Fix aarch64 builds.

CMakeLists.txt

@@ -52,6 +52,9 @@ message(STATUS "LPCNet version: ${LPCNET_VERSION}")
 # Set default flags
 set(CMAKE_C_FLAGS "-Wall -W -Wextra -Wno-unused-function -O3 -g -I. -MD ${CMAKE_C_FLAGS} -DENABLE_ASSERTIONS")
 
+# Arch specific stuff here
+message(STATUS "Host system arch is: ${CMAKE_SYSTEM_PROCESSOR}")
+
 # Detection of available CPU optimizations
 if(NOT DISABLE_CPU_OPTIMIZATION)
     if(UNIX AND NOT APPLE)
@@ -87,7 +90,7 @@ elseif(${AVX} OR ${AVX} GREATER 0)
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx")
 endif()
 
-# RPi
+# RPi / ARM 32bit
 if(${NEON} OR ${NEON} GREATER 0)
     message(STATUS "neon processor flags found or enabled.")
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfpu=neon -march=armv8-a -mtune=cortex-a53")

src/CMakeLists.txt

@@ -49,7 +49,7 @@ target_link_libraries(dump_data lpcnetfreedv m codec2)
 add_executable(test_lpcnet test_lpcnet.c)
 target_link_libraries(test_lpcnet lpcnetfreedv m codec2)
 
-if(AVX OR AVX2)
+if(AVX OR AVX2 OR CMAKE_SYSTEM_PROCESSOR STREQUAL "aarch64")
     add_executable(test_vec test_vec.c)
     target_link_libraries(test_vec m)
 else()

src/nnet.c

@@ -43,7 +43,7 @@
 
 #ifdef __AVX__
 #include "vec_avx.h"
-#elif __ARM_NEON__
+#elif __ARM_NEON__ || __aarch64__
 #include "vec_neon.h"
 #else
 #warning Compiling without any vectorization. This code will be very slow

src/test_vec.c

@@ -26,7 +26,7 @@ const char simd[]="AVX2";
 #else
 const char simd[]="AVX";
 #endif
-#elif __ARM_NEON__
+#elif __ARM_NEON__ || __aarch64__
 #include "vec_neon.h"
 const char simd[]="NEON";
 #else