diff --git a/src/util/arch/arm/simd_utils.h b/src/util/arch/arm/simd_utils.h
index 232ca76f..c918eced 100644
--- a/src/util/arch/arm/simd_utils.h
+++ b/src/util/arch/arm/simd_utils.h
@@ -151,13 +151,13 @@ static really_inline u32 movd(const m128 in) {
 }
 
 static really_inline u64a movq(const m128 in) {
-    return vgetq_lane_u64((uint64x2_t) in, 1);
+    return vgetq_lane_u64((uint64x2_t) in, 0);
 }
 
 /* another form of movq */
 static really_inline
 m128 load_m128_from_u64a(const u64a *p) {
-    return (m128) vdupq_n_u64(*p);
+    return (m128) vsetq_lane_u64(*p, zeroes128(), 0);
 }
 
 static really_inline u32 extract32from128(const m128 in, unsigned imm) {
diff --git a/unit/internal/simd_utils.cpp b/unit/internal/simd_utils.cpp
index 5c0e0b40..bc1426b1 100644
--- a/unit/internal/simd_utils.cpp
+++ b/unit/internal/simd_utils.cpp
@@ -667,7 +667,7 @@ TEST(SimdUtilsTest, movq) {
 #if defined(ARCH_IA32) || defined(ARCH_X86_64)
     simd = _mm_set_epi64x(~0LL, 0x123456789abcdef);
 #elif defined(ARCH_ARM32) || defined(ARCH_AARCH64)
-    int64x2_t a = { ~0LL, 0x123456789abcdefLL };
+    int64x2_t a = { 0x123456789abcdefLL, ~0LL };
     simd = vreinterpretq_s64_s8(a);
 #endif
 #endif