Updated supported CUDA verison to 12.8

.gitmodules

@@ -1,6 +1,3 @@
 [submodule "deps/gtest"]
 	path = deps/gtest
 	url = https://github.com/google/googletest.git
-[submodule "deps/cub"]
-	path = deps/cub
-	url = https://github.com/NVIDIA/cub.git

CMakeLists.txt

@@ -11,6 +11,8 @@ option(CUBOOL_WITH_NAIVE         "Build library with naive and naive-shared dens
 option(CUBOOL_BUILD_TESTS        "Build project unit-tests with gtest" ON)
 option(CUBOOL_COPY_TO_PY_PACKAGE "Copy compiled shared library into python package folder (for package use purposes)" ON)
 
+option(CUBOOL_USE_NSPARSE_MERGE_FUNCTOR "Use nsparse optimiztion for matrix addition function" ON)
+
 set(CUBOOL_VERSION_MAJOR 1)
 set(CUBOOL_VERSION_MINOR 0)
 set(CUBOOL_VERSION_SUB 0)
@@ -32,14 +34,6 @@ endif()
 
 # Configure cuda dependencies
 if (CUBOOL_WITH_CUDA)
-    message(STATUS "Add cub as cuda utility")
-    set(CUB_ENABLE_HEADER_TESTING OFF CACHE BOOL "" FORCE)
-    set(CUB_ENABLE_TESTING OFF CACHE BOOL "" FORCE)
-    set(CUB_ENABLE_EXAMPLES OFF CACHE BOOL "" FORCE)
-    add_subdirectory(deps/cub)
-    add_library(cub INTERFACE IMPORTED)
-    target_link_libraries(cub INTERFACE CUB::CUB)
-
     message(STATUS "Add nsparse library as crs matrix multiplication backend")
     add_subdirectory(deps/nsparse-um)
 endif()
@@ -61,4 +55,4 @@ add_subdirectory(cubool)
 file(COPY scripts DESTINATION ${CMAKE_BINARY_DIR}/)
 
 # Copy python related stuff
-file(COPY python DESTINATION ${CMAKE_BINARY_DIR}/)
+file(COPY python DESTINATION ${CMAKE_BINARY_DIR}/)

cubool/CMakeLists.txt

@@ -87,6 +87,7 @@ set(CUBOOL_C_API_SOURCES
     sources/cuBool_Matrix_Reduce2.cpp
     sources/cuBool_Matrix_EWiseAdd.cpp
     sources/cuBool_Matrix_EWiseMult.cpp
+    sources/cuBool_Matrix_EWiseMultInverted.cpp
     sources/cuBool_Vector_New.cpp
     sources/cuBool_Vector_Build.cpp
     sources/cuBool_Vector_SetElement.cpp
@@ -125,6 +126,7 @@ if (CUBOOL_WITH_CUDA)
         sources/cuda/cuda_matrix.cu
         sources/cuda/cuda_matrix_ewiseadd.cu
         sources/cuda/cuda_matrix_ewisemult.cu
+        sources/cuda/cuda_matrix_ewisemult_inverted.cu
         sources/cuda/cuda_matrix_kronecker.cu
         sources/cuda/cuda_matrix_multiply.cu
         sources/cuda/cuda_matrix_transpose.cu
@@ -147,6 +149,7 @@ if (CUBOOL_WITH_CUDA)
         sources/cuda/kernels/spgemv_t.cuh
         sources/cuda/kernels/spewiseadd.cuh
         sources/cuda/kernels/spewisemult.cuh
+        sources/cuda/kernels/spewisemultinverted.cuh
         sources/cuda/kernels/sptranspose.cuh
         sources/cuda/kernels/sptranspose2.cuh
         sources/cuda/kernels/spkron.cuh
@@ -173,6 +176,8 @@ if (CUBOOL_WITH_SEQUENTIAL)
         sources/sequential/sq_ewiseadd.hpp
         sources/sequential/sq_ewisemult.cpp
         sources/sequential/sq_ewisemult.hpp
+        sources/sequential/sq_ewisemultinverted.cpp
+        sources/sequential/sq_ewisemultinverted.hpp
         sources/sequential/sq_spgemm.cpp
         sources/sequential/sq_spgemm.hpp
         sources/sequential/sq_spgemv.cpp
@@ -201,11 +206,9 @@ target_compile_definitions(cubool PRIVATE CUBOOL_VERSION_MAJOR=${CUBOOL_VERSION_
 target_compile_definitions(cubool PRIVATE CUBOOL_VERSION_MINOR=${CUBOOL_VERSION_MINOR})
 target_compile_definitions(cubool PRIVATE CUBOOL_VERSION_SUB=${CUBOOL_VERSION_SUB})
 
-target_compile_features(cubool PUBLIC cxx_std_14)
+target_compile_definitions(cubool PRIVATE CUBOOL_USE_NSPARSE_MERGE_FUNCTOR=$<BOOL:${CUBOOL_USE_NSPARSE_MERGE_FUNCTOR}>)
 
-target_compile_options(cubool PRIVATE $<$<COMPILE_LANGUAGE:CXX>: -Wall>)
-target_compile_options(cubool PRIVATE $<$<AND:$<CONFIG:Debug>,$<COMPILE_LANGUAGE:CXX>>: -O2>)
-target_compile_options(cubool PRIVATE $<$<AND:$<CONFIG:Release>,$<COMPILE_LANGUAGE:CXX>>: -O0>)
+target_compile_features(cubool PUBLIC cxx_std_17)
 
 set_target_properties(cubool PROPERTIES CXX_STANDARD 17)
 set_target_properties(cubool PROPERTIES CXX_STANDARD_REQUIRED ON)
@@ -219,7 +222,6 @@ endforeach()
 if (CUBOOL_WITH_CUDA)
     set_target_properties(cubool PROPERTIES CUDA_STANDARD 14)
     set_target_properties(cubool PROPERTIES CUDA_STANDARD_REQUIRED ON)
-    set_target_properties(cubool PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
 
     # Settings: https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/
 
@@ -256,4 +258,4 @@ if (CUBOOL_COPY_TO_PY_PACKAGE)
         "${CMAKE_BINARY_DIR}/cubool/${LIBRARY_FILE_NAME}"
         "${CMAKE_BINARY_DIR}/python/pycubool"
         COMMENT "Copy ${LIBRARY_FILE_NAME} compiled lib into python folder")
-endif()
+endif()

cubool/include/cubool/cubool.h

@@ -925,4 +925,30 @@ CUBOOL_EXPORT CUBOOL_API cuBool_Status cuBool_Kronecker(
     cuBool_Hints hints
 );
 
+/**
+ * Performs result = left * ~right, where
+ *     '*' is boolean semiring 'and' operation
+ *     '~' is operation for invert matrix (0 swaps to 1 and 1 to 0)
+ *
+ * @note To perform this operation matrices must be compatible
+ *          dim(left) = M x T
+ *          dim(right) = T x N
+ *          dim(result) = M x N
+ *
+ * @note Pass `CUBOOL_HINT_TIME_CHECK` hint to measure operation time
+ *
+ * @param result[out] Destination matrix to store result
+ * @param left Source matrix to be multiplied
+ * @param right Source matrix to be inverted and multiplied
+ * @param hints Hints for the operation
+ *
+ * @return Error code on this operation
+ */
+CUBOOL_EXPORT CUBOOL_API cuBool_Status cuBool_Matrix_EWiseMulInverted(
+    cuBool_Matrix result,
+    cuBool_Matrix matrix,
+    cuBool_Matrix mask,
+    cuBool_Hints hints
+);
+
 #endif //CUBOOL_CUBOOL_H

cubool/sources/backend/matrix_base.hpp

@@ -49,6 +49,7 @@ namespace cubool {
         virtual void kronecker(const MatrixBase &aBase, const MatrixBase &bBase, bool checkTime) = 0;
         virtual void eWiseAdd(const MatrixBase &aBase, const MatrixBase &bBase, bool checkTime) = 0;
         virtual void eWiseMult(const MatrixBase &aBase, const MatrixBase &bBase, bool checkTime) = 0;
+        virtual void eWiseMultInverted(const MatrixBase &matrix, const MatrixBase &mask, bool checkTime) = 0;
 
         virtual index getNrows() const = 0;
         virtual index getNcols() const = 0;
@@ -59,4 +60,4 @@ namespace cubool {
 
 }
 
-#endif //CUBOOL_MATRIX_BASE_HPP
+#endif //CUBOOL_MATRIX_BASE_HPP

cubool/sources/core/matrix.cpp

@@ -349,6 +349,46 @@ namespace cubool {
         mHnd->eWiseMult(*a->mHnd, *b->mHnd, false);
     }
 
+    void Matrix::eWiseMultInverted(const MatrixBase &matrix, const MatrixBase &mask, bool checkTime) {
+
+        const auto* mat = dynamic_cast<const Matrix*>(&matrix);
+        const auto* msk = dynamic_cast<const Matrix*>(&mask);
+
+        CHECK_RAISE_ERROR(mat != nullptr, InvalidArgument, "Passed matrix does not belong to core matrix class");
+        CHECK_RAISE_ERROR(msk != nullptr, InvalidArgument, "Passed matrix does not belong to core matrix class");
+
+        index M = mat->getNrows();
+        index N = msk->getNcols();
+
+        CHECK_RAISE_ERROR(M == msk->getNrows(), InvalidArgument, "Passed matrices have incompatible size");
+        CHECK_RAISE_ERROR(N == msk->getNcols(), InvalidArgument, "Passed matrices have incompatible size");
+
+        CHECK_RAISE_ERROR(M == this->getNrows(), InvalidArgument, "Matrix has incompatible size for operation result");
+        CHECK_RAISE_ERROR(N == this->getNcols(), InvalidArgument, "Matrix has incompatible size for operation result");
+
+        mat->commitCache();
+        msk->commitCache();
+        this->releaseCache();
+
+        if (checkTime) {
+            TIMER_ACTION(timer, mHnd->eWiseMultInverted(*mat->mHnd, *msk->mHnd, false));
+
+            LogStream stream(*Library::getLogger());
+            stream << Logger::Level::Info
+                   << "Time: " << timer.getElapsedTimeMs() << " ms "
+                   << "Matrix::eWiseMultInverted: "
+                   << this->getDebugMarker() << " = "
+                   << mat->getDebugMarker() << " + "
+                   << msk->getDebugMarker() << LogStream::cmt;
+
+            return;
+        }
+
+        mHnd->eWiseMultInverted(*mat->mHnd, *msk->mHnd, false);
+    }
+
+
+
     index Matrix::getNrows() const {
         return mHnd->getNrows();
     }
@@ -396,4 +436,4 @@ namespace cubool {
         // Clear arrays
         releaseCache();
     }
-}
+}

cubool/sources/core/matrix.hpp

@@ -56,6 +56,7 @@ namespace cubool {
         void kronecker(const MatrixBase &aBase, const MatrixBase &bBase, bool checkTime) override;
         void eWiseAdd(const MatrixBase &aBase, const MatrixBase &bBase, bool checkTime) override;
         void eWiseMult(const MatrixBase &a, const MatrixBase &b, bool checkTime) override;
+        void eWiseMultInverted(const MatrixBase &matrix, const MatrixBase &mask, bool checkTime) override;
 
         index getNrows() const override;
         index getNcols() const override;
@@ -77,4 +78,4 @@ namespace cubool {
 
 }
 
-#endif //CUBOOL_MATRIX_HPP
+#endif //CUBOOL_MATRIX_HPP

cubool/sources/cuBool_Matrix_EWiseMultInverted.cpp

@@ -0,0 +1,44 @@
+/**********************************************************************************/
+/* MIT License                                                                    */
+/*                                                                                */
+/* Copyright (c) 2020, 2021 JetBrains-Research                                    */
+/*                                                                                */
+/* Permission is hereby granted, free of charge, to any person obtaining a copy   */
+/* of this software and associated documentation files (the "Software"), to deal  */
+/* in the Software without restriction, including without limitation the rights   */
+/* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell      */
+/* copies of the Software, and to permit persons to whom the Software is          */
+/* furnished to do so, subject to the following conditions:                       */
+/*                                                                                */
+/* The above copyright notice and this permission notice shall be included in all */
+/* copies or substantial portions of the Software.                                */
+/*                                                                                */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR     */
+/* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,       */
+/* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE    */
+/* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER         */
+/* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,  */
+/* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE  */
+/* SOFTWARE.                                                                      */
+/**********************************************************************************/
+
+#include <cuBool_Common.hpp>
+
+CUBOOL_EXPORT CUBOOL_API cuBool_Status cuBool_Matrix_EWiseMulInverted(
+    cuBool_Matrix result,
+    cuBool_Matrix matrix,
+    cuBool_Matrix mask,
+    cuBool_Hints hints
+) {
+    CUBOOL_BEGIN_BODY
+        CUBOOL_VALIDATE_LIBRARY
+        CUBOOL_ARG_NOT_NULL(result)
+        CUBOOL_ARG_NOT_NULL(matrix)
+        CUBOOL_ARG_NOT_NULL(mask)
+        auto resultM = (cubool::Matrix *) result;
+        auto matrixM = (cubool::Matrix *) matrix;
+        auto maskM = (cubool::Matrix *) mask;
+
+        resultM->eWiseMultInverted(*matrixM, *maskM, hints & CUBOOL_HINT_TIME_CHECK);
+    CUBOOL_END_BODY
+}

cubool/sources/cuda/cuda_matrix.hpp

@@ -56,6 +56,7 @@ namespace cubool {
         void kronecker(const MatrixBase &a, const MatrixBase &b, bool checkTime) override;
         void eWiseAdd(const MatrixBase &a, const MatrixBase &b, bool checkTime) override;
         void eWiseMult(const MatrixBase &a, const MatrixBase &b, bool checkTime) override;
+        void eWiseMultInverted(const MatrixBase &a, const MatrixBase &b, bool checkTime) override;
 
         index getNrows() const override;
         index getNcols() const override;
@@ -79,4 +80,4 @@ namespace cubool {
     };
 };
 
-#endif //CUBOOL_CUDA_MATRIX_HPP
+#endif //CUBOOL_CUDA_MATRIX_HPP

cubool/sources/cuda/cuda_matrix_ewisemult_inverted.cu

@@ -0,0 +1,63 @@
+/**********************************************************************************/
+/* MIT License                                                                    */
+/*                                                                                */
+/* Copyright (c) 2020, 2021 JetBrains-Research                                    */
+/*                                                                                */
+/* Permission is hereby granted, free of charge, to any person obtaining a copy   */
+/* of this software and associated documentation files (the "Software"), to deal  */
+/* in the Software without restriction, including without limitation the rights   */
+/* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell      */
+/* copies of the Software, and to permit persons to whom the Software is          */
+/* furnished to do so, subject to the following conditions:                       */
+/*                                                                                */
+/* The above copyright notice and this permission notice shall be included in all */
+/* copies or substantial portions of the Software.                                */
+/*                                                                                */
+/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR     */
+/* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,       */
+/* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE    */
+/* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER         */
+/* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,  */
+/* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE  */
+/* SOFTWARE.                                                                      */
+/**********************************************************************************/
+
+#include <cuda/cuda_matrix.hpp>
+#include <cuda/kernels/spewisemultinverted.cuh>
+
+namespace cubool {
+
+    void CudaMatrix::eWiseMultInverted(const MatrixBase &aBase, const MatrixBase &bBase, bool checkTime) {
+        auto a = dynamic_cast<const CudaMatrix*>(&aBase);
+        auto b = dynamic_cast<const CudaMatrix*>(&bBase);
+
+        CHECK_RAISE_ERROR(a != nullptr, InvalidArgument, "Passed matrix does not belong to csr matrix class");
+        CHECK_RAISE_ERROR(b != nullptr, InvalidArgument, "Passed matrix does not belong to csr matrix class");
+
+        index M = this->getNrows();
+        index N = this->getNcols();
+
+
+        assert(a->getNrows() == M);
+        assert(a->getNcols() == N);
+
+        assert(b->getNrows() == M);
+        assert(b->getNcols() == N);
+
+        if (a->isMatrixEmpty() || b->isMatrixEmpty()) {
+            this->clearAndResizeStorageToDim();
+            return;
+        }
+
+        // Ensure csr proper csr format even if empty
+        a->resizeStorageToDim();
+        b->resizeStorageToDim();
+
+        kernels::SpVectorEWiseMultInverted<index, DeviceAlloc<index>> spFunctor;
+        auto result = spFunctor(a->mMatrixImpl, b->mMatrixImpl);
+
+        // Assign the actual impl result to this storage
+        this->mMatrixImpl = std::move(result);
+    }
+
+}

cubool/sources/cuda/cuda_vector.cu

@@ -29,6 +29,8 @@
 #include <utils/data_utils.hpp>
 #include <limits>
 
+#include <thrust/sort.h>
+
 namespace cubool {
 
     CudaVector::CudaVector(size_t nrows, CudaInstance &instance)

cubool/sources/cuda/kernels/spewiseadd.cuh

@@ -27,6 +27,8 @@
 
 #include <cuda/details/sp_vector.hpp>
 
+#include <thrust/set_operations.h>
+
 namespace cubool {
     namespace kernels {
 

cubool/sources/cuda/kernels/spewisemult.cuh

@@ -30,6 +30,8 @@
 #include <nsparse/matrix.h>
 #include <cmath>
 
+#include <thrust/set_operations.h>
+
 namespace cubool {
     namespace kernels {