Eh? The "ondemand" governor couldn't care less what underlying userspace is running (be it Android, your desktop machine, or a server box), and only cares about overall CPU load. Plus, nothing about CPU frequency factors into garbage collection nor process time-slice (other than more run-time can be done for a given execution period).
Cite, with real data?
I'm not going to apologize - I am going to repost what I said for inspection:
That stuff that loads up in the background and gets well-managed anyway?
Yes, true on my old Moment.
The Evo - different story.
The Android resource manager is said, in some forums, to not be terribly effective with the higher speed phones, and my experience confirms that.
1. That's purely anecdotal and I identified it as such: The Android resource manager is said, in some forums, to not be terribly effective with the higher speed phones
2. I stated my rationale for bringing it up: my experience confirms that
3. I showed that I had two points to raise the issue.
If I'd had data to reference, I'd have used that in the first place. I'm still finding my way through the Android repository, so I don't always know the right terms for what I'm describing.
I will however, thank you for giving me that little bit to help me start searching on - OnDemand governor. This was an interesting starting point -
Android Rooted: setCPU and CPU Governance
It's been around for years, I've never been aware of it before:
The Ondemand Governor | LinuxInsight
You said: Plus, nothing about CPU frequency factors into garbage collection - and as that's so plain that even the word obvious fails, I'd like to know why you brought that up - I didn't, or least I didn't intend to. I did say that memory grew, but given that I know that some things do no garbage collection (to speak of), some do it quite well, and some do it well but in unusual ways (I'm looking at you, Tk data caching) - I presumed that the Java tools for Android were sufficient for the job, and that any garbage collection issues would be more a case of (really) poor app design for that to happen. My assumption is merely that some developers of some apps may well be kinda hogging and internally caching. In such a case, I would like plenty of free memory for them to do that, without bloatware impacting my system in any way.
You also said: Plus, nothing about CPU frequency factors into ... process time-slice (other than more run-time can be done for a given execution period). That's equally plain and again even the word obvious is too much. I'll take responsibility for using such unclear terminology that you must have been very frustrated to have posted that. You're describing process quanta.
However - that's not the issue. The OnDemand Governor simply scales CPU frequency based on load. I noted that from direct observation. I also noted that as frequency goes up so does power consumption.
It's not the OnDemand Governor I'm questioning - I'm questioning process management.
A quick (probably unnecessary for you) review of Linux process management is here:
Understanding the Linux Kernel: Chapter 10: Process Scheduling
Anatomy of Linux process management
However - from Android Code Day: gateway to a new kind of application - O'Reilly ONLamp Blog:
However, process management is lazy. A process that terminates is allowed to stay in memory in case the user restarts it in the near future, but the process can be purged to make way for new ones.
So, when are processes purged? Is it so inconceivable that unknown issues have crept in that impact Android's implementation of things that have a processor dependency not at issue in Linux?
I don't know and fully take responsibility for not being clear and would like to correct that:
1. I absolutely do not believe in task killers and that the Android processes should normally be trusted (that's at the top of my own thread on this).
2. That said, I encountered a forum statement that there may be issue with that on the newer high-speed processors. I dismissed it at the time, so I didn't check at that moment if I was looking at a rumor or at a qualified statement by a member of the dev community.
3. Later, after switching from my Moment to my Evo, I found myself dissatisfied. Even avoiding Sense, the bloat that didn't kill me on my Moment was giving me grief on my Evo. (Sprint) Anecdotally, I correlated that to battery usage and a target for improvement.
4. I researched available apps, and experimented with Startup Cleaner 2.0 and Application Task Cleaner Pro, and found a definite improvement in my battery and my overall enjoyment of responsiveness with the Evo.
5. This led me to conclude to try all reasonable strategies down that path. Unlike the Moment, I arrived at the conclusion that if I wanted to continue to respect the idea of not externally (from Android) process managing and get the performance I want, full root and bloat removal was my only option.
6. I invited others to see if this works for them.
I found nothing extreme in ATCPro (not to be confused with any other task manager) management of things, my phone's battery and responsiveness improved. I recommend it as a good first step or an alternative to those unwilling to root.
Now - I could be wrong in accepting that the Android process management is less effective because of the EVO's higher processor speed.
I am often wrong.
It could be that the EVO's larger memory is the issue. It could be a combination of factors.
But - there is an issue.
Detailed specs follow for fast reference:
HTC EVO 4G A9292 (HTC Supersonic) Specs | Technical Specifications | PDAdb.net - Comprehensive Database of PDA, PDA Phone, Smartphone, PNA & Mobile Device Specifications
Samsung SPH-M900 Moment Specs | Technical Specifications | PDAdb.net - Comprehensive Database of PDA, PDA Phone, Smartphone, PNA & Mobile Device Specifications
Upvote
0