On 09/03/2018 10:24, Leo Yan wrote:
Hi Daniel,
On Fri, Mar 09, 2018 at 09:59:48AM +0100, Daniel Lezcano wrote:
[...]
Just want to remind, for the boosting tasks, it's quite tricky to say should place it on big or LITTLE CPUs. The reason is, if we place one boosted task on a big CPU, and this big CPU has its saperate clock from LITTLE CPUs, then this boosted task doesn't interfere other tasks on LITTLE CPUs; on the other hand if we place this boosted task on LITTLE CPUs, this is easily to increase frequency on LITTLE CPUs and all other tasks on LITTLE CPUs also run at higher frequency and consume more power. So if can find some specific boost margin, we can observe power regression after we place boosted task on LITTLE CPU.
Hi Leo,
I'm missing the point about a separate clock line for bL. If the task is placed on a big CPUs, the other big CPUs will also be impacted and we will see a frequency increase on the big cluster and the same power regression, no ?
Yeah, you are right. But I should meantion two extra things for different cases:
Usually big CPU OPP starts from higher capacity than LITTLE CPU, so it's hard to be impacted by small boost margin; e.g. using Hikey960 as example, boost margin = 20% so the boosted util = 1024*20% = 205; this is easily to increase CA53 CPUs increase from 533MHz (cap=133) to 990MHz (cap=250), but CA73 can stay at lowest frequency 903MHz (cap=390);
IIRC, e.g. the video case have many backgroud tasks, so these tasks will be placed on LITTLE CPUs with middle workload; but if boosted tasks (foreground tasks, like surfaceflinger, etc) are placed on LITTLE CPUs and it's easily set LITTLE CPU to highest two OPPs (1.4GHz and 1.7GHz); as result the power data might be worse than place boosted tasks on big CPUs at 903MHz.
Ok, got it.
Thanks for the clarification.
-- http://www.linaro.org/ Linaro.org │ Open source software for ARM SoCs
Follow Linaro: http://www.facebook.com/pages/Linaro Facebook | http://twitter.com/#!/linaroorg Twitter | http://www.linaro.org/linaro-blog/ Blog