bthread: fix simd memory align issue

Author: icexin

src/bthread/task_group.cpp

@@ -94,7 +94,7 @@ AtomicInteger128::Value AtomicInteger128::load() const {
 
 #if __x86_64__ || __ARM_NEON
 #ifdef __x86_64__
-    __m128i value = _mm_load_si128(reinterpret_cast<const __m128i*>(&_value));
+    __m128i value = _mm_loadu_si128(reinterpret_cast<const __m128i*>(&_value));
 #else // __ARM_NEON
     int64x2_t value = vld1q_s64(reinterpret_cast<const int64_t*>(&_value));
 #endif // __x86_64__
@@ -107,8 +107,8 @@ AtomicInteger128::Value AtomicInteger128::load() const {
 
 void AtomicInteger128::store(Value value) {
 #if __x86_64__
-    __m128i v = _mm_load_si128(reinterpret_cast<__m128i*>(&value));
-    _mm_store_si128(reinterpret_cast<__m128i*>(&_value), v);
+    __m128i v = _mm_loadu_si128(reinterpret_cast<__m128i*>(&value));
+    _mm_storeu_si128(reinterpret_cast<__m128i*>(&_value), v);
 #elif __ARM_NEON
     int64x2_t v = vld1q_s64(reinterpret_cast<int64_t*>(&value));
     vst1q_s64(reinterpret_cast<int64_t*>(&_value), v);