sched: round a bit better

round a tiny bit better in high-frequency rescheduling scenarios,
by rounding around zero instead of rounding down.

(this is pretty theoretical though)

Signed-off-by: Ingo Molnar <mingo@elte.hu>

kernel/sched.c

@@ -638,25 +638,29 @@ static u64 div64_likely32(u64 divident, unsigned long divisor)
 
 #define WMULT_SHIFT	32
 
+/*
+ * Shift right and round:
+ */
+#define RSR(x, y) (((x) + (1UL << ((y) - 1))) >> (y))
+
 static unsigned long
 calc_delta_mine(unsigned long delta_exec, unsigned long weight,
 		struct load_weight *lw)
 {
 	u64 tmp;
 
 	if (unlikely(!lw->inv_weight))
-		lw->inv_weight = WMULT_CONST / lw->weight;
+		lw->inv_weight = (WMULT_CONST - lw->weight/2) / lw->weight + 1;
 
 	tmp = (u64)delta_exec * weight;
 	/*
 	 * Check whether we'd overflow the 64-bit multiplication:
 	 */
-	if (unlikely(tmp > WMULT_CONST)) {
-		tmp = ((tmp >> WMULT_SHIFT/2) * lw->inv_weight)
-				>> (WMULT_SHIFT/2);
-	} else {
-		tmp = (tmp * lw->inv_weight) >> WMULT_SHIFT;
-	}
+	if (unlikely(tmp > WMULT_CONST))
+		tmp = RSR(RSR(tmp, WMULT_SHIFT/2) * lw->inv_weight,
+			WMULT_SHIFT/2);
+	else
+		tmp = RSR(tmp * lw->inv_weight, WMULT_SHIFT);
 
 	return (unsigned long)min(tmp, (u64)(unsigned long)LONG_MAX);
 }