switch C code prefix to NIF

2026-09-03 20:04:20 +08:00
parent 7c7101ee48
commit 41ca31fd9a
51 changed files with 1047 additions and 1030 deletions
@@ -75,7 +75,7 @@ namespace Nitload.Common.HighPerformance.ProcessorAbstraction
            NumaFreeNative(address, size, numaId);
        }

-        [LibraryImport(Native.LibName, EntryPoint = "sn_get_numa_each")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_get_numa_each")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        public static partial int GetNumaEach(
@@ -84,7 +84,7 @@ namespace Nitload.Common.HighPerformance.ProcessorAbstraction
                                  Span<uint> globalCpuIdInNuma
        );

-        [LibraryImport(Native.LibName, EntryPoint = "sn_get_numa_overall")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_get_numa_overall")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        public static partial int GetNumaOverall(
@@ -119,7 +119,7 @@ namespace Nitload.Common.HighPerformance.ProcessorAbstraction
            }
        }

-        [LibraryImport(Native.LibName, EntryPoint = "sn_set_thread_affinity")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_set_thread_affinity")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        internal static partial void SetThreadAffinity(
@@ -138,7 +138,7 @@ namespace Nitload.Common.HighPerformance.ProcessorAbstraction
            return new(numaId, count, globalIndex);
        }

-        [LibraryImport(Native.LibName, EntryPoint = "sn_numa_alloc")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_numa_alloc")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        private static unsafe partial void* NumaAllocNative(
@@ -146,7 +146,7 @@ namespace Nitload.Common.HighPerformance.ProcessorAbstraction
                                            int numaId
        );

-        [LibraryImport(Native.LibName, EntryPoint = "sn_numa_free")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_numa_free")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        private static unsafe partial void NumaFreeNative(
@@ -51,7 +51,7 @@ namespace Nitload.Threading

 #pragma warning disable CA5393

-        [LibraryImport(Native.LibName, EntryPoint = "sn_set_thread_affinity")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_set_thread_affinity")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        private static partial void SetThreadAffinity(
@@ -47,13 +47,13 @@ namespace Nitload.Graphics.OpenGL

        /// <summary>
        /// <para>Forwards to the native OpenGL entry point.</para>
-        /// <para><c>void glGetShaderPrecisionFormat(GLenum shadertype, GLenum precisiontype, GLint *range, GLint *precision);</c></para>
+        /// <para><c>void glGetShaderPrecisionFormat(GLenum shaderType, GLenum precisiontype, GLint *range, GLint *precision);</c></para>
        /// <para>See <see href="https://docs.gl/gl4/glGetShaderPrecisionFormat">docs.gl</see>.</para>
        /// </summary>
        [MethodImpl(MethodImplOptions.AggressiveInlining)]
-        public void GetShaderPrecisionFormat(ShaderType shadertype, PrecisionType precisiontype, int* range, int* precision)
+        public void GetShaderPrecisionFormat(ShaderType shaderType, PrecisionType precisiontype, int* range, int* precision)
        {
-            __context->glGetShaderPrecisionFormat((uint)shadertype, (uint)precisiontype, range, precision);
+            __context->glGetShaderPrecisionFormat((uint)shaderType, (uint)precisiontype, range, precision);
        }

        /// <summary>
@@ -176,24 +176,24 @@ namespace Nitload.Graphics.OpenGL

        /// <summary>
        /// <para>Forwards to the native OpenGL entry point.</para>
-        /// <para><c>void glMultiDrawArraysIndirect(GLenum mode, const void *indirect, GLsizei drawcount, GLsizei stride);</c></para>
+        /// <para><c>void glMultiDrawArraysIndirect(GLenum mode, const void *indirect, GLsizei drawCount, GLsizei stride);</c></para>
        /// <para>See <see href="https://docs.gl/gl4/glMultiDrawArraysIndirect">docs.gl</see>.</para>
        /// </summary>
        [MethodImpl(MethodImplOptions.AggressiveInlining)]
-        public void MultiDrawArraysIndirect(PrimitiveType mode, void* indirect, int drawcount, int stride)
+        public void MultiDrawArraysIndirect(PrimitiveType mode, void* indirect, int drawCount, int stride)
        {
-            __context->glMultiDrawArraysIndirect((uint)mode, indirect, drawcount, stride);
+            __context->glMultiDrawArraysIndirect((uint)mode, indirect, drawCount, stride);
        }

        /// <summary>
        /// <para>Forwards to the native OpenGL entry point.</para>
-        /// <para><c>void glMultiDrawElementsIndirect(GLenum mode, GLenum type, const void *indirect, GLsizei drawcount, GLsizei stride);</c></para>
+        /// <para><c>void glMultiDrawElementsIndirect(GLenum mode, GLenum type, const void *indirect, GLsizei drawCount, GLsizei stride);</c></para>
        /// <para>See <see href="https://docs.gl/gl4/glMultiDrawElementsIndirect">docs.gl</see>.</para>
        /// </summary>
        [MethodImpl(MethodImplOptions.AggressiveInlining)]
-        public void MultiDrawElementsIndirect(PrimitiveType mode, DrawElementsType type, void* indirect, int drawcount, int stride)
+        public void MultiDrawElementsIndirect(PrimitiveType mode, DrawElementsType type, void* indirect, int drawCount, int stride)
        {
-            __context->glMultiDrawElementsIndirect((uint)mode, (uint)type, indirect, drawcount, stride);
+            __context->glMultiDrawElementsIndirect((uint)mode, (uint)type, indirect, drawCount, stride);
        }

        /// <summary>
@@ -43,7 +43,7 @@ namespace Nitload.Graphics

        public static GLFWFunction glfwContext => _glfw;

-        [LibraryImport(Native.LibName, EntryPoint = "load_glfw_functions")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_load_glfw_functions")]
        internal static unsafe partial void glfwLoadLibrary(
                                            GLFWFunction* context
        );
@@ -170,7 +170,7 @@ namespace Nitload.Mathematics.Numeric

        #region C invoke

-        [LibraryImport(Native.LibName, EntryPoint = "sn_compare_simd")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_compare_simd")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        private static partial int CompareSIMDID(
@@ -178,7 +178,7 @@ namespace Nitload.Mathematics.Numeric
                                   SIMDID right
        );

-        [LibraryImport(Native.LibName, EntryPoint = "load_intrinsic_function")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_load_intrinsic_function")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        private static partial void LoadIntrinsicFunction(
@@ -186,7 +186,7 @@ namespace Nitload.Mathematics.Numeric
                                    ulong id
        );

-        [LibraryImport(Native.LibName, EntryPoint = "sn_get_unite_simd")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_get_unite_simd")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        private static partial SIMDID UniteSIMDID(
@@ -155,12 +155,12 @@ namespace Nitload.Mathematics.Numeric
        }

 #pragma warning disable CA5393
-        [LibraryImport(Native.LibName, EntryPoint = "sn_get_simd_level_local")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_get_simd_level_local")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        internal static partial ulong GetSimdId();

-        [LibraryImport(Native.LibName, EntryPoint = "load_intrinsic_function")]
+        [LibraryImport(Native.LibName, EntryPoint = "nif_load_intrinsic_function")]
        [UnmanagedCallConv(CallConvs = [typeof(CallConvCdecl)])]
        [DefaultDllImportSearchPaths(DllImportSearchPath.AssemblyDirectory)]
        private static unsafe partial void LoadIntrinsicFunction(
@@ -2,8 +2,8 @@
 ---
 AccessModifierOffset: -4
 AlignAfterOpenBracket: Align
-AlignConsecutiveAssignments: false
-AlignConsecutiveDeclarations: false
+AlignConsecutiveAssignments: true
+AlignConsecutiveDeclarations: true
 AlignEscapedNewlines: Left
 AlignOperands: true
 AlignTrailingComments: false
@@ -1,2 +1,2 @@
 CompileFlags:
-  CompilationDatabase: .
+  CompilationDatabase: ../..
@@ -3,94 +3,112 @@ project(Nitload.Native LANGUAGES C)

 include(${CMAKE_CURRENT_LIST_DIR}/ExternalLibs.cmake OPTIONAL)

-set(SN_TARGET_OS "auto" CACHE STRING "Target OS identifier: auto, windows, linux, macos, ios, android, unknown")
-set_property(CACHE SN_TARGET_OS PROPERTY STRINGS auto windows linux macos ios android unknown)
+set(NIF_TARGET_OS "auto" CACHE STRING "Target OS identifier: auto, windows, linux, macos, ios, android, unknown")
+set_property(CACHE NIF_TARGET_OS PROPERTY STRINGS auto windows linux macos ios android unknown)

-set(SN_TARGET_ARCH "auto" CACHE STRING "Target ISA identifier: auto, x86_64, arm64, riscv64, ppc64le")
-set_property(CACHE SN_TARGET_ARCH PROPERTY STRINGS auto x86_64 arm64 riscv64 ppc64le)
+set(NIF_TARGET_ARCH "auto" CACHE STRING "Target ISA identifier: auto, x86_64, arm64, riscv64, ppc64le")
+set_property(CACHE NIF_TARGET_ARCH PROPERTY STRINGS auto x86_64 arm64 riscv64 ppc64le)

-sn_resolve_target_identifiers(
-  "${SN_TARGET_OS}"
-  "${SN_TARGET_ARCH}"
-  SN_RESOLVED_OS_NAME
-  SN_RESOLVED_OS_DEFINE
-  SN_RESOLVED_ARCH_NAME
-  SN_RESOLVED_ARCH_DEFINE)
+nif_resolve_target_identifiers(
+  "${NIF_TARGET_OS}"
+  "${NIF_TARGET_ARCH}"
+  NIF_RESOLVED_OS_NAME
+  NIF_RESOLVED_OS_DEFINE
+  NIF_RESOLVED_ARCH_NAME
+  NIF_RESOLVED_ARCH_DEFINE)

-file(GLOB STGSHARP_NATIVE_SOURCES CONFIGURE_DEPENDS
-  ${CMAKE_CURRENT_LIST_DIR}/sn-src/*.c)
+set(NIF_NATIVE_SOURCES
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/dllmain.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/load_glfw.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/load_intrinsic.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/m.bufop.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/m.solve.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/m.transpose.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/mk.fma.pack.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/mk.fma.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/mk.p_scan.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/s.factorial.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/s.hash.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/simdid_compare.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/simdid.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/sk.fma.x86.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/thread.linux.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/thread.windows.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/vector.c
+  ${CMAKE_CURRENT_LIST_DIR}/nif-src/vector_dot.c)

-file(GLOB STGSHARP_NATIVE_HEADERS CONFIGURE_DEPENDS
-  ${CMAKE_CURRENT_LIST_DIR}/sn-head/*.h)
+set(NIF_NATIVE_HEADERS
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/glfw_function.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/nif_internal.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/nif_intrinsic.context.base.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/nif_intrinsic.context.x86.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/nif_intrinsic.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/nif_intrinsic.matkernel.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/nif_intrinsic.std.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/nif_target.h
+  ${CMAKE_CURRENT_LIST_DIR}/nif-head/NitloadFrameWork.h)

-file(GLOB STGSHARP_NATIVE_TEMPLATES CONFIGURE_DEPENDS
-  ${CMAKE_CURRENT_LIST_DIR}/sn-template/*.h
-  ${CMAKE_CURRENT_LIST_DIR}/sn-template/*.inl
-  ${CMAKE_CURRENT_LIST_DIR}/sn-template/*.inc
-  ${CMAKE_CURRENT_LIST_DIR}/sn-template/*.ipp)
-
-if(NOT STGSHARP_NATIVE_SOURCES)
-  message(FATAL_ERROR "No source files found under sn-src/.")
-endif()
+set(NIF_NATIVE_TEMPLATES
+  ${CMAKE_CURRENT_LIST_DIR}/nif-template/nif_template.fma.x86.h)

 # Nitload.Native now lives under <root>/src/Nitload.Native, so walk up two
 # levels (project dir -> src -> solution root) to reach the solution root.
-get_filename_component(STGSHARP_NATIVE_SRC_ROOT ${CMAKE_CURRENT_LIST_DIR} DIRECTORY)
-get_filename_component(STGSHARP_SOLUTION_ROOT ${STGSHARP_NATIVE_SRC_ROOT} DIRECTORY)
+get_filename_component(NIF_NATIVE_SRC_ROOT ${CMAKE_CURRENT_LIST_DIR} DIRECTORY)
+get_filename_component(NIF_SOLUTION_ROOT ${NIF_NATIVE_SRC_ROOT} DIRECTORY)

-if(SN_RESOLVED_OS_NAME STREQUAL "windows")
-  set(STGSHARP_OS_TAG win)
-elseif(SN_RESOLVED_OS_NAME STREQUAL "linux")
-  set(STGSHARP_OS_TAG linux)
+if(NIF_RESOLVED_OS_NAME STREQUAL "windows")
+  set(NIF_OS_TAG win)
+elseif(NIF_RESOLVED_OS_NAME STREQUAL "linux")
+  set(NIF_OS_TAG linux)
 else()
-  set(STGSHARP_OS_TAG unknown)
+  set(NIF_OS_TAG unknown)
 endif()

-if(SN_RESOLVED_ARCH_NAME STREQUAL "x86_64")
-  set(STGSHARP_ARCH_TAG x64)
+if(NIF_RESOLVED_ARCH_NAME STREQUAL "x86_64")
+  set(NIF_ARCH_TAG x64)
 else()
-  set(STGSHARP_ARCH_TAG ${SN_RESOLVED_ARCH_NAME})
+  set(NIF_ARCH_TAG ${NIF_RESOLVED_ARCH_NAME})
 endif()

-set(STGSHARP_BUILD_OUTPUT_DIR ${STGSHARP_SOLUTION_ROOT}/bin/${STGSHARP_OS_TAG}-${STGSHARP_ARCH_TAG})
+set(NIF_BUILD_OUTPUT_DIR ${NIF_SOLUTION_ROOT}/bin/${NIF_OS_TAG}-${NIF_ARCH_TAG})
 # Compile-time intermediates (native PDB / .obj side files) live inside the CMake
 # build tree (cmake_build/<preset>) so they get cleaned up with it.
-set(STGSHARP_OBJECT_ROOT_DIR ${CMAKE_CURRENT_BINARY_DIR}/obj/${PROJECT_NAME})
-set(STGSHARP_NATIVE_PROJECT_FILES
-  ${STGSHARP_NATIVE_SOURCES}
-  ${STGSHARP_NATIVE_HEADERS}
-  ${STGSHARP_NATIVE_TEMPLATES})
+set(NIF_OBJECT_ROOT_DIR ${CMAKE_CURRENT_BINARY_DIR}/obj/${PROJECT_NAME})
+set(NIF_NATIVE_PROJECT_FILES
+  ${NIF_NATIVE_SOURCES}
+  ${NIF_NATIVE_HEADERS}
+  ${NIF_NATIVE_TEMPLATES})

-add_library(${PROJECT_NAME} SHARED ${STGSHARP_NATIVE_PROJECT_FILES})
+add_library(${PROJECT_NAME} SHARED ${NIF_NATIVE_PROJECT_FILES})

-source_group(TREE ${CMAKE_CURRENT_LIST_DIR} FILES ${STGSHARP_NATIVE_PROJECT_FILES})
+source_group(TREE ${CMAKE_CURRENT_LIST_DIR} FILES ${NIF_NATIVE_PROJECT_FILES})

 set_target_properties(${PROJECT_NAME} PROPERTIES
  C_STANDARD 11
  C_STANDARD_REQUIRED ON
-  RUNTIME_OUTPUT_DIRECTORY ${STGSHARP_BUILD_OUTPUT_DIR}
-  LIBRARY_OUTPUT_DIRECTORY ${STGSHARP_BUILD_OUTPUT_DIR}
-  ARCHIVE_OUTPUT_DIRECTORY ${STGSHARP_BUILD_OUTPUT_DIR})
+  RUNTIME_OUTPUT_DIRECTORY ${NIF_BUILD_OUTPUT_DIR}
+  LIBRARY_OUTPUT_DIRECTORY ${NIF_BUILD_OUTPUT_DIR}
+  ARCHIVE_OUTPUT_DIRECTORY ${NIF_BUILD_OUTPUT_DIR})

 foreach(_config Debug Release RelWithDebInfo MinSizeRel)
  string(TOUPPER ${_config} _config_upper)
  set_target_properties(${PROJECT_NAME} PROPERTIES
-    RUNTIME_OUTPUT_DIRECTORY_${_config_upper} ${STGSHARP_BUILD_OUTPUT_DIR}
-    LIBRARY_OUTPUT_DIRECTORY_${_config_upper} ${STGSHARP_BUILD_OUTPUT_DIR}
-    ARCHIVE_OUTPUT_DIRECTORY_${_config_upper} ${STGSHARP_BUILD_OUTPUT_DIR}
-    PDB_OUTPUT_DIRECTORY_${_config_upper} ${STGSHARP_BUILD_OUTPUT_DIR}
-    COMPILE_PDB_OUTPUT_DIRECTORY_${_config_upper} ${STGSHARP_OBJECT_ROOT_DIR}/${_config}/${STGSHARP_ARCH_TAG})
+    RUNTIME_OUTPUT_DIRECTORY_${_config_upper} ${NIF_BUILD_OUTPUT_DIR}
+    LIBRARY_OUTPUT_DIRECTORY_${_config_upper} ${NIF_BUILD_OUTPUT_DIR}
+    ARCHIVE_OUTPUT_DIRECTORY_${_config_upper} ${NIF_BUILD_OUTPUT_DIR}
+    PDB_OUTPUT_DIRECTORY_${_config_upper} ${NIF_BUILD_OUTPUT_DIR}
+    COMPILE_PDB_OUTPUT_DIRECTORY_${_config_upper} ${NIF_OBJECT_ROOT_DIR}/${_config}/${NIF_ARCH_TAG})
 endforeach()

 set_target_properties(${PROJECT_NAME} PROPERTIES
-  COMPILE_PDB_OUTPUT_DIRECTORY ${STGSHARP_OBJECT_ROOT_DIR}/$<CONFIG>/${STGSHARP_ARCH_TAG})
+  COMPILE_PDB_OUTPUT_DIRECTORY ${NIF_OBJECT_ROOT_DIR}/$<CONFIG>/${NIF_ARCH_TAG})

 if(MSVC OR (CMAKE_C_COMPILER_ID STREQUAL "Clang" AND CMAKE_C_COMPILER_FRONTEND_VARIANT STREQUAL "MSVC"))
  target_compile_options(${PROJECT_NAME} PRIVATE
-    /Fd${STGSHARP_OBJECT_ROOT_DIR}/$<CONFIG>/${STGSHARP_ARCH_TAG}/)
+    /Fd${NIF_OBJECT_ROOT_DIR}/$<CONFIG>/${NIF_ARCH_TAG}/)
 endif()

-if(CMAKE_C_COMPILER_ID STREQUAL "Clang" AND SN_RESOLVED_ARCH_NAME STREQUAL "x86_64")
+if(CMAKE_C_COMPILER_ID STREQUAL "Clang" AND NIF_RESOLVED_ARCH_NAME STREQUAL "x86_64")
  # Keep the intrinsic type system and builtin declarations visible to all TUs.
  # The runtime dispatch contract still follows T_arch/T_feature at the function level.
  if(CMAKE_C_COMPILER_FRONTEND_VARIANT STREQUAL "MSVC")
@@ -120,19 +138,19 @@ endif()

 target_include_directories(${PROJECT_NAME}
  PRIVATE
-    ${CMAKE_CURRENT_LIST_DIR}/sn-head
-    ${CMAKE_CURRENT_LIST_DIR}/sn-template
+    ${CMAKE_CURRENT_LIST_DIR}/nif-head
+    ${CMAKE_CURRENT_LIST_DIR}/nif-template
    ${CMAKE_CURRENT_LIST_DIR}/lib)

 target_compile_definitions(${PROJECT_NAME}
  PRIVATE
-    SN_EXTERN_TARGET
-    SN_OS_TARGET=${SN_RESOLVED_OS_DEFINE}
-    SN_ARCH_TARGET=${SN_RESOLVED_ARCH_DEFINE})
+    NIF_EXTERN_TARGET
+    NIF_OS_TARGET=${NIF_RESOLVED_OS_DEFINE}
+    NIF_ARCH_TARGET=${NIF_RESOLVED_ARCH_DEFINE})

-set(STGSHARP_NATIVE_EXTERNAL_TARGET ${PROJECT_NAME})
+set(NIF_NATIVE_EXTERNAL_TARGET ${PROJECT_NAME})
 add_subdirectory(${CMAKE_CURRENT_LIST_DIR}/lib)

-message(STATUS "Nitload.Native target OS: ${SN_RESOLVED_OS_NAME} (${SN_RESOLVED_OS_DEFINE})")
-message(STATUS "Nitload.Native target ISA: ${SN_RESOLVED_ARCH_NAME} (${SN_RESOLVED_ARCH_DEFINE})")
-message(STATUS "Nitload.Native output dir: ${STGSHARP_BUILD_OUTPUT_DIR}")
+message(STATUS "Nitload.Native target OS: ${NIF_RESOLVED_OS_NAME} (${NIF_RESOLVED_OS_DEFINE})")
+message(STATUS "Nitload.Native target ISA: ${NIF_RESOLVED_ARCH_NAME} (${NIF_RESOLVED_ARCH_DEFINE})")
+message(STATUS "Nitload.Native output dir: ${NIF_BUILD_OUTPUT_DIR}")
@@ -24,18 +24,6 @@
                "CMAKE_C_COMPILER": "clang",
                "CMAKE_CXX_COMPILER": "clang++"
            }
-        },
-        {
-            "name": "使用工具链文件配置预设",
-            "displayName": "使用工具链文件配置预设",
-            "description": "设置 Ninja 生成器、版本和安装目录",
-            "generator": "Ninja",
-            "binaryDir": "${sourceDir}/../../cmake_build/${presetName}",
-            "cacheVariables": {
-                "CMAKE_BUILD_TYPE": "Debug",
-                "CMAKE_TOOLCHAIN_FILE": "",
-                "CMAKE_INSTALL_PREFIX": "${sourceDir}/out/install/${presetName}"
-            }
        }
    ],
    "buildPresets": [
@@ -48,4 +36,4 @@
            "configurePreset": "clang-release"
        }
    ]
-}
+}
@@ -1,74 +1,74 @@
-function(sn_normalize_os_identifier INPUT_VALUE OUTPUT_NAME OUTPUT_DEFINE)
-    string(TOLOWER "${INPUT_VALUE}" _sn_os)
+function(nif_normalize_os_identifier INPUT_VALUE OUTPUT_NAME OUTPUT_DEFINE)
+    string(TOLOWER "${INPUT_VALUE}" _nif_os)

-    if(_sn_os STREQUAL "auto")
-        string(TOLOWER "${CMAKE_SYSTEM_NAME}" _sn_system_name)
-        if(_sn_system_name STREQUAL "windows")
-            set(_sn_os "windows")
-        elseif(_sn_system_name STREQUAL "linux")
-            set(_sn_os "linux")
+    if(_nif_os STREQUAL "auto")
+        string(TOLOWER "${CMAKE_SYSTEM_NAME}" _nif_system_name)
+        if(_nif_system_name STREQUAL "windows")
+            set(_nif_os "windows")
+        elseif(_nif_system_name STREQUAL "linux")
+            set(_nif_os "linux")
        endif()
    endif()

-    if(_sn_os STREQUAL "windows")
-        set(_sn_define SN_OS_WINDOWS)
-    elseif(_sn_os STREQUAL "linux")
-        set(_sn_define SN_OS_LINUX)
+    if(_nif_os STREQUAL "windows")
+        set(_nif_define NIF_OS_WINDOWS)
+    elseif(_nif_os STREQUAL "linux")
+        set(_nif_define NIF_OS_LINUX)
    else()
-        message(FATAL_ERROR "Unsupported SN_TARGET_OS='${INPUT_VALUE}'.")
+        message(FATAL_ERROR "Unsupported NIF_TARGET_OS='${INPUT_VALUE}'.")
    endif()

-    set(${OUTPUT_NAME} "${_sn_os}" PARENT_SCOPE)
-    set(${OUTPUT_DEFINE} "${_sn_define}" PARENT_SCOPE)
+    set(${OUTPUT_NAME} "${_nif_os}" PARENT_SCOPE)
+    set(${OUTPUT_DEFINE} "${_nif_define}" PARENT_SCOPE)
 endfunction()

-function(sn_normalize_arch_identifier INPUT_VALUE OUTPUT_NAME OUTPUT_DEFINE)
-    string(TOLOWER "${INPUT_VALUE}" _sn_arch)
+function(nif_normalize_arch_identifier INPUT_VALUE OUTPUT_NAME OUTPUT_DEFINE)
+    string(TOLOWER "${INPUT_VALUE}" _nif_arch)

-    if(_sn_arch STREQUAL "auto")
-        string(TOLOWER "${CMAKE_SYSTEM_PROCESSOR}" _sn_processor)
-        if(_sn_processor MATCHES "^(amd64|x86_64|x64)$")
-            set(_sn_arch "x86_64")
-        elseif(_sn_processor MATCHES "^(arm64|aarch64)$")
-            set(_sn_arch "arm64")
-        elseif(_sn_processor STREQUAL "riscv64")
-            set(_sn_arch "riscv64")
-        elseif(_sn_processor MATCHES "^(ppc64le|powerpc64le)$")
-            set(_sn_arch "ppc64le")
+    if(_nif_arch STREQUAL "auto")
+        string(TOLOWER "${CMAKE_SYSTEM_PROCESSOR}" _nif_processor)
+        if(_nif_processor MATCHES "^(amd64|x86_64|x64)$")
+            set(_nif_arch "x86_64")
+        elseif(_nif_processor MATCHES "^(arm64|aarch64)$")
+            set(_nif_arch "arm64")
+        elseif(_nif_processor STREQUAL "riscv64")
+            set(_nif_arch "riscv64")
+        elseif(_nif_processor MATCHES "^(ppc64le|powerpc64le)$")
+            set(_nif_arch "ppc64le")
        else()
-            message(FATAL_ERROR "Unable to infer ISA from CMAKE_SYSTEM_PROCESSOR='${CMAKE_SYSTEM_PROCESSOR}'. Set SN_TARGET_ARCH explicitly.")
+            message(FATAL_ERROR "Unable to infer ISA from CMAKE_SYSTEM_PROCESSOR='${CMAKE_SYSTEM_PROCESSOR}'. Set NIF_TARGET_ARCH explicitly.")
        endif()
    endif()

-    if(_sn_arch STREQUAL "x86_64")
-        set(_sn_define SN_ARCH_X86_64)
-    elseif(_sn_arch STREQUAL "arm64")
-        set(_sn_define SN_ARCH_ARM64)
-    elseif(_sn_arch STREQUAL "riscv64")
-        set(_sn_define SN_ARCH_RISCV64)
-    elseif(_sn_arch STREQUAL "ppc64le")
-        set(_sn_define SN_ARCH_PPC64le)
+    if(_nif_arch STREQUAL "x86_64")
+        set(_nif_define NIF_ARCH_X86_64)
+    elseif(_nif_arch STREQUAL "arm64")
+        set(_nif_define NIF_ARCH_ARM64)
+    elseif(_nif_arch STREQUAL "riscv64")
+        set(_nif_define NIF_ARCH_RISCV64)
+    elseif(_nif_arch STREQUAL "ppc64le")
+        set(_nif_define NIF_ARCH_PPC64le)
    else()
-        message(FATAL_ERROR "Unsupported SN_TARGET_ARCH='${INPUT_VALUE}'.")
+        message(FATAL_ERROR "Unsupported NIF_TARGET_ARCH='${INPUT_VALUE}'.")
    endif()

-    set(${OUTPUT_NAME} "${_sn_arch}" PARENT_SCOPE)
-    set(${OUTPUT_DEFINE} "${_sn_define}" PARENT_SCOPE)
+    set(${OUTPUT_NAME} "${_nif_arch}" PARENT_SCOPE)
+    set(${OUTPUT_DEFINE} "${_nif_define}" PARENT_SCOPE)
 endfunction()

-function(sn_resolve_target_identifiers INPUT_OS INPUT_ARCH OUTPUT_OS_NAME OUTPUT_OS_DEFINE OUTPUT_ARCH_NAME OUTPUT_ARCH_DEFINE)
-    sn_normalize_os_identifier("${INPUT_OS}" _sn_os_name _sn_os_define)
-    sn_normalize_arch_identifier("${INPUT_ARCH}" _sn_arch_name _sn_arch_define)
+function(nif_resolve_target_identifiers INPUT_OS INPUT_ARCH OUTPUT_OS_NAME OUTPUT_OS_DEFINE OUTPUT_ARCH_NAME OUTPUT_ARCH_DEFINE)
+    nif_normalize_os_identifier("${INPUT_OS}" _nif_os_name _nif_os_define)
+    nif_normalize_arch_identifier("${INPUT_ARCH}" _nif_arch_name _nif_arch_define)

-    set(${OUTPUT_OS_NAME} "${_sn_os_name}" PARENT_SCOPE)
-    set(${OUTPUT_OS_DEFINE} "${_sn_os_define}" PARENT_SCOPE)
-    set(${OUTPUT_ARCH_NAME} "${_sn_arch_name}" PARENT_SCOPE)
-    set(${OUTPUT_ARCH_DEFINE} "${_sn_arch_define}" PARENT_SCOPE)
+    set(${OUTPUT_OS_NAME} "${_nif_os_name}" PARENT_SCOPE)
+    set(${OUTPUT_OS_DEFINE} "${_nif_os_define}" PARENT_SCOPE)
+    set(${OUTPUT_ARCH_NAME} "${_nif_arch_name}" PARENT_SCOPE)
+    set(${OUTPUT_ARCH_DEFINE} "${_nif_arch_define}" PARENT_SCOPE)
 endfunction()

-function(sn_register_external_package TARGET PACKAGE_NAME PACKAGE_DIR)
+function(nif_register_external_package TARGET PACKAGE_NAME PACKAGE_DIR)
    if(NOT TARGET ${TARGET})
-        message(FATAL_ERROR "sn_register_external_package: target '${TARGET}' does not exist.")
+        message(FATAL_ERROR "nif_register_external_package: target '${TARGET}' does not exist.")
    endif()

    if(EXISTS "${PACKAGE_DIR}")
@@ -79,7 +79,7 @@ function(sn_register_external_package TARGET PACKAGE_NAME PACKAGE_DIR)
    endif()
 endfunction()

-function(sn_link_external_binary_if_exists TARGET LIBRARY_PATH)
+function(nif_link_external_binary_if_exists TARGET LIBRARY_PATH)
    if(EXISTS "${LIBRARY_PATH}")
        target_link_libraries(${TARGET} PRIVATE "${LIBRARY_PATH}")
        message(STATUS "Linked external binary: ${LIBRARY_PATH}")
@@ -110,7 +110,7 @@
      <LanguageStandard_C>stdc11</LanguageStandard_C>
      <MultiProcessorCompilation>true</MultiProcessorCompilation>
      <FavorSizeOrSpeed>Speed</FavorSizeOrSpeed>
-      <AdditionalIncludeDirectories>$(ProjectDir)lib;$(ProjectDir)sn-head;$(ProjectDir)sn-template;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
+      <AdditionalIncludeDirectories>$(ProjectDir)lib;$(ProjectDir)nif-head;$(ProjectDir)nif-template;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
    </ClCompile>
    <Link>
      <!-- DLL 应该用 Windows，不是 Console -->
@@ -172,23 +172,23 @@
  <!--  source files                                            -->
  <!-- ════════════════════════════════════════════════════════ -->
  <ItemGroup>
-    <ClCompile Include="sn-src\dllmain.c" />
-    <ClCompile Include="sn-src\m.bufop.x86.c" />
-    <ClCompile Include="sn-src\m.solve.x86.c" />
-    <ClCompile Include="sn-src\m.transpose.x86.c" />
-    <ClCompile Include="sn-src\mk.fma.x86.c" />
-    <ClCompile Include="sn-src\mk.p_scan.x86.c" />
-    <ClCompile Include="sn-src\s.factorial.x86.c" />
-    <ClCompile Include="sn-src\sk.fma.x86.c" />
-    <ClCompile Include="sn-src\s.hash.x86.c" />
-    <ClCompile Include="sn-src\load_intrinsic.x86.c" />
-    <ClCompile Include="sn-src\load_glfw.c" />
-    <ClCompile Include="sn-src\simdid.x86.c" />
-    <ClCompile Include="sn-src\simdid_compare.x86.c" />
-    <ClCompile Include="sn-src\thread.linux.c" />
-    <ClCompile Include="sn-src\thread.windows.c" />
-    <ClCompile Include="sn-src\vector.c" />
-    <ClCompile Include="sn-src\vector_dot.c" />
+    <ClCompile Include="nif-src\dllmain.c" />
+    <ClCompile Include="nif-src\m.bufop.x86.c" />
+    <ClCompile Include="nif-src\m.solve.x86.c" />
+    <ClCompile Include="nif-src\m.transpose.x86.c" />
+    <ClCompile Include="nif-src\mk.fma.x86.c" />
+    <ClCompile Include="nif-src\mk.p_scan.x86.c" />
+    <ClCompile Include="nif-src\s.factorial.x86.c" />
+    <ClCompile Include="nif-src\sk.fma.x86.c" />
+    <ClCompile Include="nif-src\s.hash.x86.c" />
+    <ClCompile Include="nif-src\load_intrinsic.x86.c" />
+    <ClCompile Include="nif-src\load_glfw.c" />
+    <ClCompile Include="nif-src\simdid.x86.c" />
+    <ClCompile Include="nif-src\simdid_compare.x86.c" />
+    <ClCompile Include="nif-src\thread.linux.c" />
+    <ClCompile Include="nif-src\thread.windows.c" />
+    <ClCompile Include="nif-src\vector.c" />
+    <ClCompile Include="nif-src\vector_dot.c" />
  </ItemGroup>
  <!-- ════════════════════════════════════════════════════════ -->
  <!--  header files                                            -->
@@ -197,15 +197,15 @@
    <ClInclude Include="lib\glfw\glfw3.h" />
    <ClInclude Include="lib\glfw\glfw3native.h" />
    <ClInclude Include="lib\stbi\stb_image.h" />
-    <ClInclude Include="sn-head\glfw_function.h" />
-    <ClInclude Include="sn-head\sn_internal.h" />
-    <ClInclude Include="sn-head\sn_intrinsic.context.x86.h" />
-    <ClInclude Include="sn-head\sn_intrinsic.std.h" />
-    <ClInclude Include="sn-head\sn_intrinsic.h" />
-    <ClInclude Include="sn-head\sn_intrinsic.matkernel.h" />
-    <ClInclude Include="sn-head\sn_target.h" />
-    <ClInclude Include="sn-head\StgSharpNative.h" />
-    <ClInclude Include="sn-template\sn_template.fma.x86.h" />
+    <ClInclude Include="nif-head\glfw_function.h" />
+    <ClInclude Include="nif-head\nif_internal.h" />
+    <ClInclude Include="nif-head\nif_intrinsic.context.x86.h" />
+    <ClInclude Include="nif-head\nif_intrinsic.std.h" />
+    <ClInclude Include="nif-head\nif_intrinsic.h" />
+    <ClInclude Include="nif-head\nif_intrinsic.matkernel.h" />
+    <ClInclude Include="nif-head\nif_target.h" />
+    <ClInclude Include="nif-head\NitloadFrameWork.h" />
+    <ClInclude Include="nif-template\nif_template.fma.x86.h" />
  </ItemGroup>
  <ItemGroup>
    <None Include="naming.md" />
@@ -21,69 +21,69 @@
    </Filter>
  </ItemGroup>
  <ItemGroup>
-    <ClCompile Include="sn-src\dllmain.c">
+    <ClCompile Include="nif-src\dllmain.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\s.factorial.x86.c">
+    <ClCompile Include="nif-src\s.factorial.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\s.hash.x86.c">
+    <ClCompile Include="nif-src\s.hash.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\vector.c">
+    <ClCompile Include="nif-src\vector.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\vector_dot.c">
+    <ClCompile Include="nif-src\vector_dot.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\load_glfw.c">
+    <ClCompile Include="nif-src\load_glfw.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\sk.fma.x86.c">
+    <ClCompile Include="nif-src\sk.fma.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\load_intrinsic.x86.c">
+    <ClCompile Include="nif-src\load_intrinsic.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\m.bufop.x86.c">
+    <ClCompile Include="nif-src\m.bufop.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\m.solve.x86.c">
+    <ClCompile Include="nif-src\m.solve.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\simdid.x86.c">
+    <ClCompile Include="nif-src\simdid.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\mk.fma.x86.c">
+    <ClCompile Include="nif-src\mk.fma.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\simdid_compare.x86.c">
+    <ClCompile Include="nif-src\simdid_compare.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\thread.linux.c">
+    <ClCompile Include="nif-src\thread.linux.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\thread.windows.c">
+    <ClCompile Include="nif-src\thread.windows.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\m.transpose.x86.c">
+    <ClCompile Include="nif-src\m.transpose.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
-    <ClCompile Include="sn-src\mk.p_scan.x86.c">
+    <ClCompile Include="nif-src\mk.p_scan.x86.c">
      <Filter>源文件</Filter>
    </ClCompile>
  </ItemGroup>
  <ItemGroup>
-    <ClInclude Include="sn-head\sn_internal.h">
+    <ClInclude Include="nif-head\nif_internal.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-head\StgSharpNative.h">
+    <ClInclude Include="nif-head\NitloadFrameWork.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-head\glfw_function.h">
+    <ClInclude Include="nif-head\glfw_function.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-head\sn_intrinsic.matkernel.h">
+    <ClInclude Include="nif-head\nif_intrinsic.matkernel.h">
      <Filter>头文件</Filter>
    </ClInclude>
    <ClInclude Include="lib\glfw\glfw3.h">
@@ -95,19 +95,19 @@
    <ClInclude Include="lib\stbi\stb_image.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-head\sn_target.h">
+    <ClInclude Include="nif-head\nif_target.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-head\sn_intrinsic.context.x86.h">
+    <ClInclude Include="nif-head\nif_intrinsic.context.x86.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-head\sn_intrinsic.h">
+    <ClInclude Include="nif-head\nif_intrinsic.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-head\sn_intrinsic.std.h">
+    <ClInclude Include="nif-head\nif_intrinsic.std.h">
      <Filter>头文件</Filter>
    </ClInclude>
-    <ClInclude Include="sn-template\sn_template.fma.x86.h">
+    <ClInclude Include="nif-template\nif_template.fma.x86.h">
      <Filter>模板文件</Filter>
    </ClInclude>
  </ItemGroup>
@@ -217,7 +217,7 @@ Detection:
 - **Unknown hybrid** (flag set but core type unrecognised): leave topology at 00.
 - **SIMD_DUAL**: set when any of the following: `AVX Feature DUAL=1` (glued YMM), or `AVX512_IMPL=DUAL`.

-`sn_get_unite_simd` computes topology bits [1..0] with bitwise **AND** (conservative intersection — weaker topology wins) and SIMD_DUAL with bitwise **OR** (dual if either core is dual). `sn_compare_simd` compares topology value numerically as step 6 and SIMD_DUAL as the final (step 7) tie-break.
+`nif_get_unite_simd` computes topology bits [1..0] with bitwise **AND** (conservative intersection — weaker topology wins) and SIMD_DUAL with bitwise **OR** (dual if either core is dual). `nif_compare_simd` compares topology value numerically as step 6 and SIMD_DUAL as the final (step 7) tie-break.

 ---

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.16)

-if(NOT DEFINED STGSHARP_NATIVE_EXTERNAL_TARGET)
-  message(FATAL_ERROR "STGSHARP_NATIVE_EXTERNAL_TARGET must be set before adding lib/.")
+if(NOT DEFINED NIF_NATIVE_EXTERNAL_TARGET)
+  message(FATAL_ERROR "NIF_NATIVE_EXTERNAL_TARGET must be set before adding lib/.")
 endif()

 include(${CMAKE_CURRENT_LIST_DIR}/../ExternalLibs.cmake OPTIONAL)
@@ -17,6 +17,6 @@ foreach(_pkg ${LIB_PACKAGE_DIRS})
  if(EXISTS ${_pkg_dir}/CMakeLists.txt)
    add_subdirectory(${_pkg_dir})
  else()
-    sn_register_external_package(${STGSHARP_NATIVE_EXTERNAL_TARGET} ${_pkg} ${_pkg_dir})
+    nif_register_external_package(${NIF_NATIVE_EXTERNAL_TARGET} ${_pkg} ${_pkg_dir})
  endif()
 endforeach()
@@ -1,11 +1,11 @@
 cmake_minimum_required(VERSION 3.16)

-sn_register_external_package(${STGSHARP_NATIVE_EXTERNAL_TARGET} glfw ${CMAKE_CURRENT_LIST_DIR})
+nif_register_external_package(${NIF_NATIVE_EXTERNAL_TARGET} glfw ${CMAKE_CURRENT_LIST_DIR})

 if(WIN32)
  if(NOT EXISTS ${CMAKE_CURRENT_LIST_DIR}/glfw3_mt.lib)
    message(FATAL_ERROR "Required GLFW static library not found: ${CMAKE_CURRENT_LIST_DIR}/glfw3_mt.lib")
  endif()

-  sn_link_external_binary_if_exists(${STGSHARP_NATIVE_EXTERNAL_TARGET} ${CMAKE_CURRENT_LIST_DIR}/glfw3_mt.lib)
+  nif_link_external_binary_if_exists(${NIF_NATIVE_EXTERNAL_TARGET} ${CMAKE_CURRENT_LIST_DIR}/glfw3_mt.lib)
 endif()
@@ -1,3 +1,3 @@
 cmake_minimum_required(VERSION 3.16)

-sn_register_external_package(${STGSHARP_NATIVE_EXTERNAL_TARGET} stbi ${CMAKE_CURRENT_LIST_DIR})
+nif_register_external_package(${NIF_NATIVE_EXTERNAL_TARGET} stbi ${CMAKE_CURRENT_LIST_DIR})
@@ -0,0 +1,35 @@
+#ifndef NIF_NATIVE_H
+#define NIF_NATIVE_H
+
+#ifdef __cplusplus
+#error "This header is not meant to be included in C++ code. Nitload.Native is a C library and should be used with C linkage."
+#endif
+
+#include "nif_internal.h"
+#include "nif_target.h"
+
+#include <immintrin.h>
+#include <assert.h>
+#include <glfw_function.h>
+
+#define byte char
+#define cs_char uint16_t
+#define nif_null_assert(ptr) assert(ptr != NULL)
+
+#ifndef NIF_
+#endif
+
+NIF_API void NIF_DECL nif_load_glfw_functions(GLFWFunctionTable *table);
+NIF_API void NIF_DECL nif_load_intrinsic_function(void *intrinsic_context, uint64_t id);
+NIF_API SIMDID NIF_DECL nif_get_simd_level_global();
+/* Per-core variant: always re-executes CPUID, never caches.
+ * Must be called on the target core (after affinity is set). */
+NIF_API SIMDID NIF_DECL nif_get_simd_level_local();
+NIF_API SIMDID NIF_DECL nif_get_unite_simd(SIMDID left, SIMDID right);
+NIF_API int NIF_DECL nif_compare_simd(SIMDID left, SIMDID right);
+NIF_API void NIF_DECL nif_set_thread_affinity(int core_id, int numa_id);
+NIF_API void* NIF_DECL nif_numa_alloc(size_t size, int numa_id);
+NIF_API void NIF_DECL nif_numa_free(void *ptr, size_t size, int numa_id);
+NIF_API int NIF_DECL nif_get_numa_each(int numa_id, uint64_t *internal, uint32_t *global);
+NIF_API int NIF_DECL nif_get_numa_overall(int *out_numa_count, int *count_of_each_numa);
+#endif
@@ -1,6 +1,6 @@

-#ifndef SN_INTERNAL
-#define SN_INTERNAL
+#ifndef NIF_INTERNAL
+#define NIF_INTERNAL

 #include <xmmintrin.h>

@@ -9,33 +9,33 @@
 #ifdef _MSC_VER
 #define _CRT_SECURE_NO_WARNINGS 1
 #define WIN32_LEAN_AND_MEAN
-#define SN_DECL __cdecl
-#define SN_MAT_DECL __cdecl __forceinline
-#define SN_API __declspec(dllexport)
+#define NIF_DECL __cdecl
+#define NIF_MAT_DECL __cdecl __forceinline
+#define NIF_API __declspec(dllexport)
 #define FORCEINLINE __forceinline
 #define NOINLINE __declspec(noinline)
 #include <windows.h>
 #elif defined(__clang__) || defined(__GNUC__)
-#define SN_DECL __attribute__((cdecl))
-#define SN_MAT_DECL __attribute__((cdecl)) __attribute__((always_inline))
-#define SN_API __attribute__((dllexport))
+#define NIF_DECL __attribute__((cdecl))
+#define NIF_MAT_DECL __attribute__((cdecl)) __attribute__((always_inline))
+#define NIF_API __attribute__((dllexport))
 #define FORCEINLINE __attribute__((always_inline))
 #define NOINLINE __attribute__((noinline))
 #else
-#define SN_API
-#define SN_DECL
-#define SN_MAT_DECL
+#define NIF_API
+#define NIF_DECL
+#define NIF_MAT_DECL
 #define FORCEINLINE
 #define NOINLINE
 #endif

 #if defined(_MSC_VER)
-#define SN_ALIGNAS(n) __declspec(align(n))
+#define NIF_ALIGNAS(n) __declspec(align(n))
 #else
-#define SN_ALIGNAS(n) __attribute__((aligned(n)))
+#define NIF_ALIGNAS(n) __attribute__((aligned(n)))
 #endif

-#define SN_APIDEF
+#define NIF_APIDEF

 #define PRIVATE static
 #define INTERNAL extern
@@ -1,5 +1,5 @@
-#ifndef SN_INTRINSIC_CONTEXT_BASE
-#define SN_INTRINSIC_CONTEXT_BASE
+#ifndef NIF_INTRINSIC_CONTEXT_BASE
+#define NIF_INTRINSIC_CONTEXT_BASE


 // -----------------------------------------------------------------------------
@@ -1,20 +1,20 @@
-#ifndef SN_INTRINSIC_CONTEXT_X86_H
-#define SN_INTRINSIC_CONTEXT_X86_H
+#ifndef NIF_INTRINSIC_CONTEXT_X86_H
+#define NIF_INTRINSIC_CONTEXT_X86_H

-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "sn_intrinsic.h"
-#include "sn_intrinsic.std.h"
-#include "sn_intrinsic.context.base.h"
+#include "nif_intrinsic.h"
+#include "nif_intrinsic.std.h"
+#include "nif_intrinsic.context.base.h"

-#ifndef SN_X86_TARGET_ATTR_sse
-#define SN_STD_TARGET_ATTR_IMPL(T_arch, T_feature) SN_X86_TARGET_ATTR_##T_arch##T_feature
-#define SN_X86_TARGET_ATTR_sse SN_CLANG_TARGET_ATTR("sse4.2")
-#define SN_X86_TARGET_ATTR_avx SN_CLANG_TARGET_ATTR("avx")
-#define SN_X86_TARGET_ATTR_avx_fma SN_CLANG_TARGET_ATTR("avx,avx2,fma")
-#define SN_X86_TARGET_ATTR_512 SN_CLANG_TARGET_ATTR("avx512f,avx512vl")
+#ifndef NIF_X86_TARGET_ATTR_sse
+#define NIF_STD_TARGET_ATTR_IMPL(T_arch, T_feature) NIF_X86_TARGET_ATTR_##T_arch##T_feature
+#define NIF_X86_TARGET_ATTR_sse NIF_CLANG_TARGET_ATTR("sse4.2")
+#define NIF_X86_TARGET_ATTR_avx NIF_CLANG_TARGET_ATTR("avx")
+#define NIF_X86_TARGET_ATTR_avx_fma NIF_CLANG_TARGET_ATTR("avx,avx2,fma")
+#define NIF_X86_TARGET_ATTR_512 NIF_CLANG_TARGET_ATTR("avx512f,avx512vl")
 #endif

 // -----------------------------------------------------------------------------
@@ -111,45 +111,45 @@

 INTERNAL void f32_normalize(VEC_SEGMENT(float) * source, VEC_SEGMENT(float) * target);

-SN_MK_PROC_DECL_STD(float, sse, , k_quality);
+NIF_MK_PROC_DECL_STD(float, sse, , k_quality);

-SN_MK_PROC_DECL_STD(float, sse, , add);
-SN_MK_PROC_DECL_STD(float, avx, , add);
-SN_MK_PROC_DECL_STD(float, 512, , add);
+NIF_MK_PROC_DECL_STD(float, sse, , add);
+NIF_MK_PROC_DECL_STD(float, avx, , add);
+NIF_MK_PROC_DECL_STD(float, 512, , add);

-SN_MK_PROC_DECL_STD(double, sse, , add);
-SN_MK_PROC_DECL_STD(double, avx, , add);
-SN_MK_PROC_DECL_STD(double, 512, , add);
+NIF_MK_PROC_DECL_STD(double, sse, , add);
+NIF_MK_PROC_DECL_STD(double, avx, , add);
+NIF_MK_PROC_DECL_STD(double, 512, , add);

-SN_MK_PROC_DECL_STD(float, sse, , sub);
-SN_MK_PROC_DECL_STD(float, avx, , sub);
-SN_MK_PROC_DECL_STD(float, 512, , sub);
+NIF_MK_PROC_DECL_STD(float, sse, , sub);
+NIF_MK_PROC_DECL_STD(float, avx, , sub);
+NIF_MK_PROC_DECL_STD(float, 512, , sub);

-SN_MK_PROC_DECL_STD(double, sse, , sub);
-SN_MK_PROC_DECL_STD(double, avx, , sub);
-SN_MK_PROC_DECL_STD(double, 512, , sub);
+NIF_MK_PROC_DECL_STD(double, sse, , sub);
+NIF_MK_PROC_DECL_STD(double, avx, , sub);
+NIF_MK_PROC_DECL_STD(double, 512, , sub);

-SN_MK_PROC_DECL_STD(float, sse, , transpose);
-SN_MK_PROC_DECL_STD(float, avx, , transpose);
-SN_MK_PROC_DECL_STD(float, 512, , transpose);
+NIF_MK_PROC_DECL_STD(float, sse, , transpose);
+NIF_MK_PROC_DECL_STD(float, avx, , transpose);
+NIF_MK_PROC_DECL_STD(float, 512, , transpose);

-SN_MK_PROC_DECL_STD(float, sse, , scalar_mul);
-SN_MK_PROC_DECL_STD(float, avx, , scalar_mul);
-SN_MK_PROC_DECL_STD(float, 512, , scalar_mul);
+NIF_MK_PROC_DECL_STD(float, sse, , scalar_mul);
+NIF_MK_PROC_DECL_STD(float, avx, , scalar_mul);
+NIF_MK_PROC_DECL_STD(float, 512, , scalar_mul);

-SN_MK_PROC_DECL_STD(double, sse, , scalar_mul);
-SN_MK_PROC_DECL_STD(double, avx, , scalar_mul);
-SN_MK_PROC_DECL_STD(double, 512, , scalar_mul);
+NIF_MK_PROC_DECL_STD(double, sse, , scalar_mul);
+NIF_MK_PROC_DECL_STD(double, avx, , scalar_mul);
+NIF_MK_PROC_DECL_STD(double, 512, , scalar_mul);

-SN_MK_PROC_DECL_STD(float, sse, , fill);
-SN_MK_PROC_DECL_STD(float, avx, , fill);
-SN_MK_PROC_DECL_STD(float, 512, , fill);
+NIF_MK_PROC_DECL_STD(float, sse, , fill);
+NIF_MK_PROC_DECL_STD(float, avx, , fill);
+NIF_MK_PROC_DECL_STD(float, 512, , fill);

-SN_MK_PROC_DECL_STD(double, sse, , fill);
-SN_MK_PROC_DECL_STD(double, avx, , fill);
-SN_MK_PROC_DECL_STD(double, 512, , fill);
+NIF_MK_PROC_DECL_STD(double, sse, , fill);
+NIF_MK_PROC_DECL_STD(double, avx, , fill);
+NIF_MK_PROC_DECL_STD(double, 512, , fill);

-SN_MK_PROC_DECL_STD(float, sse, , quality);
+NIF_MK_PROC_DECL_STD(float, sse, , quality);

 /*
 * pack left matrix data to L4 buffer
@@ -166,9 +166,9 @@ SN_MK_PROC_DECL_STD(float, sse, , quality);
 * scalar pack carries column length and offset data:
 * [column length, empty, empty, empty]
 */
-SN_MK_PROC_DECL_STD(float, sse, , fma_pack_left);
-SN_MK_PROC_DECL_STD(float, avx, , fma_pack_left);
-SN_MK_PROC_DECL_STD(float, 512, , fma_pack_left);
+NIF_MK_PROC_DECL_STD(float, sse, , fma_pack_left);
+NIF_MK_PROC_DECL_STD(float, avx, , fma_pack_left);
+NIF_MK_PROC_DECL_STD(float, 512, , fma_pack_left);

 /*
 * pack right matrix data to L4 buffer
@@ -185,104 +185,104 @@ SN_MK_PROC_DECL_STD(float, 512, , fma_pack_left);
 * scalar pack carries column length and offset data:
 * [column length, empty, empty, empty]
 */
-SN_MK_PROC_DECL_STD(float, sse, , fma_pack_right);
-SN_MK_PROC_DECL_STD(float, avx, , fma_pack_right);
-SN_MK_PROC_DECL_STD(float, 512, , fma_pack_right);
+NIF_MK_PROC_DECL_STD(float, sse, , fma_pack_right);
+NIF_MK_PROC_DECL_STD(float, avx, , fma_pack_right);
+NIF_MK_PROC_DECL_STD(float, 512, , fma_pack_right);

-SN_MK_PROC_DECL_STD(float, sse, , fma);
-SN_MK_PROC_DECL_STD(float, avx, , fma);
-SN_MK_PROC_DECL_STD(float, avx, _fma, fma);
-SN_MK_PROC_DECL_STD(float, 512, , fma);
+NIF_MK_PROC_DECL_STD(float, sse, , fma);
+NIF_MK_PROC_DECL_STD(float, avx, , fma);
+NIF_MK_PROC_DECL_STD(float, avx, _fma, fma);
+NIF_MK_PROC_DECL_STD(float, 512, , fma);

-SN_MK_PROC_DECL_STD(float, sse, , fma_unpack);
-SN_MK_PROC_DECL_STD(float, avx, , fma_unpack);
-SN_MK_PROC_DECL_STD(float, 512, , fma_unpack);
+NIF_MK_PROC_DECL_STD(float, sse, , fma_unpack);
+NIF_MK_PROC_DECL_STD(float, avx, , fma_unpack);
+NIF_MK_PROC_DECL_STD(float, 512, , fma_unpack);

-SN_MK_PROC_DECL_STD(double, sse, , fma_pack_left);
-SN_MK_PROC_DECL_STD(double, avx, , fma_pack_left);
-SN_MK_PROC_DECL_STD(double, 512, , fma_pack_left);
+NIF_MK_PROC_DECL_STD(double, sse, , fma_pack_left);
+NIF_MK_PROC_DECL_STD(double, avx, , fma_pack_left);
+NIF_MK_PROC_DECL_STD(double, 512, , fma_pack_left);

-SN_MK_PROC_DECL_STD(double, sse, , fma);
-SN_MK_PROC_DECL_STD(double, avx, , fma);
-SN_MK_PROC_DECL_STD(double, avx, _fma, fma);
-SN_MK_PROC_DECL_STD(double, 512, , fma);
+NIF_MK_PROC_DECL_STD(double, sse, , fma);
+NIF_MK_PROC_DECL_STD(double, avx, , fma);
+NIF_MK_PROC_DECL_STD(double, avx, _fma, fma);
+NIF_MK_PROC_DECL_STD(double, 512, , fma);

-SN_MK_PROC_DECL_STD(double, sse, , fma_unpack);
-SN_MK_PROC_DECL_STD(double, avx, , fma_unpack);
-SN_MK_PROC_DECL_STD(double, 512, , fma_unpack);
+NIF_MK_PROC_DECL_STD(double, sse, , fma_unpack);
+NIF_MK_PROC_DECL_STD(double, avx, , fma_unpack);
+NIF_MK_PROC_DECL_STD(double, 512, , fma_unpack);

-SN_SK_PROC_DECL_STD(float, avx, , add);
-SN_SK_PROC_DECL_STD(float, 512, , add);
+NIF_SK_PROC_DECL_STD(float, avx, , add);
+NIF_SK_PROC_DECL_STD(float, 512, , add);

-SN_SK_PROC_DECL_STD(double, sse, , add);
-SN_SK_PROC_DECL_STD(double, avx, , add);
-SN_SK_PROC_DECL_STD(double, 512, , add);
+NIF_SK_PROC_DECL_STD(double, sse, , add);
+NIF_SK_PROC_DECL_STD(double, avx, , add);
+NIF_SK_PROC_DECL_STD(double, 512, , add);

-SN_SK_PROC_DECL_STD(float, sse, , sub);
-SN_SK_PROC_DECL_STD(float, avx, , sub);
-SN_SK_PROC_DECL_STD(float, 512, , sub);
+NIF_SK_PROC_DECL_STD(float, sse, , sub);
+NIF_SK_PROC_DECL_STD(float, avx, , sub);
+NIF_SK_PROC_DECL_STD(float, 512, , sub);

-SN_SK_PROC_DECL_STD(double, sse, , sub);
-SN_SK_PROC_DECL_STD(double, avx, , sub);
-SN_SK_PROC_DECL_STD(double, 512, , sub);
+NIF_SK_PROC_DECL_STD(double, sse, , sub);
+NIF_SK_PROC_DECL_STD(double, avx, , sub);
+NIF_SK_PROC_DECL_STD(double, 512, , sub);

-SN_SK_PROC_DECL_STD(float, sse, , transpose);
-SN_SK_PROC_DECL_STD(float, avx, , transpose);
-SN_SK_PROC_DECL_STD(float, 512, , transpose);
+NIF_SK_PROC_DECL_STD(float, sse, , transpose);
+NIF_SK_PROC_DECL_STD(float, avx, , transpose);
+NIF_SK_PROC_DECL_STD(float, 512, , transpose);

-SN_SK_PROC_DECL_STD(float, sse, , scalar_mul);
-SN_SK_PROC_DECL_STD(float, avx, , scalar_mul);
-SN_SK_PROC_DECL_STD(float, 512, , scalar_mul);
+NIF_SK_PROC_DECL_STD(float, sse, , scalar_mul);
+NIF_SK_PROC_DECL_STD(float, avx, , scalar_mul);
+NIF_SK_PROC_DECL_STD(float, 512, , scalar_mul);

-SN_SK_PROC_DECL_STD(double, sse, , scalar_mul);
-SN_SK_PROC_DECL_STD(double, avx, , scalar_mul);
-SN_SK_PROC_DECL_STD(double, 512, , scalar_mul);
+NIF_SK_PROC_DECL_STD(double, sse, , scalar_mul);
+NIF_SK_PROC_DECL_STD(double, avx, , scalar_mul);
+NIF_SK_PROC_DECL_STD(double, 512, , scalar_mul);

-SN_SK_PROC_DECL_STD(float, sse, , fill);
-SN_SK_PROC_DECL_STD(float, avx, , fill);
-SN_SK_PROC_DECL_STD(float, 512, , fill);
+NIF_SK_PROC_DECL_STD(float, sse, , fill);
+NIF_SK_PROC_DECL_STD(float, avx, , fill);
+NIF_SK_PROC_DECL_STD(float, 512, , fill);

-SN_SK_PROC_DECL_STD(double, sse, , fill);
-SN_SK_PROC_DECL_STD(double, avx, , fill);
-SN_SK_PROC_DECL_STD(double, 512, , fill);
+NIF_SK_PROC_DECL_STD(double, sse, , fill);
+NIF_SK_PROC_DECL_STD(double, avx, , fill);
+NIF_SK_PROC_DECL_STD(double, 512, , fill);

-SN_SK_PROC_DECL_STD(float, sse, , quality);
+NIF_SK_PROC_DECL_STD(float, sse, , quality);

-SN_SK_PROC_DECL_STD(float, sse, , fma);
-SN_SK_PROC_DECL_STD(float, avx, , fma);
-SN_SK_PROC_DECL_STD(float, avx, _fma, fma);
-SN_SK_PROC_DECL_STD(float, 512, , fma);
+NIF_SK_PROC_DECL_STD(float, sse, , fma);
+NIF_SK_PROC_DECL_STD(float, avx, , fma);
+NIF_SK_PROC_DECL_STD(float, avx, _fma, fma);
+NIF_SK_PROC_DECL_STD(float, 512, , fma);

-SN_SK_PROC_DECL_STD(double, sse, , fma);
-SN_SK_PROC_DECL_STD(double, avx, , fma);
-SN_SK_PROC_DECL_STD(double, avx, _fma, fma);
-SN_SK_PROC_DECL_STD(double, 512, , fma);
+NIF_SK_PROC_DECL_STD(double, sse, , fma);
+NIF_SK_PROC_DECL_STD(double, avx, , fma);
+NIF_SK_PROC_DECL_STD(double, avx, _fma, fma);
+NIF_SK_PROC_DECL_STD(double, 512, , fma);

 #pragma endregion

 #pragma region scaler_simd

-INTERNAL uint64_t SN_DECL factorial_simd_sse(int n);
+INTERNAL uint64_t NIF_DECL factorial_simd_sse(int n);

 #pragma endregion

 #pragma region dot_product

-INTERNAL void FORCEINLINE SN_DECL dot_41_sse(MAT_KERNEL(float) const *transpose,
+INTERNAL void FORCEINLINE NIF_DECL dot_41_sse(MAT_KERNEL(float) const *transpose,
                                             __m128 const *vector, __m128 *ans);
-INTERNAL void SN_DECL dot_41_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
+INTERNAL void NIF_DECL dot_41_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
                                 __m128 *ans);
-INTERNAL void SN_DECL dot_42_sse(MAT_KERNEL(float) const *transpose, __m128 const *vector,
+INTERNAL void NIF_DECL dot_42_sse(MAT_KERNEL(float) const *transpose, __m128 const *vector,
                                 __m128 *ans);
-INTERNAL void SN_DECL dot_42_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
+INTERNAL void NIF_DECL dot_42_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
                                 __m128 *ans);

 #pragma endregion

 #pragma region string

-INTERNAL int SN_DECL city_hash_simplify_sse(byte const *str, int length);
-INTERNAL int SN_DECL index_pair_sse(short const *stram, uint32_t target, int length);
+INTERNAL int NIF_DECL city_hash_simplify_sse(byte const *str, int length);
+INTERNAL int NIF_DECL index_pair_sse(short const *stram, uint32_t target, int length);

 #pragma endregion

@@ -3,12 +3,12 @@

 #define SSC_INTRIN_HEAD

-#include "StgSharpNative.h"
-#include "sn_target.h"
+#include "NitloadFrameWork.h"
+#include "nif_target.h"
 #include <xmmintrin.h>
 #include <immintrin.h>
 #include <smmintrin.h>
-#include "sn_intrinsic.matkernel.h"
+#include "nif_intrinsic.matkernel.h"
 #include <stdint.h>

 #ifdef _MSC_VER
@@ -28,7 +28,7 @@
        ((LASTTWO(s3) << 6) | (LASTTWO(s2) << 4) | (LASTTWO(s1) << 2) | (LASTTWO(s0)))
 #define BIT_CVRT(pack, index, T) *(T *)(pack->data + index)

-#if SN_IS_ARCH(x86_64)
+#if NIF_IS_ARCH(x86_64)

 #define vec_128 __m128
 #define vec_256 __m256
@@ -66,7 +66,7 @@ typedef union uint64_u {
 typedef struct mat_kernel mat_kernel;

 typedef union M128{
-        #if SN_IS_ARCH(x86_64)
+        #if NIF_IS_ARCH(x86_64)
        __m128 m128;
        #endif
        float f32[4];
@@ -76,7 +76,7 @@ typedef union M128{
 }M128;

 typedef union M256{
-        #if SN_IS_ARCH(x86_64)
+        #if NIF_IS_ARCH(x86_64)
        __m256 m256;
        #endif
        float f32[8];
@@ -86,7 +86,7 @@ typedef union M256{
 }M256;

 typedef union M512{
-        #if SN_IS_ARCH(x86_64)
+        #if NIF_IS_ARCH(x86_64)
        __m512 m512;
        #endif
        uint64_t u64[8];
@@ -97,24 +97,24 @@ typedef union M512{
        int64_t i64[8];
 }M512;

-typedef struct sn_L4_cache_head {
+typedef struct nif_L4_cache_head {
        uint64_t reverse_mask_0;
        uint32_t position_mask;
        uint32_t size;
        uint64_t reverse_mask_1;
        uint64_t reverse_mask_2;
-} sn_L4_cache_head;
+} nif_L4_cache_head;

-typedef struct sn_L4_handle {
+typedef struct nif_L4_handle {
        void *ptr;
-        sn_L4_cache_head* head;
-} sn_L4_handle;
+        nif_L4_cache_head* head;
+} nif_L4_handle;

 typedef struct matrix_task {
-        sn_L4_handle buffer_0;
-        sn_L4_handle buffer_1;
-        sn_L4_handle buffer_2;
-        sn_L4_handle buffer_3;
+        nif_L4_handle buffer_0;
+        nif_L4_handle buffer_1;
+        nif_L4_handle buffer_2;
+        nif_L4_handle buffer_3;
        union {
                struct {
                        int32_t reverse_0;
@@ -144,18 +144,18 @@ static_assert(sizeof(matrix_task) == 128, "matrix_parallel_task_package size mis
 static_assert(sizeof(matrix_task) == 128, "matrix_parallel_task_package size mismatch");
 #endif

-typedef void(SN_DECL *UNI_MK_PROC)(matrix_task *task);
-typedef void(SN_DECL *UNI_SK_PROC)(mat_kernel const *left, mat_kernel const *right,
+typedef void(NIF_DECL *UNI_MK_PROC)(matrix_task *task);
+typedef void(NIF_DECL *UNI_SK_PROC)(mat_kernel const *left, mat_kernel const *right,
                                   mat_kernel *restrict ans, uint64_t scalar);

-typedef void(SN_DECL *VECTORNORMALIZEPROC)(VEC_SEGMENT(float) *source, VEC_SEGMENT(float) *target);
-typedef void(SN_DECL *DOTPROC)(void *transpose, __m128 *vector, __m128 *ans);
-typedef uint64_t(SN_DECL *FACTORIALROC)(int n);
+typedef void(NIF_DECL *VECTORNORMALIZEPROC)(VEC_SEGMENT(float) *source, VEC_SEGMENT(float) *target);
+typedef void(NIF_DECL *DOTPROC)(void *transpose, __m128 *vector, __m128 *ans);
+typedef uint64_t(NIF_DECL *FACTORIALROC)(int n);

 #pragma region string

-typedef int(SN_DECL *HASH)(byte const *str, int length);
-typedef int(SN_DECL *INDEX_PAIR)(short const *stram, uint32_t target, int length);
+typedef int(NIF_DECL *HASH)(byte const *str, int length);
+typedef int(NIF_DECL *INDEX_PAIR)(short const *stram, uint32_t target, int length);

 #pragma endregion

@@ -168,13 +168,13 @@ typedef struct mat_sk_intrinsic {
 } mat_sk_intrinsic;

 // Mirrors C# IntrinsicContext field order so managed/unmanaged layouts stay in sync.
-typedef struct sn_intrinsic {
+typedef struct nif_intrinsic {
        HASH city_hash_simplify;
        FACTORIALROC factorial_simd;
        VECTORNORMALIZEPROC vec_f32_normalize_3;
        mat_mk_intrinsic mat_mk[0x0a];
        mat_sk_intrinsic mat_sk[0x0a];
-} sn_intrinsic;
+} nif_intrinsic;


 typedef enum mat_proc_index {
@@ -0,0 +1,54 @@
+#ifndef NIF_MATKERNEL
+#define NIF_MATKERNEL
+
+#include <immintrin.h>
+#include <stdint.h>
+
+#include "nif_internal.h"
+
+/* Kernel storage unions for a fixed 4x4 block. */
+#define ELEMENTSIZE(T) sizeof(T)
+#define XMMCOUNT(T) (sizeof(T))
+#define YMMCOUNT(T) (sizeof(T) / 2)
+#define ZMMCOUNT(T) (sizeof(T) / 4)
+
+#define NIF_MAT_KERNEL_MEMBER(T)        \
+        {                              \
+                __m128 x[XMMCOUNT(T)]; \
+                __m256 y[YMMCOUNT(T)]; \
+                __m512 z[ZMMCOUNT(T)]; \
+                T m[4][4];             \
+        }
+        
+#define NIF_MAT_KERNEL_DEFINE(T) \
+        typedef union NIF_MAT_KERNEL_TYPE_##T NIF_MAT_KERNEL_MEMBER(T) NIF_MAT_KERNEL_TYPE_##T
+
+#define MAT_KERNEL(T) NIF_MAT_KERNEL_TYPE(T)
+#define NIF_MAT_KERNEL_TYPE(T) NIF_MAT_KERNEL_TYPE_##T
+
+NIF_MAT_KERNEL_DEFINE(float);
+NIF_MAT_KERNEL_DEFINE(double);
+NIF_MAT_KERNEL_DEFINE(uint64_t);
+
+
+#define NIF_VEC_SEGMENT_MEMBER(T)               \
+        {                                \
+                __m128 f32_x[XMMCOUNT(T) / 4]; \
+                __m256 f32_y[YMMCOUNT(T) / 4]; \
+                __m512 f32_z[ZMMCOUNT(T) / 4]; \
+                T v[4];                        \
+        }
+
+#define NIF_VEC_SEGMENT_DEFINE(T) \
+        typedef union NIF_VEC_SEGMENT_TYPE_##T NIF_VEC_SEGMENT_MEMBER(T) NIF_VEC_SEGMENT_TYPE_##T
+
+#define VEC_SEGMENT(T) NIF_VEC_SEGMENT_TYPE(T)
+#define NIF_VEC_SEGMENT_TYPE(T) NIF_VEC_SEGMENT_TYPE_##T
+
+NIF_VEC_SEGMENT_DEFINE(float);
+NIF_VEC_SEGMENT_DEFINE(double);
+
+/* Global zero kernel declaration (defined in dllmain.c), aligned to 64 bytes */
+extern __declspec(align(64)) MAT_KERNEL(u64) NIF_ZERO_KERNEL;
+
+#endif /* NIF_MATKERNEL */
@@ -1,23 +1,23 @@
-#ifndef SN_INTRINSIC_CALLCONV_STD
-#define SN_INTRINSIC_CALLCONV_STD
+#ifndef NIF_INTRINSIC_CALLCONV_STD
+#define NIF_INTRINSIC_CALLCONV_STD

-#include "sn_internal.h"
-#include "sn_intrinsic.h"
+#include "nif_internal.h"
+#include "nif_intrinsic.h"

 #ifndef static_assert
 #define static_assert _Static_assert
 #endif

-#define SN_STD_TARGET_ATTR(T_arch, T_feature) SN_STD_TARGET_ATTR_IMPL(T_arch, T_feature)
+#define NIF_STD_TARGET_ATTR(T_arch, T_feature) NIF_STD_TARGET_ATTR_IMPL(T_arch, T_feature)

 #pragma region mk_proc attributes
 #define MK_PROC_NAME_STD(T_type, T_arch, T_feature, T_op) \
        T_type##_multi_kernel_proc_##T_op##_##T_arch##_##T_feature
 #define GET_MK_PROC_STD(T_type, T_arch, T_feature, T_op) \
        ((UNI_MK_PROC)MK_PROC_NAME_STD(T_type, T_arch, T_feature, T_op))
-#define SN_MK_PROC_DECL(T_type, P_name) INTERNAL void SN_DECL P_name(MAT_TASK(T_type) const *task)
-#define SN_MK_PROC_DECL_STD(T_type, T_arch, T_feature, T_op)                                   \
-        INTERNAL NOINLINE SN_STD_TARGET_ATTR(T_arch, T_feature) void SN_DECL MK_PROC_NAME_STD( \
+#define NIF_MK_PROC_DECL(T_type, P_name) INTERNAL void NIF_DECL P_name(MAT_TASK(T_type) const *task)
+#define NIF_MK_PROC_DECL_STD(T_type, T_arch, T_feature, T_op)                                   \
+        INTERNAL NOINLINE NIF_STD_TARGET_ATTR(T_arch, T_feature) void NIF_DECL MK_PROC_NAME_STD( \
                T_type, T_arch, T_feature, T_op)(MAT_TASK(T_type) const *task)
 #pragma endregion

@@ -26,8 +26,8 @@
        T_type##_single_kernel_proc_##T_op##_##T_arch##_##T_feature
 #define GET_SK_PROC_STD(T_type, T_arch, T_feature, T_op) \
        ((UNI_SK_PROC)SK_PROC_NAME_STD(T_type, T_arch, T_feature, T_op))
-#define SN_SK_PROC_DECL_STD(T_type, T_arch, T_feature, T_op)                                   \
-        INTERNAL NOINLINE SN_STD_TARGET_ATTR(T_arch, T_feature) void SN_DECL SK_PROC_NAME_STD( \
+#define NIF_SK_PROC_DECL_STD(T_type, T_arch, T_feature, T_op)                                   \
+        INTERNAL NOINLINE NIF_STD_TARGET_ATTR(T_arch, T_feature) void NIF_DECL SK_PROC_NAME_STD( \
                T_type, T_arch, T_feature,                                                     \
                T_op)(MAT_KERNEL(T_type) const *left, MAT_KERNEL(T_type) const *right,         \
                      MAT_KERNEL(T_type) *restrict ans, uint64_t scalar)
@@ -35,8 +35,8 @@

 #define _SN_STD__PROBE_0 0
 #define _SN_STD__PROBE_1 1
-#define SN_STD__PROBE_PRINT(x) _SN_STD__PROBE_##x
-#define SN_STD__PROBE(x) SN_STD__PROBE_PRINT(x)
+#define NIF_STD__PROBE_PRINT(x) _SN_STD__PROBE_##x
+#define NIF_STD__PROBE(x) NIF_STD__PROBE_PRINT(x)

 #pragma region param styles

@@ -51,37 +51,37 @@
                int count_##i = task->count_##i;\

 #define _SN_MK_EXPAND_RIGHT_ANS                                                             \
-        SN_MAT_KERNEL_LOCAL const *right = (SN_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
-        SN_MAT_KERNEL_LOCAL *restrict ans = (SN_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;\
+        NIF_MAT_KERNEL_LOCAL const *right = (NIF_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
+        NIF_MAT_KERNEL_LOCAL *restrict ans = (NIF_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;\
        _SN_MK_EXPAND_BUFFER_SIZE(1); \
        _SN_MK_EXPAND_BUFFER_SIZE(2);

 #define _SN_MK_EXPAND_LEFT_RIGHT_ANS                                                        \
-        SN_MAT_KERNEL_LOCAL const *left = (SN_MAT_KERNEL_LOCAL const *)task->buffer_0.ptr;  \
-        SN_MAT_KERNEL_LOCAL const *right = (SN_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
-        SN_MAT_KERNEL_LOCAL *restrict ans = (SN_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;\
+        NIF_MAT_KERNEL_LOCAL const *left = (NIF_MAT_KERNEL_LOCAL const *)task->buffer_0.ptr;  \
+        NIF_MAT_KERNEL_LOCAL const *right = (NIF_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
+        NIF_MAT_KERNEL_LOCAL *restrict ans = (NIF_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;\
        _SN_MK_EXPAND_BUFFER_SIZE(0); \
        _SN_MK_EXPAND_BUFFER_SIZE(1);\
        _SN_MK_EXPAND_BUFFER_SIZE(2);

 #define _SN_MK_EXPAND_RIGHT_ANS_SCALAR                                                      \
-        SN_MAT_KERNEL_LOCAL const *right = (SN_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
-        SN_MAT_KERNEL_LOCAL *restrict ans = (SN_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;      \
+        NIF_MAT_KERNEL_LOCAL const *right = (NIF_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
+        NIF_MAT_KERNEL_LOCAL *restrict ans = (NIF_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;      \
        M128 scalar = task->scalar_pack;\
        _SN_MK_EXPAND_BUFFER_SIZE(1);\
        _SN_MK_EXPAND_BUFFER_SIZE(2);

 #define _SN_MK_EXPAND_LEFT_RIGHT_ANS_SCALAR                                                 \
-        SN_MAT_KERNEL_LOCAL const *left = (SN_MAT_KERNEL_LOCAL const *)task->buffer_0.ptr;  \
-        SN_MAT_KERNEL_LOCAL const *right = (SN_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
-        SN_MAT_KERNEL_LOCAL *restrict ans = (SN_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;      \
+        NIF_MAT_KERNEL_LOCAL const *left = (NIF_MAT_KERNEL_LOCAL const *)task->buffer_0.ptr;  \
+        NIF_MAT_KERNEL_LOCAL const *right = (NIF_MAT_KERNEL_LOCAL const *)task->buffer_1.ptr; \
+        NIF_MAT_KERNEL_LOCAL *restrict ans = (NIF_MAT_KERNEL_LOCAL *)task->buffer_2.ptr;      \
        M128 scalar = task->scalar_pack;\
        _SN_MK_EXPAND_BUFFER_SIZE(0);\
        _SN_MK_EXPAND_BUFFER_SIZE(1);\
        _SN_MK_EXPAND_BUFFER_SIZE(2);

 #define _SN_MK_EXPAND_ANS_SCALAR                                                       \
-        SN_MAT_KERNEL_LOCAL *restrict ans = (SN_MAT_KERNEL_LOCAL *)task->buffer_2.ptr; \
+        NIF_MAT_KERNEL_LOCAL *restrict ans = (NIF_MAT_KERNEL_LOCAL *)task->buffer_2.ptr; \
        M128 scalar = task->scalar_pack;\
        _SN_MK_EXPAND_BUFFER_SIZE(2);

@@ -91,11 +91,11 @@
 #pragma endregion

 #define __mk_param_std(param, T)                               \
-        typedef MAT_KERNEL(T) SN_MAT_KERNEL_LOCAL;             \
+        typedef MAT_KERNEL(T) NIF_MAT_KERNEL_LOCAL;             \
        matrix_task const *_ = (_SN_PARAM_STYLE(param), task); \
        _SN_MK_EXPAND_PARAM(param)

 #define __sk_param_std(T)                          \
-        typedef MAT_KERNEL(T) SN_MAT_KERNEL_LOCAL; \
+        typedef MAT_KERNEL(T) NIF_MAT_KERNEL_LOCAL; \
        (void)scalar
-#endif // !SN_INTRINSIC_PARAM_EXPAND
+#endif // !NIF_INTRINSIC_PARAM_EXPAND
@@ -0,0 +1,94 @@
+#ifndef NIF_TARGET
+#define NIF_TARGET
+
+#include "nif_internal.h"
+#include <stdint.h>
+
+#define NIF_OS_WINDOWS 0
+#define NIF_OS_LINUX 1
+#define NIF_OS_MACOS 2
+#define NIF_OS_IOS 3
+#define NIF_OS_ANDROID 4
+#define NIF_OS_UNKNOWN 5
+
+#define NIF_ARCH_X86_64 0
+#define NIF_ARCH_ARM64 1
+#define NIF_ARCH_RISCV64 2
+#define NIF_ARCH_PPC64le 3
+
+#if defined(_M_X64) || defined(_M_AMD64) || defined(__x86_64__) || defined(__amd64__)
+#define NIF_CURRENT_ARCH NIF_ARCH_X86_64
+
+#elif defined(_M_ARM64) || defined(__aarch64__)
+#define NIF_CURRENT_ARCH NIF_ARCH_ARM64
+
+#elif defined(_M_ARM) || defined(__arm__)
+#define NIF_CURRENT_ARCH NIF_ARCH_ARM32
+
+#elif defined(_M_IX86) || defined(__i386__)
+#define NIF_CURRENT_ARCH NIF_ARCH_X86_32
+
+#elif defined(__powerpc64__) && defined(__LITTLE_ENDIAN__)
+#define NIF_CURRENT_ARCH NIF_ARCH_PPC64LE
+
+#elif defined(__riscv) && (__riscv_xlen == 64)
+#define NIF_CURRENT_ARCH NIF_ARCH_RISCV64
+
+#else
+#error "SN: Unsupported target architecture"
+#endif
+
+/*
+ * Target selection policy:
+ * - Default (no NIF_EXTERN_TARGET): fixed Windows + x86_64.
+ * - External (NIF_EXTERN_TARGET): build system must provide NIF_OS and NIF_ARCH.
+ */
+#ifndef NIF_EXTERN_TARGET
+
+#define NIF_OS_TARGET NIF_OS_WINDOWS
+#define NIF_ARCH_TARGET NIF_ARCH_X86_64
+
+#else
+
+#ifndef NIF_OS_TARGET
+#error "NIF_EXTERN_TARGET is defined but NIF_OS is not provided."
+#endif
+
+#ifndef NIF_ARCH_TARGET
+#error "NIF_EXTERN_TARGET is defined but NIF_ARCH is not provided."
+#endif
+
+/* Validate externally supplied values are within known enum constants. */
+#if (NIF_OS_TARGET != NIF_OS_WINDOWS) && (NIF_OS_TARGET != NIF_OS_LINUX) && \
+        (NIF_OS_TARGET != NIF_OS_MACOS) && (NIF_OS_TARGET != NIF_OS_IOS) && \
+        (NIF_OS_TARGET != NIF_OS_ANDROID) && (NIF_OS_TARGET != NIF_OS_UNKNOWN)
+#error "NIF_OS has an invalid value. Use NIF_OS_<...> enum constants."
+#endif
+
+#if (NIF_ARCH_TARGET != NIF_ARCH_X86_64) && (NIF_ARCH_TARGET != NIF_ARCH_ARM64) && \
+        (NIF_ARCH_TARGET != NIF_ARCH_RISCV64) && (NIF_ARCH_TARGET != NIF_ARCH_PPC64le)
+#error "NIF_ARCH has an invalid value. Use NIF_ARCH_<...> enum constants."
+#endif
+
+#endif
+
+#define NIF_IS_OS(os) ((NIF_OS_TARGET) == (os))
+
+#define NIF_IS_ARCH(arch) ((NIF_ARCH_TARGET) == (arch))
+
+/*
+ * Per-function ISA selection is required for runtime-dispatched kernels that
+ * mix multiple x86 feature levels in one build target.
+ */
+#if defined(__clang__) && ((NIF_ARCH_TARGET) == NIF_ARCH_X86_64)
+#define NIF_CLANG_TARGET_ATTR(options) __attribute__((target(options)))
+#else
+#define NIF_CLANG_TARGET_ATTR(options)
+#endif
+
+typedef union SIMDID {
+        uint64_t mask;
+        char make_byte[8];
+} SIMDID;
+
+#endif // NIF_TARGET
@@ -1,18 +1,18 @@
-#include "StgSharpNative.h"
-#include "sn_intrinsic.matkernel.h"
+#include "NitloadFrameWork.h"
+#include "nif_intrinsic.matkernel.h"
 #include <stdint.h>

 #ifdef _WIN32

 /* Define and initialize a global zero kernel aligned to 64 bytes for universal ISA usage */

-__declspec(align(64)) MAT_KERNEL(u64) SN_ZERO_KERNEL;
+__declspec(align(64)) MAT_KERNEL(u64) NIF_ZERO_KERNEL;

 static void init_zero_kernel(void)
 {
        for (int c = 0; c < 4; ++c) {
                for (int r = 0; r < 4; ++r) {
-                        SN_ZERO_KERNEL.m[r][c] = 0;
+                        NIF_ZERO_KERNEL.m[r][c] = 0;
                }
        }
 }
@@ -1,6 +1,6 @@
-#include "StgSharpNative.h"
+#include "NitloadFrameWork.h"

-SN_API void SN_DECL load_glfw_functions(GLFWFunctionTable *table)
+NIF_API void NIF_DECL nif_load_glfw_functions(GLFWFunctionTable *table)
 {
        table->glfwCreateCursor = glfwCreateCursor;
        table->glfwCreateStandardCursor = glfwCreateStandardCursor;
@@ -1,28 +1,28 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "StgSharpNative.h"
-#include "sn_intrinsic.h"
-#include "sn_intrinsic.std.h"
-#include "sn_intrinsic.context.x86.h"
+#include "NitloadFrameWork.h"
+#include "nif_intrinsic.h"
+#include "nif_intrinsic.std.h"
+#include "nif_intrinsic.context.x86.h"

-PRIVATE void load_intrinsic_sse(sn_intrinsic *context, SIMDID id);
-PRIVATE void load_intrinsic_avx(sn_intrinsic *context, SIMDID id);
-PRIVATE void load_intrinsic_512(sn_intrinsic *context, SIMDID id);
+PRIVATE void load_intrinsic_sse(nif_intrinsic *context, SIMDID id);
+PRIVATE void load_intrinsic_avx(nif_intrinsic *context, SIMDID id);
+PRIVATE void load_intrinsic_512(nif_intrinsic *context, SIMDID id);

-SN_API void SN_DECL load_intrinsic_function(void *intrinsic_context, uint64_t id)
+NIF_API void NIF_DECL nif_load_intrinsic_function(void *intrinsic_context, uint64_t id)
 {
-        sn_null_assert(intrinsic_context);
+        nif_null_assert(intrinsic_context);

        SIMDID simdid;
        if (id) {
                simdid.mask = id;
        } else {
-                simdid = sn_get_simd_level_global();
+                simdid = nif_get_simd_level_global();
        }

-        sn_intrinsic *context = (sn_intrinsic *)intrinsic_context;
+        nif_intrinsic *context = (nif_intrinsic *)intrinsic_context;

        switch (simdid.make_byte[1]) {
        case SIMDID_MAIN_LVL_SSE:
@@ -38,7 +38,7 @@ SN_API void SN_DECL load_intrinsic_function(void *intrinsic_context, uint64_t id
        }
 }

-void load_intrinsic_sse(sn_intrinsic *context, SIMDID id)
+void load_intrinsic_sse(nif_intrinsic *context, SIMDID id)
 {
        context->city_hash_simplify = city_hash_simplify_sse;
        context->factorial_simd = factorial_simd_sse;
@@ -62,7 +62,7 @@ void load_intrinsic_sse(sn_intrinsic *context, SIMDID id)
        context->vec_f32_normalize_3 = f32_normalize;
 }

-void load_intrinsic_avx(sn_intrinsic *context, SIMDID id)
+void load_intrinsic_avx(nif_intrinsic *context, SIMDID id)
 {
        char avx_seg = id.make_byte[4];
        char is_fma = (avx_seg & SIMDID_AVX_FMA) != 0;
@@ -91,7 +91,7 @@ void load_intrinsic_avx(sn_intrinsic *context, SIMDID id)
        context->vec_f32_normalize_3 = f32_normalize;
 }

-void load_intrinsic_512(sn_intrinsic *context, SIMDID id)
+void load_intrinsic_512(nif_intrinsic *context, SIMDID id)
 {
        context->city_hash_simplify = city_hash_simplify_sse;
        context->factorial_simd = factorial_simd_sse;
@@ -1,13 +1,13 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

 #include <immintrin.h>
-#include "sn_intrinsic.h"
-#include "sn_intrinsic.std.h"
-#include "sn_intrinsic.context.x86.h"
+#include "nif_intrinsic.h"
+#include "nif_intrinsic.std.h"
+#include "nif_intrinsic.context.x86.h"

-SN_MK_PROC_DECL_STD(float, sse, , fill)
+NIF_MK_PROC_DECL_STD(float, sse, , fill)
 {
        __mk_param_std(ANS_SCALAR, float);
        register __m128 vec = _mm_set_ps1(scalar.f32[0]);
@@ -19,7 +19,7 @@ SN_MK_PROC_DECL_STD(float, sse, , fill)
        }
 }

-SN_MK_PROC_DECL_STD(float, avx, , fill)
+NIF_MK_PROC_DECL_STD(float, avx, , fill)
 {
        __mk_param_std(ANS_SCALAR, float);
        register __m256 vec = _mm256_broadcast_ss(&scalar.f32[0]);
@@ -29,7 +29,7 @@ SN_MK_PROC_DECL_STD(float, avx, , fill)
        }
 }

-SN_MK_PROC_DECL_STD(float, 512, , fill)
+NIF_MK_PROC_DECL_STD(float, 512, , fill)
 {
        __mk_param_std(ANS_SCALAR, float);
        register __m512 vec = _mm512_broadcastss_ps(scalar.m128);
@@ -38,7 +38,7 @@ SN_MK_PROC_DECL_STD(float, 512, , fill)
        }
 }

-SN_MK_PROC_DECL_STD(double, sse, , fill)
+NIF_MK_PROC_DECL_STD(double, sse, , fill)
 {
        __mk_param_std(ANS_SCALAR, double);
        register __m128d vec_d = _mm_set1_pd(scalar.f64[0]);
@@ -55,7 +55,7 @@ SN_MK_PROC_DECL_STD(double, sse, , fill)
        }
 }

-SN_MK_PROC_DECL_STD(double, avx, , fill)
+NIF_MK_PROC_DECL_STD(double, avx, , fill)
 {
        __mk_param_std(ANS_SCALAR, double);
        register __m256d vec = _mm256_broadcast_sd(&scalar.f64[0]);
@@ -67,7 +67,7 @@ SN_MK_PROC_DECL_STD(double, avx, , fill)
        }
 }

-SN_MK_PROC_DECL_STD(double, 512, , fill)
+NIF_MK_PROC_DECL_STD(double, 512, , fill)
 {
        __mk_param_std(ANS_SCALAR, double);
        register __m512d vec = _mm512_set1_pd(scalar.f64[0]);
@@ -77,7 +77,7 @@ SN_MK_PROC_DECL_STD(double, 512, , fill)
        }
 }

-SN_MK_PROC_DECL_STD(float, sse, , add)
+NIF_MK_PROC_DECL_STD(float, sse, , add)
 {
        __mk_param_std(LEFT_RIGHT_ANS, float);

@@ -94,7 +94,7 @@ SN_MK_PROC_DECL_STD(float, sse, , add)
        }
 }

-SN_MK_PROC_DECL_STD(float, avx, , add)
+NIF_MK_PROC_DECL_STD(float, avx, , add)
 {
        __mk_param_std(LEFT_RIGHT_ANS, float);

@@ -107,7 +107,7 @@ SN_MK_PROC_DECL_STD(float, avx, , add)
        }
 }

-SN_MK_PROC_DECL_STD(float, 512, , add)
+NIF_MK_PROC_DECL_STD(float, 512, , add)
 {
        __mk_param_std(LEFT_RIGHT_ANS, float);

@@ -116,7 +116,7 @@ SN_MK_PROC_DECL_STD(float, 512, , add)
                ans[i].z[0] = c0;
        }
 }
-SN_MK_PROC_DECL_STD(double, sse, , add)
+NIF_MK_PROC_DECL_STD(double, sse, , add)
 {
        __mk_param_std( LEFT_RIGHT_ANS, double);

@@ -149,7 +149,7 @@ SN_MK_PROC_DECL_STD(double, sse, , add)
        }
 }

-SN_MK_PROC_DECL_STD(double, avx, , add)
+NIF_MK_PROC_DECL_STD(double, avx, , add)
 {
        __mk_param_std( LEFT_RIGHT_ANS, double);

@@ -166,7 +166,7 @@ SN_MK_PROC_DECL_STD(double, avx, , add)
        }
 }

-SN_MK_PROC_DECL_STD(double, 512, , add)
+NIF_MK_PROC_DECL_STD(double, 512, , add)
 {
        __mk_param_std( LEFT_RIGHT_ANS, double);

@@ -178,7 +178,7 @@ SN_MK_PROC_DECL_STD(double, 512, , add)
        }
 }

-SN_MK_PROC_DECL_STD(float, sse, , sub)
+NIF_MK_PROC_DECL_STD(float, sse, , sub)
 {
        __mk_param_std( LEFT_RIGHT_ANS, float);

@@ -195,7 +195,7 @@ SN_MK_PROC_DECL_STD(float, sse, , sub)
        }
 }

-SN_MK_PROC_DECL_STD(float, avx, , sub)
+NIF_MK_PROC_DECL_STD(float, avx, , sub)
 {
        __mk_param_std( LEFT_RIGHT_ANS, float);

@@ -207,7 +207,7 @@ SN_MK_PROC_DECL_STD(float, avx, , sub)
        }
 }

-SN_MK_PROC_DECL_STD(float, 512, , sub)
+NIF_MK_PROC_DECL_STD(float, 512, , sub)
 {
        __mk_param_std( LEFT_RIGHT_ANS, float);

@@ -217,7 +217,7 @@ SN_MK_PROC_DECL_STD(float, 512, , sub)
        }
 }

-SN_MK_PROC_DECL_STD(double, sse, , sub)
+NIF_MK_PROC_DECL_STD(double, sse, , sub)
 {
        __mk_param_std( LEFT_RIGHT_ANS, double);

@@ -242,7 +242,7 @@ SN_MK_PROC_DECL_STD(double, sse, , sub)
        }
 }

-SN_MK_PROC_DECL_STD(double, avx, , sub)
+NIF_MK_PROC_DECL_STD(double, avx, , sub)
 {
        __mk_param_std( LEFT_RIGHT_ANS, double);

@@ -254,7 +254,7 @@ SN_MK_PROC_DECL_STD(double, avx, , sub)
        }
 }

-SN_MK_PROC_DECL_STD(double, 512, , sub)
+NIF_MK_PROC_DECL_STD(double, 512, , sub)
 {
        __mk_param_std( LEFT_RIGHT_ANS, double);

@@ -264,7 +264,7 @@ SN_MK_PROC_DECL_STD(double, 512, , sub)
        }
 }

-SN_MK_PROC_DECL_STD(float, sse, , scalar_mul)
+NIF_MK_PROC_DECL_STD(float, sse, , scalar_mul)
 {
        __mk_param_std( RIGHT_ANS_SCALAR, float);

@@ -282,7 +282,7 @@ SN_MK_PROC_DECL_STD(float, sse, , scalar_mul)
        }
 }

-SN_MK_PROC_DECL_STD(float, avx, , scalar_mul)
+NIF_MK_PROC_DECL_STD(float, avx, , scalar_mul)
 {
        __mk_param_std( RIGHT_ANS_SCALAR, float);

@@ -296,7 +296,7 @@ SN_MK_PROC_DECL_STD(float, avx, , scalar_mul)
        }
 }

-SN_MK_PROC_DECL_STD(float, 512, , scalar_mul)
+NIF_MK_PROC_DECL_STD(float, 512, , scalar_mul)
 {
        __mk_param_std( RIGHT_ANS_SCALAR, float);

@@ -308,7 +308,7 @@ SN_MK_PROC_DECL_STD(float, 512, , scalar_mul)
        }
 }

-SN_MK_PROC_DECL_STD(double, sse, , scalar_mul)
+NIF_MK_PROC_DECL_STD(double, sse, , scalar_mul)
 {
        __mk_param_std( RIGHT_ANS_SCALAR, double);

@@ -334,7 +334,7 @@ SN_MK_PROC_DECL_STD(double, sse, , scalar_mul)
        }
 }

-SN_MK_PROC_DECL_STD(double, avx, , scalar_mul)
+NIF_MK_PROC_DECL_STD(double, avx, , scalar_mul)
 {
        __mk_param_std( RIGHT_ANS_SCALAR, double);

@@ -352,7 +352,7 @@ SN_MK_PROC_DECL_STD(double, avx, , scalar_mul)
        }
 }

-SN_MK_PROC_DECL_STD(double, 512, , scalar_mul)
+NIF_MK_PROC_DECL_STD(double, 512, , scalar_mul)
 {
        __mk_param_std( RIGHT_ANS_SCALAR, double);

@@ -1,15 +1,15 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "sn_intrinsic.h"
-#include "sn_intrinsic.std.h"
-#include "sn_intrinsic.context.x86.h"
+#include "nif_intrinsic.h"
+#include "nif_intrinsic.std.h"
+#include "nif_intrinsic.context.x86.h"
 #include <math.h>

 static const float quality_threshold = 1e-6f;

-SN_MK_PROC_DECL_STD(float, sse, , try_plu)
+NIF_MK_PROC_DECL_STD(float, sse, , try_plu)
 {
        __mk_param_std(LEFT_RIGHT_ANS_SCALAR, float);

@@ -115,7 +115,7 @@ SN_MK_PROC_DECL_STD(float, sse, , try_plu)
 * layout of scalarpack:
 * 
 */
-SN_MK_PROC_DECL_STD(float, sse, , quality)
+NIF_MK_PROC_DECL_STD(float, sse, , quality)
 {
        __mk_param_std(RIGHT_ANS, float);

@@ -126,7 +126,7 @@ SN_MK_PROC_DECL_STD(float, sse, , quality)
        register __m128 col2 = _mm_load_ps((float *)&right->x[2]);
        register __m128 col3 = _mm_load_ps((float *)&right->x[3]);

-        SN_ALIGNAS(16) float self_dot_src[4] = { 0.0f, 0.0f, 0.0f, 0.0f };
+        NIF_ALIGNAS(16) float self_dot_src[4] = { 0.0f, 0.0f, 0.0f, 0.0f };
        // try in reigister mix later
        DOT_F32_SSE(col0, col0, self_dot_src[0]);
        DOT_F32_SSE(col1, col1, self_dot_src[1]);
@@ -153,7 +153,7 @@ SN_MK_PROC_DECL_STD(float, sse, , quality)
                return;
        }

-        SN_ALIGNAS(16) float cross_dot_src[2][4] = { 0 };
+        NIF_ALIGNAS(16) float cross_dot_src[2][4] = { 0 };
        DOT_F32_SSE(col0, col1, cross_dot_src[0][0]);
        DOT_F32_SSE(col0, col2, cross_dot_src[0][1]);
        DOT_F32_SSE(col0, col3, cross_dot_src[0][2]);
@@ -1,13 +1,13 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "sn_intrinsic.h"
-#include "sn_intrinsic.context.x86.h"
-#include "sn_intrinsic.std.h"
+#include "nif_intrinsic.h"
+#include "nif_intrinsic.context.x86.h"
+#include "nif_intrinsic.std.h"

 // Matrix transpose implementations
-SN_MK_PROC_DECL_STD(float, sse, , transpose)
+NIF_MK_PROC_DECL_STD(float, sse, , transpose)
 {
        __mk_param_std(RIGHT_ANS, float);

@@ -26,7 +26,7 @@ SN_MK_PROC_DECL_STD(float, sse, , transpose)
        ans->x[3] = _mm_movehl_ps(tmp3, tmp1);
 }

-SN_MK_PROC_DECL_STD(float, avx, , transpose)
+NIF_MK_PROC_DECL_STD(float, avx, , transpose)
 {
        __mk_param_std(RIGHT_ANS, float);
        register __m256 col_low = right->y[0];
@@ -42,7 +42,7 @@ SN_MK_PROC_DECL_STD(float, avx, , transpose)
        ans->y[1] = _mm256_shuffle_ps(col_high, col_low, _MM_SHUFFLE(3, 1, 3, 1));
 }

-SN_MK_PROC_DECL_STD(float, 512, , transpose)
+NIF_MK_PROC_DECL_STD(float, 512, , transpose)
 {
        __mk_param_std(RIGHT_ANS, float);
        register __m512 matrix = right->z[0];
@@ -1,8 +1,8 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "sn_intrinsic.context.x86.h"
+#include "nif_intrinsic.context.x86.h"

 /*
 * pack left matrix data to L4 buffer
@@ -19,14 +19,14 @@
 * scalar pack carries column length and offset data:
 * [column length, empty, empty, empty]
 */
-SN_MK_PROC_DECL_STD(float, sse, , fma_pack_left)
+NIF_MK_PROC_DECL_STD(float, sse, , fma_pack_left)
 {
        __mk_param_std(RIGHT_ANS_SCALAR, float);
        // for SSE branch, no need for rounding, a direct copy is sufficient as long as the buffer is large enough to hold the packed data
        memcpy(ans, right + offset_1, sizeof(MAT_KERNEL(float)) * count_2);
 }

-SN_MK_PROC_DECL_STD(float, avx, , fma_pack_left)
+NIF_MK_PROC_DECL_STD(float, avx, , fma_pack_left)
 {
        __mk_param_std(RIGHT_ANS_SCALAR, float);
        size_t i = 0;
@@ -61,7 +61,7 @@ SN_MK_PROC_DECL_STD(float, avx, , fma_pack_left)
        }
 }

-SN_MK_PROC_DECL_STD(float, 512, , fma_pack_left)
+NIF_MK_PROC_DECL_STD(float, 512, , fma_pack_left)
 {
        __mk_param_std(RIGHT_ANS_SCALAR, float);
        size_t i = 0;
@@ -154,7 +154,7 @@ SN_MK_PROC_DECL_STD(float, 512, , fma_pack_left)
 * scalar pack carries column length and offset data:
 * [column length, empty, empty, empty]
 */
-SN_MK_PROC_DECL_STD(float, sse, , fma_pack_right)
+NIF_MK_PROC_DECL_STD(float, sse, , fma_pack_right)
 {
        __mk_param_std(RIGHT_ANS_SCALAR, float);
        int col_length = scalar.i32[0];
@@ -163,7 +163,7 @@ SN_MK_PROC_DECL_STD(float, sse, , fma_pack_right)
        }
 }

-SN_MK_PROC_DECL_STD(float, avx, , fma_pack_right)
+NIF_MK_PROC_DECL_STD(float, avx, , fma_pack_right)
 {
        __mk_param_std(RIGHT_ANS_SCALAR, float);
        int col_length = scalar.i32[0];
@@ -205,7 +205,7 @@ SN_MK_PROC_DECL_STD(float, avx, , fma_pack_right)
        }
 }

-SN_MK_PROC_DECL_STD(float, 512, , fma_pack_right)
+NIF_MK_PROC_DECL_STD(float, 512, , fma_pack_right)
 {
        __mk_param_std(RIGHT_ANS_SCALAR, float);
        int col_length = scalar.i32[0];
@@ -1,22 +1,22 @@
-#include "sn_target.h"
+#include "nif_target.h"
 #include <immintrin.h>

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-// #include "sn_intrinsic.h"
-#include "sn_template.fma.x86.h"
-#include "sn_intrinsic.std.h"
-#include "sn_intrinsic.context.x86.h"
+// #include "nif_intrinsic.h"
+#include "nif_template.fma.x86.h"
+#include "nif_intrinsic.std.h"
+#include "nif_intrinsic.context.x86.h"

 #include <stdio.h>

 #define _FMA_SELECT(i) _MM_SHUFFLE(i, i, i, i)

-SN_MK_PROC_DECL_STD(float, sse, , fma)
+NIF_MK_PROC_DECL_STD(float, sse, , fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, float);

-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;

        register __m128 c0 = _mm_load_ps((float *)&ans_cur->x[0]);
        register __m128 c1 = _mm_load_ps((float *)&ans_cur->x[1]);
@@ -25,8 +25,8 @@ SN_MK_PROC_DECL_STD(float, sse, , fma)
        register __m128 bcol;

        for (size_t z = 0; z < count_0; z++) {
-                SN_MAT_KERNEL_LOCAL const *right_base = right;
-                SN_MAT_KERNEL_LOCAL const *left_base = left;
+                NIF_MAT_KERNEL_LOCAL const *right_base = right;
+                NIF_MAT_KERNEL_LOCAL const *left_base = left;

                const __m128 a0 = _mm_load_ps((float *)&left_base->x[0]);
                const __m128 a1 = _mm_load_ps((float *)&left_base->x[1]);
@@ -68,18 +68,18 @@ SN_MK_PROC_DECL_STD(float, sse, , fma)
        _mm_store_ps((float *)&ans_cur->x[3], c3);
 }

-SN_MK_PROC_DECL_STD(float, avx, , fma)
+NIF_MK_PROC_DECL_STD(float, avx, , fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, float);
-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
        register __m256 c0 = _mm256_load_ps((float *)&ans_cur->y[0]);
        register __m256 c1 = _mm256_load_ps((float *)&ans_cur->y[1]);
        register __m256 c2 = _mm256_load_ps((float *)&ans_cur->y[2]);
        register __m256 c3 = _mm256_load_ps((float *)&ans_cur->y[3]);

        for (size_t z = 0; z < count_0; z++) {
-                SN_MAT_KERNEL_LOCAL const *b_base = right;
-                SN_MAT_KERNEL_LOCAL const *a_base = left;
+                NIF_MAT_KERNEL_LOCAL const *b_base = right;
+                NIF_MAT_KERNEL_LOCAL const *a_base = left;

                register __m256 b0 = _mm256_load_ps((float *)&b_base->y[0]);
                register __m256 b1 = _mm256_load_ps((float *)&b_base->y[1]);
@@ -91,10 +91,10 @@ SN_MK_PROC_DECL_STD(float, avx, , fma)
                register __m256 a2 = _mm256_load_ps((float *)&a_base->y[2]);
                register __m256 a3 = _mm256_load_ps((float *)&a_base->y[3]);

-                SN_MK_FMA_F32_AVX2_CYCLE(0);
-                SN_MK_FMA_F32_AVX2_CYCLE(1);
-                SN_MK_FMA_F32_AVX2_CYCLE(2);
-                SN_MK_FMA_F32_AVX2_CYCLE(3);
+                NIF_MK_FMA_F32_AVX2_CYCLE(0);
+                NIF_MK_FMA_F32_AVX2_CYCLE(1);
+                NIF_MK_FMA_F32_AVX2_CYCLE(2);
+                NIF_MK_FMA_F32_AVX2_CYCLE(3);
        }
        _mm256_store_ps((float *)&(ans_cur->y[0]), c0);
        _mm256_store_ps((float *)&(ans_cur->y[1]), c1);
@@ -102,18 +102,18 @@ SN_MK_PROC_DECL_STD(float, avx, , fma)
        _mm256_store_ps((float *)&(ans_cur->y[3]), c3);
 }

-SN_MK_PROC_DECL_STD(float, avx, _fma, fma)
+NIF_MK_PROC_DECL_STD(float, avx, _fma, fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, float);
-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
        register __m256 c0 = _mm256_load_ps((float *)&ans_cur->y[0]);
        register __m256 c1 = _mm256_load_ps((float *)&ans_cur->y[1]);
        register __m256 c2 = _mm256_load_ps((float *)&ans_cur->y[2]);
        register __m256 c3 = _mm256_load_ps((float *)&ans_cur->y[3]);

        for (size_t z = 0; z < count_0; z += 2) {
-                const SN_MAT_KERNEL_LOCAL *a_base = left + z;
-                const SN_MAT_KERNEL_LOCAL *b_base = right + z;
+                const NIF_MAT_KERNEL_LOCAL *a_base = left + z;
+                const NIF_MAT_KERNEL_LOCAL *b_base = right + z;

                register __m256 b0 = _mm256_load_ps((float *)&b_base->y[0]);
                register __m256 b1 = _mm256_load_ps((float *)&b_base->y[1]);
@@ -125,10 +125,10 @@ SN_MK_PROC_DECL_STD(float, avx, _fma, fma)
                register __m256 a2 = _mm256_load_ps((float *)&a_base->y[2]);
                register __m256 a3 = _mm256_load_ps((float *)&a_base->y[3]);

-                SN_MK_FMA_F32_AVX2_CYCLE(0);
-                SN_MK_FMA_F32_AVX2_CYCLE(1);
-                SN_MK_FMA_F32_AVX2_CYCLE(2);
-                SN_MK_FMA_F32_AVX2_CYCLE(3);
+                NIF_MK_FMA_F32_AVX2_CYCLE(0);
+                NIF_MK_FMA_F32_AVX2_CYCLE(1);
+                NIF_MK_FMA_F32_AVX2_CYCLE(2);
+                NIF_MK_FMA_F32_AVX2_CYCLE(3);
        }
        _mm256_store_ps((float *)&(ans_cur->y[0]), c0);
        _mm256_store_ps((float *)&(ans_cur->y[1]), c1);
@@ -136,18 +136,18 @@ SN_MK_PROC_DECL_STD(float, avx, _fma, fma)
        _mm256_store_ps((float *)&(ans_cur->y[3]), c3);
 }

-SN_MK_PROC_DECL_STD(float, 512, , fma)
+NIF_MK_PROC_DECL_STD(float, 512, , fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, float);
-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
        register __m512 c0 = _mm512_load_ps((float *)&ans_cur->z[0]);
        register __m512 c1 = _mm512_load_ps((float *)&ans_cur->z[0]);
        register __m512 c2 = _mm512_load_ps((float *)&ans_cur->z[0]);
        register __m512 c3 = _mm512_load_ps((float *)&ans_cur->z[0]);

        for (size_t z = 0; z < count_0; z += 4) {
-                const SN_MAT_KERNEL_LOCAL *a_base = left + z;
-                const SN_MAT_KERNEL_LOCAL *b_base = right + z;
+                const NIF_MAT_KERNEL_LOCAL *a_base = left + z;
+                const NIF_MAT_KERNEL_LOCAL *b_base = right + z;

                register __m512 b0 = _mm512_load_ps((float *)&b_base->z[0]);
                register __m512 b1 = _mm512_load_ps((float *)&b_base->z[1]);
@@ -159,10 +159,10 @@ SN_MK_PROC_DECL_STD(float, 512, , fma)
                register __m512 a2 = _mm512_load_ps((float *)&a_base->z[2]);
                register __m512 a3 = _mm512_load_ps((float *)&a_base->z[3]);

-                SN_MK_FMA_F32_512_CYCLE(0);
-                SN_MK_FMA_F32_512_CYCLE(1);
-                SN_MK_FMA_F32_512_CYCLE(2);
-                SN_MK_FMA_F32_512_CYCLE(3);
+                NIF_MK_FMA_F32_512_CYCLE(0);
+                NIF_MK_FMA_F32_512_CYCLE(1);
+                NIF_MK_FMA_F32_512_CYCLE(2);
+                NIF_MK_FMA_F32_512_CYCLE(3);
        }
        _mm512_store_ps((float *)&ans_cur->z[0], c0);
        _mm512_store_ps((float *)&ans_cur->z[1], c1);
@@ -170,11 +170,11 @@ SN_MK_PROC_DECL_STD(float, 512, , fma)
        _mm512_store_ps((float *)&ans_cur->z[3], c3);
 }

-SN_MK_PROC_DECL_STD(double, sse, , fma)
+NIF_MK_PROC_DECL_STD(double, sse, , fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, double);

-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;

        register __m128d c0_lo = ans_cur->x[0];
        register __m128d c0_hi = ans_cur->x[1];
@@ -186,18 +186,18 @@ SN_MK_PROC_DECL_STD(double, sse, , fma)
        register __m128d c3_hi = ans_cur->x[7];

        for (size_t z = 0; z < count_0; z++) {
-                SN_MAT_KERNEL_LOCAL const *right_base = right;
-                SN_MAT_KERNEL_LOCAL const *left_base = left;
+                NIF_MAT_KERNEL_LOCAL const *right_base = right;
+                NIF_MAT_KERNEL_LOCAL const *left_base = left;
                const double *right_values = &right_base->m[0][0];
                const double *left_values = &left_base->m[0][0];

-                SN_MK_FMA_F64_SSE_PHASE(0);
+                NIF_MK_FMA_F64_SSE_PHASE(0);

-                SN_MK_FMA_F64_SSE_PHASE(1);
+                NIF_MK_FMA_F64_SSE_PHASE(1);

-                SN_MK_FMA_F64_SSE_PHASE(2);
+                NIF_MK_FMA_F64_SSE_PHASE(2);

-                SN_MK_FMA_F64_SSE_PHASE(3);
+                NIF_MK_FMA_F64_SSE_PHASE(3);
        }

        ans_cur->x[0] = c0_lo;
@@ -210,11 +210,11 @@ SN_MK_PROC_DECL_STD(double, sse, , fma)
        ans_cur->x[7] = c3_hi;
 }

-SN_MK_PROC_DECL_STD(double, avx, , fma)
+NIF_MK_PROC_DECL_STD(double, avx, , fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, double);

-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;

        register __m256d c0 = _mm256_load_pd((double *)&ans_cur->y[0]);
        register __m256d c1 = _mm256_load_pd((double *)&ans_cur->y[1]);
@@ -223,8 +223,8 @@ SN_MK_PROC_DECL_STD(double, avx, , fma)
        register __m256d bcol;

        for (size_t z = 0; z < count_0; z++) {
-                SN_MAT_KERNEL_LOCAL const *right_base = right;
-                SN_MAT_KERNEL_LOCAL const *left_base = left;
+                NIF_MAT_KERNEL_LOCAL const *right_base = right;
+                NIF_MAT_KERNEL_LOCAL const *left_base = left;

                const __m256d a0 = _mm256_load_pd((double *)&left_base->y[0]);
                const __m256d a1 = _mm256_load_pd((double *)&left_base->y[1]);
@@ -282,11 +282,11 @@ SN_MK_PROC_DECL_STD(double, avx, , fma)
        _mm256_store_pd((double *)&ans_cur->y[3], c3);
 }

-SN_MK_PROC_DECL_STD(double, avx, _fma, fma)
+NIF_MK_PROC_DECL_STD(double, avx, _fma, fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, double);

-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;

        register __m256d c0 = _mm256_load_pd((double *)&ans_cur->y[0]);
        register __m256d c1 = _mm256_load_pd((double *)&ans_cur->y[1]);
@@ -295,8 +295,8 @@ SN_MK_PROC_DECL_STD(double, avx, _fma, fma)
        register __m256d bcol;

        for (size_t z = 0; z < count_0; z++) {
-                SN_MAT_KERNEL_LOCAL const *right_base = right;
-                SN_MAT_KERNEL_LOCAL const *left_base = left;
+                NIF_MAT_KERNEL_LOCAL const *right_base = right;
+                NIF_MAT_KERNEL_LOCAL const *left_base = left;

                const __m256d a0 = _mm256_load_pd((double *)&left_base->y[0]);
                const __m256d a1 = _mm256_load_pd((double *)&left_base->y[1]);
@@ -338,10 +338,10 @@ SN_MK_PROC_DECL_STD(double, avx, _fma, fma)
        _mm256_store_pd((double *)&ans_cur->y[3], c3);
 }

-SN_MK_PROC_DECL_STD(double, 512, , fma)
+NIF_MK_PROC_DECL_STD(double, 512, , fma)
 {
        __mk_param_std(LEFT_RIGHT_ANS, double);
-        SN_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;
+        NIF_MAT_KERNEL_LOCAL *restrict ans_cur = ans + offset_2;

        register __m512d c0 = _mm512_load_pd((double *)&ans_cur->z[0]);
        register __m512d c1 = _mm512_load_pd((double *)&ans_cur->z[1]);
@@ -349,8 +349,8 @@ SN_MK_PROC_DECL_STD(double, 512, , fma)
        register __m512d c3 = _mm512_load_pd((double *)&ans_cur->z[3]);

        for (size_t z = 0; z < count_0; z += 2) {
-                SN_MAT_KERNEL_LOCAL const *right_base = right + z;
-                SN_MAT_KERNEL_LOCAL const *left_base = left + z;
+                NIF_MAT_KERNEL_LOCAL const *right_base = right + z;
+                NIF_MAT_KERNEL_LOCAL const *left_base = left + z;

                register __m512d a0 = _mm512_load_pd((double *)&left_base->z[0]);
                register __m512d a1 = _mm512_load_pd((double *)&left_base->z[1]);
@@ -362,10 +362,10 @@ SN_MK_PROC_DECL_STD(double, 512, , fma)
                register __m512d b2 = _mm512_load_pd((double *)&right_base->z[2]);
                register __m512d b3 = _mm512_load_pd((double *)&right_base->z[3]);

-                SN_MK_FMA_F64_AVX2_CYCLE(0);
-                SN_MK_FMA_F64_AVX2_CYCLE(1);
-                SN_MK_FMA_F64_AVX2_CYCLE(2);
-                SN_MK_FMA_F64_AVX2_CYCLE(3);
+                NIF_MK_FMA_F64_AVX2_CYCLE(0);
+                NIF_MK_FMA_F64_AVX2_CYCLE(1);
+                NIF_MK_FMA_F64_AVX2_CYCLE(2);
+                NIF_MK_FMA_F64_AVX2_CYCLE(3);
        }
        _mm512_store_pd((double *)&ans_cur->z[0], c0);
        _mm512_store_pd((double *)&ans_cur->z[1], c1);
@@ -1,25 +1,25 @@

-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "sn_internal.h"
-#include "sn_intrinsic.context.x86.h"
-#include "sn_intrinsic.std.h"
-#include "sn_intrinsic.matkernel.h"
+#include "nif_internal.h"
+#include "nif_intrinsic.context.x86.h"
+#include "nif_intrinsic.std.h"
+#include "nif_intrinsic.matkernel.h"
 #include <emmintrin.h>
 #include <immintrin.h>
 #include <pmmintrin.h>

-SN_MK_PROC_DECL_STD(float, sse, , k_quality)
+NIF_MK_PROC_DECL_STD(float, sse, , k_quality)
 {
        __mk_param_std(RIGHT_ANS_SCALAR, float);
 }

-INTERNAL SN_STD_TARGET_ATTR(sse, ) void kp_check_f32(mat_kernel *kernel)
+INTERNAL NIF_STD_TARGET_ATTR(sse, ) void kp_check_f32(mat_kernel *kernel)
 {
-        typedef MAT_KERNEL(float) SN_MAT;
-        SN_MAT const *k = (SN_MAT const *)kernel;
+        typedef MAT_KERNEL(float) NIF_MAT;
+        NIF_MAT const *k = (NIF_MAT const *)kernel;
        float cross[4];
        cross[0] = k->m[0][0];
        cross[1] = k->m[1][1];
@@ -1,14 +1,14 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "sn_internal.h"
-#include "sn_intrinsic.h"
+#include "nif_internal.h"
+#include "nif_intrinsic.h"

 static __m128i facIncrement;
 static __m128i facBase;

-INTERNAL uint64_t SN_DECL factorial_simd_sse(int n)
+INTERNAL uint64_t NIF_DECL factorial_simd_sse(int n)
 {
        facIncrement = _mm_set1_epi32(4);
        facBase = _mm_setr_epi32(1, 2, 3, 4);
@@ -20,7 +20,7 @@ INTERNAL uint64_t SN_DECL factorial_simd_sse(int n)
                facBaseLocal = _mm_add_epi32(facBaseLocal, facIncrement);
                baseResult = _mm_mullo_epi32(baseResult, facBaseLocal);
        }
-        SN_ALIGNAS(16) int32_t lanes[4];
+        NIF_ALIGNAS(16) int32_t lanes[4];
        _mm_store_si128((__m128i *)lanes, baseResult);
        uint64_t result = (uint64_t)lanes[0] * (uint64_t)lanes[1] * (uint64_t)lanes[2] *
                          (uint64_t)lanes[3];
@@ -1,10 +1,10 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "StgSharpNative.h"
+#include "NitloadFrameWork.h"

-INTERNAL int SN_DECL city_hash_simplify_sse(byte const *str, int const length)
+INTERNAL int NIF_DECL city_hash_simplify_sse(byte const *str, int const length)
 {
        uint32_t hash = length, size = length << 1;
        byte const *end = str + size;
@@ -1,13 +1,13 @@
-#include "sn_target.h"
+#include "nif_target.h"

 /* clang-format off */

-#if SN_IS_ARCH(SN_ARCH_X86_64)
-#include "sn_intrinsic.h"
-#include "StgSharpNative.h"
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)
+#include "nif_intrinsic.h"
+#include "NitloadFrameWork.h"
 static SIMDID global_simdid = { 0 };

-#include "sn_intrinsic.context.x86.h"
+#include "nif_intrinsic.context.x86.h"

 #include <immintrin.h>
 #include <string.h>
@@ -126,7 +126,7 @@ static int is_amd_glue_ymm(const char *vendor, int cpuid1_eax)
        return (family == 0x17) && (model < 0x30);
 }

-SN_API SIMDID SN_DECL sn_get_simd_level_local()
+NIF_API SIMDID NIF_DECL nif_get_simd_level_local()
 {
        SIMDID simdid = { 0 };

@@ -314,13 +314,13 @@ SN_API SIMDID SN_DECL sn_get_simd_level_local()
        return simdid;
 }

-SN_API SIMDID SN_DECL sn_get_simd_level_global()
+NIF_API SIMDID NIF_DECL nif_get_simd_level_global()
 {
        if (global_simdid.mask) {
                return global_simdid;
        }

-        global_simdid = sn_get_simd_level_local();
+        global_simdid = nif_get_simd_level_local();
        return global_simdid;
 }

@@ -1,9 +1,9 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "StgSharpNative.h"
-#include "sn_intrinsic.context.x86.h"
+#include "NitloadFrameWork.h"
+#include "nif_intrinsic.context.x86.h"

 /* -----------------------------------------------------------------------
 * Field extraction helpers.
@@ -49,7 +49,7 @@ static int impl_strength(uint16_t avx512)
 }

 /* -----------------------------------------------------------------------
- * sn_get_unite_simd
+ * nif_get_unite_simd
 *
 * Produce the conservative intersection of two SIMDIDs — the strongest
 * capability set that is safe to use on BOTH cores:
@@ -60,7 +60,7 @@ static int impl_strength(uint16_t avx512)
 *   - Root / Manu     : keep left (must be same arch in practice)
 *   - uArchHi         : bitwise OR (HYBRID_P/E union; SIMD_DUAL set if either core is dual)
 * ----------------------------------------------------------------------- */
-SN_API SIMDID SN_DECL sn_get_unite_simd(SIMDID left, SIMDID right)
+NIF_API SIMDID NIF_DECL nif_get_unite_simd(SIMDID left, SIMDID right)
 {
        SIMDID out;
        out.mask = 0;
@@ -117,7 +117,7 @@ SN_API SIMDID SN_DECL sn_get_unite_simd(SIMDID left, SIMDID right)
 }

 /* -----------------------------------------------------------------------
- * sn_compare_simd
+ * nif_compare_simd
 *
 * Compare two SIMDIDs by ISA strength. Returns:
 *   +1  if left is strictly stronger than right
@@ -154,7 +154,7 @@ SN_API SIMDID SN_DECL sn_get_unite_simd(SIMDID left, SIMDID right)
 *   7. uArchHi SIMD_DUAL (final tie-break)
 *      Non-dual (SIMD_DUAL=0) > dual (SIMD_DUAL=1).
 * ----------------------------------------------------------------------- */
-SN_API int SN_DECL sn_compare_simd(SIMDID left, SIMDID right)
+NIF_API int NIF_DECL nif_compare_simd(SIMDID left, SIMDID right)
 {
        /* 1. MainLevel */
        unsigned lm = FIELD_MAIN(left);
@@ -1,10 +1,10 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#include "sn_intrinsic.h"
-#include "sn_intrinsic.std.h"
-#include "sn_intrinsic.context.x86.h"
+#include "nif_intrinsic.h"
+#include "nif_intrinsic.std.h"
+#include "nif_intrinsic.context.x86.h"

 #define _FMA_SELECT(i) _MM_SHUFFLE(i, i, i, i)

@@ -12,7 +12,7 @@

 // float

-SN_SK_PROC_DECL_STD(float, sse, , fma)
+NIF_SK_PROC_DECL_STD(float, sse, , fma)
 {
        __sk_param_std(float);
        register __m128 c0 = _mm_load_ps((float *)&ans->x[0]);
@@ -60,7 +60,7 @@ SN_SK_PROC_DECL_STD(float, sse, , fma)
        _mm_store_ps((float *)&ans->x[3], c3);
 }

-SN_SK_PROC_DECL_STD(float, avx, , fma)
+NIF_SK_PROC_DECL_STD(float, avx, , fma)
 {
        __sk_param_std(float);
        register __m256 c01 = _mm256_load_ps((float *)&ans->y[0]);
@@ -86,7 +86,7 @@ SN_SK_PROC_DECL_STD(float, avx, , fma)
        _mm256_store_ps((float *)&ans->y[1], c23);
 }

-SN_SK_PROC_DECL_STD(float, avx, _fma, fma)
+NIF_SK_PROC_DECL_STD(float, avx, _fma, fma)
 {
        __sk_param_std(float);
        register __m256 c01 = _mm256_load_ps((float *)&ans->y[0]);
@@ -112,7 +112,7 @@ SN_SK_PROC_DECL_STD(float, avx, _fma, fma)
        _mm256_store_ps((float *)&ans->y[1], c23);
 }

-SN_SK_PROC_DECL_STD(float, 512, , fma)
+NIF_SK_PROC_DECL_STD(float, 512, , fma)
 {
        __sk_param_std(float);
        register __m512 c = _mm512_load_ps((float *)ans);
@@ -173,7 +173,7 @@ SN_SK_PROC_DECL_STD(float, 512, , fma)

 // clang-format on

-SN_SK_PROC_DECL_STD(double, sse, , fma)
+NIF_SK_PROC_DECL_STD(double, sse, , fma)
 {
        __sk_param_std(double);
        const double *left_values = &left->m[0][0];
@@ -204,7 +204,7 @@ SN_SK_PROC_DECL_STD(double, sse, , fma)
 }

 // clang-format off
-SN_SK_PROC_DECL_STD(double, avx, , fma)
+NIF_SK_PROC_DECL_STD(double, avx, , fma)
 {
        __sk_param_std(double);
        register __m256d c0 = _mm256_load_pd((double *)&ans->y[0]);
@@ -253,7 +253,7 @@ SN_SK_PROC_DECL_STD(double, avx, , fma)
 }
 // clang-format on

-SN_SK_PROC_DECL_STD(double, avx, _fma, fma)
+NIF_SK_PROC_DECL_STD(double, avx, _fma, fma)
 {
        __sk_param_std(double);
        register __m256d c0 = _mm256_load_pd((double *)&ans->y[0]);
@@ -301,7 +301,7 @@ SN_SK_PROC_DECL_STD(double, avx, _fma, fma)
        _mm256_store_pd((double *)&ans->y[3], c3);
 }

-SN_SK_PROC_DECL_STD(double, 512, , fma)
+NIF_SK_PROC_DECL_STD(double, 512, , fma)
 {
        __sk_param_std(double);
        register __m512d c01 = _mm512_load_pd((double *)&ans->z[0]);
@@ -1,8 +1,8 @@
 #define _GNU_SOURCE

-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_OS(SN_OS_LINUX)
+#if NIF_IS_OS(NIF_OS_LINUX)
 #include <errno.h>
 #include <inttypes.h>
 #include <limits.h>
@@ -16,27 +16,27 @@
 #include <sys/syscall.h>
 #include <unistd.h>

-typedef int (*sn_linux_id_visitor)(int value, void *context);
+typedef int (*nif_linux_id_visitor)(int value, void *context);

-typedef struct sn_linux_counter_context {
+typedef struct nif_linux_counter_context {
        int count;
-} sn_linux_counter_context;
+} nif_linux_counter_context;

-typedef struct sn_linux_max_context {
+typedef struct nif_linux_max_context {
        int max_value;
-} sn_linux_max_context;
+} nif_linux_max_context;

-typedef struct sn_linux_fill_cpu_context {
+typedef struct nif_linux_fill_cpu_context {
        uint64_t *internal;
        uint32_t *global;
        int index;
-} sn_linux_fill_cpu_context;
+} nif_linux_fill_cpu_context;

-typedef struct sn_linux_fill_node_count_context {
+typedef struct nif_linux_fill_node_count_context {
        int *count_of_each_numa;
-} sn_linux_fill_node_count_context;
+} nif_linux_fill_node_count_context;

-static int sn_linux_read_first_line(char const *path, char *buffer, size_t buffer_size)
+static int nif_linux_read_first_line(char const *path, char *buffer, size_t buffer_size)
 {
        FILE *file = fopen(path, "r");
        if (file == NULL)
@@ -47,7 +47,7 @@ static int sn_linux_read_first_line(char const *path, char *buffer, size_t buffe
        return line != NULL;
 }

-static int sn_linux_parse_nonnegative_int(char const **cursor, int *value)
+static int nif_linux_parse_nonnegative_int(char const **cursor, int *value)
 {
        char *end = NULL;
        uintmax_t parsed_value = 0;
@@ -65,7 +65,7 @@ static int sn_linux_parse_nonnegative_int(char const **cursor, int *value)
        return 1;
 }

-static int sn_linux_visit_id_list(char const *text, sn_linux_id_visitor visitor, void *context)
+static int nif_linux_visit_id_list(char const *text, nif_linux_id_visitor visitor, void *context)
 {
        char const *cursor = text;

@@ -77,14 +77,14 @@ static int sn_linux_visit_id_list(char const *text, sn_linux_id_visitor visitor,
                        break;

                int begin = 0;
-                if (!sn_linux_parse_nonnegative_int(&cursor, &begin))
+                if (!nif_linux_parse_nonnegative_int(&cursor, &begin))
                        return 0;

                int finish = begin;

                if (*cursor == '-') {
                        cursor++;
-                        if (!sn_linux_parse_nonnegative_int(&cursor, &finish))
+                        if (!nif_linux_parse_nonnegative_int(&cursor, &finish))
                                return 0;
                }

@@ -103,24 +103,24 @@ static int sn_linux_visit_id_list(char const *text, sn_linux_id_visitor visitor,
        return 1;
 }

-static int sn_linux_count_id(int value, void *context)
+static int nif_linux_count_id(int value, void *context)
 {
        (void)value;
-        ((sn_linux_counter_context *)context)->count++;
+        ((nif_linux_counter_context *)context)->count++;
        return 1;
 }

-static int sn_linux_track_max_id(int value, void *context)
+static int nif_linux_track_max_id(int value, void *context)
 {
-        sn_linux_max_context *max_context = (sn_linux_max_context *)context;
+        nif_linux_max_context *max_context = (nif_linux_max_context *)context;
        if (value > max_context->max_value)
                max_context->max_value = value;
        return 1;
 }

-static int sn_linux_fill_cpu_id(int value, void *context)
+static int nif_linux_fill_cpu_id(int value, void *context)
 {
-        sn_linux_fill_cpu_context *fill_context = (sn_linux_fill_cpu_context *)context;
+        nif_linux_fill_cpu_context *fill_context = (nif_linux_fill_cpu_context *)context;

        if (fill_context->internal != NULL)
                fill_context->internal[fill_context->index] = (uint64_t)fill_context->index;
@@ -131,37 +131,37 @@ static int sn_linux_fill_cpu_id(int value, void *context)
        return 1;
 }

-static int sn_linux_count_cpu_list(char const *path)
+static int nif_linux_count_cpu_list(char const *path)
 {
        char buffer[256];
-        if (!sn_linux_read_first_line(path, buffer, sizeof(buffer)))
+        if (!nif_linux_read_first_line(path, buffer, sizeof(buffer)))
                return 0;

-        sn_linux_counter_context counter = { 0 };
-        if (!sn_linux_visit_id_list(buffer, sn_linux_count_id, &counter))
+        nif_linux_counter_context counter = { 0 };
+        if (!nif_linux_visit_id_list(buffer, nif_linux_count_id, &counter))
                return 0;

        return counter.count;
 }

-static int sn_linux_fill_node_cpu_count(int value, void *context)
+static int nif_linux_fill_node_cpu_count(int value, void *context)
 {
        char path[128];
        snprintf(path, sizeof(path), "/sys/devices/system/node/node%d/cpulist", value);

-        ((sn_linux_fill_node_count_context *)context)->count_of_each_numa[value] =
-                sn_linux_count_cpu_list(path);
+        ((nif_linux_fill_node_count_context *)context)->count_of_each_numa[value] =
+                nif_linux_count_cpu_list(path);
        return 1;
 }

-static int sn_linux_node_exists(int numa_id)
+static int nif_linux_node_exists(int numa_id)
 {
        char path[128];
        snprintf(path, sizeof(path), "/sys/devices/system/node/node%d", numa_id);
        return access(path, F_OK) == 0;
 }

-static size_t sn_linux_round_up_to_page_size(size_t size)
+static size_t nif_linux_round_up_to_page_size(size_t size)
 {
        int64_t page_size_value = (int64_t)sysconf(_SC_PAGESIZE);
        size_t effective_page_size = (page_size_value > 0) ? (size_t)page_size_value : 4096u;
@@ -173,7 +173,7 @@ static size_t sn_linux_round_up_to_page_size(size_t size)
        return rounded - (rounded % effective_page_size);
 }

-static int sn_linux_bind_memory_to_node(void *address, size_t size, int numa_id, uint32_t flags)
+static int nif_linux_bind_memory_to_node(void *address, size_t size, int numa_id, uint32_t flags)
 {
 #if defined(SYS_mbind)
        size_t bits_per_word = sizeof(unsigned long) * 8u;
@@ -197,7 +197,7 @@ static int sn_linux_bind_memory_to_node(void *address, size_t size, int numa_id,
 #endif
 }

-static int sn_linux_populate_writable_mapping(void *address, size_t size)
+static int nif_linux_populate_writable_mapping(void *address, size_t size)
 {
 #if defined(MADV_POPULATE_WRITE)
        if (madvise(address, size, MADV_POPULATE_WRITE) == 0)
@@ -214,7 +214,7 @@ static int sn_linux_populate_writable_mapping(void *address, size_t size)
 /* Pin the calling thread to a single logical processor.
 * core_id >= 0 : bind to that logical processor (global index, across groups).
 * core_id <  0 : clear affinity constraint within the current group. */
-SN_API void SN_DECL sn_set_thread_affinity(int core_id, int numa_id)
+NIF_API void NIF_DECL nif_set_thread_affinity(int core_id, int numa_id)
 {
        int cpu_count = (int)sysconf(_SC_NPROCESSORS_ONLN);
        if (cpu_count <= 0)
@@ -238,14 +238,14 @@ SN_API void SN_DECL sn_set_thread_affinity(int core_id, int numa_id)
        CPU_FREE(set);
 }

-SN_API int SN_DECL sn_get_numa_overall(int *out_numa_count, int *count_of_each_numa)
+NIF_API int NIF_DECL nif_get_numa_overall(int *out_numa_count, int *count_of_each_numa)
 {
        char buffer[256];
-        if (!sn_linux_read_first_line("/sys/devices/system/node/online", buffer, sizeof(buffer)))
+        if (!nif_linux_read_first_line("/sys/devices/system/node/online", buffer, sizeof(buffer)))
                return 0;

-        sn_linux_max_context max_context = { -1 };
-        if (!sn_linux_visit_id_list(buffer, sn_linux_track_max_id, &max_context) ||
+        nif_linux_max_context max_context = { -1 };
+        if (!nif_linux_visit_id_list(buffer, nif_linux_track_max_id, &max_context) ||
            max_context.max_value < 0)
                return 0;

@@ -256,15 +256,15 @@ SN_API int SN_DECL sn_get_numa_overall(int *out_numa_count, int *count_of_each_n
                for (int index = 0; index <= max_context.max_value; index++)
                        count_of_each_numa[index] = 0;

-                sn_linux_fill_node_count_context fill_context = { count_of_each_numa };
-                if (!sn_linux_visit_id_list(buffer, sn_linux_fill_node_cpu_count, &fill_context))
+                nif_linux_fill_node_count_context fill_context = { count_of_each_numa };
+                if (!nif_linux_visit_id_list(buffer, nif_linux_fill_node_cpu_count, &fill_context))
                        return 0;
        }

        return 1;
 }

-SN_API int SN_DECL sn_get_numa_each(int numa_id, uint64_t *internal, uint32_t *global)
+NIF_API int NIF_DECL nif_get_numa_each(int numa_id, uint64_t *internal, uint32_t *global)
 {
        if (numa_id < 0)
                return 0;
@@ -272,16 +272,16 @@ SN_API int SN_DECL sn_get_numa_each(int numa_id, uint64_t *internal, uint32_t *g
        char path[128];
        char buffer[256];
        snprintf(path, sizeof(path), "/sys/devices/system/node/node%d/cpulist", numa_id);
-        if (!sn_linux_read_first_line(path, buffer, sizeof(buffer)))
+        if (!nif_linux_read_first_line(path, buffer, sizeof(buffer)))
                return 0;

-        sn_linux_fill_cpu_context fill_context = { internal, global, 0 };
-        return sn_linux_visit_id_list(buffer, sn_linux_fill_cpu_id, &fill_context);
+        nif_linux_fill_cpu_context fill_context = { internal, global, 0 };
+        return nif_linux_visit_id_list(buffer, nif_linux_fill_cpu_id, &fill_context);
 }

-SN_API void *SN_DECL sn_numa_alloc(size_t size, int numa_id)
+NIF_API void *NIF_DECL nif_numa_alloc(size_t size, int numa_id)
 {
-        size_t allocation_size = sn_linux_round_up_to_page_size(size);
+        size_t allocation_size = nif_linux_round_up_to_page_size(size);
        if (allocation_size == 0)
                return NULL;

@@ -303,9 +303,9 @@ SN_API void *SN_DECL sn_numa_alloc(size_t size, int numa_id)
        bind_flags = MPOL_MF_MOVE | MPOL_MF_STRICT;
 #endif

-        if (!sn_linux_node_exists(numa_id) ||
-            !sn_linux_bind_memory_to_node(memory, allocation_size, numa_id, bind_flags) ||
-            !sn_linux_populate_writable_mapping(memory, allocation_size)) {
+        if (!nif_linux_node_exists(numa_id) ||
+            !nif_linux_bind_memory_to_node(memory, allocation_size, numa_id, bind_flags) ||
+            !nif_linux_populate_writable_mapping(memory, allocation_size)) {
                munmap(memory, allocation_size);
                return NULL;
        }
@@ -313,14 +313,14 @@ SN_API void *SN_DECL sn_numa_alloc(size_t size, int numa_id)
        return memory;
 }

-SN_API void SN_DECL sn_numa_free(void *ptr, size_t size, int numa_id)
+NIF_API void NIF_DECL nif_numa_free(void *ptr, size_t size, int numa_id)
 {
        (void)numa_id;

        if (ptr == NULL || size == 0)
                return;

-        size_t allocation_size = sn_linux_round_up_to_page_size(size);
+        size_t allocation_size = nif_linux_round_up_to_page_size(size);
        if (allocation_size != 0)
                munmap(ptr, allocation_size);
 }
@@ -1,12 +1,12 @@
-#include "sn_target.h"
-#include "StgSharpNative.h"
+#include "nif_target.h"
+#include "NitloadFrameWork.h"

-#if SN_IS_OS(SN_OS_WINDOWS)
+#if NIF_IS_OS(NIF_OS_WINDOWS)

 /* Pin the calling thread to a single logical processor.
 * core_id >= 0 : bind to that logical processor (global index, across groups).
 * core_id <  0 : clear affinity constraint within the current group. */
-SN_API void SN_DECL sn_set_thread_affinity(int core_id, int numa_id)
+NIF_API void NIF_DECL nif_set_thread_affinity(int core_id, int numa_id)
 {
        if (core_id < 0 || numa_id < 0) {
                /* Query which group the thread currently belongs to */
@@ -32,7 +32,7 @@ SN_API void SN_DECL sn_set_thread_affinity(int core_id, int numa_id)
        SetThreadGroupAffinity(GetCurrentThread(), &ga, NULL);
 }

-SN_API int SN_DECL sn_get_numa_overall(int *out_numa_count, int *count_of_each_numa)
+NIF_API int NIF_DECL nif_get_numa_overall(int *out_numa_count, int *count_of_each_numa)
 {
        ULONG highest;
        if (!GetNumaHighestNodeNumber(&highest))
@@ -47,7 +47,7 @@ SN_API int SN_DECL sn_get_numa_overall(int *out_numa_count, int *count_of_each_n
        return 1;
 }

-SN_API int SN_DECL sn_get_numa_each(int numa_id, uint64_t *internal, uint32_t *global)
+NIF_API int NIF_DECL nif_get_numa_each(int numa_id, uint64_t *internal, uint32_t *global)
 {
        ULONGLONG mask = 0;
        if (!GetNumaNodeProcessorMask((UCHAR)numa_id, &mask))
@@ -67,7 +67,7 @@ SN_API int SN_DECL sn_get_numa_each(int numa_id, uint64_t *internal, uint32_t *g
 }


-SN_API void *SN_DECL sn_numa_alloc(size_t size, int numa_id)
+NIF_API void *NIF_DECL nif_numa_alloc(size_t size, int numa_id)
 {
        if (size == 0)
                return NULL;
@@ -83,7 +83,7 @@ SN_API void *SN_DECL sn_numa_alloc(size_t size, int numa_id)
                                  PAGE_READWRITE, (DWORD)numa_id);
 }

-SN_API void SN_DECL sn_numa_free(void *ptr, size_t size, int numa_id)
+NIF_API void NIF_DECL nif_numa_free(void *ptr, size_t size, int numa_id)
 {
        (void)size;
        (void)numa_id;
@@ -1,7 +1,7 @@
-#include "sn_intrinsic.h"
-#include "sn_internal.h"
+#include "nif_intrinsic.h"
+#include "nif_internal.h"

-INTERNAL void SN_DECL f32_normalize(VEC_SEGMENT(float) *restrict source, VEC_SEGMENT(float) *restrict target)
+INTERNAL void NIF_DECL f32_normalize(VEC_SEGMENT(float) *restrict source, VEC_SEGMENT(float) *restrict target)
 {
        register __m128 s = source->f32_x[0];
        register __m128 a_length_sq = _mm_mul_ps(s, s);
@@ -1,9 +1,9 @@
-#include "sn_intrinsic.h"
-#include "sn_internal.h"
+#include "nif_intrinsic.h"
+#include "nif_internal.h"

 static const int avxdot_indices[] = { 0, 4, 1, 5, 2, 3, 6, 7 };

-INTERNAL void FORCEINLINE SN_DECL dot_41_sse(MAT_KERNEL(float) const *transpose,
+INTERNAL void FORCEINLINE NIF_DECL dot_41_sse(MAT_KERNEL(float) const *transpose,
                                             __m128 const *vector, __m128 *ans)
 {
        *ans = _mm_dp_ps(transpose->x[0], *vector, 0B11110001);
@@ -15,7 +15,7 @@ INTERNAL void FORCEINLINE SN_DECL dot_41_sse(MAT_KERNEL(float) const *transpose,
                             SSE_INSERT_SIMPLE(1, 4));
 }

-INTERNAL void SN_DECL dot_41_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
+INTERNAL void NIF_DECL dot_41_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
                                 __m128 *ans)
 {
        __m256 _double = _mm256_set_m128(vector[0], vector[0]);
@@ -26,7 +26,7 @@ INTERNAL void SN_DECL dot_41_avx(MAT_KERNEL(float) const *transpose, __m128 cons
                _mm256_permutevar8x32_ps(tmp, _mm256_loadu_epi32(avxdot_indices)));
 }

-INTERNAL void SN_DECL dot_42_sse(MAT_KERNEL(float) const *transpose, __m128 const *vector,
+INTERNAL void NIF_DECL dot_42_sse(MAT_KERNEL(float) const *transpose, __m128 const *vector,
                                 __m128 *ans)
 {
        ans[0] = _mm_dp_ps(transpose->x[0], vector[0], 0B11110001);
@@ -46,7 +46,7 @@ INTERNAL void SN_DECL dot_42_sse(MAT_KERNEL(float) const *transpose, __m128 cons
                               SSE_INSERT_SIMPLE(1, 4));
 }

-INTERNAL void SN_DECL dot_42_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
+INTERNAL void NIF_DECL dot_42_avx(MAT_KERNEL(float) const *transpose, __m128 const *vector,
                                 __m128 *ans)
 {
        __m256 temp = transpose->y[0];
@@ -1,8 +1,8 @@
-#include "sn_target.h"
+#include "nif_target.h"

-#if SN_IS_ARCH(SN_ARCH_X86_64)
+#if NIF_IS_ARCH(NIF_ARCH_X86_64)

-#define SN_MK_FMA_F32_512_CYCLE(i)                                                       \
+#define NIF_MK_FMA_F32_512_CYCLE(i)                                                       \
        do {                                                                             \
                c0 = _mm512_fmadd_ps(c0, a0, _mm512_permute_ps(b0, _FMA_SELECT(i))); \
                c1 = _mm512_fmadd_ps(c1, a1, _mm512_permute_ps(b1, _FMA_SELECT(i))); \
@@ -10,7 +10,7 @@
                c3 = _mm512_fmadd_ps(c3, a3, _mm512_permute_ps(b3, _FMA_SELECT(i))); \
        } while (0)

-#define SN_MK_FMA_F32_AVX2_CYCLE(i)                                                       \
+#define NIF_MK_FMA_F32_AVX2_CYCLE(i)                                                       \
        do {                                                                             \
                c0 = _mm256_fmadd_ps(c0, a0, _mm256_permute_ps(b0, _FMA_SELECT(i))); \
                c1 = _mm256_fmadd_ps(c1, a1, _mm256_permute_ps(b1, _FMA_SELECT(i))); \
@@ -18,7 +18,7 @@
                c3 = _mm256_fmadd_ps(c3, a3, _mm256_permute_ps(b3, _FMA_SELECT(i))); \
        } while (0)

-#define SN_MK_FMA_F64_AVX2_CYCLE(i)                                                       \
+#define NIF_MK_FMA_F64_AVX2_CYCLE(i)                                                       \
        do {                                                                             \
                c0 = _mm512_fmadd_ps(c0, a0, _mm512_permutex_pd(b0, _FMA_SELECT(i))); \
                c1 = _mm512_fmadd_ps(c1, a1, _mm512_permutex_pd(b1, _FMA_SELECT(i))); \
@@ -26,7 +26,7 @@
                c3 = _mm512_fmadd_ps(c3, a3, _mm512_permutex_pd(b3, _FMA_SELECT(i))); \
        } while (0)

-#define SN_FMA_F32_512_PACK_COLUMN(col, target)                                   \
+#define NIF_FMA_F32_512_PACK_COLUMN(col, target)                                   \
        do {                                                              \
                register __m256 low = _mm512_castps512_ps256(col);        \
                register __m256 high = _mm512_extractf32x8_ps(col, 1);    \
@@ -37,7 +37,7 @@
                _mm_store_ps(target, half128);                            \
        } while (0)

-#define SN_FMA_F32_AVX2_PACK_COLUMN(col, target)                                   \
+#define NIF_FMA_F32_AVX2_PACK_COLUMN(col, target)                                   \
        do {                                                              \
                register __m128 high128 = _mm256_extractf32x4_ps(col, 1); \
                register __m128 low128 = _mm256_castps256_ps128(col);    \
@@ -45,7 +45,7 @@
                _mm_store_ps(target, half128);                            \
        } while (0)

-#define SN_FMA_F64_512_PACK_COLUMN(col, target)                                   \
+#define NIF_FMA_F64_512_PACK_COLUMN(col, target)                                   \
        do {                                                              \
                register __m128 high128 = _mm256_extractf32x4_ps(col, 1); \
                register __m128 low128 = _mm256_castps256_ps128(col);    \
@@ -54,7 +54,7 @@
        } while (0)

 // double
-#define SN_MK_FMA_F64_SSE_PHASE(i)                                               \
+#define NIF_MK_FMA_F64_SSE_PHASE(i)                                               \
        do {                                                                     \
                /* k = 0 */                                                      \
                register __m128d b_lo = _mm_load_pd(right_values + (2 * i + 0)); \
@@ -1,35 +0,0 @@
-#ifndef SNative
-#define SNative
-
-#ifdef __cplusplus
-#error "This header is not meant to be included in C++ code. Nitload.Native is a C library and should be used with C linkage."
-#endif
-
-#include "sn_internal.h"
-#include "sn_target.h"
-
-#include <immintrin.h>
-#include <assert.h>
-#include <glfw_function.h>
-
-#define byte char
-#define cs_char uint16_t
-#define sn_null_assert(ptr) assert(ptr != NULL)
-
-#ifndef SN_
-#endif
-
-SN_API void SN_DECL load_glfw_functions(GLFWFunctionTable *table);
-SN_API void SN_DECL load_intrinsic_function(void *intrinsic_context, uint64_t id);
-SN_API SIMDID SN_DECL sn_get_simd_level_global();
-/* Per-core variant: always re-executes CPUID, never caches.
- * Must be called on the target core (after affinity is set). */
-SN_API SIMDID SN_DECL sn_get_simd_level_local();
-SN_API SIMDID SN_DECL sn_get_unite_simd(SIMDID left, SIMDID right);
-SN_API int SN_DECL sn_compare_simd(SIMDID left, SIMDID right);
-SN_API void SN_DECL sn_set_thread_affinity(int core_id, int numa_id);
-SN_API void* SN_DECL sn_numa_alloc(size_t size, int numa_id);
-SN_API void SN_DECL sn_numa_free(void *ptr, size_t size, int numa_id);
-SN_API int SN_DECL sn_get_numa_each(int numa_id, uint64_t *internal, uint32_t *global);
-SN_API int SN_DECL sn_get_numa_overall(int *out_numa_count, int *count_of_each_numa);
-#endif
@@ -1,54 +0,0 @@
-#ifndef SN_MATKERNEL
-#define SN_MATKERNEL
-
-#include <immintrin.h>
-#include <stdint.h>
-
-#include "sn_internal.h"
-
-/* Kernel storage unions for a fixed 4x4 block. */
-#define ELEMENTSIZE(T) sizeof(T)
-#define XMMCOUNT(T) (sizeof(T))
-#define YMMCOUNT(T) (sizeof(T) / 2)
-#define ZMMCOUNT(T) (sizeof(T) / 4)
-
-#define SN_MAT_KERNEL_MEMBER(T)        \
-        {                              \
-                __m128 x[XMMCOUNT(T)]; \
-                __m256 y[YMMCOUNT(T)]; \
-                __m512 z[ZMMCOUNT(T)]; \
-                T m[4][4];             \
-        }
-        
-#define SN_MAT_KERNEL_DEFINE(T) \
-        typedef union SN_MAT_KERNEL_TYPE_##T SN_MAT_KERNEL_MEMBER(T) SN_MAT_KERNEL_TYPE_##T
-
-#define MAT_KERNEL(T) SN_MAT_KERNEL_TYPE(T)
-#define SN_MAT_KERNEL_TYPE(T) SN_MAT_KERNEL_TYPE_##T
-
-SN_MAT_KERNEL_DEFINE(float);
-SN_MAT_KERNEL_DEFINE(double);
-SN_MAT_KERNEL_DEFINE(uint64_t);
-
-
-#define SN_VEC_SEGMENT_MEMBER(T)               \
-        {                                \
-                __m128 f32_x[XMMCOUNT(T) / 4]; \
-                __m256 f32_y[YMMCOUNT(T) / 4]; \
-                __m512 f32_z[ZMMCOUNT(T) / 4]; \
-                T v[4];                        \
-        }
-
-#define SN_VEC_SEGMENT_DEFINE(T) \
-        typedef union SN_VEC_SEGMENT_TYPE_##T SN_VEC_SEGMENT_MEMBER(T) SN_VEC_SEGMENT_TYPE_##T
-
-#define VEC_SEGMENT(T) SN_VEC_SEGMENT_TYPE(T)
-#define SN_VEC_SEGMENT_TYPE(T) SN_VEC_SEGMENT_TYPE_##T
-
-SN_VEC_SEGMENT_DEFINE(float);
-SN_VEC_SEGMENT_DEFINE(double);
-
-/* Global zero kernel declaration (defined in dllmain.c), aligned to 64 bytes */
-extern __declspec(align(64)) MAT_KERNEL(u64) SN_ZERO_KERNEL;
-
-#endif /* SN_MATKERNEL */
@@ -1,94 +0,0 @@
-#ifndef SN_TARGET
-#define SN_TARGET
-
-#include "sn_internal.h"
-#include <stdint.h>
-
-#define SN_OS_WINDOWS 0
-#define SN_OS_LINUX 1
-#define SN_OS_MACOS 2
-#define SN_OS_IOS 3
-#define SN_OS_ANDROID 4
-#define SN_OS_UNKNOWN 5
-
-#define SN_ARCH_X86_64 0
-#define SN_ARCH_ARM64 1
-#define SN_ARCH_RISCV64 2
-#define SN_ARCH_PPC64le 3
-
-#if defined(_M_X64) || defined(_M_AMD64) || defined(__x86_64__) || defined(__amd64__)
-#define SN_CURRENT_ARCH SN_ARCH_X86_64
-
-#elif defined(_M_ARM64) || defined(__aarch64__)
-#define SN_CURRENT_ARCH SN_ARCH_ARM64
-
-#elif defined(_M_ARM) || defined(__arm__)
-#define SN_CURRENT_ARCH SN_ARCH_ARM32
-
-#elif defined(_M_IX86) || defined(__i386__)
-#define SN_CURRENT_ARCH SN_ARCH_X86_32
-
-#elif defined(__powerpc64__) && defined(__LITTLE_ENDIAN__)
-#define SN_CURRENT_ARCH SN_ARCH_PPC64LE
-
-#elif defined(__riscv) && (__riscv_xlen == 64)
-#define SN_CURRENT_ARCH SN_ARCH_RISCV64
-
-#else
-#error "SN: Unsupported target architecture"
-#endif
-
-/*
- * Target selection policy:
- * - Default (no SN_EXTERN_TARGET): fixed Windows + x86_64.
- * - External (SN_EXTERN_TARGET): build system must provide SN_OS and SN_ARCH.
- */
-#ifndef SN_EXTERN_TARGET
-
-#define SN_OS_TARGET SN_OS_WINDOWS
-#define SN_ARCH_TARGET SN_ARCH_X86_64
-
-#else
-
-#ifndef SN_OS_TARGET
-#error "SN_EXTERN_TARGET is defined but SN_OS is not provided."
-#endif
-
-#ifndef SN_ARCH_TARGET
-#error "SN_EXTERN_TARGET is defined but SN_ARCH is not provided."
-#endif
-
-/* Validate externally supplied values are within known enum constants. */
-#if (SN_OS_TARGET != SN_OS_WINDOWS) && (SN_OS_TARGET != SN_OS_LINUX) && \
-        (SN_OS_TARGET != SN_OS_MACOS) && (SN_OS_TARGET != SN_OS_IOS) && \
-        (SN_OS_TARGET != SN_OS_ANDROID) && (SN_OS_TARGET != SN_OS_UNKNOWN)
-#error "SN_OS has an invalid value. Use SN_OS_<...> enum constants."
-#endif
-
-#if (SN_ARCH_TARGET != SN_ARCH_X86_64) && (SN_ARCH_TARGET != SN_ARCH_ARM64) && \
-        (SN_ARCH_TARGET != SN_ARCH_RISCV64) && (SN_ARCH_TARGET != SN_ARCH_PPC64le)
-#error "SN_ARCH has an invalid value. Use SN_ARCH_<...> enum constants."
-#endif
-
-#endif
-
-#define SN_IS_OS(os) ((SN_OS_TARGET) == (os))
-
-#define SN_IS_ARCH(arch) ((SN_ARCH_TARGET) == (arch))
-
-/*
- * Per-function ISA selection is required for runtime-dispatched kernels that
- * mix multiple x86 feature levels in one build target.
- */
-#if defined(__clang__) && ((SN_ARCH_TARGET) == SN_ARCH_X86_64)
-#define SN_CLANG_TARGET_ATTR(options) __attribute__((target(options)))
-#else
-#define SN_CLANG_TARGET_ATTR(options)
-#endif
-
-typedef union SIMDID {
-        uint64_t mask;
-        char make_byte[8];
-} SIMDID;
-
-#endif // SN_TARGET
@@ -1,204 +0,0 @@
-# NGRA `COUNT(UNIT)` 生成说明
-
-本文记录 `TextRegexSourceGen.GenerateCount` 中单字符重复分支的当前设计，方便后续继续实现。
-
-相关代码：[`Text/TextRegexSourceGen.GenClass.cs`](Text/TextRegexSourceGen.GenClass.cs)。
-
-## 目标
-
-对于 `a{min,max}`，不要把必需的 `min` 次匹配生成成逐字符循环。生成器先把字符重复 `min` 次，形成一个字符串字面量，并使用一次 `StartsWith` 验证下界：
-
-```csharp
-if (__remain.StartsWith("aaa"))
-{
-    // 已经确认至少存在 min 个 a。
-}
-else
-{
-    // 当前工作 region 失败。
-}
-```
-
-匹配成功后一次性消费这个前缀，再只处理区间中可选的部分：
-
-```text
-必需前缀：min 次
-可选尾部：max - min 次
-```
-
-`max < 0` 表示尾部没有上限。
-
-## AST 数据归属
-
-`COUNT` 配置属于操作符节点，重复的字符属于其子节点：
-
-```csharp
-RegexCountPayload payload = node.PayloadAs<RegexCountPayload>();
-RegexAstNode value = RegexAstNode.IsNullOrEmpty(node.Left)
-    ? node.Right
-    : node.Left;
-char expected = value.Value[0];
-```
-
-不要从 `value` 读取 `RegexCountPayload`。`value` 携带的是 `RegexLiteralPayload` 或其他被重复元素的 payload。
-
-字符应从 `value.Value` 读取，因为它是解码后的值。使用 `value.Source.Source[0]` 会把 `\t` 等转义序列错误地识别成 `'\\'`。
-
-## 量词语义
-
-| 量词 | `min` | `max` | 尾部处理 |
-| --- | ---: | ---: | --- |
-| `a{n}` | `n` | `n` | 无可选尾部 |
-| `a{m,n}` | `m` | `n` | 最多再消费 `n - m` 次 |
-| `a*` | `0` | `-1` | 无上限 |
-| `a+` | `1` | `-1` | 验证一次后无上限 |
-| `a?` | `0` | `1` | 最多消费一次 |
-| `a{m,}` | `m` | `-1` | 验证 `m` 次后无上限 |
-
-有限上限可以生成：
-
-```csharp
-__remain = __remain.Slice(min);
-
-for (int i = 0;
-     i < max - min &&
-     !__remain.IsEmpty &&
-     __remain[0] == expected;
-     i++)
-{
-    __remain = __remain.Slice(1);
-}
-```
-
-无上限不能计算 `max - min`，否则结果为负数、循环不会执行。应单独生成：
-
-```csharp
-__remain = __remain.Slice(min);
-
-while (!__remain.IsEmpty && __remain[0] == expected)
-{
-    __remain = __remain.Slice(1);
-}
-```
-
-`!__remain.IsEmpty` 必须位于 `__remain[0]` 之前。输入恰好在必需前缀后耗尽时，直接索引会抛出 `IndexOutOfRangeException`。
-
-## 贪婪与惰性
-
-贪婪量词可以先消费尽可能多的可选尾部，再尝试后续 continuation。如果后续失败，需要逐次回退消费数量并重试。
-
-惰性量词只应先消费 `min`，立即尝试后续 continuation；后续失败时，才逐次多消费一个匹配字符。
-
-因此 `is_greedy` 不能只改变扫描循环，它最终需要与后续节点的 continuation 和回溯状态一起实现。若 COUNT 位于匹配末尾，贪婪模式可直接扫描至上限，惰性模式保留在 `min` 即可。
-
-## 失败工作 region
-
-当前试点使用 `do { } while (false)`，使局部失败可以写成 `break`：
-
-```csharp
-bool regionSucceeded = false;
-ReadOnlySpan<char> checkpoint = __remain;
-
-do
-{
-    if (!__remain.StartsWith(requiredPrefix))
-    {
-        break;
-    }
-
-    __remain = __remain.Slice(min);
-    // 匹配可选尾部和后续节点。
-
-    regionSucceeded = true;
-}
-while (false);
-
-if (!regionSucceeded)
-{
-    __remain = checkpoint;
-    // 向父 region 传播失败，或者在顶层返回失败结果。
-}
-```
-
-只有 COUNT 自己被 `do/while(false)` 包裹时，`break` 结束后仍会执行下一个 CONCAT 节点，不能真正表达匹配失败。最终 region 应由以下之一持有：
-
- 包含后续 continuation 的 `CONCAT`；
- 一条完整的 ALT 候选路径；
- 顶层匹配方法。
-
-父 region 负责保存 cursor 和 capture checkpoint。子节点失败后退出父 region，由父 region 统一恢复状态。ALT 和贪婪/惰性 COUNT 还需要保存多个可重试位置，不能只依赖一次 `break`。
-
-## 摊平阈值
-
-不要无条件执行：
-
-```csharp
-Span<char> required = stackalloc char[min];
-```
-
-量词来自用户输入，`a{1000000}` 会让源生成器尝试进行巨大的栈分配，并生成巨型字符串字面量。建议设置较小阈值，例如 32 或 64：
-
- `min <= threshold`：摊平为字面量并调用 `StartsWith`；
- `min > threshold`：循环比较固定长度的摊平块，最后比较不足一个块的余数；
- 非法或超过生成器限制的数量：报告 Roslyn diagnostic，而不是让生成器崩溃。
-
-以阈值 64 为例，`a{150}` 被拆成两个 64 字符块和一个 22 字符余数块。运行时使用局部 span 逐块调用 `StartsWith`，全部成功后才一次性推进正式 cursor。这样失败时无需恢复已经提交的 `__remain`：
-
-```csharp
-if (__remain.Length < 150)
-{
-    break;
-}
-
-ReadOnlySpan<char> __requiredRest_0 = __remain;
-bool __requiredMatched_0 = true;
-
-for (int __requiredChunkIndex_0 = 0; __requiredChunkIndex_0 < 2; __requiredChunkIndex_0++)
-{
-    if (!__requiredRest_0.StartsWith("aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa"))
-    {
-        __requiredMatched_0 = false;
-        break;
-    }
-
-    __requiredRest_0 = __requiredRest_0.Slice(64);
-}
-
-if (__requiredMatched_0 && !__requiredRest_0.StartsWith("aaaaaaaaaaaaaaaaaaaaaa"))
-{
-    __requiredMatched_0 = false;
-}
-
-if (!__requiredMatched_0)
-{
-    break;
-}
-
-__remain = __remain.Slice(150);
-```
-
-## 后续实现顺序
-
-1. 从 `node` 读取 `RegexCountPayload`，并用 Empty sentinel 安全选择子节点。
-2. 改用 `value.Value[0]` 获取解码字符。
-3. 为所有尾部循环加入 `!__remain.IsEmpty`。
-4. 分开生成有限上限和 `max < 0` 两种循环。
-5. 增加摊平阈值，避免巨大 `stackalloc` 和生成源膨胀。
-6. 将工作 region 上移到 CONCAT/候选路径，完成失败传播和 checkpoint 恢复。
-7. 接入 greedy/lazy continuation 回溯。
-8. 最后处理 capture 状态的保存与恢复。
-
-## 最小测试矩阵
-
-至少覆盖以下模式和输入：
-
- `a{0}`：空输入、`a`；
- `a{3}`：`aa`、`aaa`、`aaaa`；
- `a{2,4}`：长度 1 到 5；
- `a*`、`a+`、`a?`、`a{2,}`；
- `\t{2}`：确认使用解码字符；
- `a{2,4}ab`：验证贪婪回退；
- `a{2,4}?ab`：验证惰性扩展；
- COUNT 位于 CONCAT、ALT 和捕获 GROUP 内的失败恢复；
- 输入恰好在 `min` 处结束，确认不会索引空 span；
- 超大 `min`，确认不会导致生成器栈溢出或生成巨型源码。