_count;
+      if (pos < 0) {
+        pos += vox_size * ((-pos / vox_size) + 1);
+      }
+      pos = pos % vox_size;
+      conv += in[pos].Value * kernel[ik].Value;
+      ksum += kernel[ik].Value;
+    }
+    out[index].Value = conv / ksum;
+  }
+}
+#endif
 
 template <typename VoxelT>
-class VoxFilterAlgorithmLinear :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithmLinear<VoxelT> > {
+class VoxFilterAlgorithmLinear
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithmLinear<VoxelT>> {
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmLinear<VoxelT> > BaseClass;
-    VoxFilterAlgorithmLinear(const Vector3i &size) : BaseClass(size) {}
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmLinear<VoxelT>> BaseClass;
+  VoxFilterAlgorithmLinear(const Vector3i &size) : BaseClass(size) {}
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        const std::vector<VoxelT> &vbuf = buffer.ConstData();
-        const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-        int vox_size = vbuf.size();
-        int ker_size = vker.size();
-        int pos;
-        float conv = 0, ksum = 0;
-        for (int ik = 0; ik < ker_size; ++ik) {
-            pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + vox_size) % vox_size;
-            conv += vbuf[pos].Value * vker[ik].Value;
-            ksum += vker[ik].Value;
-        }
-        return conv / ksum;
+#if defined(USE_CUDA) && defined(__CUDACC__)
+  void Run() {
+    if (this->m_Image->Data().GetDevice() == MemoryDevice::VRAM ||
+        this->m_KernelData.Data().GetDevice() == MemoryDevice::VRAM) {
+
+      this->m_Image->Data().MoveToVRAM();
+      this->m_KernelData.Data().MoveToVRAM();
+
+      VoxImage<VoxelT> buffer = *(this->m_Image);
+      buffer.Data().MoveToVRAM();
+
+      int vox_size = buffer.Data().size();
+      int ker_size = this->m_KernelData.Data().size();
+
+      VoxelT *d_img_out = this->m_Image->Data().GetVRAMData();
+      const VoxelT *d_img_in = buffer.Data().GetVRAMData();
+      const VoxelT *d_kernel = this->m_KernelData.Data().GetVRAMData();
+      int center_count =
+          this->m_KernelData[this->m_KernelData.GetCenterData()].Count;
+
+      int threadsPerBlock = 256;
+      int blocksPerGrid = (vox_size + threadsPerBlock - 1) / threadsPerBlock;
+
+      LinearFilterKernel<<<blocksPerGrid, threadsPerBlock>>>(
+          d_img_in, d_img_out, d_kernel, vox_size, ker_size, center_count);
+      cudaDeviceSynchronize();
+    } else {
+      BaseClass::Run();
     }
+  }
+#endif
+
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+    const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+    int vox_size = vbuf.size();
+    int ker_size = vker.size();
+    int pos;
+    float conv = 0, ksum = 0;
+    for (int ik = 0; ik < ker_size; ++ik) {
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + vox_size) % vox_size;
+      conv += vbuf[pos].Value * vker[ik].Value;
+      ksum += vker[ik].Value;
+    }
+    return conv / ksum;
+  }
 };
 
-}
+} // namespace uLib
 
 #endif // VOXIMAGEFILTERLINEAR_HPP

src/Math/VoxImageFilterMedian.hpp

@@ -23,14 +23,12 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXIMAGEFILTERMEDIAN_HPP
 #define VOXIMAGEFILTERMEDIAN_HPP
 
-#include <Math/Dense.h>
 #include "Math/VoxImage.h"
 #include "VoxImageFilter.h"
+#include <Math/Dense.h>
 
 ////////////////////////////////////////////////////////////////////////////////
 /////  VOXIMAGE FILTER MEDIAN  /////////////////////////////////////////////////
@@ -39,37 +37,38 @@
 namespace uLib {
 
 template <typename VoxelT>
-class VoxFilterAlgorithmMedian :
-        public VoxImageFilter<VoxelT, VoxFilterAlgorithmMedian<VoxelT> > {
+class VoxFilterAlgorithmMedian
+    : public VoxImageFilter<VoxelT, VoxFilterAlgorithmMedian<VoxelT>> {
 public:
-    typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmMedian<VoxelT> > BaseClass;
-    VoxFilterAlgorithmMedian(const Vector3i &size) : BaseClass(size) {}
+  typedef VoxImageFilter<VoxelT, VoxFilterAlgorithmMedian<VoxelT>> BaseClass;
+  VoxFilterAlgorithmMedian(const Vector3i &size) : BaseClass(size) {}
 
-    float Evaluate(const VoxImage<VoxelT> &buffer, int index)
-    {
-        const std::vector<VoxelT> &vbuf = buffer.ConstData();
-        const std::vector<VoxelT> &vker = this->m_KernelData.ConstData();
-        int vox_size = vbuf.size();
-        int ker_size = vker.size();
-        int pos;
+  float Evaluate(const VoxImage<VoxelT> &buffer, int index) {
+    const DataAllocator<VoxelT> &vbuf = buffer.ConstData();
+    const DataAllocator<VoxelT> &vker = this->m_KernelData.ConstData();
+    int vox_size = vbuf.size();
+    int ker_size = vker.size();
+    int pos;
 
-        std::vector<float> mfh(ker_size);
-        for (int ik = 0; ik < ker_size; ik++) {
-            pos = index + vker[ik].Count - vker[this->m_KernelData.GetCenterData()].Count;
-            pos = (pos + vox_size) % vox_size;
-            mfh[ik] = vbuf[pos].Value * vker[ik].Value;
-        }
-        std::sort(mfh.begin(), mfh.end());
-        pos = 0;
-        // count zeroes in filter kernel to move it out of median //
-        for (int i = 0; i < ker_size; ++i)
-            if (vker[i].Value == 0.0) pos++;
-        // median //
-        pos += (ker_size - pos) / 2;
-        return mfh[pos];
+    std::vector<float> mfh(ker_size);
+    for (int ik = 0; ik < ker_size; ik++) {
+      pos = index + vker[ik].Count -
+            vker[this->m_KernelData.GetCenterData()].Count;
+      pos = (pos + vox_size) % vox_size;
+      mfh[ik] = vbuf[pos].Value * vker[ik].Value;
     }
+    std::sort(mfh.begin(), mfh.end());
+    pos = 0;
+    // count zeroes in filter kernel to move it out of median //
+    for (int i = 0; i < ker_size; ++i)
+      if (vker[i].Value == 0.0)
+        pos++;
+    // median //
+    pos += (ker_size - pos) / 2;
+    return mfh[pos];
+  }
 };
 
-}
+} // namespace uLib
 
 #endif // VOXIMAGEFILTERMEDIAN_HPP

src/Math/VoxRaytracer.cpp

@@ -39,48 +39,48 @@ namespace uLib {
 ////////////////////////////////////////////////////////////////////////////////
 
 void VoxRaytracer::RayData::AddElement(Id_t id, float L) {
+  if (m_Count >= m_Data.size()) {
+    size_t new_size = m_Data.size() == 0 ? 128 : m_Data.size() * 2;
+    m_Data.resize(new_size);
+  }
   Element el = {id, L};
-  m_Data.push_back(el);
+  m_Data[m_Count] = el;
+  m_Count++;
   m_TotalLength += L;
 }
 
 void VoxRaytracer::RayData::AppendRay(const VoxRaytracer::RayData &in) {
-  if (unlikely(!in.m_Data.size())) {
+  if (unlikely(in.m_Count == 0)) {
     std::cout << "Warinig: PoCA on exit border!\n";
     return;
-  } else if (unlikely(!m_Data.size())) {
-    m_Data = in.m_Data;
+  } else if (unlikely(m_Count == 0)) {
+    m_Data.resize(in.m_Count);
+    for (size_t i = 0; i < in.m_Count; ++i) {
+      m_Data[i] = in.m_Data[i];
+    }
+    m_Count = in.m_Count;
+    m_TotalLength = in.m_TotalLength;
     std::cout << "Warinig: PoCA on entrance border!\n";
     return;
   } else {
     // Opzione 1) un voxel in piu' //
-    if (in.m_Data.size() > 0) {
-      m_Data.insert(m_Data.end(), in.m_Data.begin(), in.m_Data.end());
+    if (in.m_Count > 0) {
+      if (m_Count + in.m_Count > m_Data.size()) {
+        m_Data.resize(m_Count + in.m_Count);
+      }
+      for (size_t i = 0; i < in.m_Count; ++i) {
+        m_Data[m_Count + i] = in.m_Data[i];
+      }
+      m_Count += in.m_Count;
     }
-    // Opzione 2) merge dei voxel nel poca.
-    //        RayData::Element &e1 = m_Data.back();
-    //        const RayData::Element &e2 = in.m_Data.front();
-    //        if(e1.vox_id == e2.vox_id)
-    //        {
-    //            m_Data.reserve(m_Data.size() + in.m_Data.size() - 1);
-    //            e1.L += e2.L; //fix//
-    //            m_Data.insert(m_Data.end(), in.m_Data.begin()+1,
-    //            in.m_Data.end());
-    //        }
-    //        else {
-    //            m_Data.reserve(m_Data.size() + in.m_Data.size());
-    //            m_Data.insert(m_Data.end(), in.m_Data.begin(),
-    //            in.m_Data.end());
-    //        }
     m_TotalLength += in.m_TotalLength;
   }
 }
 
 void VoxRaytracer::RayData::PrintSelf(std::ostream &o) {
   o << "Ray: total lenght " << m_TotalLength << "\n";
-  std::vector<Element>::iterator it;
-  for (it = m_Data.begin(); it < m_Data.end(); ++it)
-    o << "[ " << (*it).vox_id << ", " << (*it).L << "] \n";
+  for (size_t i = 0; i < m_Count; ++i)
+    o << "[ " << m_Data[i].vox_id << ", " << m_Data[i].L << "] \n";
 }
 
 ////////////////////////////////////////////////////////////////////////////////
@@ -144,14 +144,21 @@ VoxRaytracer::RayData
 VoxRaytracer::TraceBetweenPoints(const HPoint3f &in,
                                  const HPoint3f &out) const {
   RayData ray;
+
+  // get the local points and the direction vector
+  // local to image means in the normalized voxel space where the size
+  // of the voxel is 1 in all dimensions
   Vector4f pt1 = m_Image->GetLocalPoint(in);
   Vector4f pt2 = m_Image->GetLocalPoint(out);
   Vector4f s = pt2 - pt1;
 
+  // l is the total length of the ray in normalized voxel space
   float l = s.head(3).norm();
+
+  // L is the length of the ray between two grid lines in grid
   Vector3f L(l / s(0), l / s(1), l / s(2));
 
-  // Vector3f scale; // FIXXX
+  // Vector3f scale; // TODO: FIX Scaling
   // scale << (m_Image->GetWorldMatrix() * Vector4f(1,0,0,0)).norm(),
   //          (m_Image->GetWorldMatrix() * Vector4f(0,1,0,0)).norm(),
   //          (m_Image->GetWorldMatrix() * Vector4f(0,0,1,0)).norm();
@@ -174,21 +181,23 @@ VoxRaytracer::TraceBetweenPoints(const HPoint3f &in,
   float d;
   while (l > 0) {
 
+    // find which is the minimum of the offsets to the next grid line
+    // it will be also the actual normalized voxel ray length
     d = offset.minCoeff(&id);
 
+    // see if the voxel is inside the grid (we are still inside image)
     if (m_Image->IsInsideGrid(vid)) {
+      // add the voxel to the ray with mapping id and length scaled
       ray.AddElement(m_Image->Map(vid), d * m_scale(id));
     }
 
-    // nan check //
-    //        if(unlikely(!isFinite(d * scale(id)))) {
-    //            std:: cout << "NAN in raytracer\n";
-    //            exit(1);
-    //        }
-
+    // move to the next voxel
     vid(id) += (int)fast_sign(s(id));
 
+    // update the remaining length
     l -= d;
+
+    // update the offsets
     offset.array() -= d;
     offset(id) = fmin(L(id), l);
   }

src/Math/VoxRaytracer.h

@@ -23,71 +23,108 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #ifndef VOXRAYTRACER_H
 #define VOXRAYTRACER_H
 
+#include <Core/DataAllocator.h>
+#include <Core/Vector.h>
 #include <math.h>
 #include <vector>
 
 #include "Math/StructuredGrid.h"
+#include "Math/VoxImage.h"
+
+#ifdef USE_CUDA
+#include <cuda_runtime.h>
+#endif
 
 namespace uLib {
 
 class VoxRaytracer {
 
 public:
-    class RayData {
-    public:
-        RayData() : m_TotalLength(0) {}
+  class RayData {
+  public:
+    RayData() : m_TotalLength(0), m_Count(0) {}
 
-        typedef struct {
-            Id_t vox_id;
-            Scalarf L;
-        } Element;
-
-        inline void AddElement(Id_t id, float L);
-
-        void AppendRay ( const RayData &in);
-
-        inline const std::vector<Element>& Data() const { return this->m_Data; }
-
-        inline const Scalarf& TotalLength() const { return this->m_TotalLength; }
-
-        void PrintSelf(std::ostream &o);
-
-    private:
-        std::vector<Element> m_Data;
-        Scalarf              m_TotalLength;
+    struct Element {
+      Id_t vox_id;
+      Scalarf L;
+      ~Element() {}
     };
 
+    inline void AddElement(Id_t id, float L);
 
- public:
- VoxRaytracer(StructuredGrid &image) : m_Image(&image) {
-      m_scale <<
-	(m_Image->GetWorldMatrix() * Vector4f(1,0,0,0)).norm(),
-	(m_Image->GetWorldMatrix() * Vector4f(0,1,0,0)).norm(),
-	(m_Image->GetWorldMatrix() * Vector4f(0,0,1,0)).norm();
+    void AppendRay(const RayData &in);
+
+    inline uLib::Vector<Element> &Data() { return this->m_Data; }
+
+    inline const uLib::Vector<Element> &Data() const { return this->m_Data; }
+
+    inline size_t Count() const { return this->m_Count; }
+
+    inline size_t size() const { return this->m_Count; }
+
+    inline const Scalarf &TotalLength() const { return this->m_TotalLength; }
+
+    inline void SetCount(size_t c) {
+      this->m_Count = c;
+      if (this->m_Data.size() != c) {
+        this->m_Data.resize(c);
+      }
     }
 
-    bool GetEntryPoint(const HLine3f &line, HPoint3f &pt);
+    inline void SetTotalLength(Scalarf tl) { this->m_TotalLength = tl; }
 
-    bool GetExitPoint(const HLine3f &line, HPoint3f &pt);
+    void PrintSelf(std::ostream &o);
 
-    RayData TraceBetweenPoints(const HPoint3f &in, const HPoint3f &out) const;
+  private:
+    uLib::Vector<Element> m_Data;
+    Scalarf m_TotalLength;
+    size_t m_Count;
+  };
 
-    RayData TraceLine(const HLine3f &line) const;
+public:
+  VoxRaytracer(StructuredGrid &image) : m_Image(&image) {
+    m_scale << (m_Image->GetWorldMatrix() * Vector4f(1, 0, 0, 0)).norm(),
+        (m_Image->GetWorldMatrix() * Vector4f(0, 1, 0, 0)).norm(),
+        (m_Image->GetWorldMatrix() * Vector4f(0, 0, 1, 0)).norm();
+  }
 
-    inline StructuredGrid* GetImage() const { return this->m_Image; }
+  bool GetEntryPoint(const HLine3f &line, HPoint3f &pt);
+
+  bool GetExitPoint(const HLine3f &line, HPoint3f &pt);
+
+  RayData TraceBetweenPoints(const HPoint3f &in, const HPoint3f &out) const;
+
+  RayData TraceLine(const HLine3f &line) const;
+
+  inline StructuredGrid *GetImage() const { return this->m_Image; }
+
+#ifdef USE_CUDA
+  template <typename VoxelT>
+  void AccumulateLinesCUDA(const HLine3f *lines, size_t num_lines,
+                           VoxImage<VoxelT> &image);
+
+  void TraceLineCUDA(const HLine3f *lines, size_t num_lines, RayData *out_rays,
+                     int max_elements_per_ray = 128,
+                     float *kernel_time_ms = nullptr);
+
+  void TraceBetweenPointsCUDA(const HPoint3f *in_pts, const HPoint3f *out_pts,
+                              size_t num_lines, RayData *out_rays,
+                              int max_elements_per_ray = 128,
+                              float *kernel_time_ms = nullptr);
+#endif
 
 private:
-    StructuredGrid *m_Image;
-    Vector3f        m_scale;
+  StructuredGrid *m_Image;
+  Vector3f m_scale;
 };
 
-}
-
+} // namespace uLib
 
+#ifdef USE_CUDA
+#include "Math/VoxRaytracerCUDA.hpp"
+#endif
 
 #endif // VOXRAYTRACER_H

src/Math/VoxRaytracerCUDA.hpp

@@ -0,0 +1,548 @@
+#ifndef VOXRAYTRACERCUDA_H
+#define VOXRAYTRACERCUDA_H
+
+#ifdef USE_CUDA
+
+#include "Math/VoxImage.h"
+#include "Math/VoxRaytracer.h"
+#include <cuda_runtime.h>
+
+namespace uLib {
+
+#ifdef __CUDACC__
+template <typename VoxelT>
+__global__ void
+RaytraceAccumulateKernel(const float *lines_data, int num_lines,
+                         VoxelT *d_image, int dim0, int dim1, int dim2,
+                         const float *inv_world_matrix_data, float scale0,
+                         float scale1, float scale2) {
+  int idx = blockIdx.x * blockDim.x + threadIdx.x;
+  if (idx >= num_lines)
+    return;
+
+  const float *line_ptr = &lines_data[idx * 8];
+
+  float o_vec[4] = {line_ptr[0], line_ptr[1], line_ptr[2], line_ptr[3]};
+  float d_vec[4] = {line_ptr[4], line_ptr[5], line_ptr[6], line_ptr[7]};
+
+  float pt[4] = {0, 0, 0, 0};
+  float s[4] = {0, 0, 0, 0};
+
+  for (int i = 0; i < 4; ++i) {
+    for (int j = 0; j < 4; ++j) {
+      float m_val = inv_world_matrix_data[i + j * 4];
+      pt[i] += m_val * o_vec[j];
+      s[i] += m_val * d_vec[j];
+    }
+  }
+
+  float l = sqrtf(s[0] * s[0] + s[1] * s[1] + s[2] * s[2]);
+  if (l == 0)
+    return;
+
+  float L[3];
+  L[0] = l / s[0];
+  L[1] = l / s[1];
+  L[2] = l / s[2];
+
+  float offset[3];
+  for (int i = 0; i < 3; ++i) {
+    float fpt_i = floorf(pt[i]);
+    offset[i] = (s[i] >= 0) ? (1.0f - (pt[i] - fpt_i)) : (pt[i] - fpt_i);
+    offset[i] = fabsf(offset[i] * L[i]);
+    L[i] = fabsf(L[i]);
+  }
+
+  int id;
+  float d;
+  int vid[3] = {(int)floorf(pt[0]), (int)floorf(pt[1]), (int)floorf(pt[2])};
+
+  float scale_arr[3] = {scale0, scale1, scale2};
+
+  while (vid[0] >= 0 && vid[0] < dim0 && vid[1] >= 0 && vid[1] < dim1 &&
+         vid[2] >= 0 && vid[2] < dim2) {
+
+    d = offset[0];
+    id = 0;
+    if (offset[1] < d) {
+      d = offset[1];
+      id = 1;
+    }
+    if (offset[2] < d) {
+      d = offset[2];
+      id = 2;
+    }
+
+    float L_intersect = d * scale_arr[id];
+    size_t vox_index = vid[0] * dim1 * dim2 + vid[1] * dim2 + vid[2];
+
+    atomicAdd(&(d_image[vox_index].Value), L_intersect);
+
+    float sign_s = (s[id] >= 0) ? 1.0f : -1.0f;
+    vid[id] += (int)sign_s;
+
+    offset[0] -= d;
+    offset[1] -= d;
+    offset[2] -= d;
+    offset[id] = L[id];
+  }
+}
+#endif
+
+template <typename VoxelT>
+void VoxRaytracer::AccumulateLinesCUDA(const HLine3f *lines, size_t num_lines,
+                                       VoxImage<VoxelT> &image) {
+  if (num_lines == 0)
+    return;
+
+  image.Data().MoveToVRAM();
+
+  float *d_lines = nullptr;
+  size_t lines_size = num_lines * sizeof(HLine3f);
+  cudaMalloc(&d_lines, lines_size);
+  cudaMemcpy(d_lines, lines, lines_size, cudaMemcpyHostToDevice);
+  int threadsPerBlock = 256;
+  int blocksPerGrid = (num_lines + threadsPerBlock - 1) / threadsPerBlock;
+
+  Vector3i dims = image.GetDims();
+  Matrix4f inv_world_matrix = image.GetWorldMatrix().inverse();
+
+  float *d_inv_world;
+  cudaMalloc(&d_inv_world, 16 * sizeof(float));
+  cudaMemcpy(d_inv_world, inv_world_matrix.data(), 16 * sizeof(float),
+             cudaMemcpyHostToDevice);
+
+#ifdef __CUDACC__
+  RaytraceAccumulateKernel<<<blocksPerGrid, threadsPerBlock>>>(
+      d_lines, num_lines, image.Data().GetVRAMData(), dims(0), dims(1), dims(2),
+      d_inv_world, m_scale(0), m_scale(1), m_scale(2));
+  cudaDeviceSynchronize();
+
+  cudaError_t err = cudaGetLastError();
+  if (err != cudaSuccess) {
+    std::cerr << "CUDA Error in AccumulateLinesCUDA: "
+              << cudaGetErrorString(err) << std::endl;
+  }
+#else
+  std::cerr << "RaytraceAccumulateKernel requires NVCC!" << std::endl;
+#endif
+
+  cudaFree(d_lines);
+  cudaFree(d_inv_world);
+}
+
+#ifdef __CUDACC__
+
+__global__ void TraceBetweenPointsKernel(
+    const float *in_pts_data, const float *out_pts_data, int num_lines,
+    VoxRaytracer::RayData::Element **d_out_elements, size_t *d_out_counts,
+    float *d_out_lengths, int max_elements, int dim0, int dim1, int dim2,
+    const float *inv_world_matrix_data, float scale0, float scale1,
+    float scale2, int inc0, int inc1, int inc2) {
+  int idx = blockIdx.x * blockDim.x + threadIdx.x;
+  if (idx >= num_lines)
+    return;
+
+  VoxRaytracer::RayData::Element *ray_out = d_out_elements[idx];
+  size_t count = 0;
+  float tot_len = 0.0f;
+
+  const float *in_ptr = &in_pts_data[idx * 4];
+  const float *out_ptr = &out_pts_data[idx * 4];
+
+  float pt1[4] = {0, 0, 0, 0}, pt2[4] = {0, 0, 0, 0};
+  for (int i = 0; i < 4; ++i) {
+    for (int j = 0; j < 4; ++j) {
+      float m_val = inv_world_matrix_data[i + j * 4];
+      pt1[i] += m_val * in_ptr[j];
+      pt2[i] += m_val * out_ptr[j];
+    }
+  }
+
+  float s[4];
+  for (int i = 0; i < 4; ++i)
+    s[i] = pt2[i] - pt1[i];
+
+  float l = sqrtf(s[0] * s[0] + s[1] * s[1] + s[2] * s[2]);
+  if (l == 0) {
+    d_out_counts[idx] = count;
+    d_out_lengths[idx] = tot_len;
+    return;
+  }
+
+  float L[3];
+  L[0] = fabsf(l / s[0]);
+  L[1] = fabsf(l / s[1]);
+  L[2] = fabsf(l / s[2]);
+
+  float offset[3];
+  for (int i = 0; i < 3; ++i) {
+    float fpt_i = floorf(pt1[i]);
+    offset[i] = (s[i] >= 0) ? (1.0f - (pt1[i] - fpt_i)) : (pt1[i] - fpt_i);
+    offset[i] = fabsf(offset[i] * L[i]);
+  }
+
+  int vid[3] = {(int)floorf(pt1[0]), (int)floorf(pt1[1]), (int)floorf(pt1[2])};
+  int vid_out[3] = {(int)floorf(pt2[0]), (int)floorf(pt2[1]),
+                    (int)floorf(pt2[2])};
+
+  float scale_arr[3] = {scale0, scale1, scale2};
+
+  if (vid[0] == vid_out[0] && vid[1] == vid_out[1] && vid[2] == vid_out[2]) {
+    if (vid[0] >= 0 && vid[0] < dim0 && vid[1] >= 0 && vid[1] < dim1 &&
+        vid[2] >= 0 && vid[2] < dim2) {
+      if (count < max_elements) {
+        int map_id = vid[0] * inc0 + vid[1] * inc1 + vid[2] * inc2;
+        ray_out[count].vox_id = map_id;
+        ray_out[count].L = l;
+        tot_len += l;
+        count++;
+      }
+    }
+    d_out_counts[idx] = count;
+    d_out_lengths[idx] = tot_len;
+    return;
+  }
+
+  int id;
+  float d;
+  while (l > 0) {
+    d = offset[0];
+    id = 0;
+    if (offset[1] < d) {
+      d = offset[1];
+      id = 1;
+    }
+    if (offset[2] < d) {
+      d = offset[2];
+      id = 2;
+    }
+
+    if (vid[0] >= 0 && vid[0] < dim0 && vid[1] >= 0 && vid[1] < dim1 &&
+        vid[2] >= 0 && vid[2] < dim2) {
+      if (count < max_elements) {
+        int map_id = vid[0] * inc0 + vid[1] * inc1 + vid[2] * inc2;
+        ray_out[count].vox_id = map_id;
+        ray_out[count].L = d * scale_arr[id];
+        tot_len += d * scale_arr[id];
+        count++;
+      }
+    }
+
+    float sign_s = (s[id] >= 0) ? 1.0f : -1.0f;
+    vid[id] += (int)sign_s;
+
+    l -= d;
+
+    offset[0] -= d;
+    offset[1] -= d;
+    offset[2] -= d;
+    offset[id] = fminf(L[id], l);
+  }
+  d_out_counts[idx] = count;
+  d_out_lengths[idx] = tot_len;
+}
+
+__global__ void TraceLineKernel(const float *lines_data, int num_lines,
+                                VoxRaytracer::RayData::Element **d_out_elements,
+                                size_t *d_out_counts, float *d_out_lengths,
+                                int max_elements, int dim0, int dim1, int dim2,
+                                const float *inv_world_matrix_data,
+                                float scale0, float scale1, float scale2,
+                                int inc0, int inc1, int inc2) {
+  int idx = blockIdx.x * blockDim.x + threadIdx.x;
+  if (idx >= num_lines)
+    return;
+
+  VoxRaytracer::RayData::Element *ray_out = d_out_elements[idx];
+  size_t count = 0;
+  float tot_len = 0.0f;
+
+  const float *line_ptr = &lines_data[idx * 8];
+  float o_vec[4] = {line_ptr[0], line_ptr[1], line_ptr[2], line_ptr[3]};
+  float d_vec[4] = {line_ptr[4], line_ptr[5], line_ptr[6], line_ptr[7]};
+
+  float pt[4] = {0, 0, 0, 0}, s[4] = {0, 0, 0, 0};
+  for (int i = 0; i < 4; ++i) {
+    for (int j = 0; j < 4; ++j) {
+      float m_val = inv_world_matrix_data[i + j * 4];
+      pt[i] += m_val * o_vec[j];
+      s[i] += m_val * d_vec[j];
+    }
+  }
+
+  float l = sqrtf(s[0] * s[0] + s[1] * s[1] + s[2] * s[2]);
+  if (l == 0) {
+    d_out_counts[idx] = count;
+    d_out_lengths[idx] = tot_len;
+    return;
+  }
+
+  float L[3];
+  L[0] = fabsf(l / s[0]);
+  L[1] = fabsf(l / s[1]);
+  L[2] = fabsf(l / s[2]);
+
+  float offset[3];
+  for (int i = 0; i < 3; ++i) {
+    float fpt_i = floorf(pt[i]);
+    offset[i] = (s[i] >= 0) ? (1.0f - (pt[i] - fpt_i)) : (pt[i] - fpt_i);
+    offset[i] = fabsf(offset[i] * L[i]);
+  }
+
+  int id;
+  float d;
+  int vid[3] = {(int)floorf(pt[0]), (int)floorf(pt[1]), (int)floorf(pt[2])};
+  float scale_arr[3] = {scale0, scale1, scale2};
+
+  while (vid[0] >= 0 && vid[0] < dim0 && vid[1] >= 0 && vid[1] < dim1 &&
+         vid[2] >= 0 && vid[2] < dim2) {
+    d = offset[0];
+    id = 0;
+    if (offset[1] < d) {
+      d = offset[1];
+      id = 1;
+    }
+    if (offset[2] < d) {
+      d = offset[2];
+      id = 2;
+    }
+
+    if (count < max_elements) {
+      int map_id = vid[0] * inc0 + vid[1] * inc1 + vid[2] * inc2;
+      ray_out[count].vox_id = map_id;
+      ray_out[count].L = d * scale_arr[id];
+      tot_len += d * scale_arr[id];
+      count++;
+    }
+
+    float sign_s = (s[id] >= 0) ? 1.0f : -1.0f;
+    vid[id] += (int)sign_s;
+
+    offset[0] -= d;
+    offset[1] -= d;
+    offset[2] -= d;
+    offset[id] = L[id];
+  }
+  d_out_counts[idx] = count;
+  d_out_lengths[idx] = tot_len;
+}
+
+#endif // __CUDACC__
+
+inline void VoxRaytracer::TraceLineCUDA(const HLine3f *lines, size_t num_lines,
+                                        RayData *out_rays,
+                                        int max_elements_per_ray,
+                                        float *kernel_time_ms) {
+  if (num_lines == 0)
+    return;
+
+  float *d_lines = nullptr;
+  bool alloc_lines = false;
+
+  cudaPointerAttributes ptr_attr;
+  cudaPointerGetAttributes(&ptr_attr, lines);
+  if (ptr_attr.type == cudaMemoryTypeDevice) {
+    d_lines = (float *)lines;
+  } else {
+    alloc_lines = true;
+    size_t lines_size = num_lines * sizeof(HLine3f);
+    cudaMalloc(&d_lines, lines_size);
+    cudaMemcpy(d_lines, lines, lines_size, cudaMemcpyHostToDevice);
+  }
+
+  std::vector<RayData::Element *> h_out_elements(num_lines);
+  for (size_t i = 0; i < num_lines; ++i) {
+    out_rays[i].Data().resize(max_elements_per_ray);
+    out_rays[i].Data().MoveToVRAM();
+    h_out_elements[i] = out_rays[i].Data().GetVRAMData();
+  }
+
+  RayData::Element **d_out_elements;
+  cudaMalloc(&d_out_elements, num_lines * sizeof(RayData::Element *));
+  cudaMemcpy(d_out_elements, h_out_elements.data(),
+             num_lines * sizeof(RayData::Element *), cudaMemcpyHostToDevice);
+
+  size_t *d_out_counts;
+  float *d_out_lengths;
+  cudaMalloc(&d_out_counts, num_lines * sizeof(size_t));
+  cudaMalloc(&d_out_lengths, num_lines * sizeof(float));
+
+  int threadsPerBlock = 256;
+  int blocksPerGrid = (num_lines + threadsPerBlock - 1) / threadsPerBlock;
+
+  Vector3i dims = m_Image->GetDims();
+  Vector3i incs = m_Image->GetIncrements();
+  Matrix4f inv_world_matrix = m_Image->GetWorldMatrix().inverse();
+
+  float *d_inv_world;
+  cudaMalloc(&d_inv_world, 16 * sizeof(float));
+  cudaMemcpy(d_inv_world, inv_world_matrix.data(), 16 * sizeof(float),
+             cudaMemcpyHostToDevice);
+
+#ifdef __CUDACC__
+  cudaEvent_t start, stop;
+  if (kernel_time_ms) {
+    cudaEventCreate(&start);
+    cudaEventCreate(&stop);
+    cudaEventRecord(start);
+  }
+
+  TraceLineKernel<<<blocksPerGrid, threadsPerBlock>>>(
+      d_lines, num_lines, d_out_elements, d_out_counts, d_out_lengths,
+      max_elements_per_ray, dims(0), dims(1), dims(2), d_inv_world, m_scale(0),
+      m_scale(1), m_scale(2), incs(0), incs(1), incs(2));
+
+  if (kernel_time_ms) {
+    cudaEventRecord(stop);
+    cudaEventSynchronize(stop);
+    cudaEventElapsedTime(kernel_time_ms, start, stop);
+    cudaEventDestroy(start);
+    cudaEventDestroy(stop);
+  } else {
+    cudaDeviceSynchronize();
+  }
+
+  cudaError_t err = cudaGetLastError();
+  if (err != cudaSuccess) {
+    std::cerr << "CUDA Error in TraceLineCUDA: " << cudaGetErrorString(err)
+              << std::endl;
+  }
+#else
+  std::cerr << "TraceLineKernel requires NVCC!" << std::endl;
+#endif
+
+  std::vector<size_t> h_out_counts(num_lines);
+  std::vector<float> h_out_lengths(num_lines);
+  cudaMemcpy(h_out_counts.data(), d_out_counts, num_lines * sizeof(size_t),
+             cudaMemcpyDeviceToHost);
+  cudaMemcpy(h_out_lengths.data(), d_out_lengths, num_lines * sizeof(float),
+             cudaMemcpyDeviceToHost);
+
+  for (size_t i = 0; i < num_lines; ++i) {
+    out_rays[i].SetCount(h_out_counts[i]);
+    out_rays[i].SetTotalLength(h_out_lengths[i]);
+  }
+
+  if (alloc_lines) {
+    cudaFree(d_lines);
+  }
+  cudaFree(d_out_elements);
+  cudaFree(d_out_counts);
+  cudaFree(d_out_lengths);
+  cudaFree(d_inv_world);
+}
+
+inline void VoxRaytracer::TraceBetweenPointsCUDA(
+    const HPoint3f *in_pts, const HPoint3f *out_pts, size_t num_lines,
+    RayData *out_rays, int max_elements_per_ray, float *kernel_time_ms) {
+  if (num_lines == 0)
+    return;
+
+  float *d_in_pts = nullptr;
+  float *d_out_pts = nullptr;
+  bool alloc_pts = false;
+
+  cudaPointerAttributes ptr_attr;
+  cudaPointerGetAttributes(&ptr_attr, in_pts);
+  if (ptr_attr.type == cudaMemoryTypeDevice) {
+    d_in_pts = (float *)in_pts;
+    d_out_pts = (float *)out_pts;
+  } else {
+    alloc_pts = true;
+    size_t pts_size = num_lines * sizeof(HPoint3f);
+    cudaMalloc(&d_in_pts, pts_size);
+    cudaMalloc(&d_out_pts, pts_size);
+    cudaMemcpy(d_in_pts, in_pts, pts_size, cudaMemcpyHostToDevice);
+    cudaMemcpy(d_out_pts, out_pts, pts_size, cudaMemcpyHostToDevice);
+  }
+
+  std::vector<RayData::Element *> h_out_elements(num_lines);
+  for (size_t i = 0; i < num_lines; ++i) {
+    out_rays[i].Data().resize(max_elements_per_ray);
+    out_rays[i].Data().MoveToVRAM();
+    h_out_elements[i] = out_rays[i].Data().GetVRAMData();
+  }
+
+  RayData::Element **d_out_elements;
+  cudaMalloc(&d_out_elements, num_lines * sizeof(RayData::Element *));
+  cudaMemcpy(d_out_elements, h_out_elements.data(),
+             num_lines * sizeof(RayData::Element *), cudaMemcpyHostToDevice);
+
+  size_t *d_out_counts;
+  float *d_out_lengths;
+  cudaMalloc(&d_out_counts, num_lines * sizeof(size_t));
+  cudaMalloc(&d_out_lengths, num_lines * sizeof(float));
+
+  int threadsPerBlock = 256;
+  int blocksPerGrid = (num_lines + threadsPerBlock - 1) / threadsPerBlock;
+
+  Vector3i dims = m_Image->GetDims();
+  Vector3i incs = m_Image->GetIncrements();
+  Matrix4f inv_world_matrix = m_Image->GetWorldMatrix().inverse();
+
+  float *d_inv_world;
+  cudaMalloc(&d_inv_world, 16 * sizeof(float));
+  cudaMemcpy(d_inv_world, inv_world_matrix.data(), 16 * sizeof(float),
+             cudaMemcpyHostToDevice);
+
+#ifdef __CUDACC__
+  cudaEvent_t start, stop;
+  if (kernel_time_ms) {
+    cudaEventCreate(&start);
+    cudaEventCreate(&stop);
+    cudaEventRecord(start);
+  }
+
+  TraceBetweenPointsKernel<<<blocksPerGrid, threadsPerBlock>>>(
+      d_in_pts, d_out_pts, num_lines, d_out_elements, d_out_counts,
+      d_out_lengths, max_elements_per_ray, dims(0), dims(1), dims(2),
+      d_inv_world, m_scale(0), m_scale(1), m_scale(2), incs(0), incs(1),
+      incs(2));
+
+  if (kernel_time_ms) {
+    cudaEventRecord(stop);
+    cudaEventSynchronize(stop);
+    cudaEventElapsedTime(kernel_time_ms, start, stop);
+    cudaEventDestroy(start);
+    cudaEventDestroy(stop);
+  } else {
+    cudaDeviceSynchronize();
+  }
+
+  cudaError_t err = cudaGetLastError();
+  if (err != cudaSuccess) {
+    std::cerr << "CUDA Error in TraceBetweenPointsCUDA: "
+              << cudaGetErrorString(err) << std::endl;
+  }
+#else
+  std::cerr << "TraceBetweenPointsKernel requires NVCC!" << std::endl;
+#endif
+
+  std::vector<size_t> h_out_counts(num_lines);
+  std::vector<float> h_out_lengths(num_lines);
+  cudaMemcpy(h_out_counts.data(), d_out_counts, num_lines * sizeof(size_t),
+             cudaMemcpyDeviceToHost);
+  cudaMemcpy(h_out_lengths.data(), d_out_lengths, num_lines * sizeof(float),
+             cudaMemcpyDeviceToHost);
+
+  for (size_t i = 0; i < num_lines; ++i) {
+    out_rays[i].SetCount(h_out_counts[i]);
+    out_rays[i].SetTotalLength(h_out_lengths[i]);
+  }
+
+  if (alloc_pts) {
+    cudaFree(d_in_pts);
+    cudaFree(d_out_pts);
+  }
+  cudaFree(d_out_elements);
+  cudaFree(d_out_counts);
+  cudaFree(d_out_lengths);
+  cudaFree(d_inv_world);
+}
+
+} // namespace uLib
+
+#endif // USE_CUDA
+
+#endif // VOXRAYTRACERCUDA_H

src/Math/testing/CMakeLists.txt

@@ -5,6 +5,7 @@ set(TESTS
         ContainerBoxTest
         VoxImageTest
         VoxRaytracerTest
+        VoxRaytracerTestExtended
         StructuredDataTest
         VoxImageFilterTest
         PolicyTest
@@ -22,3 +23,8 @@ set(LIBRARIES
 )
 
 uLib_add_tests(Math)
+
+if(USE_CUDA)
+    set_source_files_properties(VoxImageTest.cpp VoxImageCopyTest.cpp VoxImageFilterTest.cpp VoxRaytracerTest.cpp VoxRaytracerTestExtended.cpp PROPERTIES LANGUAGE CUDA)
+    set_source_files_properties(VoxRaytracerTest.cpp VoxRaytracerTestExtended.cpp PROPERTIES CXX_STANDARD 17 CUDA_STANDARD 17)
+endif()

src/Math/testing/ContainerBoxTest.cpp

@@ -31,6 +31,7 @@
 #include "Math/Dense.h"
 #include "Math/ContainerBox.h"
 
+#include <cmath>
 #include <iostream>
 #include <math.h>
 
@@ -52,41 +53,82 @@ int main()
 
   BEGIN_TESTING(Math ContainerBox);
 
-  ContainerBox Cnt;
-
-//  // Local transform:
-  Cnt.SetOrigin(Vector3f(-1,-1,-1));
-  Cnt.SetSize(Vector3f(2,2,2)); // scaling //
-  std::cout << "Container scale is: " << Cnt.GetSize().transpose() << "\n";
-  std::cout << "Container scale is: " << Cnt.GetSize().transpose() << "\n";
-  TEST0( Vector4f0(Cnt.GetSize().homogeneous() - HVector3f(2,2,2)) );
+  {
+    ContainerBox Cnt;
+    Cnt.SetOrigin(Vector3f(0,0,0));
+    Cnt.SetSize(Vector3f(2,2,2));
+    TEST0( Vector4f0(Cnt.GetOrigin().homogeneous() - HVector3f(0,0,0)) );
+    TEST0( Vector4f0(Cnt.GetSize().homogeneous() - HVector3f(2,2,2)) );
 
 
-  ContainerBox Box;
+    HPoint3f pt = Cnt.GetLocalPoint(HPoint3f(0,0,0));
+    HPoint3f wp = Cnt.GetWorldPoint(pt);  
+    TEST0( Vector4f0(wp - HPoint3f(0,0,0)) );
 
-  Box.SetPosition(Vector3f(1,1,1));
-  Box.SetSize(Vector3f(2,2,2));
-  Box.EulerYZYRotate(Vector3f(0,0,0));
-  HPoint3f pt = Box.GetLocalPoint(HPoint3f(2,3,2));
-  HPoint3f wp = Box.GetWorldPoint(pt);
-  TEST0( Vector4f0(wp - HPoint3f(2,3,2)) );
+    HPoint3f pt2 = Cnt.GetLocalPoint(HPoint3f(2,2,2));
+    HPoint3f wp2 = Cnt.GetWorldPoint(pt2);  
+    TEST0( Vector4f0(wp2 - HPoint3f(2,2,2)) );
 
+    HPoint3f pt3 = Cnt.GetLocalPoint(HPoint3f(1,1,1));
+    HPoint3f wp3 = Cnt.GetWorldPoint(pt3);  
+    TEST0( Vector4f0(wp3 - HPoint3f(1,1,1)) );
 
-////  // Global
-//  Cnt.SetPosition(Vector3f(1,1,1));
-//  Cnt.EulerYZYRotate(Vector3f(M_PI_2,M_PI_2,0));
-//  HPoint3f p = Cnt.GetWorldPoint(1,1,1);
-//  //std::cout << p.transpose() << "\n";
-//  TEST0( Vector4f0(p - HVector3f(2,1,2)) );
-//  p = Cnt.GetWorldPoint(1,2,3);
-//  //std::cout << p.transpose() << "\n";
-//  TEST0( Vector4f0(p - HVector3f(4,1,3)) );
+    HPoint3f pt4 = Cnt.GetLocalPoint(HPoint3f(1,2,3));
+    HPoint3f wp4 = Cnt.GetWorldPoint(pt4);  
+    TEST0( Vector4f0(wp4 - HPoint3f(1,2,3)) );
+  }
 
+  {
+    ContainerBox Cnt;
+    Cnt.SetOrigin(Vector3f(0,0,0));
+    Cnt.SetSize(Vector3f(2,2,2));
+    Cnt.EulerYZYRotate(Vector3f(M_PI,0,0));
+    
+    HPoint3f pt = Cnt.GetLocalPoint(HPoint3f(0,0,0));
+    HPoint3f wp = Cnt.GetWorldPoint(pt);  
+    TEST0( Vector4f0(wp - HPoint3f(0,0,0)) );
 
-//  // scaling //
+    HPoint3f pt2 = Cnt.GetLocalPoint(HPoint3f(2,2,2));
+    HPoint3f wp2 = Cnt.GetWorldPoint(pt2);  
+    TEST0( Vector4f0(wp2 - HPoint3f(2,2,2)) );
 
+    pt2 = HPoint3f(1,1,1);
+    wp2 = Cnt.GetWorldPoint(pt2);  
+    TEST0( Vector4f0(wp2 - HPoint3f(-2,2,-2)) );
 
+    pt2 = HPoint3f(1,2,3);
+    wp2 = Cnt.GetWorldPoint(pt2);  
+    TEST0( Vector4f0(wp2 - HPoint3f(-2,4,-6)) );
+  }
 
+  {
+    ContainerBox Cnt;
+    Cnt.SetOrigin(Vector3f(-1,-1,-1));
+    Cnt.SetSize(Vector3f(2,2,2)); // scaling //
+
+    HPoint3f pt2 = HPoint3f(.5,.5,.5);
+    HPoint3f wp2 = Cnt.GetWorldPoint(pt2);  
+    TEST0( Vector4f0(wp2 - HPoint3f(0,0,0)) );
+
+    pt2 = HPoint3f(0,0,0);
+    wp2 = Cnt.GetWorldPoint(pt2);  
+    TEST0( Vector4f0(wp2 - HPoint3f(-1,-1,-1)) );
+
+    Cnt.EulerYZYRotate(Vector3f(M_PI,0,0));
+    pt2 = HPoint3f(0,0,0);
+    wp2 = Cnt.GetWorldPoint(pt2);  
+    TEST0( Vector4f0(wp2 - HPoint3f(1,-1,1)) );
+  }
+
+  {
+    ContainerBox Box;
+    Box.SetPosition(Vector3f(1,1,1));
+    Box.SetSize(Vector3f(2,2,2));
+    Box.EulerYZYRotate(Vector3f(0,0,0));
+    HPoint3f pt = Box.GetLocalPoint(HPoint3f(2,3,2));
+    HPoint3f wp = Box.GetWorldPoint(pt);
+    TEST0( Vector4f0(wp - HPoint3f(2,3,2)) );
+  }
 
   END_TESTING;
 }

src/Math/testing/VoxImageCopyTest.cpp

@@ -23,55 +23,44 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
 #include "testing-prototype.h"
 
 #include "Math/VoxImage.h"
 
 using namespace uLib;
 
-
 struct TestVoxel {
-    Scalarf Value;
-    unsigned int Count;
+  Scalarf Value;
+  unsigned int Count;
 };
 
 int main() {
-    BEGIN_TESTING(Math VoxImage Copy);
+  BEGIN_TESTING(Math VoxImage Copy);
 
-    {
-        VoxImage<TestVoxel> img(Vector3i(10,10,10));
-        TestVoxel zero = {0,0};
-        img.InitVoxels(zero);
-        TestVoxel nonzero = {5.552368, 0};
-        img[Vector3i(5,1,7)] = nonzero;
-        img[img.Find(HPoint3f(3,3,3))].Value = 5.552369;
-        TEST1( img.GetValue(Vector3i(5,1,7)) == 5.552368f );
+  {
+    VoxImage<TestVoxel> img(Vector3i(10, 10, 10));
+    TestVoxel zero = {0.f, 0};
+    img.InitVoxels(zero);
+    TestVoxel nonzero = {5.552368f, 0};
+    img[Vector3i(5, 1, 7)] = nonzero;
+    img[img.Find(HPoint3f(3, 3, 3))].Value = 5.552369;
+    TEST1(img.GetValue(Vector3i(5, 1, 7)) == 5.552368f);
 
+    img.SetOrigin(Vector3f(4, 5, 6));
 
-        img.SetOrigin(Vector3f(4,5,6));
+    std::cout << "\n";
 
-        std::cout << "\n";
+    img.PrintSelf(std::cout);
 
-        img.PrintSelf(std::cout);
+    VoxImage<TestVoxel> img2 = img;
+    img2.PrintSelf(std::cout);
 
-        VoxImage<TestVoxel> img2 = img;
-        img2.PrintSelf(std::cout);
+    TEST1(img.GetOrigin() == img2.GetOrigin());
+    TEST1(img.GetSpacing() == img2.GetSpacing());
 
-        TEST1( img.GetOrigin() == img2.GetOrigin() );
-        TEST1( img.GetSpacing() == img2.GetSpacing() );
+    img2 = img;
+  }
 
-        img2 = img;
-
-    }
-
-
-
-
-
-
-
-    std::cout << "returns " << _fail << "\n";
-    END_TESTING;
+  std::cout << "returns " << _fail << "\n";
+  END_TESTING;
 }

src/Math/testing/VoxImageFilterTest.cpp

@@ -23,128 +23,191 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
-
-#include "testing-prototype.h"
 #include "Math/StructuredGrid.h"
+#include "testing-prototype.h"
 
 #include "Math/VoxImage.h"
 #include "Math/VoxImageFilter.h"
 
-
-
 using namespace uLib;
 
 struct TestVoxel {
-    Scalarf Value;
-    unsigned int Count;
+  Scalarf Value;
+  unsigned int Count;
 };
 
-float GaussianShape(float d)
-{
-    // normalized manually .. fix //
-    return 4.5 * exp(-d * 4.5);
+float GaussianShape(float d) {
+  // normalized manually .. fix //
+  return 4.5 * exp(-d * 4.5);
 }
 
-
 class GaussianShapeClass : public Interface::VoxImageFilterShape {
 public:
-    GaussianShapeClass(float sigma) :
-        m_sigma(sigma)
-    {}
+  GaussianShapeClass(float sigma) : m_sigma(sigma) {}
 
-    float operator ()(float d) {
-        return (1/m_sigma) * exp(-d/m_sigma);
-    }
+  float operator()(float d) { return (1 / m_sigma) * exp(-d / m_sigma); }
 
 private:
-    float m_sigma;
+  float m_sigma;
 };
 
-
-static float MaxInVector(const std::vector<float> &v)
-{
-    float max = 0;
-    for(int i=0; i<v.size(); ++i)
-        if(v.at(i) > max) max = v.at(i);
-    return max;
+static float MaxInVector(const std::vector<float> &v) {
+  float max = 0;
+  for (int i = 0; i < v.size(); ++i)
+    if (v.at(i) > max)
+      max = v.at(i);
+  return max;
 }
 
+int main() {
+  BEGIN_TESTING(VoxImageFilters);
 
-int main()
-{
-    BEGIN_TESTING(VoxImageFilters);
+  VoxImage<TestVoxel> image(Vector3i(20, 30, 40));
+  image[Vector3i(10, 10, 10)].Value = 1;
+  // image[Vector3i(10,10,8)].Value = 1;
+  image.ExportToVtk("test_filter_original.vtk", 0);
 
-    VoxImage<TestVoxel> image(Vector3i(20,30,40));
-    image[Vector3i(10,10,10)].Value = 1;
-    //image[Vector3i(10,10,8)].Value = 1;
-    image.ExportToVtk("test_filter_original.vtk",0);
+  ////////////////////////////////////////////////////////////////////////////
+  ////////////////////////////////////////////////////////////////////////////
+  ////////////////////////////////////////////////////////////////////////////
+  // RPS //
 
+  {
+    VoxFilterAlgorithmSPR<TestVoxel> filter(Vector3i(2, 3, 4));
 
+    VoxImage<TestVoxel> filtered = image;
 
-    ////////////////////////////////////////////////////////////////////////////
-    ////////////////////////////////////////////////////////////////////////////
-    ////////////////////////////////////////////////////////////////////////////
-    // RPS //
+    std::vector<float> values;
+    for (int i = 0; i < filter.GetKernelData().GetDims().prod(); ++i) {
+      values.push_back(1.);
+      std::cout << values[i] << " ";
+    }
+    std::cout << "\n";
 
-    {
-        VoxFilterAlgorithmSPR<TestVoxel> filter(Vector3i(2,3,4));
+    filter.SetImage(&filtered);
 
-        VoxImage<TestVoxel> filtered = image;
+    filter.SetKernelNumericXZY(values);
 
-        std::vector<float> values;
-        for(int i=0; i < filter.GetKernelData().GetDims().prod(); ++i) {
-            values.push_back(1.);
-            std::cout << values[i] << " ";
-        }
-        std::cout << "\n";
+    filter.SetABTrim(0, 2);
 
-        filter.SetImage(&filtered);
+    filter.GetKernelData().PrintSelf(std::cout);
 
-        filter.SetKernelNumericXZY(values);
+    filter.Run();
 
-        filter.SetABTrim(0,2);
+    filtered.ExportToVtk("filter_RPS_out.vtk", 0);
+  }
 
-        filter.GetKernelData().PrintSelf(std::cout);
+  {
 
-        filter.Run();
+    VoxImage<TestVoxel> image(Vector3i(20, 30, 40));
+    image[Vector3i(10, 10, 10)].Value = 1;
+    image[Vector3i(9, 10, 8)].Value = 2;
+    image.ExportToVtk("test_filter_max_original.vtk", 0);
 
-        filtered.ExportToVtk("filter_RPS_out.vtk",0);
+    VoxFilterAlgorithmCustom<TestVoxel> filter(Vector3i(3, 3, 4));
+
+    std::vector<float> values;
+    for (int i = 0; i < filter.GetKernelData().GetDims().prod(); ++i) {
+      values.push_back(static_cast<float>(1));
     }
 
+    filter.SetImage(&image);
 
+    filter.SetKernelNumericXZY(values);
 
+    filter.SetCustomEvaluate(MaxInVector);
 
+    filter.Run();
 
-    {
+    image.ExportToVtk("test_filter_max.vtk", 0);
+  }
 
-        VoxImage<TestVoxel> image(Vector3i(20,30,40));
-        image[Vector3i(10,10,10)].Value = 1;
-        image[Vector3i(9,10,8)].Value = 2;
-        image.ExportToVtk("test_filter_max_original.vtk",0);
+  ////////////////////////////////////////////////////////////////////////////
+  // CUDA Allocator Transfer Test //
+  {
+    VoxImage<TestVoxel> image(Vector3i(10, 10, 10));
+    image[Vector3i(5, 5, 5)].Value = 1;
 
+    VoxFilterAlgorithmLinear<TestVoxel> filter(Vector3i(3, 3, 3));
 
-
-        VoxFilterAlgorithmCustom<TestVoxel> filter(Vector3i(3,3,4));
-
-        std::vector<float> values;
-        for(int i=0; i < filter.GetKernelData().GetDims().prod(); ++i) {
-            values.push_back(static_cast<float>(1));
-        }
-
-        filter.SetImage(&image);
-
-        filter.SetKernelNumericXZY(values);
-
-        filter.SetCustomEvaluate(MaxInVector);
-
-        filter.Run();
-
-        image.ExportToVtk("test_filter_max.vtk",0);
+    std::vector<float> values;
+    for (int i = 0; i < filter.GetKernelData().GetDims().prod(); ++i) {
+      values.push_back(1.0f);
     }
 
+    filter.SetImage(&image);
+    filter.SetKernelNumericXZY(values);
 
+    // Move the kernel data and image data to VRAM to simulate CUDA transfer
+    filter.GetKernelData().Data().MoveToVRAM();
+    image.Data().MoveToVRAM();
 
-    END_TESTING;
+    // Validate devices
+    if (filter.GetKernelData().Data().GetDevice() != MemoryDevice::VRAM ||
+        image.Data().GetDevice() != MemoryDevice::VRAM) {
+#ifdef USE_CUDA
+      std::cerr << "Failed to move memory to VRAM." << std::endl;
+#else
+      std::cout << "DataAllocator correctly simulates VRAM without crashing."
+                << std::endl;
+#endif
+    }
+
+    // Run the filter; The fallback CPU filter will trigger MoveToRAM
+    // behind the scenes inside Convolve / Evaluate.
+    filter.Run();
+
+    // Assert it came back to RAM if evaluated on CPU
+    if (image.Data().GetDevice() != MemoryDevice::RAM) {
+#ifdef USE_CUDA
+      std::cout << "Data correctly stayed in VRAM after CUDA execution!"
+                << std::endl;
+#else
+      std::cout << "Data correctly stayed in RAM simulation." << std::endl;
+#endif
+    }
+
+    image.ExportToVtk("test_filter_cuda_transfer.vtk", 0);
+  }
+
+  ////////////////////////////////////////////////////////////////////////////
+  // CUDA ABTrim Allocator Transfer Test //
+  {
+    VoxImage<TestVoxel> image(Vector3i(10, 10, 10));
+    image[Vector3i(5, 5, 5)].Value = 10;
+    image[Vector3i(5, 5, 6)].Value = 2; // Test trimming
+
+    VoxFilterAlgorithmAbtrim<TestVoxel> filter(Vector3i(3, 3, 3));
+
+    std::vector<float> values;
+    for (int i = 0; i < filter.GetKernelData().GetDims().prod(); ++i) {
+      values.push_back(1.0f);
+    }
+
+    filter.SetImage(&image);
+    filter.SetKernelNumericXZY(values);
+    filter.SetABTrim(1, 1); // trim highest and lowest
+
+    // Move the kernel data and image data to VRAM to simulate CUDA transfer
+    filter.GetKernelData().Data().MoveToVRAM();
+    image.Data().MoveToVRAM();
+
+    // Run the filter
+    filter.Run();
+
+    // Ensure data stays on device if CUDA was toggled
+    if (image.Data().GetDevice() != MemoryDevice::RAM) {
+#ifdef USE_CUDA
+      std::cout << "ABTrim correctly stayed in VRAM after CUDA execution!"
+                << std::endl;
+#else
+      std::cout << "ABTrim Data correctly stayed in RAM simulation."
+                << std::endl;
+#endif
+    }
+
+    image.ExportToVtk("test_filter_abtrim_cuda_transfer.vtk", 0);
+  }
+
+  END_TESTING;
 }

src/Math/testing/VoxImageTest.cpp

@@ -23,99 +23,91 @@
 
 //////////////////////////////////////////////////////////////////////////////*/
 
-
-
-
-#include "testing-prototype.h"
-#include "Math/StructuredGrid.h"
 #include "Math/VoxImage.h"
+#include "Math/StructuredGrid.h"
+#include "testing-prototype.h"
 
 using namespace uLib;
 
-
 struct TestVoxel {
-    Scalarf Value;
-    unsigned int Count;
+  Scalarf Value;
+  unsigned int Count;
 };
 
 int main() {
-    BEGIN_TESTING(Math StructuredGrid);
+  BEGIN_TESTING(Math StructuredGrid);
 
-    {   // SIMPLE TESTS //
-        StructuredGrid img(Vector3i(10,10,10));
-        img.SetSpacing(Vector3f(3,3,3));
-        TEST1( img.GetWorldPoint(2,0,0) == HPoint3f(6,0,0) );
-        TEST1( img.GetWorldPoint(1,1,1) == HPoint3f(3,3,3) );
+  { // SIMPLE TESTS //
+    StructuredGrid img(Vector3i(10, 10, 10));
+    img.SetSpacing(Vector3f(3, 3, 3));
+    TEST1(img.GetWorldPoint(2, 0, 0) == HPoint3f(6, 0, 0));
+    TEST1(img.GetWorldPoint(1, 1, 1) == HPoint3f(3, 3, 3));
 
-        img.SetPosition(Vector3f(1,1,1));
-        TEST1( img.GetWorldPoint(1,1,1) == HPoint3f(4,4,4) );
-        TEST1( img.GetLocalPoint(4,4,4) == HPoint3f(1,1,1) );
+    img.SetPosition(Vector3f(1, 1, 1));
+    TEST1(img.GetWorldPoint(1, 1, 1) == HPoint3f(4, 4, 4));
+    TEST1(img.GetLocalPoint(4, 4, 4) == HPoint3f(1, 1, 1));
 
-        TEST0( img.IsInsideBounds(HPoint3f(5,33,-5)));
-        TEST0( img.IsInsideBounds(HPoint3f(0,0,0)));
-        TEST1( img.IsInsideBounds(HPoint3f(1,1,1)));
-    }
+    TEST0(img.IsInsideBounds(HPoint3f(5, 33, -5)));
+    TEST0(img.IsInsideBounds(HPoint3f(0, 0, 0)));
+    TEST1(img.IsInsideBounds(HPoint3f(1, 1, 1)));
+  }
 
-    {   // TEST WITH ORIGIN //
-        StructuredGrid img(Vector3i(10,10,10));
-        img.SetSpacing(Vector3f(3,3,3));
-        img.SetOrigin(Vector3f(-1,1,-1));
-        img.SetPosition(Vector3f(1,1,1));
-        TEST1( img.GetWorldPoint(1,1,1) == HPoint3f(3,5,3) );
-    }
+  { // TEST WITH ORIGIN //
+    StructuredGrid img(Vector3i(10, 10, 10));
+    img.SetSpacing(Vector3f(3, 3, 3));
+    img.SetOrigin(Vector3f(-1, 1, -1));
+    img.SetPosition(Vector3f(1, 1, 1));
+    TEST1(img.GetWorldPoint(1, 1, 1) == HPoint3f(3, 5, 3));
+  }
 
+  {
+    VoxImage<TestVoxel> img(Vector3i(10, 10, 10));
+    TestVoxel zero = {0.f, 0};
+    img.InitVoxels(zero);
+    TestVoxel nonzero = {5.552368f, 0};
+    img[Vector3i(5, 1, 7)] = nonzero;
+    img[img.Find(HPoint3f(3, 3, 3))].Value = 5.552369;
+    img.ExportToVtk("./test_vox_image.vtk", 0);
+    img.ExportToVtkXml("./test_vox_image.vti", 0);
+    TEST1(img.GetValue(Vector3i(5, 1, 7)) == 5.552368f);
+  }
 
-    {
-        VoxImage<TestVoxel> img(Vector3i(10,10,10));
-        TestVoxel zero = {0,0};
-        img.InitVoxels(zero);
-        TestVoxel nonzero = {5.552368, 0};
-        img[Vector3i(5,1,7)] = nonzero;
-        img[img.Find(HPoint3f(3,3,3))].Value = 5.552369;
-        img.ExportToVtk("./test_vox_image.vtk",0);
-        img.ExportToVtkXml("./test_vox_image.vti",0);
-        TEST1( img.GetValue(Vector3i(5,1,7)) == 5.552368f );
-    }
+  {
+    VoxImage<TestVoxel> img(Vector3i(4, 4, 4));
+    TestVoxel zero = {0.f, 0};
+    img.InitVoxels(zero);
+    img.SetSpacing(Vector3f(2, 2, 2));
+    img.SetPosition(Vector3f(-4, -4, -4));
+    TEST1(img.GetWorldPoint(img.GetLocalPoint(HPoint3f(5, 5, 5))) ==
+          HPoint3f(5, 5, 5));
+  }
 
-    {
-        VoxImage<TestVoxel> img(Vector3i(4,4,4));
-        TestVoxel zero = {0,0};
-        img.InitVoxels(zero);
-        img.SetSpacing(Vector3f(2,2,2));
-        img.SetPosition(Vector3f(-4,-4,-4));
-        TEST1( img.GetWorldPoint(img.GetLocalPoint(HPoint3f(5,5,5))) == HPoint3f(5,5,5));
-    }
+  {
+    VoxImage<TestVoxel> imgR(Vector3i(0, 0, 0));
+    imgR.ImportFromVtk("./test_vox_image.vtk");
+    imgR.ExportToVtk("./read_and_saved.vtk");
+  }
 
-    {
-        VoxImage<TestVoxel> imgR(Vector3i(0,0,0));
-        imgR.ImportFromVtk("./test_vox_image.vtk");
-        imgR.ExportToVtk("./read_and_saved.vtk");
-    }
-
-    {
-        VoxImage<TestVoxel> img(Vector3i(4,4,4));
-        img.InitVoxels({0,0});
-        for (int i=0; i<4; i++) {
-            for (int j=0; j<4; j++) {
-                for (int k=0; k<4; k++) {
-                    img[Vector3i(i,j,k)] = {i+j+k,0};
-                }
-            }
+  {
+    VoxImage<TestVoxel> img(Vector3i(4, 4, 4));
+    img.InitVoxels({0.f, 0});
+    for (int i = 0; i < 4; i++) {
+      for (int j = 0; j < 4; j++) {
+        for (int k = 0; k < 4; k++) {
+          img[Vector3i(i, j, k)] = {static_cast<float>(i + j + k), 0};
         }
-        img.ExportToVti("./vti_saved.vti",0,1);
-        // img.ImportFromVtkXml("./test_vox_image.vti");
+      }
     }
+    img.ExportToVti("./vti_saved.vti", 0, 1);
+    // img.ImportFromVtkXml("./test_vox_image.vti");
+  }
 
+  {
+    VoxImage<TestVoxel> img1(Vector3i(5, 5, 5));
+    VoxImage<TestVoxel> img2;
+    img2 = img1;
+    TEST1(img1.GetDims() == img2.GetDims());
+  }
 
-    {
-        VoxImage<TestVoxel> img1(Vector3i(5,5,5));
-        VoxImage<TestVoxel> img2;
-        img2 = img1;
-        TEST1( img1.GetDims() == img2.GetDims() );
-    }
-
-
-
-
-    END_TESTING
+  END_TESTING
 }

src/Math/testing/VoxRaytracerTest.cpp

@@ -48,6 +48,11 @@ int Vector4f0(Vector4f c) {
 
 typedef VoxRaytracer Raytracer;
 
+struct TestVoxel {
+  float Value;
+  int Count;
+};
+
 int main() {
   BEGIN_TESTING(Math VoxRaytracer);
 
@@ -89,7 +94,8 @@ int main() {
 
     Raytracer::RayData rdata =
         ray.TraceBetweenPoints(HPoint3f(-3, -3, -3), HPoint3f(3, 3, 3));
-    for (const Raytracer::RayData::Element &el : rdata.Data()) {
+    for (size_t i = 0; i < rdata.Count(); ++i) {
+      const Raytracer::RayData::Element &el = rdata.Data()[i];
       std::cout << " " << el.vox_id << " , " << el.L << "\n";
     }
   }
@@ -100,7 +106,7 @@ int main() {
     Raytracer rt(img);
 
     Raytracer::RayData ray = rt.TraceBetweenPoints(pt1, pt2);
-    TEST1(ray.Data().size() == 2);
+    TEST1(ray.Count() == 2);
     TEST1(ray.Data().at(0).vox_id == 6);
     TEST1(ray.Data().at(1).vox_id == 7);
     ray.PrintSelf(std::cout);
@@ -112,7 +118,7 @@ int main() {
     Raytracer rt(img);
 
     Raytracer::RayData ray = rt.TraceBetweenPoints(pt1, pt2);
-    TEST1(ray.Data().size() == 2);
+    TEST1(ray.Count() == 2);
     TEST1(ray.Data().at(0).vox_id == 6);
     TEST1(ray.Data().at(1).vox_id == 4);
     ray.PrintSelf(std::cout);
@@ -124,7 +130,7 @@ int main() {
     Raytracer rt(img);
 
     Raytracer::RayData ray = rt.TraceBetweenPoints(pt1, pt2);
-    TEST1(ray.Data().size() == 4);
+    TEST1(ray.Count() == 4);
     TEST1(ray.Data().at(0).vox_id == 6);
     TEST1(ray.Data().at(1).vox_id == 4);
     TEST1(ray.Data().at(2).vox_id == 5);
@@ -132,5 +138,81 @@ int main() {
     ray.PrintSelf(std::cout);
   }
 
+#ifdef USE_CUDA
+  {
+    std::cout << "\n--- Testing CUDA Raytracer Accumulator ---\n";
+
+    Raytracer rt(img);
+
+    {
+      HPoint3f pt1(1, -0.5, 1);
+      HPoint3f pt2(1, 4.5, 1);
+      HPoint3f pts1[1] = {pt1};
+      HPoint3f pts2[1] = {pt2};
+      Raytracer::RayData ray_cuda[1];
+      rt.TraceBetweenPointsCUDA(pts1, pts2, 1, ray_cuda);
+      TEST1(ray_cuda[0].Count() == 2);
+      TEST1(ray_cuda[0].Data().at(0).vox_id == 6);
+      TEST1(ray_cuda[0].Data().at(1).vox_id == 7);
+    }
+
+    {
+      HPoint3f pt1(5, 1, 1);
+      HPoint3f pt2(-3, 1, 1);
+      HPoint3f pts1[1] = {pt1};
+      HPoint3f pts2[1] = {pt2};
+      Raytracer::RayData ray_cuda[1];
+      rt.TraceBetweenPointsCUDA(pts1, pts2, 1, ray_cuda);
+      TEST1(ray_cuda[0].Count() == 2);
+      TEST1(ray_cuda[0].Data().at(0).vox_id == 6);
+      TEST1(ray_cuda[0].Data().at(1).vox_id == 4);
+    }
+
+    {
+      HPoint3f pt1(1, 1, 1);
+      HPoint3f pt2(-1, 3, -1);
+      HPoint3f pts1[1] = {pt1};
+      HPoint3f pts2[1] = {pt2};
+      Raytracer::RayData ray_cuda[1];
+      rt.TraceBetweenPointsCUDA(pts1, pts2, 1, ray_cuda);
+      TEST1(ray_cuda[0].Count() == 4);
+      TEST1(ray_cuda[0].Data().at(0).vox_id == 6);
+      TEST1(ray_cuda[0].Data().at(1).vox_id == 4);
+      TEST1(ray_cuda[0].Data().at(2).vox_id == 5);
+      TEST1(ray_cuda[0].Data().at(3).vox_id == 1);
+    }
+
+    VoxImage<TestVoxel> img_cuda(Vector3i(4, 4, 4));
+    img_cuda.SetSpacing(Vector3f(2, 2, 2));
+    img_cuda.SetPosition(Vector3f(-4, -4, -4));
+
+    Raytracer ray(img_cuda);
+
+    HLine3f line1;
+    line1.origin << -3, -3, -3, 1;
+    line1.direction << 1, 1, 1, 0;
+
+    HLine3f line2;
+    line2.origin << -3, -3, 1, 1;
+    line2.direction << 1, 1, -1, 0;
+
+    HLine3f lines[2] = {line1, line2};
+    // Execute CUDA kernel wrapper over target VoxImage mapped internally into
+    // VRAM
+    ray.AccumulateLinesCUDA(lines, 2, img_cuda);
+
+    // Validate device synchronization returned data correctly pulling back to
+    // host
+    TEST1(img_cuda.Data().GetDevice() !=
+          MemoryDevice::RAM); // Confirms VRAM executed
+
+    // Pull down checking values
+    float l_val = img_cuda[img_cuda.Find(Vector4f(-3, -3, -3, 1))].Value;
+    std::cout << "Accumulated Voxel test trace point length returned: " << l_val
+              << "\n";
+    TEST1(l_val > 0.1f);
+  }
+#endif
+
   END_TESTING
 }

src/Math/testing/VoxRaytracerTestExtended.cpp

@@ -0,0 +1,211 @@
+/*//////////////////////////////////////////////////////////////////////////////
+// CMT Cosmic Muon Tomography project //////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////
+
+  Copyright (c) 2014, Universita' degli Studi di Padova, INFN sez. di Padova
+  All rights reserved
+
+  Authors: Andrea Rigoni Garola < andrea.rigoni@pd.infn.it >
+
+  ------------------------------------------------------------------
+  This library is free software;  you  can  redistribute  it  and/or
+  modify it  under the  terms  of  the  GNU  Lesser  General  Public
+  License as published  by  the  Free  Software  Foundation;  either
+  version 3.0 of the License, or (at your option) any later version.
+
+  This library is  distributed in  the hope that it will  be useful,
+  but  WITHOUT ANY WARRANTY;  without  even  the implied warranty of
+  MERCHANTABILITY  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+  Lesser General Public License for more details.
+
+  You should have received a copy of  the GNU Lesser General  Public
+  License along with this library.
+
+//////////////////////////////////////////////////////////////////////////////*/
+
+#include "Math/StructuredGrid.h"
+#include "Math/VoxRaytracer.h"
+#include "testing-prototype.h"
+#include <chrono>
+#include <iostream>
+#include <random>
+
+using namespace uLib;
+
+typedef VoxRaytracer Raytracer;
+
+int main() {
+  BEGIN_TESTING(Math VoxRaytracer Extended Benchmark);
+
+  std::cout << "\n=============================================\n";
+  std::cout << "  VoxRaytracer CPU vs CUDA Benchmark Test\n";
+  std::cout << "=============================================\n\n";
+
+  // Create a 100x100x100 grid (1 million voxels)
+  StructuredGrid img(Vector3i(100, 100, 100));
+  img.SetSpacing(Vector3f(1.0f, 1.0f, 1.0f));
+  img.SetPosition(Vector3f(-50.0f, -50.0f, -50.0f));
+
+  Raytracer rt(img);
+
+  const size_t NUM_RAYS = 100000;
+  std::cout << "Generating " << NUM_RAYS
+            << " random ray pairs across a 100x100x100 grid...\n";
+
+  std::vector<HPoint3f> in_pts(NUM_RAYS);
+  std::vector<HPoint3f> out_pts(NUM_RAYS);
+
+  // Use a fixed seed for reproducible tests
+  std::random_device rd;
+  std::mt19937 gen(rd());
+  // The grid spans from -50 to 50 on each axis
+  std::uniform_real_distribution<float> dist(-49.9f, 49.9f);
+  // Pick a random face for in/out to ensure rays cross the volume
+  std::uniform_int_distribution<int> face_dist(0, 5);
+
+  for (size_t i = 0; i < NUM_RAYS; ++i) {
+    HPoint3f p1, p2;
+    // Generate point 1 on a random face
+    int f1 = face_dist(gen);
+    p1(0) = (f1 == 0) ? -50.0f : (f1 == 1) ? 50.0f : dist(gen);
+    p1(1) = (f1 == 2) ? -50.0f : (f1 == 3) ? 50.0f : dist(gen);
+    p1(2) = (f1 == 4) ? -50.0f : (f1 == 5) ? 50.0f : dist(gen);
+    p1(3) = 1.0f;
+
+    // Generate point 2 on a different face
+    int f2;
+    do {
+      f2 = face_dist(gen);
+    } while (
+        f1 == f2 ||
+        f1 / 2 ==
+            f2 / 2); // Avoid same face or opposite face trivially if desired
+
+    p2(0) = (f2 == 0) ? -50.0f : (f2 == 1) ? 50.0f : dist(gen);
+    p2(1) = (f2 == 2) ? -50.0f : (f2 == 3) ? 50.0f : dist(gen);
+    p2(2) = (f2 == 4) ? -50.0f : (f2 == 5) ? 50.0f : dist(gen);
+    p2(3) = 1.0f;
+
+    in_pts[i] = p1;
+    out_pts[i] = p2;
+  }
+
+  std::vector<Raytracer::RayData> cpu_results(NUM_RAYS);
+
+  std::cout << "\nRunning CPU Raytracing...\n";
+  auto start_cpu = std::chrono::high_resolution_clock::now();
+
+  for (size_t i = 0; i < NUM_RAYS; ++i) {
+    cpu_results[i] = rt.TraceBetweenPoints(in_pts[i], out_pts[i]);
+  }
+
+  auto end_cpu = std::chrono::high_resolution_clock::now();
+  std::chrono::duration<double, std::milli> cpu_ms = end_cpu - start_cpu;
+  std::cout << "CPU Execution Time: " << cpu_ms.count() << " ms\n";
+
+#ifdef USE_CUDA
+  std::vector<Raytracer::RayData> cuda_results(NUM_RAYS);
+  int max_elements_per_ray =
+      400; // 100x100x100 grid max trace length usually ~300 items
+
+  std::cout << "\nPre-Allocating Data to VRAM...\n";
+  // Pre-allocate input and output points to VRAM
+  HPoint3f *d_in_pts;
+  HPoint3f *d_out_pts;
+  size_t pts_size = NUM_RAYS * sizeof(HPoint3f);
+  cudaMalloc(&d_in_pts, pts_size);
+  cudaMalloc(&d_out_pts, pts_size);
+  cudaMemcpy(d_in_pts, in_pts.data(), pts_size, cudaMemcpyHostToDevice);
+  cudaMemcpy(d_out_pts, out_pts.data(), pts_size, cudaMemcpyHostToDevice);
+
+  // Pre-allocate elements output arrays in VRAM via DataAllocator
+  for (size_t i = 0; i < NUM_RAYS; ++i) {
+    cuda_results[i].Data().resize(max_elements_per_ray);
+    cuda_results[i].Data().MoveToVRAM();
+  }
+
+  std::cout << "Running CUDA Raytracing...\n";
+  auto start_cuda = std::chrono::high_resolution_clock::now();
+
+  float kernel_time_ms = 0.0f;
+  rt.TraceBetweenPointsCUDA(d_in_pts, d_out_pts, NUM_RAYS, cuda_results.data(),
+                            max_elements_per_ray, &kernel_time_ms);
+
+  auto end_cuda = std::chrono::high_resolution_clock::now();
+  std::chrono::duration<double, std::milli> cuda_ms = end_cuda - start_cuda;
+
+  // Free explicit input pointers
+  cudaFree(d_in_pts);
+  cudaFree(d_out_pts);
+
+  // Also query memory usage info
+  size_t free_byte;
+  size_t total_byte;
+  cudaMemGetInfo(&free_byte, &total_byte);
+  double free_db = (double)free_byte / (1024.0 * 1024.0);
+  double total_db = (double)total_byte / (1024.0 * 1024.0);
+  double used_db = total_db - free_db;
+
+  std::cout << "CUDA Kernel Exec Time: " << kernel_time_ms << " ms\n";
+  std::cout << "CUDA Total Time (API): " << cuda_ms.count() << " ms\n";
+  std::cout << "CUDA Total Time Spdup: " << (cpu_ms.count() / cuda_ms.count())
+            << "x\n";
+  if (kernel_time_ms > 0.0f) {
+    std::cout << "CUDA Kernel Speedup  : " << (cpu_ms.count() / kernel_time_ms)
+              << "x\n";
+  }
+  std::cout << "CUDA VRAM Usage Est. : " << used_db << " MB out of " << total_db
+            << " MB total\n";
+
+  std::cout << "\nVerifying CUDA results against CPU...\n";
+  size_t mismatches = 0;
+  for (size_t i = 0; i < NUM_RAYS; ++i) {
+    const auto &cpu_ray = cpu_results[i];
+    const auto &cuda_ray = cuda_results[i];
+
+    if (cpu_ray.Count() != cuda_ray.Count() ||
+        std::abs(cpu_ray.TotalLength() - cuda_ray.TotalLength()) > 1e-3) {
+      if (mismatches < 5) {
+        std::cout << "Mismatch at ray " << i
+                  << ": CPU count=" << cpu_ray.Count()
+                  << ", len=" << cpu_ray.TotalLength()
+                  << " vs CUDA count=" << cuda_ray.Count()
+                  << ", len=" << cuda_ray.TotalLength() << "\n";
+      }
+      mismatches++;
+      continue;
+    }
+
+    // Check elements
+    for (size_t j = 0; j < cpu_ray.Count(); ++j) {
+      if (cpu_ray.Data()[j].vox_id != cuda_ray.Data()[j].vox_id ||
+          std::abs(cpu_ray.Data()[j].L - cuda_ray.Data()[j].L) > 1e-3) {
+        if (mismatches < 5) {
+          std::cout << "Mismatch at ray " << i << ", element " << j
+                    << ": CPU id=" << cpu_ray.Data()[j].vox_id
+                    << ", L=" << cpu_ray.Data()[j].L
+                    << " vs CUDA id=" << cuda_ray.Data()[j].vox_id
+                    << ", L=" << cuda_ray.Data()[j].L << "\n";
+        }
+        mismatches++;
+        break;
+      }
+    }
+  }
+
+  if (mismatches == 0) {
+    std::cout << "SUCCESS! All " << NUM_RAYS
+              << " rays perfectly match between CPU and CUDA.\n";
+  } else {
+    std::cout << "FAILED! " << mismatches << " rays contain mismatched data.\n";
+  }
+
+  TEST1(mismatches == 0);
+
+#else
+  std::cout << "\nUSE_CUDA is not defined. Skipping CUDA benchmarking.\n";
+#endif
+
+  std::cout << "=============================================\n";
+  END_TESTING
+}

src/ParticlePhysics/Geant/Makefile.am

@@ -1,14 +0,0 @@
-include $(top_srcdir)/Common.am
-
-library_includedir = $(includedir)/libmutom-${PACKAGE_VERSION}/ParticlePhysics/Geant
-
-library_include_HEADERS =
-
-_PPGEANT_SOURCES =
-
-
-
-noinst_LTLIBRARIES = libmutomppgeant.la
-libmutomppgeant_la_SOURCES = ${_PPGEANT_SOURCES}
-
-

src/Python/CMakeLists.txt

@@ -0,0 +1,58 @@
+set(HEADERS "")
+
+set(SOURCES
+    module.cpp
+    core_bindings.cpp
+    math_bindings.cpp
+    math_filters_bindings.cpp
+)
+
+# Use pybind11 to add the python module
+pybind11_add_module(uLib_python module.cpp core_bindings.cpp math_bindings.cpp math_filters_bindings.cpp)
+
+# Link against our C++ libraries
+target_link_libraries(uLib_python PRIVATE
+    ${PACKAGE_LIBPREFIX}Core
+    ${PACKAGE_LIBPREFIX}Math
+)
+
+# Include directories from Core and Math are automatically handled if target_include_directories were set appropriately,
+# but we might need to manually include them if they aren't INTERFACE includes.
+target_include_directories(uLib_python PRIVATE
+    ${PROJECT_SOURCE_DIR}/src
+    ${PROJECT_BINARY_DIR}
+)
+
+# Install uLib_python within the uLib install target
+install(TARGETS uLib_python
+        EXPORT "uLibTargets"
+        RUNTIME DESTINATION ${INSTALL_BIN_DIR} COMPONENT bin
+        LIBRARY DESTINATION ${INSTALL_LIB_DIR} COMPONENT lib
+        ARCHIVE DESTINATION ${INSTALL_LIB_DIR} COMPONENT lib
+)
+
+# --- Python Tests ---------------------------------------------------------- #
+
+if(BUILD_TESTING)
+    find_package(Python3 COMPONENTS Interpreter REQUIRED)
+
+    add_test(NAME pybind_general
+             COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/testing/pybind_test.py)
+    set_tests_properties(pybind_general PROPERTIES 
+                         ENVIRONMENT "PYTHONPATH=$<TARGET_FILE_DIR:uLib_python>:${PROJECT_SOURCE_DIR}/src/Python")
+
+    add_test(NAME pybind_core 
+             COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/testing/core_pybind_test.py)
+    set_tests_properties(pybind_core PROPERTIES 
+                         ENVIRONMENT "PYTHONPATH=$<TARGET_FILE_DIR:uLib_python>:${PROJECT_SOURCE_DIR}/src/Python")
+
+    add_test(NAME pybind_math 
+             COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/testing/math_pybind_test.py)
+    set_tests_properties(pybind_math PROPERTIES 
+                         ENVIRONMENT "PYTHONPATH=$<TARGET_FILE_DIR:uLib_python>:${PROJECT_SOURCE_DIR}/src/Python")
+
+    add_test(NAME pybind_math_filters 
+             COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/testing/math_filters_test.py)
+    set_tests_properties(pybind_math_filters PROPERTIES 
+                         ENVIRONMENT "PYTHONPATH=$<TARGET_FILE_DIR:uLib_python>:${PROJECT_SOURCE_DIR}/src/Python")
+endif()

src/Python/core_bindings.cpp

@@ -0,0 +1,30 @@
+#include <pybind11/pybind11.h>
+#include <pybind11/stl.h>
+
+#include "Core/Object.h"
+#include "Core/Timer.h"
+#include "Core/Options.h"
+#include "Core/Uuid.h"
+
+namespace py = pybind11;
+using namespace uLib;
+
+void init_core(py::module_ &m) {
+    py::class_<Object, std::shared_ptr<Object>>(m, "Object")
+        .def(py::init<>())
+        .def("DeepCopy", &Object::DeepCopy);
+
+    py::class_<Timer>(m, "Timer")
+        .def(py::init<>())
+        .def("Start", &Timer::Start)
+        .def("StopWatch", &Timer::StopWatch);
+
+    py::class_<Options>(m, "Options")
+        .def(py::init<const char*>(), py::arg("str") = "Program options")
+        .def("parse_config_file", py::overload_cast<const char*>(&Options::parse_config_file))
+        .def("save_config_file", &Options::save_config_file)
+        .def("count", &Options::count);
+
+    py::class_<TypeRegister>(m, "TypeRegister")
+        .def_static("Controller", &TypeRegister::Controller, py::return_value_policy::reference_internal);
+}

src/Python/math_bindings.cpp

@@ -0,0 +1,384 @@
+#include <pybind11/pybind11.h>
+#include <pybind11/eigen.h>
+#include <pybind11/stl.h>
+#include <pybind11/stl_bind.h>
+
+#include <pybind11/numpy.h>
+
+#include "Math/Dense.h"
+#include "Math/Transform.h"
+#include "Math/Geometry.h"
+#include "Math/ContainerBox.h"
+#include "Math/StructuredData.h"
+#include "Math/StructuredGrid.h"
+#include "Math/Structured2DGrid.h"
+#include "Math/Structured4DGrid.h"
+#include "Math/TriangleMesh.h"
+#include "Math/VoxRaytracer.h"
+#include "Math/Accumulator.h"
+#include "Math/VoxImage.h"
+
+namespace py = pybind11;
+using namespace uLib;
+
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Scalari>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Scalarui>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Scalarl>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Scalarul>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Scalarf>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Scalard>);
+
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Vector3f>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Vector3i>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Vector4f>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Vector4i>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Vector3d>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Vector4d>);
+
+PYBIND11_MAKE_OPAQUE(uLib::Vector<Voxel>);
+PYBIND11_MAKE_OPAQUE(uLib::Vector<VoxRaytracer::RayData::Element>);
+
+template <typename MatrixType>
+void bind_eigen_type(py::module_ &m, const char *name) {
+    using Scalar = typename MatrixType::Scalar;
+    constexpr bool is_vector = MatrixType::IsVectorAtCompileTime;
+
+    // Default constructor (zeros)
+    m.def(name, []() -> MatrixType {
+        if constexpr (MatrixType::RowsAtCompileTime == Eigen::Dynamic || MatrixType::ColsAtCompileTime == Eigen::Dynamic) {
+            return MatrixType(); // Empty dynamic matrix
+        } else {
+            return MatrixType::Zero(); // Zero static matrix
+        }
+    });
+
+    // Specialized constructor for dynamic matrices
+    if constexpr (MatrixType::RowsAtCompileTime == Eigen::Dynamic || MatrixType::ColsAtCompileTime == Eigen::Dynamic) {
+        m.def(name, [](int rows, int cols) -> MatrixType {
+            MatrixType mat;
+            mat.setZero(rows, cols);
+            return mat;
+        });
+    }
+
+    // Initialize from list
+    m.def(name, [](py::list l) -> MatrixType {
+        MatrixType mat;
+        if constexpr (is_vector) {
+            mat.setZero(l.size());
+            for (size_t i = 0; i < l.size(); ++i) {
+                mat(i) = l[i].cast<Scalar>();
+            }
+        } else {
+            int rows = MatrixType::RowsAtCompileTime == Eigen::Dynamic ? (int)std::sqrt(l.size()) : MatrixType::RowsAtCompileTime;
+            int cols = MatrixType::ColsAtCompileTime == Eigen::Dynamic ? (int)std::sqrt(l.size()) : MatrixType::ColsAtCompileTime;
+            mat.setZero(rows, cols);
+            for (size_t i = 0; i < (size_t)l.size(); ++i) {
+                mat(i / cols, i % cols) = l[i].cast<Scalar>();
+            }
+        }
+        return mat;
+    });
+
+    // Initialize from py::array
+    m.def(name, [](py::array_t<Scalar, py::array::c_style | py::array::forcecast> arr) -> MatrixType {
+        auto buf = arr.request();
+        MatrixType mat;
+        if constexpr (is_vector) {
+            mat.setZero(buf.size);
+            Scalar* ptr = static_cast<Scalar*>(buf.ptr);
+            for (ssize_t i = 0; i < buf.size; ++i) mat(i) = ptr[i];
+        } else {
+            int rows = buf.shape.size() > 0 ? (int)buf.shape[0] : 1;
+            int cols = buf.shape.size() > 1 ? (int)buf.shape[1] : 1;
+            mat.setZero(rows, cols);
+            Scalar* ptr = static_cast<Scalar*>(buf.ptr);
+            for (int i = 0; i < rows; ++i) {
+                for (int j = 0; j < cols; ++j) {
+                    mat(i, j) = ptr[i * cols + j];
+                }
+            }
+        }
+        return mat;
+    });
+}
+
+void init_math(py::module_ &m) {
+
+    // 1. Basic Eigen Types (Vectors and Matrices)
+    bind_eigen_type<Vector1f>(m, "Vector1f");
+    bind_eigen_type<Vector2f>(m, "Vector2f");
+    bind_eigen_type<Vector3f>(m, "Vector3f");
+    bind_eigen_type<Vector4f>(m, "Vector4f");
+    bind_eigen_type<Vector1i>(m, "Vector1i");
+    bind_eigen_type<Vector2i>(m, "Vector2i");
+    bind_eigen_type<Vector3i>(m, "Vector3i");
+    bind_eigen_type<Vector4i>(m, "Vector4i");
+    bind_eigen_type<Vector1d>(m, "Vector1d");
+    bind_eigen_type<Vector2d>(m, "Vector2d");
+    bind_eigen_type<Vector3d>(m, "Vector3d");
+    bind_eigen_type<Vector4d>(m, "Vector4d");
+
+    bind_eigen_type<Matrix2f>(m, "Matrix2f");
+    bind_eigen_type<Matrix3f>(m, "Matrix3f");
+    bind_eigen_type<Matrix4f>(m, "Matrix4f");
+    bind_eigen_type<Matrix2i>(m, "Matrix2i");
+    bind_eigen_type<Matrix3i>(m, "Matrix3i");
+    bind_eigen_type<Matrix4i>(m, "Matrix4i");
+    bind_eigen_type<Matrix2d>(m, "Matrix2d");
+    bind_eigen_type<Matrix3d>(m, "Matrix3d");
+    bind_eigen_type<Matrix4d>(m, "Matrix4d");
+
+    bind_eigen_type<MatrixXi>(m, "MatrixXi");
+    bind_eigen_type<MatrixXf>(m, "MatrixXf");
+    bind_eigen_type<MatrixXd>(m, "MatrixXd");
+
+    // 2. Homogeneous types
+    py::class_<HPoint3f>(m, "HPoint3f")
+        .def(py::init<>())
+        .def(py::init<float, float, float>())
+        .def(py::init<Vector3f &>());
+    py::class_<HVector3f>(m, "HVector3f")
+        .def(py::init<>())
+        .def(py::init<float, float, float>())
+        .def(py::init<Vector3f &>());
+    py::class_<HLine3f>(m, "HLine3f")
+        .def(py::init<>())
+        .def_readwrite("origin", &HLine3f::origin)
+        .def_readwrite("direction", &HLine3f::direction);
+    py::class_<HError3f>(m, "HError3f")
+        .def(py::init<>())
+        .def_readwrite("position_error", &HError3f::position_error)
+        .def_readwrite("direction_error", &HError3f::direction_error);
+
+    // 3. Dynamic Vectors (uLib::Vector)
+    py::bind_vector<uLib::Vector<Scalari>>(m, "Vector_i")
+        .def("MoveToVRAM", &uLib::Vector<Scalari>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Scalari>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Scalarui>>(m, "Vector_ui")
+        .def("MoveToVRAM", &uLib::Vector<Scalarui>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Scalarui>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Scalarl>>(m, "Vector_l")
+        .def("MoveToVRAM", &uLib::Vector<Scalarl>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Scalarl>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Scalarul>>(m, "Vector_ul")
+        .def("MoveToVRAM", &uLib::Vector<Scalarul>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Scalarul>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Scalarf>>(m, "Vector_f")
+        .def("MoveToVRAM", &uLib::Vector<Scalarf>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Scalarf>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Scalard>>(m, "Vector_d")
+        .def("MoveToVRAM", &uLib::Vector<Scalard>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Scalard>::MoveToRAM);
+
+    py::bind_vector<uLib::Vector<Vector3f>>(m, "Vector_Vector3f")
+        .def("MoveToVRAM", &uLib::Vector<Vector3f>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Vector3f>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Vector3i>>(m, "Vector_Vector3i")
+        .def("MoveToVRAM", &uLib::Vector<Vector3i>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Vector3i>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Vector4f>>(m, "Vector_Vector4f")
+        .def("MoveToVRAM", &uLib::Vector<Vector4f>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Vector4f>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Vector4i>>(m, "Vector_Vector4i")
+        .def("MoveToVRAM", &uLib::Vector<Vector4i>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Vector4i>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Vector3d>>(m, "Vector_Vector3d")
+        .def("MoveToVRAM", &uLib::Vector<Vector3d>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Vector3d>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Vector4d>>(m, "Vector_Vector4d")
+        .def("MoveToVRAM", &uLib::Vector<Vector4d>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Vector4d>::MoveToRAM);
+    py::bind_vector<uLib::Vector<Voxel>>(m, "Vector_Voxel")
+        .def("MoveToVRAM", &uLib::Vector<Voxel>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<Voxel>::MoveToRAM);
+    py::bind_vector<uLib::Vector<VoxRaytracer::RayData::Element>>(m, "Vector_VoxRaytracerRayDataElement")
+        .def("MoveToVRAM", &uLib::Vector<VoxRaytracer::RayData::Element>::MoveToVRAM)
+        .def("MoveToRAM", &uLib::Vector<VoxRaytracer::RayData::Element>::MoveToRAM);
+
+    // 4. Accumulators
+    py::class_<Accumulator_Mean<float>>(m, "Accumulator_Mean_f")
+        .def(py::init<>())
+        .def("AddPass", &Accumulator_Mean<float>::AddPass)
+        .def("__call__", py::overload_cast<const float>(&Accumulator_Mean<float>::operator()))
+        .def("__call__", py::overload_cast<>(&Accumulator_Mean<float>::operator(), py::const_));
+
+    py::class_<Accumulator_Mean<double>>(m, "Accumulator_Mean_d")
+        .def(py::init<>())
+        .def("AddPass", &Accumulator_Mean<double>::AddPass)
+        .def("__call__", py::overload_cast<const double>(&Accumulator_Mean<double>::operator()))
+        .def("__call__", py::overload_cast<>(&Accumulator_Mean<double>::operator(), py::const_));
+
+    py::class_<Accumulator_ABTrim<float>>(m, "Accumulator_ABTrim_f")
+        .def(py::init<>())
+        .def("SetABTrim", &Accumulator_ABTrim<float>::SetABTrim)
+        .def("__iadd__", [](Accumulator_ABTrim<float> &self, float val) { self += val; return &self; })
+        .def("__call__", &Accumulator_ABTrim<float>::operator());
+
+    py::class_<Accumulator_ABTrim<double>>(m, "Accumulator_ABTrim_d")
+        .def(py::init<>())
+        .def("SetABTrim", &Accumulator_ABTrim<double>::SetABTrim)
+        .def("__iadd__", [](Accumulator_ABTrim<double> &self, double val) { self += val; return &self; })
+        .def("__call__", &Accumulator_ABTrim<double>::operator());
+
+    py::class_<Accumulator_ABClip<float>>(m, "Accumulator_ABClip_f")
+        .def(py::init<>())
+        .def("SetABTrim", &Accumulator_ABClip<float>::SetABTrim)
+        .def("__iadd__", [](Accumulator_ABClip<float> &self, float val) { self += val; return &self; })
+        .def("__call__", &Accumulator_ABClip<float>::operator());
+
+    py::class_<Accumulator_ABClip<double>>(m, "Accumulator_ABClip_d")
+        .def(py::init<>())
+        .def("SetABTrim", &Accumulator_ABClip<double>::SetABTrim)
+        .def("__iadd__", [](Accumulator_ABClip<double> &self, double val) { self += val; return &self; })
+        .def("__call__", &Accumulator_ABClip<double>::operator());
+
+    // 5. Core Math Structures
+    py::class_<AffineTransform>(m, "AffineTransform")
+        .def(py::init<>())
+        .def("GetWorldMatrix", &AffineTransform::GetWorldMatrix)
+        .def("SetPosition", &AffineTransform::SetPosition)
+        .def("GetPosition", &AffineTransform::GetPosition)
+        .def("Translate", &AffineTransform::Translate)
+        .def("Scale", &AffineTransform::Scale)
+        .def("SetRotation", &AffineTransform::SetRotation)
+        .def("GetRotation", &AffineTransform::GetRotation)
+        .def("Rotate", &AffineTransform::Rotate)
+        .def("EulerYZYRotate", &AffineTransform::EulerYZYRotate)
+        .def("FlipAxes", &AffineTransform::FlipAxes);
+
+    py::class_<Geometry, AffineTransform>(m, "Geometry")
+        .def(py::init<>())
+        .def("GetWorldPoint", &Geometry::GetWorldPoint)
+        .def("GetLocalPoint", &Geometry::GetLocalPoint);
+
+    py::class_<ContainerBox, AffineTransform>(m, "ContainerBox")
+        .def(py::init<>())
+        .def("SetOrigin", &ContainerBox::SetOrigin)
+        .def("GetOrigin", &ContainerBox::GetOrigin)
+        .def("SetSize", &ContainerBox::SetSize)
+        .def("GetSize", &ContainerBox::GetSize)
+        .def("GetWorldMatrix", &ContainerBox::GetWorldMatrix)
+        .def("GetWorldPoint", &ContainerBox::GetWorldPoint)
+        .def("GetLocalPoint", &ContainerBox::GetLocalPoint);
+
+    py::enum_<StructuredData::_Order>(m, "StructuredDataOrder")
+        .value("CustomOrder", StructuredData::CustomOrder)
+        .value("XYZ", StructuredData::XYZ)
+        .value("XZY", StructuredData::XZY)
+        .value("YXZ", StructuredData::YXZ)
+        .value("YZX", StructuredData::YZX)
+        .value("ZXY", StructuredData::ZXY)
+        .value("ZYX", StructuredData::ZYX)
+        .export_values();
+
+    py::class_<StructuredData>(m, "StructuredData")
+        .def(py::init<const Vector3i &>())
+        .def("GetDims", &StructuredData::GetDims)
+        .def("SetDims", &StructuredData::SetDims)
+        .def("GetIncrements", &StructuredData::GetIncrements)
+        .def("SetIncrements", &StructuredData::SetIncrements)
+        .def("SetDataOrder", &StructuredData::SetDataOrder)
+        .def("GetDataOrder", &StructuredData::GetDataOrder)
+        .def("IsInsideGrid", &StructuredData::IsInsideGrid)
+        .def("Map", &StructuredData::Map)
+        .def("UnMap", &StructuredData::UnMap);
+
+    py::class_<StructuredGrid, ContainerBox, StructuredData>(m, "StructuredGrid")
+        .def(py::init<const Vector3i &>())
+        .def("SetSpacing", &StructuredGrid::SetSpacing)
+        .def("GetSpacing", &StructuredGrid::GetSpacing)
+        .def("IsInsideBounds", &StructuredGrid::IsInsideBounds)
+        .def("Find", [](StructuredGrid &self, Vector3f pt) {
+            return self.Find(HPoint3f(pt));
+        });
+
+    py::class_<Structured2DGrid>(m, "Structured2DGrid")
+        .def(py::init<>())
+        .def("SetDims", &Structured2DGrid::SetDims)
+        .def("GetDims", &Structured2DGrid::GetDims)
+        .def("IsInsideGrid", &Structured2DGrid::IsInsideGrid)
+        .def("Map", &Structured2DGrid::Map)
+        .def("UnMap", &Structured2DGrid::UnMap)
+        .def("SetPhysicalSpace", &Structured2DGrid::SetPhysicalSpace)
+        .def("GetSpacing", &Structured2DGrid::GetSpacing)
+        .def("GetOrigin", &Structured2DGrid::GetOrigin)
+        .def("IsInsideBounds", &Structured2DGrid::IsInsideBounds)
+        .def("PhysicsToUnitSpace", &Structured2DGrid::PhysicsToUnitSpace)
+        .def("UnitToPhysicsSpace", &Structured2DGrid::UnitToPhysicsSpace)
+        .def("SetDebug", &Structured2DGrid::SetDebug);
+
+    py::class_<Structured4DGrid>(m, "Structured4DGrid")
+        .def(py::init<>())
+        .def("SetDims", &Structured4DGrid::SetDims)
+        .def("GetDims", &Structured4DGrid::GetDims)
+        .def("IsInsideGrid", &Structured4DGrid::IsInsideGrid)
+        .def("Map", &Structured4DGrid::Map)
+        .def("UnMap", &Structured4DGrid::UnMap)
+        .def("SetPhysicalSpace", &Structured4DGrid::SetPhysicalSpace)
+        .def("GetSpacing", &Structured4DGrid::GetSpacing)
+        .def("GetOrigin", &Structured4DGrid::GetOrigin)
+        .def("IsInsideBounds", &Structured4DGrid::IsInsideBounds)
+        .def("PhysicsToUnitSpace", &Structured4DGrid::PhysicsToUnitSpace)
+        .def("UnitToPhysicsSpace", &Structured4DGrid::UnitToPhysicsSpace)
+        .def("SetDebug", &Structured4DGrid::SetDebug);
+
+    // 6. High-level Structures
+    py::class_<Voxel>(m, "Voxel")
+        .def(py::init<>())
+        .def_readwrite("Value", &Voxel::Value)
+        .def_readwrite("Count", &Voxel::Count);
+
+    py::class_<Abstract::VoxImage, StructuredGrid>(m, "AbstractVoxImage")
+        .def("GetValue", py::overload_cast<const Vector3i &>(&Abstract::VoxImage::GetValue, py::const_))
+        .def("GetValue", py::overload_cast<const int>(&Abstract::VoxImage::GetValue, py::const_))
+        .def("SetValue", py::overload_cast<const Vector3i &, float>(&Abstract::VoxImage::SetValue))
+        .def("SetValue", py::overload_cast<const int, float>(&Abstract::VoxImage::SetValue))
+        .def("ExportToVtk", &Abstract::VoxImage::ExportToVtk)
+        .def("ExportToVti", &Abstract::VoxImage::ExportToVti)
+        .def("ImportFromVtk", &Abstract::VoxImage::ImportFromVtk)
+        .def("ImportFromVti", &Abstract::VoxImage::ImportFromVti);
+
+    py::class_<VoxImage<Voxel>, Abstract::VoxImage>(m, "VoxImage")
+        .def(py::init<>())
+        .def(py::init<const Vector3i &>())
+        .def("Data", &VoxImage<Voxel>::Data, py::return_value_policy::reference_internal)
+        .def("InitVoxels", &VoxImage<Voxel>::InitVoxels)
+        .def("Abs", &VoxImage<Voxel>::Abs)
+        .def("clipImage", py::overload_cast<const Vector3i, const Vector3i>(&VoxImage<Voxel>::clipImage, py::const_))
+        .def("clipImage", py::overload_cast<const HPoint3f, const HPoint3f>(&VoxImage<Voxel>::clipImage, py::const_))
+        .def("clipImage", py::overload_cast<const float>(&VoxImage<Voxel>::clipImage, py::const_))
+        .def("maskImage", py::overload_cast<const HPoint3f, const HPoint3f, float>(&VoxImage<Voxel>::maskImage, py::const_))
+        .def("maskImage", py::overload_cast<const float, float, float>(&VoxImage<Voxel>::maskImage, py::const_), py::arg("threshold"), py::arg("belowValue") = 0, py::arg("aboveValue") = 0)
+        .def("fixVoxels", py::overload_cast<const float, float>(&VoxImage<Voxel>::fixVoxels, py::const_))
+        .def("__getitem__", py::overload_cast<unsigned int>(&VoxImage<Voxel>::operator[]))
+        .def("__getitem__", py::overload_cast<const Vector3i &>(&VoxImage<Voxel>::operator[]));
+
+    py::class_<TriangleMesh>(m, "TriangleMesh")
+        .def(py::init<>())
+        .def("AddPoint", &TriangleMesh::AddPoint)
+        .def("AddTriangle", py::overload_cast<const Vector3i &>(&TriangleMesh::AddTriangle))
+        .def("Points", &TriangleMesh::Points, py::return_value_policy::reference_internal)
+        .def("Triangles", &TriangleMesh::Triangles, py::return_value_policy::reference_internal);
+
+    py::class_<VoxRaytracer::RayData::Element>(m, "VoxRaytracerRayDataElement")
+        .def(py::init<>())
+        .def_readwrite("vox_id", &VoxRaytracer::RayData::Element::vox_id)
+        .def_readwrite("L", &VoxRaytracer::RayData::Element::L);
+
+    py::class_<VoxRaytracer::RayData>(m, "VoxRaytracerRayData")
+        .def(py::init<>())
+        .def("AppendRay", &VoxRaytracer::RayData::AppendRay)
+        .def("Data", py::overload_cast<>(&VoxRaytracer::RayData::Data), py::return_value_policy::reference_internal)
+        .def("Count", &VoxRaytracer::RayData::Count)
+        .def("TotalLength", &VoxRaytracer::RayData::TotalLength)
+        .def("SetCount", &VoxRaytracer::RayData::SetCount)
+        .def("SetTotalLength", &VoxRaytracer::RayData::SetTotalLength);
+
+    py::class_<VoxRaytracer>(m, "VoxRaytracer")
+        .def(py::init<StructuredGrid &>(), py::keep_alive<1, 2>())
+        .def("GetImage", &VoxRaytracer::GetImage, py::return_value_policy::reference_internal)
+        .def("TraceLine", &VoxRaytracer::TraceLine)
+        .def("TraceBetweenPoints", &VoxRaytracer::TraceBetweenPoints);
+
+}

src/Python/math_filters_bindings.cpp

@@ -0,0 +1,100 @@
+#include <pybind11/pybind11.h>
+#include <pybind11/eigen.h>
+#include <pybind11/stl.h>
+
+#include "Math/VoxImage.h"
+#include "Math/VoxImageFilter.h"
+#include "Math/VoxImageFilterLinear.hpp"
+#include "Math/VoxImageFilterABTrim.hpp"
+#include "Math/VoxImageFilterBilateral.hpp"
+#include "Math/VoxImageFilterThreshold.hpp"
+#include "Math/VoxImageFilterMedian.hpp"
+#include "Math/VoxImageFilter2ndStat.hpp"
+#include "Math/VoxImageFilterCustom.hpp"
+
+namespace py = pybind11;
+using namespace uLib;
+
+template <typename Algorithm>
+void bind_common_filter(py::class_<Algorithm, Abstract::VoxImageFilter> &cls) {
+    cls.def(py::init<const Vector3i &>())
+       .def("Run", &Algorithm::Run)
+       .def("SetKernelNumericXZY", &Algorithm::SetKernelNumericXZY)
+       .def("GetImage", &Algorithm::GetImage, py::return_value_policy::reference_internal)
+       .def("SetImage", &Algorithm::SetImage);
+}
+
+void init_math_filters(py::module_ &m) {
+
+    // Abstract::VoxImageFilter
+    py::class_<Abstract::VoxImageFilter, std::unique_ptr<Abstract::VoxImageFilter, py::nodelete>>(m, "AbstractVoxImageFilter")
+        .def("Run", &Abstract::VoxImageFilter::Run)
+        .def("SetImage", &Abstract::VoxImageFilter::SetImage);
+
+    // Helper macro to define standard bindings for a filter
+#define BIND_FILTER(ClassName) \
+    { \
+        auto cls = py::class_<ClassName<Voxel>, Abstract::VoxImageFilter>(m, #ClassName); \
+        bind_common_filter(cls); \
+    }
+
+    // VoxFilterAlgorithmLinear
+    {
+        auto cls = py::class_<VoxFilterAlgorithmLinear<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmLinear");
+        bind_common_filter(cls);
+    }
+
+    // VoxFilterAlgorithmAbtrim
+    {
+        auto cls = py::class_<VoxFilterAlgorithmAbtrim<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmAbtrim");
+        bind_common_filter(cls);
+        cls.def("SetABTrim", &VoxFilterAlgorithmAbtrim<Voxel>::SetABTrim);
+    }
+
+    // VoxFilterAlgorithmSPR
+    {
+        auto cls = py::class_<VoxFilterAlgorithmSPR<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmSPR");
+        bind_common_filter(cls);
+        cls.def("SetABTrim", &VoxFilterAlgorithmSPR<Voxel>::SetABTrim);
+    }
+
+    // VoxFilterAlgorithmBilateral
+    {
+        auto cls = py::class_<VoxFilterAlgorithmBilateral<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmBilateral");
+        bind_common_filter(cls);
+        cls.def("SetIntensitySigma", &VoxFilterAlgorithmBilateral<Voxel>::SetIntensitySigma);
+    }
+
+    // VoxFilterAlgorithmBilateralTrim
+    {
+        auto cls = py::class_<VoxFilterAlgorithmBilateralTrim<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmBilateralTrim");
+        bind_common_filter(cls);
+        cls.def("SetIntensitySigma", &VoxFilterAlgorithmBilateralTrim<Voxel>::SetIntensitySigma);
+        cls.def("SetABTrim", &VoxFilterAlgorithmBilateralTrim<Voxel>::SetABTrim);
+    }
+
+    // VoxFilterAlgorithmThreshold
+    {
+        auto cls = py::class_<VoxFilterAlgorithmThreshold<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmThreshold");
+        bind_common_filter(cls);
+        cls.def("SetThreshold", &VoxFilterAlgorithmThreshold<Voxel>::SetThreshold);
+    }
+
+    // VoxFilterAlgorithmMedian
+    {
+        auto cls = py::class_<VoxFilterAlgorithmMedian<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmMedian");
+        bind_common_filter(cls);
+    }
+
+    // VoxFilterAlgorithm2ndStat
+    {
+        auto cls = py::class_<VoxFilterAlgorithm2ndStat<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithm2ndStat");
+        bind_common_filter(cls);
+    }
+
+    // VoxFilterAlgorithmCustom (Omit CustomEvaluate since it uses static function ptrs)
+    {
+        auto cls = py::class_<VoxFilterAlgorithmCustom<Voxel>, Abstract::VoxImageFilter>(m, "VoxFilterAlgorithmCustom");
+        bind_common_filter(cls);
+    }
+}

src/Python/module.cpp

@@ -0,0 +1,20 @@
+#include <pybind11/pybind11.h>
+
+namespace py = pybind11;
+
+void init_core(py::module_ &m);
+void init_math(py::module_ &m);
+void init_math_filters(py::module_ &m);
+
+PYBIND11_MODULE(uLib_python, m) {
+    m.doc() = "Python bindings for uLib Core and Math libraries";
+    
+    // Core submodule
+    py::module_ core = m.def_submodule("Core", "Core library bindings");
+    init_core(core);
+
+    // Math submodule
+    py::module_ math = m.def_submodule("Math", "Math library bindings");
+    init_math(math);
+    init_math_filters(math);
+}

src/Python/testing/core_pybind_test.py

@@ -0,0 +1,33 @@
+import sys
+import os
+import unittest
+import time
+
+import uLib
+
+class TestCoreOptions(unittest.TestCase):
+    def test_options(self):
+        opt = uLib.Core.Options("Test Options")
+        
+        # Test basic config file parsing
+        with open("test_configuration.ini", "w") as f:
+            f.write("[Section]\n")
+            
+        opt.parse_config_file("test_configuration.ini")
+        os.remove("test_configuration.ini")
+
+class TestCoreObject(unittest.TestCase):
+    def test_object(self):
+        obj = uLib.Core.Object()
+        self.assertIsNotNone(obj)
+
+class TestCoreTimer(unittest.TestCase):
+    def test_timer(self):
+        timer = uLib.Core.Timer()
+        timer.Start()
+        time.sleep(0.1)
+        val = timer.StopWatch()
+        self.assertGreater(val, 0.09)
+
+if __name__ == '__main__':
+    unittest.main()

src/Python/testing/math_filters_test.py

@@ -0,0 +1,151 @@
+import unittest
+import numpy as np
+import os
+import sys
+
+# Ensure PYTHONPATH is correct if run from root
+sys.path.append(os.path.join(os.getcwd(), 'src', 'Python'))
+
+import uLib
+
+class TestMathFilters(unittest.TestCase):
+    def test_filter_creation(self):
+        # 1. Linear Filter
+        dims = [10, 10, 10]
+        v_dims = uLib.Math.Vector3i(dims)
+        
+        linear_filter = uLib.Math.VoxFilterAlgorithmLinear(v_dims)
+        self.assertIsNotNone(linear_filter)
+        
+        # 2. ABTrim Filter
+        abtrim_filter = uLib.Math.VoxFilterAlgorithmAbtrim(v_dims)
+        self.assertIsNotNone(abtrim_filter)
+        abtrim_filter.SetABTrim(1, 1)
+        
+        # 3. Bilateral Filter
+        bilat_filter = uLib.Math.VoxFilterAlgorithmBilateral(v_dims)
+        self.assertIsNotNone(bilat_filter)
+        bilat_filter.SetIntensitySigma(0.5)
+        
+        # 4. Threshold Filter
+        threshold_filter = uLib.Math.VoxFilterAlgorithmThreshold(v_dims)
+        self.assertIsNotNone(threshold_filter)
+        threshold_filter.SetThreshold(0.5)
+        
+        # 5. Median Filter
+        median_filter = uLib.Math.VoxFilterAlgorithmMedian(v_dims)
+        self.assertIsNotNone(median_filter)
+
+    def test_filter_run(self):
+        # Create image
+        dims = [10, 10, 10]
+        vox_img = uLib.Math.VoxImage(dims)
+        for i in range(10*10*10):
+            vox_img.SetValue(i, 1.0)
+        
+        # Linear filter
+        linear_filter = uLib.Math.VoxFilterAlgorithmLinear([3, 3, 3])
+        linear_filter.SetImage(vox_img)
+        
+        # Set kernel (simple 3x3x3 all ones)
+        # Weights are usually normalized in linear filter logic?
+        # Let's just test it runs.
+        linear_filter.SetKernelNumericXZY([1.0] * 27)
+        
+        # Run filter
+        linear_filter.Run()
+        
+        # Value should be 1.0 (mean of all 1.0 is 1.0)
+        self.assertAlmostEqual(vox_img.GetValue(0), 1.0)
+
+    def test_filter_run_abtrim(self):
+        # Create image
+        dims = [10, 10, 10]
+        vox_img = uLib.Math.VoxImage(dims)
+        for i in range(10*10*10):
+            vox_img.SetValue(i, 1.0)
+        
+        # ABTrim filter
+        abtrim_filter = uLib.Math.VoxFilterAlgorithmAbtrim([3, 3, 3])
+        abtrim_filter.SetImage(vox_img)
+        
+        # Set kernel (simple 3x3x3 all ones)
+        # Weights are usually normalized in linear filter logic?
+        # Let's just test it runs.
+        abtrim_filter.SetKernelNumericXZY([1.0] * 27)
+        
+        # Run filter
+        abtrim_filter.Run()
+        
+        # Value should be 1.0 (mean of all 1.0 is 1.0)
+        self.assertAlmostEqual(vox_img.GetValue(0), 1.0)
+
+    def test_filter_run_bilateral(self):
+        # Create image
+        dims = [10, 10, 10]
+        vox_img = uLib.Math.VoxImage(dims)
+        for i in range(10*10*10):
+            vox_img.SetValue(i, 1.0)
+        
+        # Bilateral filter
+        bilat_filter = uLib.Math.VoxFilterAlgorithmBilateral([3, 3, 3])
+        bilat_filter.SetImage(vox_img)
+        
+        # Set kernel (simple 3x3x3 all ones)
+        # Weights are usually normalized in linear filter logic?
+        # Let's just test it runs.
+        bilat_filter.SetKernelNumericXZY([1.0] * 27)
+        
+        # Run filter
+        bilat_filter.Run()
+        
+        # Value should be 1.0 (mean of all 1.0 is 1.0)
+        self.assertAlmostEqual(vox_img.GetValue(0), 1.0)
+
+    def test_filter_run_threshold(self):
+        # Create image
+        dims = [10, 10, 10]
+        vox_img = uLib.Math.VoxImage(dims)
+        for i in range(10*10*10):
+            vox_img.SetValue(i, 1.0)
+        
+        # Threshold filter
+        threshold_filter = uLib.Math.VoxFilterAlgorithmThreshold([3, 3, 3])
+        threshold_filter.SetImage(vox_img)
+        
+        # Set kernel (simple 3x3x3 all ones)
+        # Weights are usually normalized in linear filter logic?
+        # Let's just test it runs.
+        threshold_filter.SetKernelNumericXZY([1.0] * 27)
+        
+        # Run filter
+        threshold_filter.Run()
+        
+        # Value should be 1.0 (mean of all 1.0 is 1.0)
+        self.assertAlmostEqual(vox_img.GetValue(0), 1.0)
+
+    def test_filter_run_median(self):
+        # Create image
+        dims = [10, 10, 10]
+        vox_img = uLib.Math.VoxImage(dims)
+        for i in range(10*10*10):
+            vox_img.SetValue(i, 1.0)
+        
+        # Median filter
+        median_filter = uLib.Math.VoxFilterAlgorithmMedian([3, 3, 3])
+        median_filter.SetImage(vox_img)
+        
+        # Set kernel (simple 3x3x3 all ones)
+        # Weights are usually normalized in linear filter logic?
+        # Let's just test it runs.
+        median_filter.SetKernelNumericXZY([1.0] * 27)
+        
+        # Run filter
+        median_filter.Run()
+        
+        # Value should be 1.0 (mean of all 1.0 is 1.0)
+        self.assertAlmostEqual(vox_img.GetValue(0), 1.0)
+
+
+if __name__ == '__main__':
+    unittest.main()

src/Python/testing/math_pybind_test.py

@@ -0,0 +1,189 @@
+import sys
+import os
+import unittest
+import numpy as np
+
+import uLib
+
+def vector4f0(v, target):
+    diff = np.array(v) - np.array(target)
+    diff[3] = 0 # ignoring w
+    return np.all(np.abs(diff) < 0.001)
+
+
+class TestMathMatrix(unittest.TestCase):
+
+    def test_matrix(self):
+        def check_1234(m2f):
+            self.assertEqual(m2f[0, 0], 1)
+            self.assertEqual(m2f[0, 1], 2)
+            self.assertEqual(m2f[1, 0], 3)
+            self.assertEqual(m2f[1, 1], 4)
+
+        m2f = uLib.Math.Matrix2f()
+        m2f[0, 0] = 1
+        m2f[0, 1] = 2
+        m2f[1, 0] = 3
+        m2f[1, 1] = 4
+        check_1234(m2f)
+
+        m2f = uLib.Math.Matrix2f([1, 2, 3, 4])
+        check_1234(m2f)
+
+        # m2f = uLib.Math.Matrix2f([[1, 2], [3, 4]])
+        # check_1234(m2f)
+
+        m2f = uLib.Math.Matrix2f(np.array([[1, 2], [3, 4]]))
+        check_1234(m2f)
+
+    def test_vector2(self):
+        v2f = uLib.Math.Vector2f()
+        v2f[0] = 1
+        v2f[1] = 2
+        self.assertEqual(v2f[0], 1)
+        self.assertEqual(v2f[1], 2)
+        
+        v2f = uLib.Math.Vector2f([1, 2])
+        self.assertEqual(v2f[0], 1)
+        self.assertEqual(v2f[1], 2)
+        
+        v2f = uLib.Math.Vector2f(np.array([1, 2]))
+        self.assertEqual(v2f[0], 1)
+        self.assertEqual(v2f[1], 2)
+    
+    def test_vector3(self):
+        v3f = uLib.Math.Vector3f()
+        v3f[0] = 1
+        v3f[1] = 2
+        v3f[2] = 3
+        self.assertEqual(v3f[0], 1)
+        self.assertEqual(v3f[1], 2)
+        self.assertEqual(v3f[2], 3)
+        
+        v3f = uLib.Math.Vector3f([1, 2, 3])
+        self.assertEqual(v3f[0], 1)
+        self.assertEqual(v3f[1], 2)
+        self.assertEqual(v3f[2], 3)
+        
+        v3f = uLib.Math.Vector3f(np.array([1, 2, 3]))
+        self.assertEqual(v3f[0], 1)
+        self.assertEqual(v3f[1], 2)
+        self.assertEqual(v3f[2], 3)
+    
+
+class TestMathGeometry(unittest.TestCase):
+    def test_geometry(self):
+        Geo = uLib.Math.Geometry()
+        
+        Geo.SetPosition([1, 1, 1])
+        
+        pt = Geo.GetLocalPoint([2, 3, 2, 1])
+        wp = Geo.GetWorldPoint(pt)
+        
+        self.assertTrue(vector4f0(wp, [2, 3, 2, 1]))
+        
+        Geo.Scale([2, 2, 2])
+        wp = Geo.GetWorldPoint([1, 1, 1, 1])
+        self.assertTrue(vector4f0(wp, [3, 3, 3, 1]))
+        
+class TestMathContainerBox(unittest.TestCase):
+    def test_container_box_local(self):
+        Cnt = uLib.Math.ContainerBox()
+        Cnt.SetOrigin([-1, -1, -1])
+        Cnt.SetSize([2, 2, 2])
+        
+        size = Cnt.GetSize()
+        self.assertTrue(np.allclose(size, [2, 2, 2]))
+
+    def test_container_box_global(self):
+        Box = uLib.Math.ContainerBox()
+        Box.SetPosition([1, 1, 1])
+        Box.SetSize([2, 2, 2])
+        
+        pt = Box.GetLocalPoint([2, 3, 2, 1])
+        wp = Box.GetWorldPoint(pt)
+        self.assertTrue(vector4f0(wp, [2, 3, 2, 1]))
+
+class TestMathStructuredGrid(unittest.TestCase):
+    def test_structured_grid(self):
+        grid = uLib.Math.StructuredGrid([10, 10, 10])
+        grid.SetSpacing([1, 1, 1])
+        
+        spacing = grid.GetSpacing()
+        self.assertTrue(np.allclose(spacing, [1, 1, 1]))
+
+class TestMathAccumulator(unittest.TestCase):
+    def test_accumulator_mean(self):
+        acc = uLib.Math.Accumulator_Mean_f()
+        acc(10.0)
+        acc(20.0)
+        self.assertAlmostEqual(acc(), 15.0)
+
+class TestMathNewTypes(unittest.TestCase):
+    def test_eigen_vectors(self):
+        v1f = uLib.Math.Vector1f()
+        v3d = uLib.Math.Vector3d()
+        m4f = uLib.Math.Matrix4f()
+        self.assertIsNotNone(v1f)
+        self.assertIsNotNone(v3d)
+        self.assertIsNotNone(m4f)
+
+    def test_ulib_vectors(self):
+        vi = uLib.Math.Vector_i()
+        vi.append(1)
+        vi.append(2)
+        self.assertEqual(len(vi), 2)
+        self.assertEqual(vi[0], 1)
+        self.assertEqual(vi[1], 2)
+        
+        vf = uLib.Math.Vector_f()
+        vf.append(1.5)
+        self.assertAlmostEqual(vf[0], 1.5)
+
+    def test_homogeneous(self):
+        p = uLib.Math.HPoint3f(1.0, 2.0, 3.0)
+        v = uLib.Math.HVector3f(0.0, 1.0, 0.0)
+        self.assertIsNotNone(p)
+        self.assertIsNotNone(v)
+
+    def test_vox_image(self):
+        img = uLib.Math.VoxImage([2, 2, 2])
+        self.assertEqual(img.GetDims()[0], 2)
+        img.SetValue([0, 0, 0], 10.5)
+        # Note: GetValue returns float, and there might be internal scaling (1.E-6 observed in code)
+        # Actually in VoxImage.h: GetValue(id) returns At(id).Value
+        # SetValue(id, value) sets At(id).Value = value
+        self.assertAlmostEqual(img.GetValue([0, 0, 0]), 10.5)
+
+class TestMathVoxRaytracer(unittest.TestCase):
+    def test_raytracer(self):
+        grid = uLib.Math.StructuredGrid([10, 10, 10])
+        grid.SetSpacing([1, 1, 1])
+        grid.SetOrigin([0, 0, 0])
+        
+        rt = uLib.Math.VoxRaytracer(grid)
+        self.assertIsNotNone(rt)
+        
+        # Test TraceBetweenPoints
+        p1 = np.array([0.5, 0.5, -1.0, 1.0], dtype=np.float32)
+        p2 = np.array([0.5, 0.5, 11.0, 1.0], dtype=np.float32)
+        data = rt.TraceBetweenPoints(p1, p2)
+        
+        self.assertGreater(data.Count(), 0)
+        self.assertAlmostEqual(data.TotalLength(), 10.0)
+        
+        # Check elements
+        elements = data.Data()
+        for i in range(data.Count()):
+            self.assertGreaterEqual(elements[i].vox_id, 0)
+            self.assertGreater(elements[i].L, 0)
+
+    def test_ray_data(self):
+        data = uLib.Math.VoxRaytracerRayData()
+        data.SetCount(10)
+        data.SetTotalLength(5.5)
+        self.assertEqual(data.Count(), 10)
+        self.assertAlmostEqual(data.TotalLength(), 5.5)
+
+if __name__ == '__main__':
+    unittest.main()