Messages in this thread | | | Subject | Re: [PATCH 1/3] sched/fair: Prepare variables for increased precision of EAS estimated energy | From | Lukasz Luba <> | Date | Wed, 7 Jul 2021 09:23:54 +0100 |
| |
On 7/7/21 9:00 AM, Vincent Guittot wrote: > On Wed, 7 Jul 2021 at 09:49, Lukasz Luba <lukasz.luba@arm.com> wrote: >> >> >> >> On 7/7/21 8:07 AM, Vincent Guittot wrote: >>> On Fri, 25 Jun 2021 at 17:26, Lukasz Luba <lukasz.luba@arm.com> wrote: >>>> >>>> The Energy Aware Scheduler (EAS) tries to find best CPU for a waking up >>>> task. It probes many possibilities and compares the estimated energy values >>>> for different scenarios. For calculating those energy values it relies on >>>> Energy Model (EM) data and em_cpu_energy(). The precision which is used in >>>> EM data is in milli-Watts (or abstract scale), which sometimes is not >>>> sufficient. In some cases it might happen that two CPUs from different >>>> Performance Domains (PDs) get the same calculated value for a given task >>>> placement, but in more precised scale, they might differ. This rounding >>>> error has to be addressed. This patch prepares EAS code for better >>>> precision in the coming EM improvements. >>> >>> Could you explain why 32bits results are not enough and you need to >>> move to 64bits ? >>> >>> Right now the result is in the range [0..2^32[ mW. If you need more >>> precision and you want to return uW instead, you will have a result in >>> the range [0..4kW[ which seems to be still enough >>> >> >> Currently we have the max value limit for 'power' in EM which is >> EM_MAX_POWER 0xffff (64k - 1). We allow to register such big power >> values ~64k mW (~64Watts) for an OPP. Then based on 'power' we >> pre-calculate 'cost' fields: >> cost[i] = power[i] * freq_max / freq[i] >> So, for max freq the cost == power. Let's use that in the example. >> >> Then the em_cpu_energy() calculates as follow: >> cost * sum_util / scale_cpu >> We are interested in the first part - the value of multiplication. > > But all these are internal computations of the energy model. At the > end, the computed energy that is returned by compute_energy() and > em_cpu_energy(), fits in a long
Let's take a look at existing *10000 precision for x CPUs: cost * sum_util / scale_cpu = (64k *10000) * (x * 800) / 1024 which is: x * ~500mln
So to be close to overflowing u32 the 'x' has to be > (?=) 8 (depends on sum_util).
> >> >> The sum_util values that we can see for x CPUs which have scale_cap=1024 >> can be close to 800, let's use it in the example: >> cost * sum_util = 64k * (x * 800), where >> x=4: ~200mln >> x=8: ~400mln >> x=16: ~800mln >> x=64: ~3200mln (last one which would fit in u32) >> >> When we increase the precision by even 100, then the above values won't >> fit in the u32. Even a max cost of e.g. 10k mW and 100 precision has >> issues: >> cost * sum_util = (10k *100) * (x * 800), where >> x=4: ~3200mln >> x=8: ~6400mln >> >> For *1000 precision even a power of 1Watt becomes an issue: >> cost * sum_util = (1k *1000) * (x * 800), where >> x=4: ~3200mln >> x=8: ~6400mln >> >> That's why to make the code safe for bigger power values, I had to use >> the u64 on 32bit machines.
| |