diff --git a/config.d/compiling.m4 b/config.d/compiling.m4
index 463dd8795..81cca85bf 100644
--- a/config.d/compiling.m4
+++ b/config.d/compiling.m4
@@ -98,37 +98,41 @@ AC_ARG_ENABLE(optimize,
 	optimize=yes
 )
 
-AC_ARG_ENABLE(simd,
-	AS_HELP_STRING([--enable-simd@<:@=arg@:.@],
-		[enable SIMD support (default auto)]),
-	[],
-	[enable_simd=yes]
-)
+if test "x$host_cpu" = xaarch64; then
+	simd=neon
+else
+	AC_ARG_ENABLE(simd,
+		AS_HELP_STRING([--enable-simd@<:@=arg@:.@],
+			[enable SIMD support (default auto)]),
+		[],
+		[enable_simd=yes]
+	)
 
-case "$enable_simd" in
-	no)
-		simd=no
-		;;
-	sse|sse2|avx|avx2)
-		QF_CC_OPTION(-m$enable_simd)
-		simd=$enable_simd
-		;;
-	yes)
-		for simd in avx2 avx sse2 sse; do
-			if lscpu | grep -q -w $simd; then
-				QF_CC_OPTION(-m$simd)
-				break
-			fi
-		done
-		;;
-esac
-case "$simd" in
-	avx*)
-		;;
-	*)
-		QF_CC_OPTION(-Wno-psabi)
-		;;
-esac
+	case "$enable_simd" in
+		no)
+			simd=no
+			;;
+		sse|sse2|avx|avx2)
+			QF_CC_OPTION(-m$enable_simd)
+			simd=$enable_simd
+			;;
+		yes)
+			for simd in avx2 avx sse2 sse; do
+				if lscpu | grep -q -w $simd; then
+					QF_CC_OPTION(-m$simd)
+					break
+				fi
+			done
+			;;
+	esac
+	case "$simd" in
+		avx*)
+			;;
+		*)
+			QF_CC_OPTION(-Wno-psabi)
+			;;
+	esac
+fi
 
 AC_MSG_CHECKING(for optimization)
 if test "x$optimize" = xyes -a "x$leave_cflags_alone" != "xyes"; then
diff --git a/include/QF/simd/mat4f.h b/include/QF/simd/mat4f.h
index 981d71d22..ce061edd0 100644
--- a/include/QF/simd/mat4f.h
+++ b/include/QF/simd/mat4f.h
@@ -28,8 +28,6 @@
 #ifndef __QF_simd_mat4f_h
 #define __QF_simd_mat4f_h
 
-#include <immintrin.h>
-
 #include "QF/simd/types.h"
 
 GNU89INLINE inline void maddf (mat4f_t c, const mat4f_t a, const mat4f_t b);
@@ -172,27 +170,27 @@ mat4fquat (mat4f_t m, vec4f_t q)
 #undef m
 	{
 		vec4f_t a = xq;
-		vec4f_t b = _mm_xor_ps (shuff103 (yq), (__m128) mpm);
-		vec4f_t c = _mm_xor_ps (shuff230 (zq), (__m128) pmm);
-		vec4f_t d = _mm_xor_ps (shuff321 (wq), (__m128) mmp);
+		vec4f_t b = (vec4f_t) ((vec4i_t) shuff103 (yq) ^ mpm);
+		vec4f_t c = (vec4f_t) ((vec4i_t) shuff230 (zq) ^ pmm);
+		vec4f_t d = (vec4f_t) ((vec4i_t) shuff321 (wq) ^ mmp);
 		// column: ww + xx - yy - zz // 2xy + 2wz // 2zx - 2wy // 0
-		m[0] = _mm_and_ps (a + b - c - d, (__m128) mask);
+		m[0] = (vec4f_t) ((vec4i_t) (a + b - c - d) & mask);
 	}
 	{
-		vec4f_t a = _mm_xor_ps (shuff103 (xq), (__m128) mpm);
+		vec4f_t a = (vec4f_t) ((vec4i_t) shuff103 (xq) ^ mpm);
 		vec4f_t b = yq;
-		vec4f_t c = _mm_xor_ps (shuff321 (zq), (__m128) mmp);
-		vec4f_t d = _mm_xor_ps (shuff230 (wq), (__m128) pmm);
+		vec4f_t c = (vec4f_t) ((vec4i_t) shuff321 (zq) ^ mmp);
+		vec4f_t d = (vec4f_t) ((vec4i_t) shuff230 (wq) ^ pmm);
 		// column: 2xy - 2wz // ww - xx + yy - zz // 2yz + 2wx // 0
-		m[1] = _mm_and_ps (b + c - a - d, (__m128) mask);
+		m[1] = (vec4f_t) ((vec4i_t) (b + c - a - d) & mask);
 	}
 	{
-		vec4f_t a = _mm_xor_ps (shuff230 (xq), (__m128) pmm);
-		vec4f_t b = _mm_xor_ps (shuff321 (yq), (__m128) mmp);
+		vec4f_t a = (vec4f_t) ((vec4i_t) shuff230 (xq) ^ pmm);
+		vec4f_t b = (vec4f_t) ((vec4i_t) shuff321 (yq) ^ mmp);
 		vec4f_t c = zq;
-		vec4f_t d = _mm_xor_ps (shuff103 (wq), (__m128) mpm);
+		vec4f_t d = (vec4f_t) ((vec4i_t) shuff103 (wq) ^ mpm);
 		// column: 2xz + 2wy // 2yz - 2wx // ww - xx - yy + zz // 0
-		m[2] = _mm_and_ps (a - b + c - d, (__m128) mask);
+		m[2] = (vec4f_t) ((vec4i_t) (a - b + c - d) & mask);
 	}
 	m[3] = (vec4f_t) { 0, 0, 0, 1 };
 }
diff --git a/include/QF/simd/types.h b/include/QF/simd/types.h
index 26aceccd0..c7c950fd0 100644
--- a/include/QF/simd/types.h
+++ b/include/QF/simd/types.h
@@ -105,30 +105,4 @@ typedef struct vspheref_s {
 	float       radius;
 } vspheref_t;
 
-#include <immintrin.h>
-#ifndef __SSE__
-#define _mm_xor_ps __qf_mm_xor_ps
-#define _mm_and_ps __qf_mm_and_ps
-GNU89INLINE inline __m128 _mm_xor_ps (__m128 a, __m128 b);
-GNU89INLINE inline __m128 _mm_and_ps (__m128 a, __m128 b);
-#ifndef IMPLEMENT_MAT4F_Funcs
-GNU89INLINE inline
-#else
-VISIBLE
-#endif
-__m128 _mm_xor_ps (__m128 a, __m128 b)
-{
-	return (__m128) ((vec4i_t) a ^ (vec4i_t) b);
-}
-#ifndef IMPLEMENT_MAT4F_Funcs
-GNU89INLINE inline
-#else
-VISIBLE
-#endif
-__m128 _mm_and_ps (__m128 a, __m128 b)
-{
-	return (__m128) ((vec4i_t) a & (vec4i_t) b);
-}
-#endif
-
 #endif//__QF_simd_types_h
diff --git a/include/QF/simd/vec2d.h b/include/QF/simd/vec2d.h
index 86be9302b..25eb263bf 100644
--- a/include/QF/simd/vec2d.h
+++ b/include/QF/simd/vec2d.h
@@ -29,7 +29,11 @@
 #ifndef __QF_simd_vec2d_h
 #define __QF_simd_vec2d_h
 
+#ifdef __aarch64__
+#include <arm_neon.h>
+#else
 #include <immintrin.h>
+#endif
 
 #include "QF/simd/types.h"
 
@@ -50,7 +54,11 @@ VISIBLE
 vec2d_t
 vsqrt2d (vec2d_t v)
 {
+#ifdef __aarch64__
+	return vsqrtq_f64 (v);
+#else
 	return _mm_sqrt_pd (v);
+#endif
 }
 
 #ifndef IMPLEMENT_VEC2D_Funcs
diff --git a/include/QF/simd/vec2f.h b/include/QF/simd/vec2f.h
index 1459dd37b..3dcc97ecc 100644
--- a/include/QF/simd/vec2f.h
+++ b/include/QF/simd/vec2f.h
@@ -29,7 +29,11 @@
 #ifndef __QF_simd_vec2f_h
 #define __QF_simd_vec2f_h
 
+#ifdef __aarch64__
+#include <arm_neon.h>
+#else
 #include <immintrin.h>
+#endif
 #include <math.h>
 
 #include "QF/simd/types.h"
@@ -67,9 +71,13 @@ VISIBLE
 vec2f_t
 vsqrt2f (vec2f_t v)
 {
+#ifdef __aarch64__
+	return vsqrt_f32 (v);
+#else
 	vec4f_t     t = { v[0], v[1], 0, 0 };
 	t = _mm_sqrt_ps (t);
 	return (vec2f_t) { t[0], t[1] };
+#endif
 }
 
 #ifndef IMPLEMENT_VEC2F_Funcs
diff --git a/include/QF/simd/vec2i.h b/include/QF/simd/vec2i.h
index 29f2cc57c..0032e3e33 100644
--- a/include/QF/simd/vec2i.h
+++ b/include/QF/simd/vec2i.h
@@ -28,8 +28,6 @@
 #ifndef __QF_simd_vec2i_h
 #define __QF_simd_vec2i_h
 
-#include <immintrin.h>
-#include <math.h>
 
 #include "QF/simd/types.h"
 
@@ -59,7 +57,7 @@ VISIBLE
 int
 any2i (vec2i_t v)
 {
-	vec2i_t     t = _m_pcmpeqd (v, (vec2i_t) {0, 0});
+	vec2i_t     t = v == (vec2i_t) {0, 0};
 #ifndef __SSSE3__
 	return (t[0] + t[1]) > -2;
 #else
@@ -75,7 +73,7 @@ VISIBLE
 int
 all2i (vec2i_t v)
 {
-	vec2i_t     t = _m_pcmpeqd (v, (vec2i_t) {0, 0});
+	vec2i_t     t = v == (vec2i_t) {0, 0};
 #ifndef __SSSE3__
 	return (t[0] + t[1]) == 0;
 #else
@@ -91,7 +89,7 @@ VISIBLE
 int
 none2i (vec2i_t v)
 {
-	vec2i_t     t = _m_pcmpeqd (v, (vec2i_t) {0, 0});
+	vec2i_t     t = v == (vec2i_t) {0, 0};
 #ifndef __SSSE3__
 	return (t[0] + t[1]) == -2;
 #else
diff --git a/include/QF/simd/vec4d.h b/include/QF/simd/vec4d.h
index 04e4b81a2..a56c70ab7 100644
--- a/include/QF/simd/vec4d.h
+++ b/include/QF/simd/vec4d.h
@@ -28,7 +28,11 @@
 #ifndef __QF_simd_vec4d_h
 #define __QF_simd_vec4d_h
 
+#ifdef __aarch64__
+#include <arm_neon.h>
+#else
 #include <immintrin.h>
+#endif
 
 #include "QF/simd/types.h"
 #include "QF/simd/vec2d.h"
diff --git a/include/QF/simd/vec4f.h b/include/QF/simd/vec4f.h
index ee92d928e..effb32864 100644
--- a/include/QF/simd/vec4f.h
+++ b/include/QF/simd/vec4f.h
@@ -28,7 +28,11 @@
 #ifndef __QF_simd_vec4f_h
 #define __QF_simd_vec4f_h
 
+#ifdef __aarch64__
+#include <arm_neon.h>
+#else
 #include <immintrin.h>
+#endif
 #include <math.h>
 
 #include "QF/simd/types.h"
@@ -121,12 +125,16 @@ VISIBLE
 vec4f_t
 vsqrt4f (vec4f_t v)
 {
+#ifdef __aarch64__
+	return vsqrtq_f32 (v);
+#else
 #ifndef __SSE__
 	vec4f_t     r = { sqrtf (v[0]), sqrtf (v[1]), sqrtf (v[2]), sqrtf (v[3]) };
 	return r;
 #else
 	return _mm_sqrt_ps (v);
 #endif
+#endif
 }
 
 #ifndef IMPLEMENT_VEC4F_Funcs
diff --git a/include/QF/simd/vec4i.h b/include/QF/simd/vec4i.h
index d3e4ac5dc..0e947a06a 100644
--- a/include/QF/simd/vec4i.h
+++ b/include/QF/simd/vec4i.h
@@ -28,7 +28,6 @@
 #ifndef __QF_simd_vec4i_h
 #define __QF_simd_vec4i_h
 
-#include <immintrin.h>
 #include <math.h>
 
 #include "QF/simd/types.h"
diff --git a/tools/qfvis/source/qfvis.c b/tools/qfvis/source/qfvis.c
index 2e64e9a0d..7630de93b 100644
--- a/tools/qfvis/source/qfvis.c
+++ b/tools/qfvis/source/qfvis.c
@@ -242,14 +242,13 @@ split_edge (const vec4f_t *points, const vec4f_t *dists,
 	// "nan" because 0x7fffffff is nan when viewed as a float
 	static const vec4i_t onenan = {0x3f800000,0x3f800000,0x3f800000,~0u >> 1};
 	static const vec4i_t nan = { ~0u >> 1, ~0u >> 1, ~0u >> 1, ~0u >> 1};
-	vec4i_t     x = _mm_and_ps (split, (__m128) nan) == (__m128) onenan;
+	vec4i_t     x = ((vec4i_t) split & nan) == onenan;
 	// plane vector has -dist in w
-	vec4f_t     y = _mm_and_ps (split, (__m128) x) * -split[3];
+	vec4f_t     y = (vec4f_t) ((vec4i_t) split & x) * -split[3];
 #ifdef __SSE3__
 	mid = _mm_blendv_ps (mid, y, (__m128) x);
 #else
-	mid = (vec4f_t) ((vec4i_t) _mm_and_ps (y, (__m128) x) |
-					 (vec4i_t) _mm_and_ps (mid, (__m128) ~x));
+	mid = (vec4f_t) (((vec4i_t) y & x) | ((vec4i_t) mid & ~x));
 #endif
 //	if (isnan (mid[0])) *(int *) 0 = 0;
 	return mid;