use sse2neon for aarch64 performance improvements

common/rfb/CMakeLists.txt

@@ -107,7 +107,12 @@ set(SCALE_DUMMY_SOURCES
   scale_dummy.cxx)
 
 if(COMPILER_SUPPORTS_SSE2)
-  set_source_files_properties(${SSE2_SOURCES} PROPERTIES COMPILE_FLAGS ${COMPILE_FLAGS} -msse2)
+  if(COMPILER_ARM)
+    # This is for Graviton2, adjust for other CPUs: -march=armv8-a+crc+crypto
+    set_source_files_properties(${SSE2_SOURCES} PROPERTIES COMPILE_FLAGS ${COMPILE_FLAGS} -march=armv8.2-a+fp16+rcpc+dotprod+crypto)
+  else()
+    set_source_files_properties(${SSE2_SOURCES} PROPERTIES COMPILE_FLAGS ${COMPILE_FLAGS} -msse2)
+  endif()
   set(RFB_SOURCES
     ${RFB_SOURCES}
     ${SSE2_SOURCES}

common/rfb/cpuid.cxx

@@ -51,7 +51,9 @@ namespace rfb {
 
 bool supportsSSE2() {
 	getcpuid();
-#if defined(__x86_64__) || defined(__i386__)
+#if defined(__aarch64__)
+        return true;
+#elif defined(__x86_64__) || defined(__i386__)
 	#define bit_SSE2        (1 << 26)
 	return cpuid[3] & bit_SSE2;
 #endif

common/rfb/scale_sse2.cxx

@@ -16,7 +16,11 @@
  * USA.
  */
 
+#ifdef __aarch64__
+#include "sse2neon.h"
+#else
 #include <emmintrin.h>
+#endif
 
 #include <rfb/scale_sse2.h>