diff --git a/Makefile b/Makefile
index df99a25..76f3709 100644
--- a/Makefile
+++ b/Makefile
@@ -56,6 +56,10 @@ ifeq ($(strip $(USE_REFERENCE_VERSION)),true)
     DEFINES += -DUSE_REFERENCE_VERSION
 endif
 
+ifeq ($(strip $(DEBUG)),true)
+    DEFINES += -DDEBUG
+endif
+
 ifneq ($(VECTOR_WIDTH),)
     DEFINES += -DVECTOR_WIDTH=$(VECTOR_WIDTH)
 endif
diff --git a/config.mk b/config.mk
index 2f16422..62206d8 100644
--- a/config.mk
+++ b/config.mk
@@ -10,6 +10,8 @@ DATA_TYPE ?= DP
 DATA_LAYOUT ?= AOS
 # Assembly syntax to generate (ATT/INTEL)
 ASM_SYNTAX ?= ATT
+# Debug
+DEBUG ?= false
 
 # Number of times to run the atoms loop on stubbed variant
 ATOMS_LOOP_RUNS ?= 1
@@ -30,7 +32,7 @@ COMPUTE_STATS ?= true
 # AOS or SOA
 CLUSTER_LAYOUT ?= SOA
 # Use reference version
-USE_REFERENCE_VERSION ?= true
+USE_REFERENCE_VERSION ?= false
 
 #Feature options
 OPTIONS =  -DALIGNMENT=64
diff --git a/gromacs/force_lj.c b/gromacs/force_lj.c
index b861019..166c280 100644
--- a/gromacs/force_lj.c
+++ b/gromacs/force_lj.c
@@ -186,10 +186,10 @@ double computeForceLJ_4xn(Parameter *param, Atom *atom, Neighbor *neighbor, Stat
             MD_SIMD_FLOAT dely3 = simd_sub(yi3_tmp, yj_tmp);
             MD_SIMD_FLOAT delz3 = simd_sub(zi3_tmp, zj_tmp);
 
-            MD_SIMD_MASK excl_mask0 = simd_mask_from_u32((unsigned int)(0xff - 0x80 * cond0 - 0x8 * cond1));
-            MD_SIMD_MASK excl_mask1 = simd_mask_from_u32((unsigned int)(0xff - 0x40 * cond0 - 0x4 * cond1));
-            MD_SIMD_MASK excl_mask2 = simd_mask_from_u32((unsigned int)(0xff - 0x20 * cond0 - 0x2 * cond1));
-            MD_SIMD_MASK excl_mask3 = simd_mask_from_u32((unsigned int)(0xff - 0x10 * cond0 - 0x1 * cond1));
+            MD_SIMD_MASK excl_mask0 = simd_mask_from_u32((unsigned int)(0xff - 0x1 * cond0 - 0x10 * cond1));
+            MD_SIMD_MASK excl_mask1 = simd_mask_from_u32((unsigned int)(0xff - 0x2 * cond0 - 0x20 * cond1));
+            MD_SIMD_MASK excl_mask2 = simd_mask_from_u32((unsigned int)(0xff - 0x4 * cond0 - 0x40 * cond1));
+            MD_SIMD_MASK excl_mask3 = simd_mask_from_u32((unsigned int)(0xff - 0x8 * cond0 - 0x80 * cond1));
 
             MD_SIMD_FLOAT rsq0 = simd_fma(delx0, delx0, simd_fma(dely0, dely0, simd_mul(delz0, delz0)));
             MD_SIMD_FLOAT rsq1 = simd_fma(delx1, delx1, simd_fma(dely1, dely1, simd_mul(delz1, delz1)));
diff --git a/gromacs/includes/simd.h b/gromacs/includes/simd.h
index 1c867df..d089188 100644
--- a/gromacs/includes/simd.h
+++ b/gromacs/includes/simd.h
@@ -21,11 +21,14 @@
  * =======================================================================================
  */
 
+#include <string.h>
 #include <immintrin.h>
 #include <zmmintrin.h>
 
-#define MD_SIMD_FLOAT   __m512d
-#define MD_SIMD_MASK    __mmask8
+#define MD_SIMD_FLOAT       __m512d
+#define MD_SIMD_MASK        __mmask8
+#define SIMD_PRINT_REAL(a)  simd_print_real(#a, a);
+#define SIMD_PRINT_MASK(a)  simd_print_mask(#a, a);
 
 static inline MD_SIMD_FLOAT simd_broadcast(double scalar) { return _mm512_set1_pd(scalar); }
 static inline MD_SIMD_FLOAT simd_zero() { return _mm512_set1_pd(0.0); }
@@ -35,7 +38,8 @@ static inline MD_SIMD_FLOAT simd_mul(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return
 static inline MD_SIMD_FLOAT simd_fma(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_FLOAT c) { return _mm512_fmadd_pd(a, b, c); }
 static inline MD_SIMD_FLOAT simd_reciprocal(MD_SIMD_FLOAT a) { return _mm512_rcp14_pd(a); }
 static inline MD_SIMD_FLOAT simd_masked_add(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b, MD_SIMD_MASK m) { return _mm512_mask_add_pd(a, m, a, b); }
-static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int m) { return _cvtu32_mask8(m); }
+static inline MD_SIMD_MASK simd_mask_from_u32(unsigned int a) { return _cvtu32_mask8(a); }
+static inline MD_SIMD_MASK simd_mask_to_u32(unsigned int a) { return _cvtmask8_u32(a); }
 static inline MD_SIMD_MASK simd_mask_and(MD_SIMD_MASK a, MD_SIMD_MASK b) { return _kand_mask8(a, b); }
 static inline MD_SIMD_MASK simd_mask_cond_lt(MD_SIMD_FLOAT a, MD_SIMD_FLOAT b) { return _mm512_cmp_pd_mask(a, b, _CMP_LT_OQ); }
 
@@ -59,3 +63,17 @@ static inline MD_FLOAT simd_horizontal_sum(MD_SIMD_FLOAT a) {
     x = _mm512_add_pd(x, _mm512_permute_pd(x, 0x01));
     return *((double *) &x);
 }
+
+static inline void simd_print_real(const char *ref, MD_SIMD_FLOAT a) {
+    double x[8];
+    memcpy(x, &a, sizeof(x));
+
+    fprintf(stdout, "%s: ", ref);
+    for(int i = 0; i < 8; i++) {
+        fprintf(stdout, "%f ", x[i]);
+    }
+
+    fprintf(stdout, "\n");
+}
+
+static inline void simd_print_mask(const char *ref, MD_SIMD_MASK a) { fprintf(stdout, "%s: %x\n", ref, simd_mask_to_u32(a)); }
diff --git a/gromacs/includes/util.h b/gromacs/includes/util.h
index c777282..197f253 100644
--- a/gromacs/includes/util.h
+++ b/gromacs/includes/util.h
@@ -33,9 +33,9 @@
 #define ABS(a) ((a) >= 0 ? (a) : -(a))
 #endif
 #ifdef DEBUG
-#define DEBUG_MESSAGE(msg)  printf
+#define DEBUG_MESSAGE   printf
 #else
-#define DEBUG_MESSAGE(msg)
+#define DEBUG_MESSAGE
 #endif
 
 #define FF_LJ   0