Merge pull request #50 from petiaccja/bump-dependencies

bump xsimd and catch2

MathterConfig.cmake.in

@@ -7,7 +7,7 @@ set_and_check(Mathter_INCLUDE_DIR "@PACKAGE_INCLUDE_INSTALL_DIR@")
 
 include(CMakeFindDependencyMacro)
 if (${MATHTER_ENABLE_SIMD})
-	find_dependency(xsimd)
+	find_dependency(xsimd 13)
 endif()
 
 include("${CMAKE_CURRENT_LIST_DIR}/MathterTargets.cmake")

benchmark/main.cpp

@@ -20,31 +20,50 @@ using namespace std;
 using namespace mathter;
 
 
+void PrintArch(std::string_view name, unsigned supported) {
+	std::cout << name << (supported ? "YES" : "NO") << std::endl;
+}
+
+
+template <class T, int Dim>
+void PrintVectorType(std::string_view name) {
+	std::cout << "  " << name << ": " << (IsBatched<T, Dim, false>() ? "YES" : "NO") << " - " << sizeof(Vector<T, Dim, false>) << " bytes" << std::endl;
+}
+
+
 void DisplayArchitectureInfo() {
 #if MATHTER_ENABLE_SIMD
 	std::cout << "Available on CPU: " << std::endl;
 	const auto architectures = xsimd::available_architectures();
-	std::cout << "  sse2: " << (architectures.sse2 ? "YES" : "NO") << std::endl;
-	std::cout << "  sse3: " << (architectures.sse3 ? "YES" : "NO") << std::endl;
-	std::cout << "  ssse3: " << (architectures.ssse3 ? "YES" : "NO") << std::endl;
-	std::cout << "  sse4_1: " << (architectures.sse4_1 ? "YES" : "NO") << std::endl;
-	std::cout << "  sse4_2: " << (architectures.sse4_2 ? "YES" : "NO") << std::endl;
-	std::cout << "  sse4a: " << (architectures.sse4a ? "YES" : "NO") << std::endl;
-	std::cout << "  fma3_sse: " << (architectures.fma3_sse ? "YES" : "NO") << std::endl;
-	std::cout << "  fma4: " << (architectures.fma4 ? "YES" : "NO") << std::endl;
-	std::cout << "  xop: " << (architectures.xop ? "YES" : "NO") << std::endl;
-	std::cout << "  avx: " << (architectures.avx ? "YES" : "NO") << std::endl;
-	std::cout << "  fma3_avx: " << (architectures.fma3_avx ? "YES" : "NO") << std::endl;
-	std::cout << "  avx2: " << (architectures.avx2 ? "YES" : "NO") << std::endl;
-	std::cout << "  fma3_avx2: " << (architectures.fma3_avx2 ? "YES" : "NO") << std::endl;
-	std::cout << "  avx512f: " << (architectures.avx512f ? "YES" : "NO") << std::endl;
-	std::cout << "  avx512cd: " << (architectures.avx512cd ? "YES" : "NO") << std::endl;
-	std::cout << "  avx512dq: " << (architectures.avx512dq ? "YES" : "NO") << std::endl;
-	std::cout << "  avx512bw: " << (architectures.avx512bw ? "YES" : "NO") << std::endl;
-	std::cout << "  neon: " << (architectures.neon ? "YES" : "NO") << std::endl;
-	std::cout << "  neon64: " << (architectures.neon64 ? "YES" : "NO") << std::endl;
-	std::cout << "  sve: " << (architectures.sve ? "YES" : "NO") << std::endl;
-	std::cout << std::endl;
+
+	PrintArch("sse2", architectures.sse2);
+	PrintArch("sse3", architectures.sse3);
+	PrintArch("ssse3", architectures.ssse3);
+	PrintArch("sse4_1", architectures.sse4_1);
+	PrintArch("sse4_2", architectures.sse4_2);
+	PrintArch("fma3_sse42", architectures.fma3_sse42);
+	PrintArch("fma4", architectures.fma4);
+	PrintArch("avx", architectures.avx);
+	PrintArch("fma3_avx", architectures.fma3_avx);
+	PrintArch("avx2", architectures.avx2);
+	PrintArch("avxvnni", architectures.avxvnni);
+	PrintArch("fma3_avx2", architectures.fma3_avx2);
+	PrintArch("avx512f", architectures.avx512f);
+	PrintArch("avx512cd", architectures.avx512cd);
+	PrintArch("avx512dq", architectures.avx512dq);
+	PrintArch("avx512bw", architectures.avx512bw);
+	PrintArch("avx512er", architectures.avx512er);
+	PrintArch("avx512pf", architectures.avx512pf);
+	PrintArch("avx512ifma", architectures.avx512ifma);
+	PrintArch("avx512vbmi", architectures.avx512vbmi);
+	PrintArch("avx512vnni_bw", architectures.avx512vnni_bw);
+	PrintArch("avx512vnni_vbmi", architectures.avx512vnni_vbmi);
+	PrintArch("neon", architectures.neon);
+	PrintArch("neon64", architectures.neon64);
+	PrintArch("i8mm_neon64", architectures.i8mm_neon64);
+	PrintArch("sve", architectures.sve);
+	PrintArch("rvv", architectures.rvv);
+	PrintArch("wasm", architectures.wasm);
 
 	std::cout << "Enabled in build: " << std::endl;
 	xsimd::all_architectures::for_each([](const auto& arch) {
@@ -63,12 +82,24 @@ int main(int argc, char* argv[]) {
 	DisplayArchitectureInfo();
 
 	std::cout << "SIMD support:" << std::endl;
-	std::cout << "  float.2: " << (IsBatched<float, 2, false>() ? "YES" : "NO") << " - " << sizeof(Vector<float, 2, false>) << " bytes" << std::endl;
-	std::cout << "  float.3: " << (IsBatched<float, 3, false>() ? "YES" : "NO") << " - " << sizeof(Vector<float, 3, false>) << " bytes" << std::endl;
-	std::cout << "  float.4: " << (IsBatched<float, 4, false>() ? "YES" : "NO") << " - " << sizeof(Vector<float, 4, false>) << " bytes" << std::endl;
-	std::cout << "  double.2: " << (IsBatched<double, 2, false>() ? "YES" : "NO") << " - " << sizeof(Vector<double, 2, false>) << " bytes" << std::endl;
-	std::cout << "  double.3: " << (IsBatched<double, 3, false>() ? "YES" : "NO") << " - " << sizeof(Vector<double, 3, false>) << " bytes" << std::endl;
-	std::cout << "  double.4: " << (IsBatched<double, 4, false>() ? "YES" : "NO") << " - " << sizeof(Vector<double, 4, false>) << " bytes" << std::endl;
+	PrintVectorType<float, 2>("float2");
+	PrintVectorType<float, 3>("float3");
+	PrintVectorType<float, 4>("float4");
+	PrintVectorType<double, 2>("double2");
+	PrintVectorType<double, 3>("double3");
+	PrintVectorType<double, 4>("double4");
+	PrintVectorType<std::complex<float>, 2>("c_float2");
+	PrintVectorType<std::complex<float>, 3>("c_float3");
+	PrintVectorType<std::complex<float>, 4>("c_float4");
+	PrintVectorType<std::complex<double>, 2>("c_double2");
+	PrintVectorType<std::complex<double>, 3>("c_double3");
+	PrintVectorType<std::complex<double>, 4>("c_double4");
+	PrintVectorType<int32_t, 2>("i32_2");
+	PrintVectorType<int32_t, 3>("i32_3");
+	PrintVectorType<int32_t, 4>("i32_4");
+	PrintVectorType<int64_t, 2>("i64_2");
+	PrintVectorType<int64_t, 3>("i64_3");
+	PrintVectorType<int64_t, 4>("i64_4");
 	std::cout << std::endl;
 
 	int ret = Catch::Session().run(argc, argv);

conanfile.txt

@@ -1,6 +1,6 @@
-[requires]
-catch2/3.5.3
-xsimd/11.1.0
+[requires]
+catch2/3.7.0
+xsimd/13.0.0
 
 [generators]
 CMakeDeps

include/Mathter/CMakeLists.txt

@@ -67,6 +67,11 @@ target_sources(Mathter
 		"Vector/SIMDUtil.hpp"
 		"Vector/Swizzle.hpp"
 		"Vector/Vector.hpp"
+		# Swizzles
+		"Vector/SwizzleInc/Swizzle1.hpp.inc"
+		"Vector/SwizzleInc/Swizzle2.hpp.inc"
+		"Vector/SwizzleInc/Swizzle3.hpp.inc"
+		"Vector/SwizzleInc/Swizzle4.hpp.inc"
 )
 
 set_target_properties(Mathter PROPERTIES VERIFY_INTERFACE_HEADER_SETS ON)
@@ -83,7 +88,7 @@ endif()
 
 if (${MATHTER_ENABLE_SIMD})
 	message("Vectorization: ON (using XSimd).")
-	find_package(xsimd REQUIRED)
+	find_package(xsimd 13 REQUIRED)
 	target_link_libraries(Mathter INTERFACE xsimd)
 	target_compile_definitions(Mathter INTERFACE MATHTER_ENABLE_SIMD=1)
 else()

include/Mathter/Vector/OperationUtil.hpp

@@ -29,7 +29,7 @@ Batch FillMasked(Batch batch, Element value) {
 	}
 	else {
 		const auto fillers = Batch{ value };
-		const auto mask = xsimd::make_batch_bool_constant<Batch, MaskGenerator>();
+		const auto mask = xsimd::make_batch_bool_constant<typename Batch::value_type, typename Batch::arch_type, MaskGenerator>();
 		return xsimd::select(mask, batch, fillers);
 	}
 #else
@@ -52,7 +52,7 @@ Batch FillMaskedWithFirst(Batch batch) {
 
 	using UintBatch = xsimd::batch<Uint, typename Batch::arch_type>;
 
-	const auto mask = xsimd::make_batch_constant<UintBatch, Generator>();
+	const auto mask = xsimd::make_batch_constant<typename UintBatch::value_type, typename UintBatch::arch_type, Generator>();
 	return xsimd::swizzle(batch, mask);
 #else
 	return batch;

include/Mathter/Vector/Swizzle.hpp

@@ -200,7 +200,7 @@ typename Swizzle<T, Dim, Packed, Indices...>::TargetStorage Swizzle<T, Dim, Pack
 		using SourceIntBatch = xsimd::batch<Integer, typename SourceBatch::arch_type>;
 
 		const auto sourceBatch = SourceBatch::load_aligned(array.data());
-		const auto mask = xsimd::make_batch_constant<SourceIntBatch, impl::LinearizationGenerator<Indices...>>();
+		const auto mask = xsimd::make_batch_constant<typename SourceIntBatch::value_type, typename SourceIntBatch::arch_type, impl::LinearizationGenerator<Indices...>>();
 		const auto linearizedBatch = xsimd::swizzle(sourceBatch, mask);
 
 		alignas(GetStorageAlignment<T, Dim, Packed>()) SourceStorage linearized;
@@ -226,7 +226,7 @@ auto Swizzle<T, Dim, Packed, Indices...>::Delinearize(const TargetStorage& targe
 		alignas(GetStorageAlignment<T, Dim, Packed>()) SourceStorage targetPadded;
 		std::copy(target.begin(), target.end(), targetPadded.begin());
 		const auto targetBatch = SourceBatch::load_aligned(targetPadded.data());
-		const auto mask = xsimd::make_batch_constant<SourceIntBatch, impl::DelinearizationGenerator<Indices...>>();
+		const auto mask = xsimd::make_batch_constant<typename SourceIntBatch::value_type, typename SourceIntBatch::arch_type, impl::DelinearizationGenerator<Indices...>>();
 		const auto delinearizedBatch = xsimd::swizzle(targetBatch, mask);
 
 		alignas(GetStorageAlignment<T, Dim, Packed>()) SourceStorage delinearized;
@@ -250,7 +250,7 @@ auto Swizzle<T, Dim, Packed, Indices...>::Blend(const SourceStorage& old, const
 	if constexpr (isSourceBatched && sizeof...(Indices) <= Dim) {
 		const auto oldBatch = SourceBatch::load_unaligned(old.data());
 		const auto freshBatch = SourceBatch::load_unaligned(fresh.data());
-		const auto mask = xsimd::make_batch_bool_constant<SourceBatch, impl::BlendGenerator<Indices...>>();
+		const auto mask = xsimd::make_batch_bool_constant<typename SourceBatch::value_type, typename SourceBatch::arch_type, impl::BlendGenerator<Indices...>>();
 		const auto blendedBatch = xsimd::select(mask, freshBatch, oldBatch);
 
 		alignas(GetStorageAlignment<T, Dim, Packed>()) SourceStorage blended;