Another try at using NEON

georgmartius · Jun 7, 2023 · 406e61d · 406e61d
1 parent 7de0fa3
commit 406e61d
Show file tree

Hide file tree

Showing 2 changed files with 11 additions and 0 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -32,6 +32,10 @@ if(NOT WIN32)
   add_definitions(-fPIC)
 endif()
 
+if(APPLE)
+  add_definitions(-march=armv8-a+fp+simd+crypto+crc)
+endif(APPLE)
+
 ### ORC is not used in any active code at the moment  ###
 # I tried it with 0.4.14
 #  0.4.10 did not work (not all opcode implemented)

diff --git a/src/motiondetect_opt.c b/src/motiondetect_opt.c
@@ -37,6 +37,13 @@
 #define SSE2_CMP_SUM_ROWS 8
 #endif
 
+#ifdef __ARM_NEON__
+#include "sse2neon.h"
+#define USE_SSE2
+#define USE_SSE2_CMP_HOR
+#define SSE2_CMP_SUM_ROWS 8
+#endif
+
 #ifdef USE_SSE2
 /**
    \see contrastSubImg using SSE2 optimization, Planar (1 byte per channel) only