diff --git a/thiersort3.h b/thiersort3.h
index afa1d99..e5df720 100644
--- a/thiersort3.h
+++ b/thiersort3.h
@@ -22,7 +22,7 @@ union th3_fu {
 typedef union th3_fu th3_fu;
 
 /** Tells from the key which bucket it is in. */
-static inline int witch_bucket3(uint32_t key) {
+static inline uint32_t witch_bucket3(uint32_t key) {
 	/* This is hackz to misuse int->float converter HEAVILY and IEE for bucketing */
 	/* https://en.wikipedia.org/wiki/Single-precision_floating-point_format */
 
@@ -35,7 +35,7 @@ static inline int witch_bucket3(uint32_t key) {
 	th3_fu as;
 	as.f = (float) key;
 
-	int witch_base = (key <= 2) ? 0 : (as.u >> 23) - 128; // 0, [127..159] -> [0..31]
+	uint32_t witch_base = (key <= 2) ? 0 : (as.u >> 23) - 128; // 0, [127..159] -> [0..31]
 	return witch_base * 8 + ((as.u >> (23 - 3)) & 7); // 0..255
 	/* All below ones were worse, because three-ways 28 bit sorting would just make unnecessary work at cost of some locality */
 	//return witch_base * 128 + ((as.u >> (23 - 7)) & 127); // 0..4095
@@ -50,6 +50,7 @@ static inline int witch_bucket3(uint32_t key) {
  * @param arr The array to sort, will contain result afterwards
  * @param temparr The temporary array with same size
  * @param n Number of elements in arr and temparr
+ * @param rstate Create with sch_rand_state rstate = schwab_rand_state(junk_uint32_t);
  */
 static inline void thiersort3(uint32_t *arr, uint32_t *temparr, int n) {
 	int bucket[256]; /* Inclusive */
@@ -76,7 +77,7 @@ static inline void thiersort3(uint32_t *arr, uint32_t *temparr, int n) {
 	}
 
 	/* Prefix sum (like in Magyarsort) */
-	int prev = 0;
+	uint32_t prev = 0;
 	#pragma GCC unroll 4
 	for (int i = 0; i < 256; i++) {
 		bucket[i] += prev;
@@ -107,7 +108,7 @@ static inline void thiersort3(uint32_t *arr, uint32_t *temparr, int n) {
 	#pragma GCC unroll 128
 	for(int i = 0; i < n; ++i) {
 		uint32_t num = arr[i];
-		int witch = witch_bucket3(num);
+		uint32_t witch = witch_bucket3(num);
 		int offset = (--bucket[witch]);
 		temparr[offset] = num;
 	}
@@ -126,7 +127,7 @@ static inline void thiersort3(uint32_t *arr, uint32_t *temparr, int n) {
 		/* Use our specialized threepass */
 		/* - which does not use up all the bits */
 		/* - which does not allocate and copy back to arr here */
-		int n = end - begin;
+		uint32_t n = end - begin;
 		uint32_t *a = &(temparr[begin]);
 		uint32_t *buf = &(arr[begin]);
 		threepass_xb(a, buf, n);
@@ -138,7 +139,7 @@ static inline void thiersort3(uint32_t *arr, uint32_t *temparr, int n) {
 	#pragma GCC unroll 64
 	for(int i = 0; i < n; ++i) {
 		uint32_t num = arr[i];
-		int witch = witch_bucket3(num);
+		uint32_t witch = witch_bucket3(num);
 		int offset = (num & (1 << 27)) ?
 			(--bucket[witch]) :
 			(bucket_leftend[witch]++);
@@ -160,7 +161,7 @@ static inline void thiersort3(uint32_t *arr, uint32_t *temparr, int n) {
 			/* Use our specialized threepass */
 			/* - which does not use up all the bits */
 			/* - which does not allocate and copy back to arr here */
-			int n = lend - lbegin;
+			uint32_t n = lend - lbegin;
 			uint32_t *a = &(temparr[lbegin]);
 			uint32_t *buf = &(arr[lbegin]);
 			threepass_xb(a, buf, n);
@@ -171,7 +172,7 @@ static inline void thiersort3(uint32_t *arr, uint32_t *temparr, int n) {
 			/* Use our specialized threepass */
 			/* - which does not use up all the bits */
 			/* - which does not allocate and copy back to arr here */
-			int n = rend - rbegin;
+			uint32_t n = rend - rbegin;
 			uint32_t *a = &(temparr[rbegin]);
 			uint32_t *buf = &(arr[rbegin]);
 			threepass_xb(a, buf, n);