Update sse2neon.h to latest for better support for Linux ARM64

* Use sse2neon.h from https://github.com/DLTcollab/sse2neon/blob/de2817727c72fc2f4ce9f54e2db6e40ce0548414/sse2neon.h
This is current master from 10.10.2023

* _mm_shuffle_pd is provided by newer sse2neon.h

https://github.com/DLTcollab/sse2neon/blob/de2817727c72fc2f4ce9f54e2db6e40ce0548414/sse2neon.h#L5118

* aarch64 does not support cpuid

Signed-off-by: Martin Tzvetanov Grigorov <mgrigorov@apache.org>

.github/workflows/ci.yaml

@@ -0,0 +1,59 @@
+name: CI
+
+on:
+  push:
+    branches:
+    - master
+  pull_request:
+    branches:
+    - master
+
+jobs:
+  build:
+    name: Build on Linux x86_64
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v4
+
+    - name: Install dependencies
+      run: |
+        sudo apt-get update
+        sudo apt-get install -y cmake file
+
+    - name: Build
+      run: |
+        mkdir build
+        cd build
+        cmake ..
+        make -j
+        file libhmsbeagle/libhmsbeagle.so.* | grep x86-64
+
+  build-aarch64:
+    name: Build on Linux aarch64
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v4
+
+    - name: Build
+      uses: uraimo/run-on-arch-action@v2
+      with:
+        arch: aarch64
+        distro: ubuntu20.04
+        githubToken: ${{ github.token }}
+        dockerRunArgs: |
+          --volume "${PWD}:/beagle-lib"
+        install: |
+          apt-get update -q -y
+          apt-get install -q -y cmake gcc g++ openjdk-11-jdk file
+        run: |
+          export JAVA_HOME=/usr/lib/jvm/java-11-openjdk-arm64
+          cd /beagle-lib
+          mkdir build
+          cd build
+          cmake ..
+          make -j
+          file libhmsbeagle/libhmsbeagle.so.* | grep aarch64

libhmsbeagle/CPU/BeagleCPUSSEPlugin.cpp

@@ -116,6 +116,8 @@ bool check_sse2()
   if (edx & bit_SSE2)
 	return true;
   return false;
+#elif defined(__aarch64__)
+  return false;
 #else // HAVE_CPUID.H
 	// Determine if cpuid supported:
     unsigned int res;

libhmsbeagle/CPU/SSEDefinitions.h

@@ -23,40 +23,12 @@
 #define DLS_USE_SSE2
 
 #if defined(DLS_USE_SSE2)
-#	if defined(__ARM64_ARCH_8__)
+#	if defined(__aarch64__)
 #		include "libhmsbeagle/CPU/sse2neon.h"
 #       define _MM_SHUFFLE2(fp1,fp0) (((fp1) << 1) | (fp0))
 #		define VEC_SHUFFLE0(a,b)	_mm_shuffle_pd(a, b, _MM_SHUFFLE2(0,0)) // vreinterpretq_f64_m128d(a)
 #		define VEC_SHUFFLE1(a,b)	_mm_shuffle_pd(a, b, _MM_SHUFFLE2(1,1)) // vreinterpretq_f64_m128d(a)
-#       if __has_builtin(__builtin_shufflevector)
-#       	define _mm_shuffle_pd(a,b,imm)                                \
-				__extension__({                                           \
-		        float64x2_t _input1 = vreinterpretq_f64_m128(a);          \
-	    	    float64x2_t _input2 = vreinterpretq_f64_m128(b);          \
-		        float64x2_t _shuf = __builtin_shufflevector(              \
-	    	        _input1, _input2, (imm) & (0x1), ((imm) >> 1) & 0x1); \
-	        	vreinterpretq_m128_f32(_shuf);                            \
-    	    })
-#		else
-#			error "Need to implement NEON translation of _mm_shuffle_pd"
-#		endif
-
-		static inline __m128 _mm_div_pd(__m128 a, __m128 b) {
-		    return vreinterpretq_m128_f64(
-        		vdivq_f64(vreinterpretq_f64_m128(a), vreinterpretq_f64_m128(b)));
-		}
-
-		static inline void _mm_store_sd(double* a, __m128 b) {
-			const auto _b = vreinterpretq_f64_m128(b);
-			a[0] = _b[0];
-		}
 
-        static inline __m128 _mm_dp_pd(__m128 lhs, __m128 rhs, const int) {
-            auto const product = vmulq_f64(vreinterpretq_f64_m128d(lhs),
-                                           vreinterpretq_f64_m128d(rhs));
-            auto const sum = product[0] + product[1]; // TODO Almost surely an more efficient way
-            return vreinterpretq_m128d_f64(vdupq_n_f64(sum));
-        }
 #   else
 #		if !defined(DLS_MACOS)
 #			include <emmintrin.h>