Now there are a lot of issues across multiple cores, so I decided to write a post dedicated cores, so to speak from a to z.1.Nachnom with the fact that such a core.
2. how to stitch
Kernel, kernel - the operating system kernel is in the form of * .tar and / or * .md5 for odin, as well as in the form of zip for Recovery. Affects the work of the entire apparatus in the most that is the fullest!
This core-operating system component that performs important functions such as memory management, process, interaction between processes, access to devices. All this is happening at a low level.
3. How to choose and compatible cores
Can sew through odin or Recovery (morecore firmware
Before insertion of any desirable to perform the nucleus ( "set") script purification
An exemplary description of nuclei with personal experience.
In the description of the kernel is usually indicated for any firmware is used. Each core is endowed with its own advantages and disadvantages and the choice of the kernel depends on what you want to get a single core for games and one for CPU overclocking, and the third to "stable" performance and chu bit more economical in power consumption. All kernelone Firmware interchangeably. I was unable to get to a certain firmware for installation on it of different nuclei are not office phone, most importantly do everything according to instructionsI used to sew through one core.
4.IMAGE EDITING nucleus.
1. Speedmod simple yet stable kernel. The latest version added a lot of interesting things, which was not in the previous. There are complaints that the latest version has increased power consumption, but the versions come out often so I think I have nothing wrong with it.
2.CF-Root as root with the standard kernel, and small changes.
3.GTO for fans of games in the latest versions removed support for video codecs and all of the power given to 3D! But in my experience a battery consumption more than other nuclei.
4.Darky very good stable core for everyday use, and with good power. Version 2.7.3 for the JVP, Version 2.8 for JVQNO CHANGES!
Now there is already a newer version, but the quality and stability of the cores are the same.
5.Talon now stands I have this nucleus. Of all chose it with acceleration up to 1200, and as a tuning voltage at 1000 (see below).For me this is the most stable and bright enough core perfect for my needs.
6.galaxian so often updated version of the kernel that does not have time to write the pros and cons. New versions every day so I advise to put only fans of the experiments, as every day to update the kernel and look for mistakes in him not everyone wants to. We are waiting for the final version ...
I tried to describe the most common core, well, as you can imagine they are developing rapidly and go out often new versions so everything about each core, I can not write. Who has any opinion on the cores send me a PM, I'll edit the post and add something or remove.
Setting one of the core described SW..//Hack Example of setting card core I will write my experience, so to speak, can someone come in handy. Part one
When you install the core system is usually converted to ext4, you can check it out in SMW. When we go there once or write any file system or check in Voodoo, it should be written lagfix enabled.
I have to configure the kernel Voltage Control Utility enough. Now many kernels support this function. You can see in the description of the core should be a lineOC / UV support.
In this program, there is such a setting as the I / O Scheduler will tell you not to bother if there is a placesio or cfq
for fans of the information description under a spoiler below.
Part two settings Continued
Complete Fair Queuing (CFQ) I / O Scheduler
The CFQ Planner uses a different approach to achieve the same goals. In cfq everyone
The process is assigned its own queue, and a time slice is assigned to each queue.
(timeslice). The I / O Scheduler rounds each queue and serves requests.
from the queue until the time limit is reached (timeslice) or remains
requests in this queue. In the latter case, the CFQ scheduler will wait by default.
10-ms, new request from the queue. If the wait was in vain, then the scheduler goes to
the next turn.
Within each process queue, synchronized requests (such as reading)
take precedence over unsynchronized requests. Thus, CFQ promotes reading
and prevents the problem of writes-starving-reads.
The CFQ scheduler works well for most tasks, making it a great first.
Noop I / O Scheduler
The NOOP Scheduler is the most basic scheduler available. He does not perform what
sorting, only major mergers. It is used for specialized devices,
which do not require sorting their queries.
Deadline I / O Scheduler
Deadline I / O Scheduler stores a queue sorted (as described above, comment), and
introduces two additional queues: a FIFO queue for reading and a FIFO queue for writing. Records
in each of these queues are sorted by arrival time (in fact, the first one entered is
first came out). Each request in the FIFO queue is assigned an end time. For the queue
read requests are 500 milliseconds. For a write request queue, this is five seconds. With
receipt of a new I / O request, it is inserted-sorted into the standard queue and
placed at the end of the corresponding (read or write) FIFO queue.
As a rule, I / O requests from the standard head are sent to the hard disk.
sorted queue. This maximizes overall throughput while minimizing
search operations and installation of heads on the disk, since the normal queue is sorted by
block number (as with Linus Elevator).
When an entry at the beginning of the list of one of the additional FIFO queues has an assigned
time, the I / O scheduler stops processing I / O requests from the standard queue, and
starts servicing requests from this FIFO queue. I / O scheduler checks and processes
requests only from the head of the queue where the oldest requests are located.
Thus, Deadline I / O Scheduler supports efficient overall throughput.
without fasting any single request is unacceptable for a long time. Problem
writes-starving-reads is reduced to a minimum.
Simple I / O scheduler (SIO)
SIO is based on the deadline, but it looks more like a mix
between noop and deadline. In other words, SIO looks like a lightweight deadline version, but it
does not perform sorting, so it is designed mainly for random access
to disk (such as SSD hard drives), where a sorting request is not necessary (since
any sector can be available at any time, regardless of its
The settingCPU Governor specify what you want from your phone. I should Mode conservative . Battery mode with long lives, and brake, I never noticed (as quadrant numbers less only). Standard mode is Ondemand
Description regimes under spoiler below
Setting the core voltage
This information was provided by fears. Its veracity and accuracy is not guaranteed.
* Conservative - works at the lowest possible CPU frequency and increases step by step, if necessary.
* Userspace - allows you to change the frequency of the CPU, either in manual mode or dynamically, at the discretion of the software.
* Powersave - lowers the frequency of the CPU to the lowest possible level.
* Ondemand - runs at the lowest possible CPU frequency and increases to the maximum when the CPU load reaches 100%.
* Performance - sets the maximum frequency of the CPU, prohibiting changing the frequency value downwards.
I attach screenshots Program
it is necessary to produce as it is written herecore voltage setting
P.S.vse your wishes opinions and criticism about my post in a PM asking. Prompt what to add or remove. I will be glad for any help. If you are experienced and made comparisons cores that also please write I will lay a small statistics and results. All that is written above is my personal experience and my opinion, so it might be a mistake and someone not happy, I will be glad to criticism.Post has been editedhutor - 26.07.11, 18:02
Reason for edit: corrected description of nuclei