--- boost/math/special_functions/detail/lanczos_sse2.hpp
+++ boost/math/special_functions/detail/lanczos_sse2.hpp
@@ -51,11 +51,11 @@
       static_cast<double>(23531376880.41075968857200767445163675473L),
       static_cast<double>(0u)
    };
-   register __m128d vx = _mm_load1_pd(&x);
-   register __m128d sum_even = _mm_load_pd(coeff);
-   register __m128d sum_odd = _mm_load_pd(coeff+2);
-   register __m128d nc_odd, nc_even;
-   register __m128d vx2 = _mm_mul_pd(vx, vx);
+   __m128d vx = _mm_load1_pd(&x);
+   __m128d sum_even = _mm_load_pd(coeff);
+   __m128d sum_odd = _mm_load_pd(coeff+2);
+   __m128d nc_odd, nc_even;
+   __m128d vx2 = _mm_mul_pd(vx, vx);
 
    sum_even = _mm_mul_pd(sum_even, vx2);
    nc_even = _mm_load_pd(coeff + 4);
@@ -136,11 +136,11 @@
          static_cast<double>(56906521.91347156388090791033559122686859L),
          static_cast<double>(0u)
    };
-   register __m128d vx = _mm_load1_pd(&x);
-   register __m128d sum_even = _mm_load_pd(coeff);
-   register __m128d sum_odd = _mm_load_pd(coeff+2);
-   register __m128d nc_odd, nc_even;
-   register __m128d vx2 = _mm_mul_pd(vx, vx);
+   __m128d vx = _mm_load1_pd(&x);
+   __m128d sum_even = _mm_load_pd(coeff);
+   __m128d sum_odd = _mm_load_pd(coeff+2);
+   __m128d nc_odd, nc_even;
+   __m128d vx2 = _mm_mul_pd(vx, vx);
 
    sum_even = _mm_mul_pd(sum_even, vx2);
    nc_even = _mm_load_pd(coeff + 4);