From 1aa2cf6edf9f2bd57a21998fceeb095cbe1f2b84 Mon Sep 17 00:00:00 2001
From: Gerdya <37900163+Gerdya@users.noreply.github.com>
Date: Sat, 26 Nov 2022 01:31:12 +0100
Subject: [PATCH] Some fixes for compilation with MSVC for Windows ARM64
 (#3690)

---
 src/include/OpenImageIO/platform.h | 10 +++++++++-
 src/include/OpenImageIO/simd.h     | 19 +++++++++++++++----
 2 files changed, 24 insertions(+), 5 deletions(-)

diff --git a/src/include/OpenImageIO/platform.h b/src/include/OpenImageIO/platform.h
index 233da35a8a..be7e3d1f9d 100644
--- a/src/include/OpenImageIO/platform.h
+++ b/src/include/OpenImageIO/platform.h
@@ -42,6 +42,11 @@
 #    include <intrin.h>
 #endif
 
+//Avoid min and max being defined
+#if defined(_WIN32)
+#define NOMINMAX
+#endif
+
 #include <OpenImageIO/oiioversion.h>
 #include <OpenImageIO/export.h>
 
@@ -568,7 +573,7 @@ inline void cpuid (int info[4], int infoType, int extra)
 {
     // Implementation cribbed from Halide (http://halide-lang.org), which
     // cribbed it from ISPC (https://github.com/ispc/ispc).
-#if (defined(_WIN32) || defined(__i386__) || defined(__x86_64__))
+#if (defined(_M_X64) || defined(_M_IX86) || defined(_M_AMD64) || defined(__i386__) || defined(__x86_64__))
 # ifdef _MSC_VER
     __cpuidex(info, infoType, extra);
 # elif defined(__x86_64__)
@@ -584,6 +589,9 @@ inline void cpuid (int info[4], int infoType, int extra)
         : "=a" (info[0]), "=r" (info[1]), "=c" (info[2]), "=d" (info[3])
         : "0" (infoType), "2" (extra));
 # endif
+#elif (defined(_M_ARM64) || defined (__aarch64__) || defined(__aarch64))
+    info[0] = 0; info[1] = 0; info[2] = 0; info[3] = 0;
+    #define __ARM_NEON__
 #else
     info[0] = 0; info[1] = 0; info[2] = 0; info[3] = 0;
 #endif
diff --git a/src/include/OpenImageIO/simd.h b/src/include/OpenImageIO/simd.h
index f82f8786b5..2a75f38f10 100644
--- a/src/include/OpenImageIO/simd.h
+++ b/src/include/OpenImageIO/simd.h
@@ -92,6 +92,11 @@
 #define OIIO_NO_SSE 1
 #endif
 
+//Disable SSE for ARM64 targets
+#if defined(_M_ARM64) || defined(__aarch64) || defined(__aarch64__)
+#define OIIO_NO_SSE 1
+#endif
+
 // Make sure to disable SSE intrinsics when compiling for Cuda.
 #if defined(__CUDA_ARCH__) && !defined(OIIO_NO_SSE)
 #define OIIO_NO_SSE 1
@@ -6750,11 +6755,17 @@ OIIO_FORCEINLINE void vfloat4::load (const float *values, int n) {
         break;
     }
 #elif OIIO_SIMD_NEON
+    //switch (n) {
+    //case 1: m_simd = vdupq_n_f32(0); m_simd[0] = values[0]; break;
+    //case 2: load (values[0], values[1], 0.0f, 0.0f);      break;
+    //case 3: load (values[0], values[1], values[2], 0.0f); break;
+    //case 4: m_simd = vld1q_f32 (values);                   break;
+    //default: break;
+    m_simd = vld1q_f32(values);
     switch (n) {
-    case 1: m_simd = vdupq_n_f32(0); m_simd[0] = values[0]; break;
-    case 2: load (values[0], values[1], 0.0f, 0.0f);      break;
-    case 3: load (values[0], values[1], values[2], 0.0f); break;
-    case 4: m_simd = vld1q_f32 (values);                   break;
+    case 1: m_simd = vsetq_lane_f32(0.0f, m_simd, 1);
+    case 2: m_simd = vsetq_lane_f32(0.0f, m_simd, 2);
+    case 3: m_simd = vsetq_lane_f32(0.0f, m_simd, 3);
     default: break;
     }
 #else