On Tue, Mar 24, 2015 at 1:58 PM, Vincent Guittot vincent.guittot@linaro.org wrote:
On 24 March 2015 at 07:34, Amit Kucheria amit.kucheria@linaro.org wrote:
On Wed, Mar 11, 2015 at 1:08 AM, Larry Bassel larry.bassel@linaro.org wrote:
root@linaro-developer:~/pm-qa/cputopology# ./cputopology_03.sh cputopology_03.0/cpu0: checking sched_domain_flags (domain 0)... Ok domain0 flag 0x102f domain0 share cpu capacity flag not set domain0 share power domain flag not set domain0 share cpu package resources flag not set cputopology_03.0/cpu1: checking sched_domain_flags (domain 0)... Ok domain0 flag 0x102f domain0 share cpu capacity flag not set domain0 share power domain flag not set domain0 share cpu package resources flag not set cputopology_03.0/cpu2: checking sched_domain_flags (domain 0)... Ok domain0 flag 0x102f domain0 share cpu capacity flag not set domain0 share power domain flag not set domain0 share cpu package resources flag not set cputopology_03.0/cpu3: checking sched_domain_flags (domain 0)... Ok domain0 flag 0x102f domain0 share cpu capacity flag not set domain0 share power domain flag not set domain0 share cpu package resources flag not set
cputopology_03: pass
I'm not sure if printing the domain0 flag is still necessary except for debugging, but I'll let others comment.
I think someone wanted all flags interpreted, personally I think this is overkill since the test IMHO should be focused on flags important to EAS, printing the flag value is a compromise between nothing and full interpretation.
One question that I may have forgotten to ask you earlier: How did you define a test failure? AFAICT, I get the impression that the test "passes" every time, or if it will skip if the sched_domain directory doesn't exist. I haven't come across an instance where a subtest fails.
Yes, I think the test isn't so much a pass/fail one than an informational one and the person running the test would need to know if the flag settings make sense. Do (any of you) have any better ideas here? Perhaps if the directory doesn't exist it should be a fail (if the EAS patches creating this directory weren't applied or were not working properly, the directory would not exist).
I think the sched_domain directory is always present. Vincent?
It's present only with debug config
Larry, this fact should be documented in the test description with the name of the config option.
I think the power domain flag is the interesting one for EAS but I'll defer to Vincent on whether we need to care about the others - e.g. we in case of cpu capacity we care about the values they're set to.
We are interested in the 3 flags that are tested by the script: share capacity,share resources and share power domain These flags reflect the soc topology