diff --git a/include/linux/sched/walt.h b/include/linux/sched/walt.h
index 87757ac85a4d..853c7c169795 100644
--- a/include/linux/sched/walt.h
+++ b/include/linux/sched/walt.h
@@ -32,7 +32,13 @@ enum task_boost_type {
 };
 
 #define WALT_NR_CPUS 8
-#define RAVG_HIST_SIZE_MAX 8
+/*
+ * RAVG_HIST_SHIFT trick can only be used if RAVG_HIST_SIZE is a power of 2.
+ */
+#define RAVG_HIST_SIZE 8
+#define RAVG_HIST_SHIFT 3
+#define RAVG_HIST_MASK (RAVG_HIST_SIZE - 1)
+
 /* wts->bucket_bitmask needs to be updated if NUM_BUSY_BUCKETS > 16 */
 #define NUM_BUSY_BUCKETS 16
 #define NUM_BUSY_BUCKETS_SHIFT 4
@@ -90,7 +96,7 @@ struct walt_task_struct {
 	u64				mark_start;
 	u32				sum, demand;
 	u32				coloc_demand;
-	u32				sum_history[RAVG_HIST_SIZE_MAX];
+	u32				sum_history[RAVG_HIST_SIZE];
 	u32				curr_window_cpu[WALT_NR_CPUS];
 	u32				prev_window_cpu[WALT_NR_CPUS];
 	u32				curr_window, prev_window;
diff --git a/kernel/sched/walt/trace.h b/kernel/sched/walt/trace.h
index b1f1eb4e0688..4b48c88a5cd2 100644
--- a/kernel/sched/walt/trace.h
+++ b/kernel/sched/walt/trace.h
@@ -83,7 +83,7 @@ TRACE_EVENT(sched_update_history,
 		__field(unsigned int,		demand)
 		__field(unsigned int,		coloc_demand)
 		__field(unsigned int,		pred_demand_scaled)
-		__array(u32,			hist, RAVG_HIST_SIZE_MAX)
+		__array(u32,			hist, RAVG_HIST_SIZE)
 		__field(unsigned int,		nr_big_tasks)
 		__field(int,			cpu)
 	),
@@ -98,7 +98,7 @@ TRACE_EVENT(sched_update_history,
 		__entry->coloc_demand	= wts->coloc_demand;
 		__entry->pred_demand_scaled	= wts->pred_demand_scaled;
 		memcpy(__entry->hist, wts->sum_history,
-					RAVG_HIST_SIZE_MAX * sizeof(u32));
+					RAVG_HIST_SIZE * sizeof(u32));
 		__entry->nr_big_tasks	= wrq->walt_stats.nr_big_tasks;
 		__entry->cpu		= rq->cpu;
 	),
diff --git a/kernel/sched/walt/walt.c b/kernel/sched/walt/walt.c
index ab0262b1e97e..78036d5df2c2 100644
--- a/kernel/sched/walt/walt.c
+++ b/kernel/sched/walt/walt.c
@@ -167,8 +167,6 @@ static inline u64 walt_rq_clock(struct rq *rq)
 
 static unsigned int walt_cpu_high_irqload;
 
-static __read_mostly unsigned int sched_ravg_hist_size = RAVG_HIST_SIZE_MAX;
-
 static __read_mostly unsigned int sched_io_is_busy = 1;
 
 /* Window size (in ns) */
@@ -1935,10 +1933,10 @@ static void update_history(struct rq *rq, struct task_struct *p,
 	/* Push new 'runtime' value onto stack */
 	for (; samples > 0; samples--) {
 		hist[wts->cidx] = runtime;
-		wts->cidx = ++(wts->cidx) % sched_ravg_hist_size;
+		wts->cidx = ++(wts->cidx) & RAVG_HIST_MASK;
 	}
 
-	for (i = 0; i < sched_ravg_hist_size; i++) {
+	for (i = 0; i < RAVG_HIST_SIZE; i++) {
 		sum += hist[i];
 		if (hist[i] > max)
 			max = hist[i];
@@ -1951,7 +1949,7 @@ static void update_history(struct rq *rq, struct task_struct *p,
 	} else if (sysctl_sched_window_stats_policy == WINDOW_STATS_MAX) {
 		demand = max;
 	} else {
-		avg = div64_u64(sum, sched_ravg_hist_size);
+		avg = sum >> RAVG_HIST_SHIFT;
 		if (sysctl_sched_window_stats_policy == WINDOW_STATS_AVG)
 			demand = avg;
 		else
@@ -1980,7 +1978,7 @@ static void update_history(struct rq *rq, struct task_struct *p,
 
 	wts->demand = demand;
 	wts->demand_scaled = demand_scaled;
-	wts->coloc_demand = div64_u64(sum, sched_ravg_hist_size);
+	wts->coloc_demand = sum >> RAVG_HIST_SHIFT;
 	wts->pred_demand_scaled = pred_demand_scaled;
 
 	if (demand_scaled > sysctl_sched_min_task_util_for_colocation)
@@ -2307,7 +2305,7 @@ static void init_new_task_load(struct task_struct *p)
 	wts->demand_scaled = init_load_windows_scaled;
 	wts->coloc_demand = init_load_windows;
 	wts->pred_demand_scaled = 0;
-	for (i = 0; i < RAVG_HIST_SIZE_MAX; ++i)
+	for (i = 0; i < RAVG_HIST_SIZE; ++i)
 		wts->sum_history[i] = init_load_windows;
 	wts->misfit = false;
 	wts->rtg_high_prio = false;
@@ -2850,7 +2848,7 @@ static void _set_preferred_cluster(struct walt_related_thread_group *grp)
 		}
 
 		if (wts->mark_start < wallclock -
-		    (sched_ravg_window * sched_ravg_hist_size))
+		    (sched_ravg_window * RAVG_HIST_SIZE))
 			continue;
 
 		combined_demand += wts->coloc_demand;
@@ -3726,7 +3724,7 @@ void walt_fill_ta_data(struct core_ctl_notif_data *data)
 
 	list_for_each_entry(wts, &grp->tasks, grp_list) {
 		if (wts->mark_start < wallclock -
-		    (sched_ravg_window * sched_ravg_hist_size))
+		    (sched_ravg_window * RAVG_HIST_SIZE))
 			continue;
 
 		total_demand += wts->coloc_demand;