diff --git a/check.c b/check.c
index cea1be3e..9ec644aa 100644
--- a/check.c
+++ b/check.c
@@ -488,7 +488,7 @@ static int state_check_process(struct snapraid_state* state, int check, int fix,
 
 	buffer = malloc_nofail_vector_align(diskmax, buffermax, state->block_size, &buffer_alloc);
 	if (!state->opt.skip_self)
-		mtest_vector(buffer, buffermax, state->block_size);
+		mtest_vector(buffermax, state->block_size, buffer);
 
 	/* fill up the zero buffer */
 	memset(buffer[buffermax-1], 0, state->block_size);
diff --git a/raid/memory.c b/raid/memory.c
index 0cfceee5..bb49ee1e 100644
--- a/raid/memory.c
+++ b/raid/memory.c
@@ -34,41 +34,53 @@ void *raid_malloc_align(size_t size, void **freeptr)
 	return ptr;
 }
 
-void **raid_malloc_vector(size_t reverse, size_t count, size_t size, void **freeptr)
+void **raid_malloc_vector(int nd, int n, size_t size, void **freeptr)
 {
 	void **v;
 	unsigned char *va;
-	size_t i;
+	int i;
 
-	assert(reverse <= count);
-
-	v = malloc(count * sizeof(void *));
+	v = malloc(n * sizeof(void *));
 	if (!v)
 		return 0;
 
-	va = raid_malloc_align(count * (size + RAID_MALLOC_DISPLACEMENT), freeptr);
+	va = raid_malloc_align(n * (size + RAID_MALLOC_DISPLACEMENT), freeptr);
 	if (!va) {
 		free(v);
 		return 0;
 	}
 
-	for (i = 0; i < count; ++i)
-		v[i] = va + i * (size + RAID_MALLOC_DISPLACEMENT);
+	for (i = 0; i < n; ++i) {
+		v[i] = va;
+		va += size + RAID_MALLOC_DISPLACEMENT;
+	}
 
-	/* reverse order for the initial ones */
-	for (i = 0; i < reverse/2; ++i) {
+	/* reverse order of the data blocks */
+	/* because they are usually accessed from the last one */
+	for (i = 0; i < nd/2; ++i) {
 		void *ptr = v[i];
-		v[i] = v[reverse - 1 - i];
-		v[reverse - 1 - i] = ptr;
+		v[i] = v[nd - 1 - i];
+		v[nd - 1 - i] = ptr;
 	}
 
 	return v;
 }
 
-int raid_mtest_vector(void **vv, size_t count, size_t size)
+void raid_mrand_vector(int n, size_t size, void **vv)
 {
 	unsigned char **v = (unsigned char **)vv;
-	size_t i;
+	int i;
+	size_t j;
+
+	for (i = 0; i < n; ++i)
+		for (j = 0; j < size; ++j)
+			v[i][j] = rand();
+}
+
+int raid_mtest_vector(int n, size_t size, void **vv)
+{
+	unsigned char **v = (unsigned char **)vv;
+	int i;
 	size_t j;
 	unsigned k;
 	unsigned char d;
@@ -76,7 +88,7 @@ int raid_mtest_vector(void **vv, size_t count, size_t size)
 
 	/* fill with 0 */
 	d = 0;
-	for (i = 0; i < count; ++i)
+	for (i = 0; i < n; ++i)
 		for (j = 0; j < size; ++j)
 			v[i][j] = d;
 
@@ -86,7 +98,7 @@ int raid_mtest_vector(void **vv, size_t count, size_t size)
 		d = k;
 
 		/* forward fill */
-		for (i = 0; i < count; ++i) {
+		for (i = 0; i < n; ++i) {
 			for (j = 0; j < size; ++j) {
 				if (v[i][j] != p)
 					return -1;
@@ -97,7 +109,7 @@ int raid_mtest_vector(void **vv, size_t count, size_t size)
 		p = d;
 		d = ~p;
 		/* backward fill with complement */
-		for (i = 0; i < count; ++i) {
+		for (i = 0; i < n; ++i) {
 			for (j = size; j > 0; --j) {
 				if (v[i][j-1] != p)
 					return -1;
@@ -109,14 +121,3 @@ int raid_mtest_vector(void **vv, size_t count, size_t size)
 	return 0;
 }
 
-void raid_mrand_vector(void **vv, size_t count, size_t size)
-{
-	unsigned char **v = (unsigned char **)vv;
-	size_t i;
-	size_t j;
-
-	for (i = 0; i < count; ++i)
-		for (j = 0; j < size; ++j)
-			v[i][j] = rand();
-}
-
diff --git a/raid/memory.h b/raid/memory.h
index 3a679709..bea3cc33 100644
--- a/raid/memory.h
+++ b/raid/memory.h
@@ -28,26 +28,27 @@
  *
  * When allocating a sequence of blocks with a size of power of 2,
  * there is the risk that the start of each block is mapped into the same
- * cache line, resulting in cache collisions if you access all the blocks
- * in parallel, from the start to the end.
+ * cache line or prefetching prediction, resulting in collisions if you
+ * access all the blocks in parallel, from the start to the end.
  *
  * The selected value was choosen empirically with some speed tests
  * with 8/12/16/20/24 data buffers of 256 KB.
  *
- * With displacement:
- * (8 buffers of 256 KB, icore5, 32 bits):
+ * These are the results with no displacement:
  *
- *        sse2
- * par1  21936 [MB/s]
- * par2  11902 [MB/s]
- * parz   5838 [MB/s]
+ *            sse2
+ *    par1   15368 [MB/s]
+ *    par2    6814 [MB/s]
+ *    parz    3033 [MB/s]
  *
- * Without displacement:
+ * These are the results with displacement resulting in improvments
+ * in the order of 20% or more:
+ *
+ *            sse2
+ *    par1   21936 [MB/s]
+ *    par2   11902 [MB/s]
+ *    parz    5838 [MB/s]
  *
- *        sse2
- * par1  15368 [MB/s]
- * par2   6814 [MB/s]
- * parz   3033 [MB/s]
  */
 #define RAID_MALLOC_DISPLACEMENT (7*256)
 
@@ -58,22 +59,22 @@ void *raid_malloc_align(size_t size, void **freeptr);
 
 /**
  * Aligned vector allocation.
- * Returns a vector of "count" pointers, each one pointing to a block of the
- * specified "size".
- * The first "reverse" pointers are put in reverse memory offset order.
+ * Returns a vector of ::n pointers, each one pointing to a block of
+ * the specified ::size.
+ * The first ::nd elements are reversed in order.
  */
-void **raid_malloc_vector(size_t reverse, size_t count, size_t size, void **freeptr);
+void **raid_malloc_vector(int nd, int n, size_t size, void **freeptr);
+
+/**
+ * Fills the memory vector with random data.
+ */
+void raid_mrand_vector(int n, size_t size, void **vv);
 
 /**
  * Tests the memory vector for RAM problems.
  * If a problem is found, it crashes.
  */
-int raid_mtest_vector(void **vv, size_t count, size_t size);
-
-/**
- * Fills the memory vector with random data.
- */
-void raid_mrand_vector(void **vv, size_t count, size_t size);
+int raid_mtest_vector(int n, size_t size, void **vv);
 
 #endif
 
diff --git a/raid/module.c b/raid/module.c
index 9b5f9843..ba185b0d 100644
--- a/raid/module.c
+++ b/raid/module.c
@@ -228,7 +228,7 @@ int raid_selftest(void)
 		/* test parity generation */
 		ret = raid_test_par(nd, np, size, v, ref);
 		if (ret != 0)
-			return ret;
+			goto bail;
 
 		/* test recovering with full broken data disks */
 		for (i = 0; i < np; ++i)
diff --git a/raid/test.c b/raid/test.c
index 3eb0143f..6a3217cb 100644
--- a/raid/test.c
+++ b/raid/test.c
@@ -148,7 +148,7 @@ int raid_test_rec(int mode, int nd, size_t size)
 	waste = v[nv-1];
 
 	/* fill data disk with random */
-	raid_mrand_vector(v, nd, size);
+	raid_mrand_vector(nd, size, v);
 
 	/* setup recov functions */
 	for (i = 0; i < np; ++i) {
@@ -253,7 +253,7 @@ int raid_test_par(int mode, int nd, size_t size)
 		return -1;
 
 	/* fill with random */
-	raid_mrand_vector(v, nv, size);
+	raid_mrand_vector(nv, size, v);
 
 	/* compute the parity */
 	raid_par_ref(nd, np, size, v);
diff --git a/scrub.c b/scrub.c
index 175fe6d9..cbf0bcbb 100644
--- a/scrub.c
+++ b/scrub.c
@@ -69,7 +69,7 @@ static int state_scrub_process(struct snapraid_state* state, struct snapraid_par
 
 	buffer = malloc_nofail_vector_align(diskmax, buffermax, state->block_size, &buffer_alloc);
 	if (!state->opt.skip_self)
-		mtest_vector(buffer, buffermax, state->block_size);
+		mtest_vector(buffermax, state->block_size, buffer);
 
 	error = 0;
 	silent_error = 0;
diff --git a/speed.c b/speed.c
index d63bb60d..2d2b0c42 100644
--- a/speed.c
+++ b/speed.c
@@ -81,7 +81,6 @@ void speed(void)
 	void *v_alloc;
 	void **v;
 
-	/* we need disk + 1 for each parity buffers + 1 zero v */
 	nv = nd + RAID_PARITY_MAX + 1;
 
 	v = malloc_nofail_vector_align(nd, nv, size, &v_alloc);
diff --git a/sync.c b/sync.c
index 376bfe46..f229af25 100644
--- a/sync.c
+++ b/sync.c
@@ -76,7 +76,7 @@ static int state_sync_process(struct snapraid_state* state, struct snapraid_pari
 
 	buffer = malloc_nofail_vector_align(diskmax, buffermax, state->block_size, &buffer_alloc);
 	if (!state->opt.skip_self)
-		mtest_vector(buffer, buffermax, state->block_size);
+		mtest_vector(buffermax, state->block_size, buffer);
 
 	error = 0;
 	silent_error = 0;
diff --git a/util.c b/util.c
index 25e5787f..67040174 100644
--- a/util.c
+++ b/util.c
@@ -1384,11 +1384,11 @@ void* malloc_nofail_align(size_t size, void** freeptr)
 	return ptr;
 }
 
-void** malloc_nofail_vector_align(size_t reverse, size_t count, size_t size, void** freeptr)
+void** malloc_nofail_vector_align(int nd, int n, size_t size, void** freeptr)
 {
 	void* ptr;
 
-	ptr = raid_malloc_vector(reverse, count, size, freeptr);
+	ptr = raid_malloc_vector(nd, n, size, freeptr);
 
 	if (!ptr) {
 		/* don't use printf to avoid any possible extra allocation */
@@ -1401,20 +1401,20 @@ void** malloc_nofail_vector_align(size_t reverse, size_t count, size_t size, voi
 	return ptr;
 }
 
-void mtest_vector(void** vv, size_t count, size_t size)
+void mrand_vector(int n , size_t size, void** vv)
 {
-	if (raid_mtest_vector(vv, count, size) != 0) {
+	raid_mrand_vector(n, size, vv);
+}
+
+void mtest_vector(int n , size_t size, void** vv)
+{
+	if (raid_mtest_vector(n, size, vv) != 0) {
 		fprintf(stderr, "DANGER! Your RAM memory is broken! DO NOT PROCEED UNTIL FIXED!\n");
 		fprintf(stderr, "Try running some memory test like http://www.memtest86.com/\n");
 		exit(EXIT_FAILURE);
 	}
 }
 
-void mrand_vector(void** vv, size_t count, size_t size)
-{
-	raid_mrand_vector(vv, count, size);
-}
-
 /****************************************************************************/
 /* byte operations */
 
diff --git a/util.h b/util.h
index aa10fcaf..e528ec63 100644
--- a/util.h
+++ b/util.h
@@ -466,22 +466,20 @@ void* malloc_nofail_align(size_t size, void** freeptr);
 
 /**
  * Safe aligned vector allocation.
- * Returns a vector of "count" pointers, each one pointing to a block of the specified "size".
- * The first "reverse" pointers are put in reverse memory offset order.
  * If no memory is available, it aborts.
  */
-void** malloc_nofail_vector_align(size_t reverse, size_t count, size_t size, void** freeptr);
+void** malloc_nofail_vector_align(int nd, int n, size_t size, void** freeptr);
+
+/**
+ * Fills the memory vector with random data.
+ */
+void mrand_vector(int n, size_t size, void** vv);
 
 /**
  * Tests the memory vector for RAM problems.
  * If a problem is found, it crashes.
  */
-void mtest_vector(void** vv, size_t count, size_t size);
-
-/**
- * Fills the memory vector with random data.
- */
-void mrand_vector(void** vv, size_t count, size_t size);
+void mtest_vector(int n, size_t size, void** vv);
 
 /**
  * Safe strdup.