Sort all includes in src/gromacs

[alexxy/gromacs.git] / src / gromacs / mdlib / nbnxn_kernels / simd_2xnn / nbnxn_kernel_simd_2xnn_outer.h

diff --git a/src/gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_outer.h b/src/gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_outer.h
index 38b0293872682a5b49ce30bc017b3eb3b37fad0c..21a63ab4aa7b6fcdd899101711d8c75690025dd6 100644 (file)
--- a/src/gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_outer.h
+++ b/src/gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_outer.h
@@ -97,7 +97,7 @@
     gmx_simd_real_t  fix_S0, fiy_S0, fiz_S0;
     gmx_simd_real_t  fix_S2, fiy_S2, fiz_S2;
     /* We use an i-force SIMD register width of 4 */
-    /* The pr4 stuff is defined in nbnxn_kernel_simd_utils.h */
+    /* The simd4 stuff might be defined in nbnxn_kernel_simd_utils.h */
     gmx_simd4_real_t fix_S, fiy_S, fiz_S;
 
     gmx_simd_real_t  diagonal_jmi_S;
@@ -201,15 +201,6 @@
     gmx_simd_real_t  rcvdw2_S;
 #endif
 
-#ifdef CALC_ENERGIES
-    /* cppcheck-suppress unassignedVariable */
-    real       tmpsum_array[2*GMX_SIMD_REAL_WIDTH], *tmpsum;
-#endif
-#ifdef CALC_SHIFTFORCES
-    /* cppcheck-suppress unassignedVariable */
-    real       shf_array[2*GMX_SIMD_REAL_WIDTH], *shf;
-#endif
-
     int ninner;
 
 #ifdef COUNT_PAIRS
@@ -366,13 +357,6 @@
     shiftvec            = shift_vec[0];
     x                   = nbat->x;
 
-#ifdef CALC_ENERGIES
-    tmpsum   = gmx_simd_align_r(tmpsum_array);
-#endif
-#ifdef CALC_SHIFTFORCES
-    shf      = gmx_simd_align_r(shf_array);
-#endif
-
 #ifdef FIX_LJ_C
     pvdw_c6  = gmx_simd_align_r(pvdw_array);
     pvdw_c12 = pvdw_c6 + UNROLLI*UNROLLJ;
@@ -619,13 +603,13 @@
 #define CHECK_EXCLS
             while (cjind < cjind1 && nbl->cj[cjind].excl != NBNXN_INTERACTION_MASK_ALL)
             {
-#include "nbnxn_kernel_simd_2xnn_inner.h"
+#include "gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_inner.h"
                 cjind++;
             }
 #undef CHECK_EXCLS
             for (; (cjind < cjind1); cjind++)
             {
-#include "nbnxn_kernel_simd_2xnn_inner.h"
+#include "gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_inner.h"
             }
 #undef HALF_LJ
 #undef CALC_COULOMB
@@ -637,13 +621,13 @@
 #define CHECK_EXCLS
             while (cjind < cjind1 && nbl->cj[cjind].excl != NBNXN_INTERACTION_MASK_ALL)
             {
-#include "nbnxn_kernel_simd_2xnn_inner.h"
+#include "gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_inner.h"
                 cjind++;
             }
 #undef CHECK_EXCLS
             for (; (cjind < cjind1); cjind++)
             {
-#include "nbnxn_kernel_simd_2xnn_inner.h"
+#include "gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_inner.h"
             }
 #undef CALC_COULOMB
         }
@@ -653,13 +637,13 @@
 #define CHECK_EXCLS
             while (cjind < cjind1 && nbl->cj[cjind].excl != NBNXN_INTERACTION_MASK_ALL)
             {
-#include "nbnxn_kernel_simd_2xnn_inner.h"
+#include "gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_inner.h"
                 cjind++;
             }
 #undef CHECK_EXCLS
             for (; (cjind < cjind1); cjind++)
             {
-#include "nbnxn_kernel_simd_2xnn_inner.h"
+#include "gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn_inner.h"
             }
         }
 #undef CALC_LJ
@@ -667,27 +651,26 @@
 
         /* Add accumulated i-forces to the force array */
         fix_S = gmx_mm_transpose_sum4h_pr(fix_S0, fix_S2);
-        gmx_simd4_store_r(f+scix, gmx_add_pr4(fix_S, gmx_simd4_load_r(f+scix)));
+        gmx_simd4_store_r(f+scix, gmx_simd4_add_r(fix_S, gmx_simd4_load_r(f+scix)));
 
         fiy_S = gmx_mm_transpose_sum4h_pr(fiy_S0, fiy_S2);
-        gmx_simd4_store_r(f+sciy, gmx_add_pr4(fiy_S, gmx_simd4_load_r(f+sciy)));
+        gmx_simd4_store_r(f+sciy, gmx_simd4_add_r(fiy_S, gmx_simd4_load_r(f+sciy)));
 
         fiz_S = gmx_mm_transpose_sum4h_pr(fiz_S0, fiz_S2);
-        gmx_simd4_store_r(f+sciz, gmx_add_pr4(fiz_S, gmx_simd4_load_r(f+sciz)));
+        gmx_simd4_store_r(f+sciz, gmx_simd4_add_r(fiz_S, gmx_simd4_load_r(f+sciz)));
 
 #ifdef CALC_SHIFTFORCES
-        fshift[ish3+0] += gmx_sum_simd4(fix_S, shf);
-        fshift[ish3+1] += gmx_sum_simd4(fiy_S, shf);
-        fshift[ish3+2] += gmx_sum_simd4(fiz_S, shf);
+        fshift[ish3+0] += gmx_simd4_reduce_r(fix_S);
+        fshift[ish3+1] += gmx_simd4_reduce_r(fiy_S);
+        fshift[ish3+2] += gmx_simd4_reduce_r(fiz_S);
 #endif
 
 #ifdef CALC_ENERGIES
         if (do_coul)
         {
-            *Vc += gmx_sum_simd(vctot_S, tmpsum);
+            *Vc += gmx_simd_reduce_r(vctot_S);
         }
-
-        *Vvdw += gmx_sum_simd(Vvdwtot_S, tmpsum);
+        *Vvdw += gmx_simd_reduce_r(Vvdwtot_S);
 #endif
 
         /* Outer loop uses 6 flops/iteration */