v2.9.0

What's Changed

• Remove use of internal default policy in algorithms by @sbaldu in #251
• Mark points from read_output as clustered by @sbaldu in #253
• Only theck input data in test_host_points by @sbaldu in #254
• Improve debug and release flags used in tests by @sbaldu in #252
• Provide CLUEstering::CLUEstering alias for CMake targets by @sbaldu in #255
• Create workflow for testing gpu versions on NGT machines by @sbaldu in #211
• Add explicit libraries link in example build by @sbaldu in #257
• Remove unused <execution> header includes by @sbaldu in #258
• Feature utility for computing silhouette score from host points by @sbaldu in #259
• Remove delta internal buffer for device points by @sbaldu in #256
• Remove exception condition in AssociationMap constructor by @sbaldu in #261
• Mark device-only functions as ALPAKA_FN_ACC by @sbaldu in #262
• Provide Point accessor for PointsView by @sbaldu in #263
• Expose AssociationMap containers by @sbaldu in #264
• Move point conversion functions to public header by @sbaldu in #265
• Add internal meta::apply function by @sbaldu in #266
• Remove duplicate template from DistanceParameter by @sbaldu in #267
• Add more test datasets by @sbaldu in #268
• Memcpy columns separately in copyToDevice by @sbaldu in #269
• Feature overload of copyTo methods returning points by @sbaldu in #270
• Add path suffixes in file path inside to_csv if absent by @sbaldu in #271
• Fix symbol ambuguity in lambda template declarations by @sbaldu in #272
• Refactor README file content by @sbaldu in #273
• Move packaging script to scripts/ folder by @sbaldu in #274
• Provide default value for dm parameter by @sbaldu in #275
• Use DistanceParameter in setParameters by @sbaldu in #277
• Add fit, fit_predict and labels methods to Python Clusterer by @sbaldu in #279
• Fix extents returned by AssociationMap by @sbaldu in #280
• Setup micro-benchmark of binary associator by @sbaldu in #281
• Use nested namespace specifiers where possible by @fopancaldi in #282
• Add CONTRIBUTING.md file by @sbaldu in #283
• Set CMAKE_EXPORT_COMPILE_COMMANDS in tests and examples by @sbaldu in #284
• Add clangd and clang-tidy configuration files by @sbaldu in #285
• Restructure tests to produce separate executables by @sbaldu in #219
• Add measurement of Python code coverage by @sbaldu in #286
• Specify repository in codecov action by @sbaldu in #289
• Run coverage workflow with only serial backend by @sbaldu in #288
• Feature n_clusters getter for device points by @sbaldu in #290
• Make points container constructors more explicit by @sbaldu in #292
• Rework docs with sphinx to include Python API by @sbaldu in #291
• Add tests of device points constructors by @sbaldu in #293
• Extend tests for Python backend by @sbaldu in #296
• Refactor benchmarks to use Google Benchmark by @sbaldu in #297
• Create CI workflows for GPU benchmarks by @sbaldu in #298
• Test python clusterer's plotting methods by @sbaldu in #301
• Add support for wrapped coordinates in Python API by @sbaldu in #302
• Remove seed buffer and properties from Python API by @sbaldu in #303
• Extend tests of the python clusterer's methods by @sbaldu in #304
• Build clusters associator with clustered point size by @sbaldu in #305
• Remove deprecated overload of make_clusters by @sbaldu in #306
• Fix set of ticks in python plotters test by @sbaldu in #307
• Add tests for distance parameters class by @sbaldu in #300
• Refactor setup functions as detail free functions by @sbaldu in #308
• Remove internal init_device function by @sbaldu in #309
• Refactor seeds buffer with runtime shape by @sbaldu in #311
• Fix offset check during tiles setup by @AuroraPerego in #314
• Implement meta::accumulate internal function by @sbaldu in #316
• Move math functions to clue::math namespace by @sbaldu in #317
• Reorder arguments of Tiles::initialize by @sbaldu in #319
• Remove unused variable reserve by @sbaldu in #321
• Conditional fetch of doctest and fmt in tests by @sbaldu in #322
• Added weight to coordiantes returned by PointView by @Electricks94 in #323
• Add contributor Markus Holzer to .zenodo.json by @Electricks94 in #324
• Fix computation of silhouette scores by @sbaldu in #325
• Use silhouette score as validation in tests by @sbaldu in #326
• Add F. Pancaldi and C. Helsens to zenodo as contributors by @sbaldu in #327
• Feature distance metrics by @sbaldu in #318
• Implement distance_metric concept by @sbaldu in #328
• Implement Chebyshev distance metrics by @sbaldu in #329
• Move convolutional_kernel concept to kernels header by @sbaldu in #332
• Add S. Lanzi to zenodo as contributor by @sbaldu in #334
• Add documentation for distance metrics by @sbaldu in #333
• Fix name of Euclidean distance metric by @sbaldu in #335
• Implement euclidean metric for periodic coordinates by @sbaldu in #337
• Update documentation for make_clusters methods by @sbaldu in #338
• Add PyPi and platforms badges to README by @sbaldu in #344
• Setup release version 2.9.0 by @sbaldu in #315

New Contributors

• @fopancaldi made their first contribution in #282
• @Electricks94 made their first contribution in #323

Full Changelog: 2.8.0...2.9.0

v2.8.0

What's Changed

• Add error handling in C++ interface public methods by @sbaldu in #199
• Feature associator cpu methods specializations by @sbaldu in #200
• Implement make_associator functions in internal namespace by @sbaldu in #204
• Implement workflow testing installation of C++ interface by @sbaldu in #205
• Fix warning in associator friend declaration by @sbaldu in #206
• Add tests compilation for the SYCL backend by @sbaldu in #202
• Provide accessor to single points in PointsHost by @sbaldu in #188
• Fix the Python installation workflow by @sbaldu in #209
• Add Francesco Giacomini to zenodo.json by @sbaldu in #212
• Use target_link_libraries for alpaka and boost in testing by @sbaldu in #213
• Fix integer sign comarison in PointsCommon by @sbaldu in #215
• Fix algorithm use in device implementation of make_associator by @sbaldu in #217
• Remove throwing check on negative access to points coordinates by @sbaldu in #218
• Feature std::threads alpaka backend by @sbaldu in #216
• Remove is_seed from public interface through PointsHost by @sbaldu in #185
• Remove unnecessary reinterpret casts in Points backend by @sbaldu in #220
• Make Clusterer::getClusters return an host_associator by @sbaldu in #203
• Remove associator cpu fill unused parameter and improve const-correctness by @sbaldu in #222
• Use std::size_t for number of dimensions template by @sbaldu in #223
• Feature generalized multi-dimensional distance parameters by @sbaldu in #214
• Set docs introduction as mainpage by @sbaldu in #224
• Implement utility for computing cluster centroids by @sbaldu in #184
• Feature separate coords buffers by @sbaldu in #226
• Use unsigned integral literals inside kernles by @sbaldu in #227
• Feature device version of get_clusters by @sbaldu in #229
• Use std::views::take in points tests by @sbaldu in #230
• Disable thrust algorithms when compiling with ALPAKA_HOST_ONLY by @sbaldu in #231
• Make AssociationMapView publicly accessible by @sbaldu in #232
• Document latest utilities and data structures by @sbaldu in #233
• Remove Tiles from the public interface by @sbaldu in #234
• Remove unused methods from Tiles by @sbaldu in #235
• Add workflow for testing debug compilation with Werror by @sbaldu in #236
• Disable Codacy and Flawfinder CI Workflows by @sbaldu in #238
• Add more pedantic flags to tests debug compilation by @sbaldu in #237
• Add lazily evaluated getters for cluster properties inside PointsHost by @sbaldu in #239
• Add boolean flag in Points for preventing output-reading before clustering by @sbaldu in #210
• Allow generic naming for columns in input dataframes by @sbaldu in #201
• Test clustering on dataset with wrapped coordinates by @sbaldu in #207
• Mark host points as clustered in CopyToHost by @sbaldu in #240
• Provide more verbose messages for failed assertions by @sbaldu in #241
• Fix partition of host and device points view by @sbaldu in #242
• Fix warnings about assert usage in PointsCommon and PointsHost by @sbaldu in #243
• Compile with -Werror in CI debug build by @sbaldu in #244
• Clean old comments and add missing documentation by @sbaldu in #245
• Remove memset call when resetting AssociationMap by @sbaldu in #246
• Remove old method from documentation by @sbaldu in #248
• Support association map with zero keys by @sbaldu in #247
• Update versions for release 2.8.0 by @sbaldu in #249

Full Changelog: 2.7.2...2.8.0

v2.7.2

Patch release

What's Changed

• Only compile for cuda architectures from sm75 onwards by @sbaldu in #192
• Define __version__ variable in __init__.py by @sbaldu in #193
• Set nearest higher to -1 for confirmed seeds by @sbaldu in #194
• Update license in zenodo.json by @sbaldu in #198

Full Changelog: 2.7.1...2.7.2

v.2.7.1

This is a patch release to fix the python package on PyPi. It also adds some of the latest changes.

What's Changed

• Use reduce instead of extrema algorithms when calculating tiles by @sbaldu in #180
• Define default execution policy for internal use of std algorithms by @sbaldu in #181
• Add configuration for codecov report and patch by @sbaldu in #183
• Implement nostd::zip_iterator for internal use by @sbaldu in #182
• Move to MPL 2.0 license by @sbaldu in #186
• Use std::reduce for computing number of clusters by @sbaldu in #187
• Update package to patch version 2.7.1 by @sbaldu in #189

Full Changelog: 2.7.0...2.7.1

v2.7.0

What's Changed

• Define make_array utility in nostd namespace by @sbaldu in #132
• Implement data generators for benchmarks of dataset size by @sbaldu in #133
• Use vectorized reduction in benchmark statistic operations by @sbaldu in #134
• Fix test of device points by @sbaldu in #129
• Use xtd library for math operations on the device by @sbaldu in #106
• Implement PointsDevice span getters by @sbaldu in #136
• Remove unneeded use of accelerator object in Tiles methods by @sbaldu in #137
• Rework followers using AssociatonMap by @sbaldu in #138
• Pass alpaka Queues by reference by @sbaldu in #139
• Use int32_t instead of uint32_t by @sbaldu in #140
• Change read_csv to return clue::PointsHost by @sbaldu in #141
• Separate compilation of python bindings from base CMake file by @sbaldu in #142
• Use alpaka 1.2.0 when testing c++ interface by @sbaldu in #144
• Provide interface for running clustering from device points by @sbaldu in #143
• Create zenodo.json by @sbaldu in #145
• Remove alpakaFwd header by @sbaldu in #146
• Allow conditional use of the host/device caching allocator by @sbaldu in #147
• Remove xtd dependency and rework internal code structure by @sbaldu in #148
• Compile internal math functions only inside device code by @sbaldu in #149
• Move Clusterer definition inside core folder by @sbaldu in #150
• Remove alpaka accelerator namespace and move definitions in clue::internal by @sbaldu in #151
• Move convolutional kernel inside clue namespace by @sbaldu in #152
• Update passing conditions for version checking workflow by @sbaldu in #153
• Separate internal data structures from the public interface by @sbaldu in #154
• Separate declaration and implementation details of public classes by @sbaldu in #155
• Provide get_device utility for choosing the device by @sbaldu in #156
• Add iterators and STL-like methods to AssociationMap by @sbaldu in #158
• Add clue::get_queue utility by @sbaldu in #159
• Adding New CI/CD workflow checking cmake file formatting by @waredjeb in #160
• Default device template parameter of PointsDevice and AssociationMap to clue::Device by @sbaldu in #163
• Add first example by @sbaldu in #161
• Document cxx interface by @sbaldu in #157
• Add workflow for automatic deployment of the documentation by @sbaldu in #164
• Setup compilation of test and benchmarks for HIP backend by @sbaldu in #166
• Include option for saving plots as files by @sbaldu in #43
• Add code coverage workflow by @sbaldu in #171
• Provide Clusterer method for setting parameters by @sbaldu in #172
• Remove view buffers and save them directly by @sbaldu in #173
• Provide a common CRTP interface for Points data structures by @sbaldu in #174
• Create cmake-format configuration file by @sbaldu in #175
• Separate calls to algorithms internal kernels by @sbaldu in #176
• Default blocksize for make_clusters to 256 by @sbaldu in #177
• Define convolutional_kernel concept by @sbaldu in #179
• Reorder make_clusters arguments and provide default value to convolutional kernel by @sbaldu in #178

New Contributors

• @waredjeb made their first contribution in #160

Full Changelog: 2.6.5...v2.7.0

2.6.5

What's Changed

• Use std::array for allocating coordinates in kernels by @sbaldu in #114
• Support CUDA sm80 and sm90 architectures by @sbaldu in #117
• Allow clustering without passing the queue externally by @sbaldu in #122
• Rework searchbox struct using std::array by @sbaldu in #121
• Use fmt::format instead of std::format in tests by @sbaldu in #123
• Use concepts instead of SFINAE in backend by @sbaldu in #124
• Use distance for wrapped coordinates in nearest higher kernel by @sbaldu in #125
• Feature installation of the C++ interface by @sbaldu in #126
• Fix bug in input plotter by @sbaldu in #128
• Use current binary and source dirs in CMake files by @sbaldu in #130

Full Changelog: 2.6.4.3...2.6.5

2.6.4.3

What's Changed

• Allow vectorization in density kernel by @sbaldu in #103
• Implement copyToHost and copyToDevice functions by @sbaldu in #113

Full Changelog: 2.6.4.2...2.6.4.3

2.6.4.2

What's Changed

• Add seed_dc parameter and reorder Clusterer constructor arguments by @sbaldu in #91

Full Changelog: 2.6.4.1...2.6.4.2

2.6.4.1

Merge pull request #112 from sbaldu/rework-clusterer

Rename "Clusterer" class and move it to clue namespace

2.6.4

What's Changed

• Test tbb in python and benchmark workflows by @sbaldu in #95
• Use span in getClusters and validation by @sbaldu in #98
• Feature Alpaka-OpenMP backend by @sbaldu in #97
• Separate cmakefiles for benchmark and profiling based on backend by @sbaldu in #100
• Allow automatic vectorization and loop unrolling by @sbaldu in #101

Full Changelog: 2.6.3...2.6.4