Fix neon headers for MSVC ARM64

MSVC uses arm64_neon.h for ARM64, arm_neon.h is ARM32 only.

Common/ColorConvNEON.cpp

@@ -18,7 +18,11 @@
 #include "ppsspp_config.h"
 #if PPSSPP_ARCH(ARM_NEON)
 
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
+#endif
 #include "ColorConvNEON.h"
 #include "Common.h"
 #include "CPUDetect.h"

Core/HW/StereoResampler.cpp

@@ -47,8 +47,12 @@
 #include <emmintrin.h>
 #endif
 #if PPSSPP_ARCH(ARM_NEON)
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
 #endif
+#endif
 
 StereoResampler::StereoResampler()
 		: m_bufsize(MAX_SAMPLES_DEFAULT)

Core/MIPS/IR/IRInterpreter.cpp

@@ -10,8 +10,12 @@
 #endif
 
 #if PPSSPP_ARCH(ARM_NEON)
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
 #endif
+#endif
 
 #include "Core/Core.h"
 #include "Core/CoreTiming.h"

Core/Util/AudioFormatNEON.cpp

@@ -18,7 +18,11 @@
 #include "ppsspp_config.h"
 #if PPSSPP_ARCH(ARM_NEON)
 
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
+#endif
 #include "Common/Common.h"
 #include "Core/Util/AudioFormat.h"
 #include "Core/Util/AudioFormatNEON.h"

GPU/Common/TextureCacheCommon.cpp

@@ -35,8 +35,12 @@
 #include <emmintrin.h>
 #endif
 #if PPSSPP_ARCH(ARM_NEON)
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
 #endif
+#endif
 
 // Videos should be updated every few frames, so we forget quickly.
 #define VIDEO_DECIMATE_AGE 4

GPU/Common/TextureDecoderNEON.cpp

@@ -21,7 +21,11 @@
 
 #include "ext/xxhash.h"
 
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
+#endif
 
 #include "GPU/GPUState.h"
 #include "GPU/Common/TextureDecoder.h"

GPU/GPUState.cpp

@@ -28,8 +28,12 @@
 #include <emmintrin.h>
 #endif
 #if PPSSPP_ARCH(ARM_NEON)
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
 #endif
+#endif
 
 // This must be aligned so that the matrices within are aligned.
 alignas(16) GPUgstate gstate;

ext/native/ext/libpng17/arm/filter_neon_intrinsics.c

@@ -17,7 +17,11 @@
 /* This code requires -mfpu=neon on the command line: */
 #if PNG_ARM_NEON_IMPLEMENTATION == 1 /* intrinsics code from pngpriv.h */
 
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
+#endif
 
 /* libpng row pointers are not necessarily aligned to any particular boundary,
  * however this code will only work with appropriate alignment.  arm/arm_init.c

ext/native/math/dataconv.h

@@ -10,8 +10,12 @@
 #include <emmintrin.h>
 #endif
 #if PPSSPP_PLATFORM(ARM_NEON)
+#if defined(_MSC_VER) && defined(_M_ARM64)
+#include <arm64_neon.h>
+#else
 #include <arm_neon.h>
 #endif
+#endif
 
 extern const float one_over_255_x4[4];
 
@@ -160,4 +164,4 @@ inline void ExpandFloat24x3ToFloat4(float dest[4], const uint32_t src[3]) {
 
 inline uint32_t BytesToUint32(uint8_t a, uint8_t b, uint8_t c, uint8_t d) {
 	return (a) | (b << 8) | (c << 16) | (d << 24);
-}
+}