[PATCH 4/7] arm: Micro-architecture invariant load tracking support

[Morten Rasmussen <morten.rasmussen@arm.com>]

Adds micro-architecture difference into scheduler load-tracking
scale-invariance correction factor for big.LITTLE systems. The factor is
now:
	(current_freq(cpu) * uarch_factor(cpu) * SCHED_CAPACITY_SCALE) /
		(max_freq(cpu) * max_uarch_factor)

Signed-off-by: Morten Rasmussen <morten.rasmussen@arm.com>
---
 arch/arm/kernel/topology.c |   17 ++++++++++++++---
 1 file changed, 14 insertions(+), 3 deletions(-)

diff --git a/arch/arm/kernel/topology.c b/arch/arm/kernel/topology.c
index 3e9a979..9f499e2 100644
--- a/arch/arm/kernel/topology.c
+++ b/arch/arm/kernel/topology.c
@@ -79,6 +79,9 @@ static unsigned long *__cpu_capacity;
 
 static unsigned long middle_capacity = 1;
 
+static unsigned long max_raw_capacity = 1;
+static DEFINE_PER_CPU(unsigned long, cpu_raw_capacity);
+
 /*
  * Iterate all CPUs' descriptor in DT and compute the efficiency
  * (as per table_efficiency). Also calculate a middle efficiency
@@ -117,6 +120,9 @@ static void __init parse_dt_topology(void)
 		if (cpu_eff->compatible == NULL)
 			continue;
 
+		per_cpu(cpu_raw_capacity, cpu) = cpu_eff->efficiency;
+		max_raw_capacity = max(max_raw_capacity, cpu_eff->efficiency);
+
 		rate = of_get_property(cn, "clock-frequency", &len);
 		if (!rate || len != 4) {
 			pr_err("%s missing clock-frequency property\n",
@@ -173,7 +179,8 @@ static void update_cpu_capacity(unsigned int cpu)
  * Scheduler load-tracking scale-invariance
  *
  * Provides the scheduler with a scale-invariance correction factor that
- * compensates for frequency scaling.
+ * compensates for frequency scaling and micro-architecture differences between
+ * cpus.
  */
 
 static DEFINE_PER_CPU(atomic_long_t, cpu_curr_freq);
@@ -195,11 +202,15 @@ unsigned long arch_scale_load_capacity(int cpu)
 {
 	unsigned long curr = atomic_long_read(&per_cpu(cpu_curr_freq, cpu));
 	unsigned long max = atomic_long_read(&per_cpu(cpu_max_freq, cpu));
+	unsigned long ret;
 
-	if (!max)
+	if (!max || !per_cpu(cpu_raw_capacity, cpu))
 		return SCHED_CAPACITY_SCALE;
 
-	return (curr * SCHED_CAPACITY_SCALE) / max;
+	ret = (curr * SCHED_CAPACITY_SCALE) / max;
+	ret = (ret * per_cpu(cpu_raw_capacity, cpu)) / max_raw_capacity;
+
+	return ret;
 }
 
 #else
-- 
1.7.9.5