Revision control

Copy as Markdown

diff --git a/src/3rdparty/meshoptimizer/src/clusterizer.cpp b/src/3rdparty/meshoptimizer/src/clusterizer.cpp
index 73cc0ab53..3ad678bdf 100644
--- a/src/3rdparty/meshoptimizer/src/clusterizer.cpp
+++ b/src/3rdparty/meshoptimizer/src/clusterizer.cpp
@@ -8,10 +8,10 @@
// The block below auto-detects SIMD ISA that can be used on the target platform
#ifndef MESHOPTIMIZER_NO_SIMD
-#if defined(__SSE2__) || (defined(_MSC_VER) && defined(_M_X64))
+#if defined(__SSE2__) || (defined(_MSC_VER) && defined(_M_X64) && !defined(_M_ARM64EC))
#define SIMD_SSE
#include <emmintrin.h>
-#elif defined(__aarch64__) || (defined(_MSC_VER) && defined(_M_ARM64) && _MSC_VER >= 1922)
+#elif defined(__aarch64__) || (defined(_MSC_VER) && (defined(_M_ARM64) || defined(_M_ARM64EC)) && _MSC_VER >= 1922)
#define SIMD_NEON
#include <arm_neon.h>
#endif
diff --git a/src/3rdparty/meshoptimizer/src/vertexcodec.cpp b/src/3rdparty/meshoptimizer/src/vertexcodec.cpp
index 7085cce32..d3a84b000 100644
--- a/src/3rdparty/meshoptimizer/src/vertexcodec.cpp
+++ b/src/3rdparty/meshoptimizer/src/vertexcodec.cpp
@@ -19,7 +19,7 @@
#endif
// MSVC supports compiling SSSE3 code regardless of compile options; we use a cpuid-based scalar fallback
-#if !defined(SIMD_SSE) && !defined(SIMD_AVX) && defined(_MSC_VER) && !defined(__clang__) && (defined(_M_IX86) || defined(_M_X64))
+#if !defined(SIMD_SSE) && !defined(SIMD_AVX) && defined(_MSC_VER) && !defined(__clang__) && (defined(_M_IX86) || (defined(_M_X64) && !defined(_M_ARM64EC)))
#define SIMD_SSE
#define SIMD_FALLBACK
#endif
@@ -37,7 +37,7 @@
#endif
// On MSVC, we assume that ARM builds always target NEON-capable devices
-#if !defined(SIMD_NEON) && defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64))
+#if !defined(SIMD_NEON) && defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64) || defined(_M_ARM64EC))
#define SIMD_NEON
#endif
diff --git a/src/3rdparty/meshoptimizer/src/vertexfilter.cpp b/src/3rdparty/meshoptimizer/src/vertexfilter.cpp
index 3fd836083..d282538c9 100644
--- a/src/3rdparty/meshoptimizer/src/vertexfilter.cpp
+++ b/src/3rdparty/meshoptimizer/src/vertexfilter.cpp
@@ -13,7 +13,7 @@
#endif
// MSVC supports compiling SSE2 code regardless of compile options; we assume all 32-bit CPUs support SSE2
-#if !defined(SIMD_SSE) && defined(_MSC_VER) && !defined(__clang__) && (defined(_M_IX86) || defined(_M_X64))
+#if !defined(SIMD_SSE) && defined(_MSC_VER) && !defined(__clang__) && (defined(_M_IX86) || (defined(_M_X64) && !defined(_M_ARM64EC)))
#define SIMD_SSE
#endif
@@ -23,7 +23,7 @@
#endif
// On MSVC, we assume that ARM builds always target NEON-capable devices
-#if !defined(SIMD_NEON) && defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64))
+#if !defined(SIMD_NEON) && defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64) || defined(_M_ARM64EC))
#define SIMD_NEON
#endif
@@ -536,7 +536,7 @@ static void decodeFilterColorSimd16(unsigned short* data, size_t count)
}
#endif
-#if defined(SIMD_NEON) && !defined(__aarch64__) && !defined(_M_ARM64)
+#if defined(SIMD_NEON) && !defined(__aarch64__) && !(defined(_M_ARM64) || defined(_M_ARM64EC))
inline float32x4_t vsqrtq_f32(float32x4_t x)
{
float32x4_t r = vrsqrteq_f32(x);
@@ -640,7 +640,7 @@ static void decodeFilterOctSimd16(short* data, size_t count)
// compute normal length & scale
float32x4_t ll = vfmaq_f32(vfmaq_f32(vmulq_f32(x, x), y, y), z, z);
-#if !defined(__aarch64__) && !defined(_M_ARM64)
+#if !defined(__aarch64__) && !(defined(_M_ARM64) || defined(_M_ARM64EC))
float32x4_t rl = vrsqrteq_f32(ll);
rl = vmulq_f32(rl, vrsqrtsq_f32(vmulq_f32(rl, ll), rl)); // refine rsqrt estimate
float32x4_t s = vmulq_f32(vdupq_n_f32(32767.f), rl);