On 05/13/2015 12:39 PM, Jon Medhurst (Tixy) wrote:
On Wed, 2015-05-13 at 11:06 -0400, Jorge Ramirez-Ortiz wrote:
From: Jorge Ramirez-Ortiz jro@xenomai.org
The current arm64 power scaling calculations deviate from the arm implementation causing scheduling problems on non bigLITTLE architectures (this could be due to baseline/merging issues).
I've never really looked at this code before but I did some comparisons of arm versus arm64 versions and LSK versus mainline and my observations are:
There is no arm64 implementation in mainline. (Is there a reason?)
The arm64 implementation in LSK closely resembles the arm mainline
version, except that mainline doesn't have the
if (min_capacity == max_capacity) return; else
That looks like an accidental side effect of commit 816a8de0017f ("ARM: topology: remove hwid/MPIDR dependency from cpu_capacity"). But it does meant that in mainline 'middle_capacity' is now set by parse_dt_cpu_power (to a useful value?) rather than being left as '1'.
- The arm version in LSK looks like the ARM64 version, in the sense
that 'middle_capacity' is not set, so is this something which needs addressing on arm as well?
I'd think so. the patch below (scaling in the case where min_capacity == max_capacity) is one way of fixing it.
The proposed workaround uses the middle_capacity weighting factor to guarantee that all CPUs report the same power (1024) on non bigLITTLE SoC.
Signed-off-by: Jorge Ramirez-Ortiz jro@xenomai.org
arch/arm64/kernel/topology.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/arm64/kernel/topology.c b/arch/arm64/kernel/topology.c index db8bb29..65837c2 100644 --- a/arch/arm64/kernel/topology.c +++ b/arch/arm64/kernel/topology.c @@ -325,7 +325,7 @@ static void __init parse_dt_cpu_power(void) * constraint explained near table_efficiency[]. */ if (min_capacity == max_capacity)
return;
middle_capacity = max_capacity >> SCHED_POWER_SHIFT;