DIY - KeK

[twobob]

On post https://www.mobileread.com/forums/sho...5&postcount=68

Originally Posted by knc1

** oops **

core2quad ~ $ cd /opt/arm-21012.03/bin

Should be cd /opt/arm-2012.03/bin

HTH.

(In the same section core2quad bin $ cd also looks suspiciously pointless)

[knc1]

cd without any arguments puts you back into ~
as the current directory.

The listing should show that the prompt changed from "bin" to "~"

= = = =

The two of us seem to be the only ones actually using (and reporting back) on any of this stuff.
Although the KeK server has shipped out hundreds of copies of the files.

[twobob]

Originally Posted by knc1

cd without any arguments puts you back into ~
as the current directory.

The listing should show that the prompt changed from "bin" to "~"

= = = =

The two of us seem to be the only ones actually using (and reporting back) on any of this stuff.
Although the KeK server has shipped out hundreds of copies of the files.

ah wasn't aware of that

Yes indeed people are reticent to spend time talking in the help of others, it's the human condition manifest in the microcosm of blogging. Or lack thereof. Sad but true.

Given the amount of people who actually search for assistance it is a shockingly low %age who would then go on to share the results for the next person. The 'Someone else will do it' or 'I'm alright Jack' attitude shall ever be fave in the lazying land of the manyana laissez-fair masses.

I just thank my lucky stars that occasionally SOMEONE takes the time to do it at all. In music if you are in it for the money or the recognition you should probably be in another profession (there are of course notable exceptions), perhaps it's similar in the sphere of didactics for dummies.

Either way, homage to those that do.

[geekmaster]

I have been studying this:
http://www.davespace.co.uk/arm/effic...arm/index.html

VERY enlightening. Some things coded in C can be horribly inefficient compiled with arm GCC, unless you do some interesting adjustments to the C code. Also check out the ARM assembler tutorial there.

No new code yet, but I plan to do a fancy audio/video player with touchscreen (and keyboard) GUI one of these days Real Soon Now.

[knc1]

Originally Posted by geekmaster

I have been studying this:
http://www.davespace.co.uk/arm/effic...arm/index.html

Which is quoting from the eighth re-printing of a 1953 book.

Originally Posted by geekmaster

VERY enlightening. Some things coded in C can be horribly inefficient compiled with arm GCC, unless you do some interesting adjustments to the C code. Also check out the ARM assembler tutorial there.

Even **if** that 2004 course mentions GCC as a specific example, GCC has change a lot in the past 8 years and many of the things no longer apply, or apply only to MIPS in-house compiler.

Originally Posted by geekmaster

No new code yet, but I plan to do a fancy audio/video player with touchscreen (and keyboard) GUI one of these days Real Soon Now.

Why even bother?

You don't publish re-usable, maintainable, code, just code that you can brag about how clever you where in eliminating whitespace.

That makes it a big waste of your very important time and a waste of time for anyone who tries to read it, learn from it, re-use it, or maintain it.

And please don't start a flame war over whitespace character counts, that is not the point.
The point is that computer programs are written in a Language, and Languages are intended to communicate among people.

Any computer program parser (other than a person) is very, very good at tossing out whitespace and comments. Those are only included when trying to communicate among people.
Which you clearly demonstrate is not the intent of your publications.

[geekmaster]

Originally Posted by knc1

Which is quoting from the eighth re-printing of a 1953 book.

Even **if** that 2004 course mentions GCC as a specific example, GCC has change a lot in the past 8 years and many of the things no longer apply, or apply only to MIPS in-house compiler.

Why even bother?
...
Any computer program parser (other than a person) is very, very good at tossing out whitespace and comments. Those are only included when trying to communicate among people. Which you clearly demonstrate is not the intent of your publications.

The course that you say is from 2004 contains updates that are only two weeks old. Also, the archive.org "Wayback Machine" has archives of that website from 2000 through 2008, which do NOT contain that course, so it was not added until after archive.org stopped archiving that site in 2009. It appears that you may be wrong about your "2004" date.

What "1953 book" are you referring to? That course is all about writing efficient C programs for ARM processors like we have in our kindles. The C language was developed in the early 1970's and the first ARM processor was developed in the mid 1980's.

Where did you get these 1953 and 2004 dates that seem to be irrelevant to the content of the links that I provided?

"Why bother?" The whole point of that tutorial is that you need to look at the assembler output of the compiler to see how well it conforms to the ARM architecture, and adjust your C code to generate efficient compiled code. Most of the GCC optimizations are x86-specific, and it is only RECENT versions of the GCC tools that do somewhat better with ARM instructions (especially with the FREE barrel shifting built into most instructions, and the ability to mix Thumb with full ARM instructions). Unfortunately, when working on kernel and driver stuff, you really need older cross tools that match your target system such as the K3 (or you have extra work finessing the headers and libraries), which means recent optimization improvements will be MISSING, requiring manual C code optimization as shown in that tutorial. It is a good read and worth learning anyway, just for its own sake. Learning is good for the brain.

Regarding coding style, I code for my OWN pleasure and my own benefit, which requires being able to view more code with less scrolling. I just choose to SHARE my code with others. As I said before, it is VERY easy to feed my "condensed" code into AStyle to convert it to WHATEVER your favorite style happens to be before reading it. AStyle is very good at ADDING all that whitespace that YOU need to be happy (but you may also want all my variable names replaced with camelCase phrases, and three lines of comments for every line of code too, before MY code has any chance of making YOU happy). Meh...

There is certainly no need for you to get so upset every time I share some of my own personal code with the MobileRead crowd. I do not plan to change how I write my code, and I do not plan to stop sharing my source code with the world just to make you happy.

Like I said before in response to your previous harsh criticisms of my coding style, just for a point of reference, where is YOUR original source code that you must be so proud of?

[knc1]

Originally Posted by knc1

Luckly, JRO (the author of auFS) and I correspond regularly.
- - - -
Besides the fact that he is the sole author, I know that he sells customized versions of the system to commercial machine manufacturers.

After an exchange of correspondence, JRO and I are clear on the subject of this "Undocumented use" of the auFS file system.

He also suggested a variation on the plan I have so far written up about using auFS as a filter.
Both variations have their specific use case(s).

JRO's variation would make building software targeted at: /usr/local/* (the common "add-on" tree in *nix) and then moving it to somewhere else (say: /mnt/us/*) quick and easy.

Hmm... After an exchange between the author and an end-user, I guess auFS filtering has left the "Undocumented use" catagory.

[knc1]

When building the client-side (ram kernel, Kindle-side) of the ATK source code package, you should be using a gcc cross-compile tool chain configured for building "Standalone" applications.

A "Standalone" application does not reference a system C library.
A "Hosted" application does reference a system C library.

Both tool-chains (for either Linux or Windows) are indexed in the KeK Reference Manual, at:
http://knetconnect.com/KeK/KeK_refer...bsection-B.1.2

You probably want both the -56 and the -57 builds.

When you unpack them, they will both create an "arm-2012.03" parent directory . . .

Although it should be possible to unpack both of them into that same file system tree, to ease future maintenance, I put them each into their own file system tree.

Unpack the -57 archive at the desired location on your system (typically /opt), then rename the parent directory: arm-2012.03-glibc

Then unpack the -56 archive at the desired location on your system (typically /opt), then rename the parent directory:
arm-2012.03-elf

A couple of notes for those building the ATK source under M$-Windows . . .

The ATK project files and make files assume a cygwin environment, but the above tool chains are "Windows Native" tool chains.
You can drop all references to cygwin in the Freescale project.

You can also drop all references to the cygwin tool chain components (such as: objcopy), because you just installed a much more recent "Windows Native" build of binutils.

The ATK source code bundle, as shipped, was using a gcc-4.1.1 based tool chain. You just installed a gcc-4.6 based tool chain.
GCC-4.6 is much more strict about the source code than GCC-4.1, meaning you may find more errors and warnings during the build which will require tweaking (correcting) the Freescale sources.

Enjoy.

[knc1]

A quick note -
The files in the "bin" directory -
objcopy.exe - that is a binutils program, linked against the bfd - both are GPL licensed.
The cygwin binaries are GPL licensed.

Yet Freescale does not include those sources, duh . . . .

Anyone want to bet that Freescale will be much quicker to defend their licensed code than they are to observe the licenses of the code they ship?

Edit:
I opened a support ticket with Freescale over the license question(s).
Will be interesting to see what (if any) response I get.

[knc1]

Originally Posted by knc1

A quick note -
The files in the "bin" directory -
objcopy.exe - that is a binutils program, linked against the bfd - both are GPL licensed.
The cygwin binaries are GPL licensed.

Yet Freescale does not include those sources, duh . . . .

Anyone want to bet that Freescale will be much quicker to defend their licensed code than they are to observe the licenses of the code they ship?

Edit:
I opened a support ticket with Freescale over the license question(s).
Will be interesting to see what (if any) response I get.

Shame on me for talking down about Freescale and license compliance . . .

Less than twelve hours elapsed between submitting the message to the "license support" desk and their replying that they have passed the information onto the (atk) development group for action.

Somebody, other than FSF and Dr. Stallman, takes the GPL license seriously.

[twobob]

Originally Posted by knc1

Shame on me for talking down about Freescale and license compliance . . .

Less than twelve hours elapsed between submitting the message to the "license support" desk and their replying that they have passed the information onto the (atk) development group for action.

Somebody, other than FSF and Dr. Stallman, takes the GPL license seriously.

I missed this.

from me. Now where is that pesky emulator... Searches top post. mumble mumble

[knc1]

Originally Posted by twobob

I missed this.

from me. Now where is that pesky emulator... Searches top post. mumble mumble

tool prefix index will help. I think I put anything that counts in there.

[twobob]

Originally Posted by twobob

Tomorrow - with a following wind - I may actually get a chance to do an end to end - 32 bit. Thanks again.

That was a very hopeful statement. ey : )

ah well perhaps tomorrow for real this time.
must - understand - emulator

need - more - tea...

[twobob]

Originally Posted by knc1

tool prefix index will help. I think I put anything that counts in there.

I'm officially dumb as sh.. I either don't see it, or don't know what I'm looking for. Or both.

\0/ Ultimate FAIL

[knc1]

Originally Posted by twobob

I'm officially dumb as sh.. I either don't see it, or don't know what I'm looking for. Or both.

\0/ Ultimate FAIL

https://www.mobileread.com/forums/sho...1&postcount=73

The first "wget" command fetches the pre-built copy of the emulated system.

I now see that post has two topics, the reference to the newly built qemu and directions on downloading the emulated system.

I fixed up the description in the tools index to be more clear.
https://wiki.mobileread.com/wiki/Tool...mulated_Native