add loongarch64 build support

Author: amazingfate

Source/GmmLib/CMakeLists.txt

@@ -167,6 +167,8 @@ endif()
 
 if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^aarch")
     set(GMMLIB_MARCH "armv8-a+fp+simd")
+elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^loongarch")
+    set(GMMLIB_MARCH "la464")
 elseif("${GMMLIB_MARCH}" STREQUAL "")
     set(GMMLIB_MARCH "corei7")
 endif()
@@ -443,6 +445,18 @@ include_directories(BEFORE ${PROJECT_SOURCE_DIR})
 
 if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^aarch")
     include_directories(${GFX_DEVELOPMENT_DIR}/third_party/sse2neon)
+elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^loongarch")
+    find_path(SIMDE_INCLUDE_DIR
+        NAMES simde/simde-common.h  # A key SIMDE header
+        PATHS /usr/include /usr/local/include  # Default paths
+        DOC "Path to SIMDE headers"
+    )
+    if(SIMDE_INCLUDE_DIR)
+        include_directories(${SIMDE_INCLUDE_DIR})
+        message(STATUS "Found SIMDE: ${SIMDE_INCLUDE_DIR}")
+    else()
+        message(FATAL_ERROR "SIMDE not found. Install it or set SIMDE_INCLUDE_DIR manually.")
+    endif()
 endif()
 
 set(headers

Source/GmmLib/Linux.cmake

@@ -54,6 +54,8 @@ SET (GMMLIB_COMPILER_FLAGS_COMMON
 
 if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^aarch")
     list(APPEND GMMLIB_COMPILER_FLAGS_COMMON "-DUSE_NEON")
+elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^loongarch")
+    list(APPEND GMMLIB_COMPILER_FLAGS_COMMON "-Wno-attributes")
 else()
     list (APPEND GMMLIB_COMPILER_FLAGS_COMMON
         -Wall
@@ -145,6 +147,9 @@ endforeach()
 if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^aarch")
     SET(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
     SET(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS}")
+elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^loongarch")
+    SET(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS}")
+    SET(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS}")
 else()
     SET(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -m${GMMLIB_ARCH}")
     SET(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} -m${GMMLIB_ARCH}")

Source/GmmLib/Utility/CpuSwizzleBlt/CpuSwizzleBlt.c

@@ -375,6 +375,9 @@ extern void CpuSwizzleBlt(CPU_SWIZZLE_BLT_SURFACE *pDest, CPU_SWIZZLE_BLT_SURFAC
     #include <intrin.h>
 #elif defined(__ARM_ARCH)
     #include <sse2neon.h>
+#elif defined(__loongarch64)
+    #define SIMDE_X86_SSE2_ENABLE_NATIVE_ALIASES
+    #include <simde/x86/sse2.h>
 #elif((defined __clang__) ||(__GNUC__ > 4) || ((__GNUC__ == 4) && (__GNUC_MINOR__ >= 5)))
     #include <cpuid.h>
     #include <x86intrin.h>
@@ -749,6 +752,9 @@ void CpuSwizzleBlt( // #########################################################
                 #elif(defined(__ARM_ARCH))
                     #define MOVNTDQA_R(Reg, Src) ((Reg) = (Reg))
                     StreamingLoadSupported = 0;
+                #elif(defined(__loongarch64))
+                    #define MOVNTDQA_R(Reg, Src) ((Reg) = (Reg))
+                    StreamingLoadSupported = 0;
                 #elif((defined __clang__) || (__GNUC__ > 4) || (__GNUC__ == 4) && (__GNUC_MINOR__ >= 5))
                     #define MOVNTDQA_R(Reg, Src) ((Reg) = _mm_stream_load_si128((__m128i *)(Src)))
                     unsigned int eax, ebx, ecx, edx;
@@ -1148,7 +1154,11 @@ void CpuSwizzleBlt( // #########################################################
 
             } // foreach(y)
 
-            _mm_sfence(); // Flush Non-Temporal Writes
+            #if(defined(__loongarch64))
+                __sync_synchronize();
+            #else
+                _mm_sfence(); // Flush Non-Temporal Writes
+            #endif
 
             #if(_MSC_VER)
                 #pragma warning(pop)