Use include_sse2neon.h in gcore/overview functionality

rouault · Mar 17, 2024 · 4e225a7 · 4e225a7
1 parent 450da77
commit 4e225a7
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 4 deletions.
diff --git a/gcore/gdalsse_priv.h b/gcore/gdalsse_priv.h
@@ -39,13 +39,18 @@
 #if (defined(__x86_64) || defined(_M_X64) || defined(USE_SSE2)) &&             \
     !defined(USE_SSE2_EMULATION)
 
+#include <string.h>
+
+#ifdef USE_NEON_OPTIMIZATIONS
+#include "include_sse2neon.h"
+#else
 /* Requires SSE2 */
 #include <emmintrin.h>
-#include <string.h>
 
 #ifdef __SSE4_1__
 #include <smmintrin.h>
 #endif
+#endif
 
 #include "gdal_priv_templates.hpp"
 

diff --git a/gcore/overview.cpp b/gcore/overview.cpp
@@ -52,9 +52,15 @@
 #include "gdal_thread_pool.h"
 #include "gdalwarper.h"
 
+#ifdef USE_NEON_OPTIMIZATIONS
+#include "include_sse2neon.h"
+#define USE_SSE2
+
+#include "gdalsse_priv.h"
+
 // Restrict to 64bit processors because they are guaranteed to have SSE2,
 // or if __AVX2__ is defined.
-#if defined(__x86_64) || defined(_M_X64) || defined(__AVX2__)
+#elif defined(__x86_64) || defined(_M_X64) || defined(__AVX2__)
 #define USE_SSE2
 
 #include "gdalsse_priv.h"
@@ -320,7 +326,7 @@ inline GUInt16 ComputeIntegerRMS_4values<GUInt16, double>(double sumSquares)
 /*                   QuadraticMeanByteSSE2OrAVX2()                      */
 /************************************************************************/
 
-#ifdef __SSE4_1__
+#if defined(__SSE4_1__) || defined(USE_NEON_OPTIMIZATIONS)
 #define sse2_packus_epi32 _mm_packus_epi32
 #else
 inline __m128i sse2_packus_epi32(__m128i a, __m128i b)
@@ -335,7 +341,7 @@ inline __m128i sse2_packus_epi32(__m128i a, __m128i b)
 }
 #endif
 
-#ifdef __SSSE3__
+#if defined(__SSSE3__) || defined(USE_NEON_OPTIMIZATIONS)
 #define sse2_hadd_epi16 _mm_hadd_epi16
 #else
 inline __m128i sse2_hadd_epi16(__m128i a, __m128i b)