Make tests compile. Will get them passing in a later commit

Author: Pencilcaseman

cmake/ArchDetect2.cmake

@@ -53,6 +53,14 @@ function(check_simd_capability FLAG_GNU FLAG_MSVC NAME TEST_SOURCE VAR)
     endif ()
 endfunction()
 
+check_simd_capability("-mmmx" "" "MMX" "
+#include <mmintrin.h>
+int main() {
+    __m64 a = _mm_set_pi32(-1, 2);
+    __m64 result = _mm_abs_pi32(a);
+    return 0;
+}" SIMD_MMX)
+
 # Check SSE2 (not a valid flag for MSVC)
 check_simd_capability("-msse2" "" "SSE2" "
 #include <emmintrin.h>
@@ -99,6 +107,16 @@ int main() {
     return 0;
 }" SIMD_SSE4_2)
 
+check_simd_capability("-mfma" "/arch:FMA" "FMA3" "
+#include <immintrin.h>
+int main() {
+    __m256 a = _mm256_set_ps(-1.0f, 2.0f, -3.0f, 4.0f, -1.0f, 2.0f, -3.0f, 4.0f);
+    __m256 b = _mm256_set_ps(1.0f, 2.0f, 3.0f, 4.0f, 1.0f, 2.0f, 3.0f, 4.0f);
+    __m256 c = _mm256_set_ps(1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f, 1.0f);
+    __m256 result = _mm256_fmadd_ps(a, b, c);
+    return 0;
+}" SIMD_FMA3)
+
 # Check AVX
 check_simd_capability("-mavx" "/arch:AVX" "AVX" "
 #include <immintrin.h>
@@ -171,6 +189,15 @@ int main() {
     return 0;
 }" SIMD_AVX512PF)
 
+check_simd_capability("-march=armv6" "" "ARMv6" "
+#include <arm_neon.h>
+int main() {
+    int32x2_t a = vdup_n_s32(1);
+    int32x2_t b = vdup_n_s32(2);
+    int32x2_t result = vadd_s32(a, b);
+    return 0;
+}" SIMD_ARMv6)
+
 # ARM
 check_simd_capability("-march=armv7-a" "" "ARMv7" "
 #include <arm_neon.h>
@@ -238,6 +265,6 @@ int main() {
 
 if (LIBRAPID_ARCH_FOUND)
     message(STATUS "[ LIBRAPID ] Architecture Flags: ${LIBRAPID_ARCH_FLAGS}")
-else()
+else ()
     message(STATUS "[ LIBRAPID ] Architecture Flags Not Found")
-endif()
+endif ()

examples/example-cuda.cpp

@@ -31,7 +31,7 @@ auto main() -> int {
 	fmt::print("Vector: \n{}\n", vector);
 	fmt::print("Matrix dot Vector^T:\n{}\n", lrc::dot(cudaArray, lrc::transpose(vector)));
 #else
-	fmt::print("OpenCL not enabled in this build of librapid\n");
+	fmt::print("CUDA not enabled in this build of librapid\n");
 	fmt::print("Check the documentation for more information on enabling OpenCL\n");
 	fmt::print("https://librapid.readthedocs.io/en/latest/cmakeIntegration.html#librapid-use-cuda\n");
 #endif // LIBRAPID_HAS_CUDA

librapid/include/librapid/array/assignOps.hpp

@@ -274,7 +274,7 @@ namespace librapid {
         LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE auto
         openCLTupleEvaluatorImpl(const detail::Function<descriptor, Functor, Args...> &function) {
             array::ArrayContainer<
-              decltype(function.shape()),
+              typename std::decay_t<decltype(function.shape())>,
               OpenCLStorage<typename detail::Function<descriptor, Functor, Args...>::Scalar>>
               result(function.shape());
             assign(result, function);
@@ -367,7 +367,7 @@ namespace librapid {
         LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE auto
         cudaTupleEvaluatorImpl(const detail::Function<descriptor, Functor, Args...> &function) {
             array::ArrayContainer<
-              decltype(function.shape()),
+              typename std::decay_t<decltype(function.shape())>,
               CudaStorage<typename detail::Function<descriptor, Functor, Args...>::Scalar>>
               result(function.shape());
             assign(result, function);

librapid/include/librapid/array/function.hpp

@@ -155,6 +155,8 @@ namespace librapid {
 
 			LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE auto size() const -> size_t;
 
+			LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE auto ndim() const -> size_t;
+
 			/// Return the shape of the Function's result
 			/// \return The shape of the Function's result
 			LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE auto shape() const -> const ShapeType &;
@@ -229,6 +231,11 @@ namespace librapid {
 			return m_size;
 		}
 
+		template<typename desc, typename Functor, typename... Args>
+		LIBRAPID_ALWAYS_INLINE auto Function<desc, Functor, Args...>::ndim() const -> size_t {
+			return m_shape.ndim();
+		}
+
 		template<typename desc, typename Functor, typename... Args>
 		LIBRAPID_ALWAYS_INLINE auto &Function<desc, Functor, Args...>::args() const {
 			return m_args;
@@ -237,7 +244,7 @@ namespace librapid {
 		template<typename desc, typename Functor, typename... Args>
 		LIBRAPID_ALWAYS_INLINE auto
 		Function<desc, Functor, Args...>::operator[](int64_t index) const {
-			return array::GeneralArrayView(*this)[index];
+			return createGeneralArrayView(*this)[index];
 		}
 
 		template<typename desc, typename Functor, typename... Args>

librapid/include/librapid/array/generalArrayView.hpp

@@ -7,7 +7,10 @@ namespace librapid {
 		struct TypeInfo<array::GeneralArrayView<T, S>> {
 			static constexpr detail::LibRapidType type = detail::LibRapidType::GeneralArrayView;
 			using Scalar							   = typename TypeInfo<std::decay_t<T>>::Scalar;
-			using Backend							 = typename TypeInfo<std::decay_t<T>>::Backend;
+			using Backend		= typename TypeInfo<std::decay_t<T>>::Backend;
+			using ArrayViewType = std::decay_t<T>;
+			using ShapeType		= typename TypeInfo<ArrayViewType>::ShapeType;
+			using StorageType	= typename TypeInfo<ArrayViewType>::StorageType;
 			static constexpr bool allowVectorisation = false;
 		};
 
@@ -21,7 +24,7 @@ namespace librapid {
 	}
 
 	template<typename ShapeType, typename T>
-	LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE auto createGeneralArrayView(T &&array) {
+	LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE auto createGeneralArrayViewShapeModifier(T &&array) {
 		return array::GeneralArrayView<T, ShapeType>(std::forward<T>(array));
 	}
 
@@ -219,7 +222,7 @@ namespace librapid {
 			  "Index {} out of bounds in ArrayContainer::operator[] with leading dimension={}",
 			  index,
 			  m_shape[0]);
-			auto view		  = createGeneralArrayView<Shape>(m_ref);
+			auto view		  = createGeneralArrayViewShapeModifier<Shape>(m_ref);
 			const auto stride = Stride(m_shape);
 			view.setShape(m_shape.subshape(1, ndim()));
 			if (ndim() == 1)
@@ -238,7 +241,7 @@ namespace librapid {
 			  "Index {} out of bounds in ArrayContainer::operator[] with leading dimension={}",
 			  index,
 			  m_shape[0]);
-			auto view		  = createGeneralArrayView<Shape>(m_ref);
+			auto view		  = createGeneralArrayViewShapeModifier<Shape>(m_ref);
 			const auto stride = Stride(m_shape);
 			view.setShape(m_shape.subshape(1, ndim()));
 			if (ndim() == 1)

librapid/include/librapid/array/linalg/transpose.hpp

@@ -7,7 +7,9 @@ namespace librapid {
 		struct TypeInfo<array::Transpose<T>> {
 			static constexpr detail::LibRapidType type = detail::LibRapidType::ArrayFunction;
 			using Scalar							   = typename TypeInfo<std::decay_t<T>>::Scalar;
-			using Backend							 = typename TypeInfo<std::decay_t<T>>::Backend;
+			using Backend	  = typename TypeInfo<std::decay_t<T>>::Backend;
+			using ShapeType	  = typename TypeInfo<std::decay_t<T>>::ShapeType;
+			using StorageType = typename TypeInfo<std::decay_t<T>>::StorageType;
 			static constexpr bool allowVectorisation = false;
 		};
 

librapid/include/librapid/array/shape.hpp

@@ -807,17 +807,25 @@ namespace librapid {
 			using Type = VectorShape;
 		};
 
-		template<typename First, typename Second, typename... Rest>
-		struct ShapeTypeHelper {
-			using FirstResult = typename ShapeTypeHelperImpl<First, Second>::Type;
-			using Type		  = typename ShapeTypeHelper<FirstResult, Rest...>::Type;
+		template<typename... Args>
+		struct ShapeTypeHelper;
+
+		template<typename First>
+		struct ShapeTypeHelper<First> {
+			using Type = First;
 		};
 
 		template<typename First, typename Second>
 		struct ShapeTypeHelper<First, Second> {
 			using Type = typename ShapeTypeHelperImpl<First, Second>::Type;
 		};
 
+		template<typename First, typename Second, typename... Rest>
+		struct ShapeTypeHelper<First, Second, Rest...> {
+			using FirstResult = typename ShapeTypeHelperImpl<First, Second>::Type;
+			using Type		  = typename ShapeTypeHelper<FirstResult, Rest...>::Type;
+		};
+
 		template<typename T>
 		struct SubscriptShapeType {
 			using Type = Shape;

librapid/include/librapid/array/storage.hpp

@@ -148,8 +148,8 @@ namespace librapid {
 		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE Pointer begin() noexcept;
 		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE Pointer end() noexcept;
 
-		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE ConstIterator begin() const noexcept;
-		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE ConstIterator end() const noexcept;
+		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE ConstPointer begin() const noexcept;
+		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE ConstPointer end() const noexcept;
 
 		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE ConstIterator cbegin() const noexcept;
 		LIBRAPID_NODISCARD LIBRAPID_ALWAYS_INLINE ConstIterator cend() const noexcept;
@@ -391,17 +391,22 @@ namespace librapid {
 			return ptr_;
 		}
 
-		template<typename T>
-		void fastCopy(T *__restrict dst, const T *__restrict src, size_t size) {
+		template<typename T, typename V>
+		void fastCopy(T *__restrict dst, const V *__restrict src, size_t size) {
 			LIBRAPID_ASSUME(size > 0);
 			LIBRAPID_ASSUME(dst != nullptr);
 			LIBRAPID_ASSUME(src != nullptr);
 
-			if (typetraits::TriviallyDefaultConstructible<T>::value) {
-				// Use a slightly faster memcpy if the type is trivially default constructible
-				std::uninitialized_copy(src, src + size, dst);
+			if constexpr (std::is_same_v<T, V>) {
+				if constexpr (typetraits::TriviallyDefaultConstructible<T>::value) {
+					// Use a slightly faster memcpy if the type is trivially default constructible
+					std::uninitialized_copy(src, src + size, dst);
+				} else {
+					// Otherwise, use the standard copy algorithm
+					std::copy(src, src + size, dst);
+				}
 			} else {
-				// Otherwise, use the standard copy algorithm
+				// Cannot use memcpy if the types are different
 				std::copy(src, src + size, dst);
 			}
 		}
@@ -441,14 +446,14 @@ namespace librapid {
 	template<typename V>
 	Storage<T>::Storage(const std::initializer_list<V> &list) :
 			m_begin(nullptr), m_size(0), m_ownsData(true) {
-		initData(list.begin(), list.end());
+		initData(static_cast<const V *>(list.begin()), static_cast<const V *>(list.end()));
 	}
 
 	template<typename T>
 	template<typename V>
 	Storage<T>::Storage(const std::vector<V> &vector) :
 			m_begin(nullptr), m_size(0), m_ownsData(true) {
-		initData(vector.begin(), vector.end());
+		initData(static_cast<const V *>(vector.data()), vector.size());
 	}
 
 	template<typename T>
@@ -620,12 +625,12 @@ namespace librapid {
 	}
 
 	template<typename T>
-	auto Storage<T>::begin() const noexcept -> ConstIterator {
+	auto Storage<T>::begin() const noexcept -> ConstPointer {
 		return m_begin;
 	}
 
 	template<typename T>
-	auto Storage<T>::end() const noexcept -> ConstIterator {
+	auto Storage<T>::end() const noexcept -> ConstPointer {
 		return m_begin + m_size;
 	}