diff --git a/common/includes/simd/avx512_double.h b/common/includes/simd/avx512_double.h
index 62a0d67..453af04 100644
--- a/common/includes/simd/avx512_double.h
+++ b/common/includes/simd/avx512_double.h
@@ -102,7 +102,8 @@ static inline MD_SIMD_FLOAT simd_gather(MD_SIMD_INT vidx, const MD_FLOAT *m, int
 static inline MD_SIMD_INT simd_int_broadcast(int scalar) { return _mm256_set1_epi32(scalar); }
 static inline MD_SIMD_INT simd_int_zero() { return _mm256_setzero_si256(); }
 static inline MD_SIMD_INT simd_int_seq() { return _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0); }
-static inline MD_SIMD_INT simd_int_load(const int *m) { return _mm256_load_epi32(m); }
+static inline MD_SIMD_INT simd_int_load(const int *m) { return _mm256_load_si256((const MD_SIMD_INT *) m); }
+//static inline MD_SIMD_INT simd_int_load(const int *m) { return _mm256_load_epi32(m); }
 static inline MD_SIMD_INT simd_int_add(MD_SIMD_INT a, MD_SIMD_INT b) { return _mm256_add_epi32(a, b); }
 static inline MD_SIMD_INT simd_int_mul(MD_SIMD_INT a, MD_SIMD_INT b) { return _mm256_mul_epi32(a, b); }
 static inline MD_SIMD_INT simd_int_mask_load(const int *m, MD_SIMD_MASK k) { return _mm256_mask_load_epi32(simd_int_zero(), k, m); }
diff --git a/include_NVCC.mk b/include_NVCC.mk
index 07edb9b..446ccbf 100644
--- a/include_NVCC.mk
+++ b/include_NVCC.mk
@@ -6,9 +6,11 @@ ANSI_CFLAGS += -std=c99
 ANSI_CFLAGS += -pedantic
 ANSI_CFLAGS += -Wextra
 
-# CFLAGS   = -O0 -g  -std=c99 -fargument-noalias
+CFLAGS   = -O3 -march=native  -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
+#CFLAGS   = -O3 -march=cascadelake  -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
+#CFLAGS   = -O3 -g # -fopenmp
+#CFLAGS   = -O0 -g  -std=c99 -fargument-noalias
 #CFLAGS   = -O3 -g -arch=sm_61 # -fopenmp
-CFLAGS   = -O3 -g # -fopenmp
 ASFLAGS  =  -masm=intel
 LFLAGS   =
 DEFINES  = -D_GNU_SOURCE -DCUDA_TARGET -DNO_ZMM_INTRIN #-DLIKWID_PERFMON
diff --git a/lammps/main.c b/lammps/main.c
index c08f1da..1cc5c46 100644
--- a/lammps/main.c
+++ b/lammps/main.c
@@ -116,6 +116,17 @@ double computeForce(Eam *eam, Parameter *param, Atom *atom, Neighbor *neighbor,
     #endif
 }
 
+void writeInput(Parameter *param, Atom *atom) {
+    FILE *fpin = fopen("input.in", "w");
+    fprintf(fpin, "0,%f,0,%f,0,%f\n", param->xprd, param->yprd, param->zprd);
+
+    for(int i = 0; i < atom->Nlocal; i++) {
+        fprintf(fpin, "1,%f,%f,%f,%f,%f,%f\n", atom_x(i), atom_y(i), atom_z(i), atom_vx(i), atom_vy(i), atom_vz(i));
+    }
+
+    fclose(fpin);
+}
+
 int main(int argc, char** argv) {
     double timer[NUMTIMER];
     Eam eam;
@@ -218,6 +229,8 @@ int main(int argc, char** argv) {
     traceAddresses(&param, &atom, &neighbor, n + 1);
     #endif
 
+    //writeInput(&param, &atom);
+
     timer[FORCE] = computeForce(&eam, &param, &atom, &neighbor, &stats);
     timer[NEIGH] = 0.0;
     timer[TOTAL] = getTimeStamp();