Merge branch 'release-4-6', adds the nbnxn functionality

[alexxy/gromacs.git] / src / gromacs / legacyheaders / gmx_math_x86_sse4_1_double.h

diff --git a/src/gromacs/legacyheaders/gmx_math_x86_sse4_1_double.h b/src/gromacs/legacyheaders/gmx_math_x86_sse4_1_double.h
index 37a9cac29b2d8272e38bd9dfc6e53bb24cbcbfc5..e742f952854e09ade2af73744da22709aacef1f0 100644 (file)
--- a/src/gromacs/legacyheaders/gmx_math_x86_sse4_1_double.h
+++ b/src/gromacs/legacyheaders/gmx_math_x86_sse4_1_double.h
@@ -24,7 +24,7 @@
 #include <stdio.h>
 #include <math.h>
 
-#include "gmx_x86_sse4.h"
+#include "gmx_x86_sse4_1.h"
 
 
 
@@ -65,11 +65,11 @@ gmx_mm_invsqrt_pair_pd(__m128d x1, __m128d x2, __m128d *invsqrt1, __m128d *invsq
     __m128d lu1,lu2;
     
     /* Do first N-R step in float for 2x throughput */
-    xf  = _mm_shuffle_ps(_mm_cvtpd_ps(x1),_mm_cvtpd_ps(x2),MM_SHUFFLE(1,0,1,0));
+    xf  = _mm_shuffle_ps(_mm_cvtpd_ps(x1),_mm_cvtpd_ps(x2),_MM_SHUFFLE(1,0,1,0));
     luf = _mm_rsqrt_ps(xf);
     luf = _mm_mul_ps(halff,_mm_mul_ps(_mm_sub_ps(threef,_mm_mul_ps(_mm_mul_ps(luf,luf),xf)),luf));
     
-    lu2 = _mm_cvtps_pd(_mm_shuffle_ps(luf,luf,MM_SHUFFLE(3,2,3,2)));
+    lu2 = _mm_cvtps_pd(_mm_shuffle_ps(luf,luf,_MM_SHUFFLE(3,2,3,2)));
     lu1 = _mm_cvtps_pd(luf);
     
     *invsqrt1 = _mm_mul_pd(half,_mm_mul_pd(_mm_sub_pd(three,_mm_mul_pd(_mm_mul_pd(lu1,lu1),x1)),lu1));
@@ -839,7 +839,7 @@ gmx_mm_erfc_pd(__m128d x)
  *    vectorial force to add to the particles.
  *
  */
-__m128d
+static __m128d
 gmx_mm_pmecorrF_pd(__m128d z2)
 {
     const __m128d  FN10     = _mm_set1_pd(-8.0072854618360083154e-14);
@@ -935,8 +935,8 @@ gmx_mm_pmecorrF_pd(__m128d z2)
  *    and you have your potential.
  *
  */
-__m128d
-gmx_mm_pmecorrV_pd(__m256d z2)
+static __m128d
+gmx_mm_pmecorrV_pd(__m128d z2)
 {
     const __m128d  VN9      = _mm_set1_pd(-9.3723776169321855475e-13);
     const __m128d  VN8      = _mm_set1_pd(1.2280156762674215741e-10);