Merge branch 'auto_range' into nd_range

CMakeLists.txt

@@ -70,7 +70,7 @@ set(PLSSVM_BASE_SOURCES
 
 ## create base library: linked against all backend libraries
 set(PLSSVM_BASE_LIBRARY_NAME plssvm-base)
-add_library(${PLSSVM_BASE_LIBRARY_NAME} STATIC ${PLSSVM_BASE_SOURCES})
+add_library(${PLSSVM_BASE_LIBRARY_NAME} SHARED ${PLSSVM_BASE_SOURCES})
 ## create all library: one target against all backends are linked
 set(PLSSVM_ALL_LIBRARY_NAME plssvm-all)
 add_library(${PLSSVM_ALL_LIBRARY_NAME} INTERFACE)
@@ -158,15 +158,15 @@ if (cxxopts_FOUND)
 else ()
     message(STATUS "Couldn't find package cxxopts. Building from source ...")
     set(PLSSVM_cxxopts_VERSION v3.1.1)
+    # set options for cxxopts
+    set(CXXOPTS_BUILD_EXAMPLES OFF CACHE INTERNAL "" FORCE)
+    set(CXXOPTS_BUILD_TESTS OFF CACHE INTERNAL "" FORCE)
+    set(CXXOPTS_ENABLE_WARNINGS OFF CACHE INTERNAL "" FORCE)
     # fetch command line parser library cxxopts
     FetchContent_Declare(cxxopts
             GIT_REPOSITORY https://github.com/jarro2783/cxxopts.git
             GIT_TAG ${PLSSVM_cxxopts_VERSION}
-            GIT_SHALLOW TRUE
             QUIET
-
-            set (CXXOPTS_BUILD_EXAMPLES OFF CACHE BOOL "" FORCE)
-            set (CXXOPTS_BUILD_TESTS OFF CACHE BOOL "" FORCE)
     )
     FetchContent_MakeAvailable(cxxopts)
     add_dependencies(${PLSSVM_BASE_LIBRARY_NAME} cxxopts)
@@ -184,16 +184,15 @@ if (fast_float_FOUND)
     target_include_directories(${PLSSVM_BASE_LIBRARY_NAME} PUBLIC ${fast_float_INCLUDE_DIR})
 else ()
     message(STATUS "Couldn't find package fast_float. Building from source ...")
-    set(PLSSVM_fast_float_VERSION v3.10.0) # TODO: 5.2.0: https://github.com/fastfloat/fast_float/issues/215
+    set(PLSSVM_fast_float_VERSION v3.10.0) # TODO: 5.3.0: https://github.com/fastfloat/fast_float/issues/215
+    # set options for fast_float
+    set(FASTFLOAT_TEST OFF CACHE INTERNAL "" FORCE)
+    set(FASTFLOAT_SANITIZE OFF CACHE INTERNAL "" FORCE)
     # fetch float parsing library fast_float
     FetchContent_Declare(fast_float
             GIT_REPOSITORY https://github.com/fastfloat/fast_float
             GIT_TAG ${PLSSVM_fast_float_VERSION}
-            GIT_SHALLOW TRUE
             QUIET
-
-            set (FASTFLOAT_TEST OFF CACHE BOOL "" FORCE)
-            set (FASTFLOAT_SANITIZE OFF CACHE BOOL "" FORCE)
     )
     FetchContent_GetProperties(fast_float)
     if (NOT fast_float_POPULATED)
@@ -217,14 +216,13 @@ if (igor_FOUND)
 else ()
     message(STATUS "Couldn't find package igor. Building from source ...")
     set(PLSSVM_igor_VERSION a5224c60d266974d3f407191583fe266cbe1c93d)
+    # set options for igor
+    set(IGOR_BUILD_TESTS OFF CACHE INTERNAL "" FORCE)
     # fetch named argument library igor
     FetchContent_Declare(igor
             GIT_REPOSITORY https://github.com/bluescarni/igor
             GIT_TAG ${PLSSVM_igor_VERSION}
-            GIT_SHALLOW TRUE
             QUIET
-
-            set (IGOR_BUILD_TESTS OFF CACHE BOOL "" FORCE)
     )
     FetchContent_GetProperties(igor)
     if (NOT igor_POPULATED)
@@ -247,24 +245,29 @@ if (fmt_FOUND)
 else ()
     message(STATUS "Couldn't find package fmt. Building from source ...")  # TODO: change to next release; bug fix commit
     set(PLSSVM_fmt_VERSION d9063baf227882da0f48c761abcbb08247eb1296)
+    # set options for fmt
     if (PLSSVM_ENABLE_STL_DEBUG_MODE)
         set(CMAKE_CXX_FLAGS_OLD "${CMAKE_CXX_FLAGS}")
         set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${PLSSVM_STL_DEBUG_MODE_FLAGS}")
     endif ()
+    set(FMT_PEDANTIC OFF CACHE INTERNAL "" FORCE)
+    set(FMT_WERROR OFF CACHE INTERNAL "" FORCE)
+    set(FMT_DOC OFF CACHE INTERNAL "" FORCE)
+    set(FMT_INSTALL OFF CACHE INTERNAL "" FORCE)
+    set(FMT_TEST OFF CACHE INTERNAL "" FORCE)
+    set(FMT_FUZZ OFF CACHE INTERNAL "" FORCE)
+    set(FMT_CUDA_TEST OFF CACHE INTERNAL "" FORCE)
+    set(FMT_MODULE OFF CACHE INTERNAL "" FORCE)
+    set(FMT_SYSTEM_HEADERS ON CACHE INTERNAL "" FORCE)
     # fetch string formatting library fmt
     FetchContent_Declare(fmt
             GIT_REPOSITORY https://github.com/fmtlib/fmt.git
             GIT_TAG ${PLSSVM_fmt_VERSION}
-#            GIT_SHALLOW TRUE
             QUIET
-
-#            set (FMT_TEST OFF CACHE BOOL "" FORCE)
-#            set (FMT_DOC OFF CACHE BOOL "" FORCE)
-#            set (FMT_INSTALL OFF CACHE BOOL "" FORCE)
-#            set (FMT_SYSTEM_HEADERS ON CACHE BOOL "" FORCE)
     )
     FetchContent_MakeAvailable(fmt)
     set_property(TARGET fmt PROPERTY POSITION_INDEPENDENT_CODE ON)
+    target_compile_definitions(fmt PRIVATE FMT_USE_FULL_CACHE_DRAGONBOX)
     add_dependencies(${PLSSVM_BASE_LIBRARY_NAME} fmt)
     target_include_directories(${PLSSVM_BASE_LIBRARY_NAME} PUBLIC
             $<BUILD_INTERFACE:${fmt_SOURCE_DIR}/include>

README.md

@@ -305,32 +305,33 @@ export LD_LIBRARY_PATH=${CMAKE_INSTALL_PREFIX}/lib:${LD_LIBRARY_PATH}
 The repository comes with a Python3 script (in the `utility_scripts/` directory) to simply generate arbitrarily large data sets.
 
 In order to use all functionality, the following Python3 modules must be installed:
-[`argparse`](https://docs.python.org/3/library/argparse.html), [`timeit`](https://docs.python.org/3/library/timeit.html),
+[`argparse`](https://docs.python.org/3/library/argparse.html), [`timeit`](https://docs.python.org/3/library/timeit.html), 
 [`numpy`](https://pypi.org/project/numpy/), [`pandas`](https://pypi.org/project/pandas/),
 [`sklearn`](https://scikit-learn.org/stable/), [`arff`](https://pypi.org/project/arff/),
 [`matplotlib`](https://pypi.org/project/matplotlib/), [`mpl_toolkits`](https://pypi.org/project/matplotlib/),
 and [`humanize`](https://pypi.org/project/humanize/).
 
 ```bash
-python3 utility_scripts/generate_data.py --help
-usage: generate_data.py [-h] --output OUTPUT --format FORMAT [--problem PROBLEM] --samples SAMPLES [--test_samples TEST_SAMPLES] --features FEATURES [--plot]
+usage: generate_data.py [-h] [--output OUTPUT] [--format FORMAT] [--problem PROBLEM] --samples SAMPLES [--test_samples TEST_SAMPLES] --features FEATURES [--classes CLASSES] [--plot]
 
-optional arguments:
+options:
   -h, --help            show this help message and exit
   --output OUTPUT       the output file to write the samples to (without extension)
-  --format FORMAT       the file format; either arff or libsvm
-  --problem PROBLEM     the problem to solve; one of: blobs, blobs_merged, planes, planes_merged, ball
+  --format FORMAT       the file format; either arff, libsvm, or csv
+  --problem PROBLEM     the problem to solve; one of: blobs, blobs_merged, planes, ball
   --samples SAMPLES     the number of training samples to generate
   --test_samples TEST_SAMPLES
                         the number of test samples to generate; default: 0
   --features FEATURES   the number of features per data point
+  --classes CLASSES     the number of classes to generate; default: 2
   --plot                plot training samples; only possible if 0 < samples <= 2000 and 1 < features <= 3
 ```
 
-An example invocation generating a data set consisting of blobs with 1000 data points with 200 features each could look like:
+An example invocation generating a data set consisting of blobs with 1000 data points with 200 features each and 
+4 classes could look like:
 
 ```bash
-python3 generate_data.py --output data_file --format libsvm --problem blobs --samples 1000 --features 200
+python3 generate_data.py --output data_file --format libsvm --problem blobs --samples 1000 --features 200 --classes 4
 ```
 
 ### Training

bindings/Python/detail/performance_tracker.cpp

@@ -32,7 +32,11 @@ void init_performance_tracker([[maybe_unused]] py::module_ &m) {
         .def(
             "resume", []() { plssvm::detail::global_tracker->resume_tracking(); }, "resume performance tracking")
         .def(
-            "save", [](const std::string &filename) { plssvm::detail::global_tracker->save(filename); }, "save the performance tracking results to the specified yaml file");
+            "save", [](const std::string &filename) { plssvm::detail::global_tracker->save(filename); }, "save the performance tracking results to the specified yaml file")
+        .def(
+            "is_tracking", []() { return plssvm::detail::global_tracker->is_tracking(); }, "check whether performance tracking is currently enabled")
+        .def(
+            "clear_tracking_entries", []() { plssvm::detail::global_tracker->clear_tracking_entries(); }, "remove all currently tracked entries from the performance tracker");
 
 #endif
 }

install/python_requirements.txt

@@ -1,15 +1,39 @@
 ### optional and required python packages
+
+## for the data set generation
+# LIBSVM
 argparse
 scikit-learn
+humanize
+
+# additionally for generating ARFF files
+numpy
+arff
+pandas
+
+# additionally for generation CSV files
+# numpy
+
+# for plotting the generated data set (only available if #features <= 3)
+matplotlib
+
+
+## for the automatic PLSSVM target platform determination
+# argparse
 py-cpuinfo
 GPUtil
 pyamdgpuinfo
 pylspci
-numpy
-pandas
-arff
-matplotlib
-humanize
+
+
+## for the performance tracker parser
+# argparse
+# matplotlib
 pyyaml
 pint
+
+
+## for the performance analysis
+# argparse
+# scikit-learn
 wrapt-timeout-decorator

src/plssvm/backends/CUDA/csvm.cu

@@ -153,7 +153,7 @@ auto csvm::run_assemble_kernel_matrix_explicit(const parameter &params, const de
                     static_cast<int>(std::ceil(static_cast<double>(num_rows_reduced) / static_cast<double>(block.y))));
 
 #if defined(PLSSVM_USE_GEMM)
-    device_ptr_type kernel_matrix_d{ num_rows_reduced * num_rows_reduced };  // store full matrix
+    device_ptr_type kernel_matrix_d{ num_rows_reduced * num_rows_reduced, devices_[0] };  // store full matrix
 #else
     device_ptr_type kernel_matrix_d{ num_rows_reduced * (num_rows_reduced + 1) / 2, devices_[0] };  // only explicitly store the upper triangular matrix
 #endif

src/plssvm/backends/HIP/csvm.hip.cpp

@@ -150,7 +150,7 @@ auto csvm::run_assemble_kernel_matrix_explicit(const parameter &params, const de
                     static_cast<int>(std::ceil(static_cast<double>(num_rows_reduced) / static_cast<double>(block.y))));
 
 #if defined(PLSSVM_USE_GEMM)
-    device_ptr_type kernel_matrix_d{ num_rows_reduced * num_rows_reduced };  // store full matrix
+    device_ptr_type kernel_matrix_d{ num_rows_reduced * num_rows_reduced, devices_[0] };  // store full matrix
 #else
     device_ptr_type kernel_matrix_d{ num_rows_reduced * (num_rows_reduced + 1) / 2, devices_[0] };  // only explicitly store the upper triangular matrix
 #endif

src/plssvm/backends/OpenCL/CMakeLists.txt

@@ -36,7 +36,7 @@ set(PLSSVM_OPENCL_SOURCES
 
 # set target properties
 set_local_and_parent(PLSSVM_OPENCL_BACKEND_LIBRARY_NAME plssvm-OpenCL)
-add_library(${PLSSVM_OPENCL_BACKEND_LIBRARY_NAME} STATIC ${PLSSVM_OPENCL_SOURCES})
+add_library(${PLSSVM_OPENCL_BACKEND_LIBRARY_NAME} SHARED ${PLSSVM_OPENCL_SOURCES})
 target_include_directories(${PLSSVM_OPENCL_BACKEND_LIBRARY_NAME} PUBLIC ${OpenCL_INCLUDE_DIRS})
 target_link_libraries(${PLSSVM_OPENCL_BACKEND_LIBRARY_NAME} PUBLIC OpenCL)
 

src/plssvm/backends/OpenCL/detail/utility.cpp

@@ -238,7 +238,7 @@ std::vector<command_queue> create_command_queues(const std::vector<context> &con
 //    ::plssvm::detail::replace_all(kernel_src_string, "THREAD_BLOCK_SIZE", fmt::format("{}", THREAD_BLOCK_SIZE));
 
     // append number of device to influence checksum calculation
-    kernel_src_string.append(fmt::format("\n// num_devices: {}\n// OpenCL library: {}", contexts[0].devices.size(), PLSSVM_OPENCL_LIBRARY));
+    kernel_src_string.append(fmt::format("\n// num_devices: {}\n// OpenCL library: {}\n// GEMM: {}", contexts[0].devices.size(), PLSSVM_OPENCL_LIBRARY, PLSSVM_IS_DEFINED(PLSSVM_USE_GEMM)));
 
     // create source code hash
     const std::string checksum = plssvm::detail::sha256{}(kernel_src_string);

src/plssvm/backends/OpenMP/CMakeLists.txt

@@ -41,7 +41,7 @@ set(PLSSVM_OPENMP_SOURCES
 
 # set target properties
 set_local_and_parent(PLSSVM_OPENMP_BACKEND_LIBRARY_NAME plssvm-OpenMP)
-add_library(${PLSSVM_OPENMP_BACKEND_LIBRARY_NAME} STATIC ${PLSSVM_OPENMP_SOURCES})
+add_library(${PLSSVM_OPENMP_BACKEND_LIBRARY_NAME} SHARED ${PLSSVM_OPENMP_SOURCES})
 target_link_libraries(${PLSSVM_OPENMP_BACKEND_LIBRARY_NAME} PUBLIC OpenMP::OpenMP_CXX)
 # special command line options for MSVC:
 # -openmp:llvm -> enables unsigned loop indexes in OpenMP parallel for loops

tests/CMakeLists.txt

@@ -22,7 +22,6 @@ else ()
     FetchContent_Declare(googletest
             GIT_REPOSITORY https://github.com/google/googletest
             GIT_TAG ${PLSSVM_googletest_VERSION}
-            GIT_SHALLOW TRUE
             QUIET
     )
     # For Windows: Prevent overriding the parent project's compiler/linker settings
@@ -89,7 +88,7 @@ set(PLSSVM_BASE_TEST_LIBRARY_NAME_SOURCES
 )
 
 # create base test library
-add_library(${PLSSVM_BASE_TEST_LIBRARY_NAME} STATIC ${PLSSVM_BASE_TEST_LIBRARY_NAME_SOURCES})
+add_library(${PLSSVM_BASE_TEST_LIBRARY_NAME} SHARED ${PLSSVM_BASE_TEST_LIBRARY_NAME_SOURCES})
 
 # set necessary include directories
 target_include_directories(${PLSSVM_BASE_TEST_LIBRARY_NAME} PUBLIC "${CMAKE_CURRENT_SOURCE_DIR}")