Use SIMDe for NEON acceleration

.gitmodules

@@ -1,3 +1,6 @@
 [submodule "enet"]
 	path = enet
 	url = https://github.com/cgutman/enet.git
+[submodule "nanors/deps/simde"]
+	path = nanors/deps/simde
+	url = https://github.com/simd-everywhere/simde-no-tests.git

CMakeLists.txt

@@ -105,8 +105,9 @@ endif()
 
 target_include_directories(moonlight-common-c SYSTEM PUBLIC src)
 
-target_include_directories(moonlight-common-c PRIVATE
+target_include_directories(moonlight-common-c SYSTEM PRIVATE
   ${CMAKE_CURRENT_SOURCE_DIR}/nanors
+  ${CMAKE_CURRENT_SOURCE_DIR}/nanors/deps
   ${CMAKE_CURRENT_SOURCE_DIR}/nanors/deps/obl
 )
 

nanors/deps/obl/autoshim.h

@@ -8,8 +8,7 @@
 #define OBLAS_SSE3
 #else
 #if defined(__aarch64__) || (defined(_MSC_VER) && defined(_M_ARM64))
-// To be replaced with SIMDe
-// #define OBLAS_NEON
+#define OBLAS_NEON
 #endif
 #endif
 #endif

nanors/deps/obl/oblas_lite.c

@@ -176,7 +176,8 @@ void obl_axpyb32_ref(u8 *a, u32 *b, u8 u, unsigned k)
 #else
 #if defined(OBLAS_SSE3) || defined(OBLAS_NEON)
 #if defined(OBLAS_NEON)
-#include "sse2neon/sse2neon.h"
+#define SIMDE_ENABLE_NATIVE_ALIASES
+#include <simde/x86/ssse3.h>
 #else
 #include <emmintrin.h>
 #include <tmmintrin.h>