Hi all,

I have currently kernel 3.8 that I am using for beaglebone black. I see that there are a lot of files that beaglebone doesn't use and funtions with same names, that makes it difficult to search.
Is there a way to get rid of all this unnecessary files?

If you build your own custom kernel on the device there's make localmodconfig. Which will create a .config with only the modules for the hardware you have. Caveat, and only the filesystems currently available on attached storage things.

The short answer is "No".

The longer answer is that the kernel build procedure only builds the ones called for by the configuration. The others are/will be checked to see if they are needed.

Read this first, there are related helpful links in it, and this one too. There are many sources and guides available in the internet.

Good luck.

Just remember - once you start deleting things from the kernel distribution you then responsible for all maintenance afterward. You won't be able to add patches very well afterward.

Most people consider this prohibitive.

Can you post some of the examples of names you wish to remove?

There are many config, .c and .h files for other architecture such as mips, powerpc. If I am building for BBB, I don't want to see them in my kernel workspace because many functions have same name in those files. What I wish to do this is I don't have any hardware debugger or any sort of method to step in the code so I have to find the files manually and guess how booting is happening and document it.

So the scenario I am looking for is
write the configuration and parse all the useful files for BBB to another directory with this configuration and then build. So I can have one untouched kernel and one another directory named as kernel-BBB inside(or outside) it.

If there is a solution in some scripting language, I just need a small example. Rest I can figure it out.

If you are using find to locate files it is relatively simple to exclude... Just exclude the architcture name.

Removing the configuration files is a good bit more complex as you have to remove the references from files that refer to them...

As I said, removing files is a complex operation.

The files in the /arch directory apply to different processor architectures. The names are identical because they are "the same files" for different types of CPUs. Only one of them applies to your system.

As my Spanish friend would say: "¡No joda!"

Don't mess around with the content or the arrangement of the kernel source files. Yeah, there are a whole lot of 'em, delicately arranged.

Ok.
So its a complex task basically and same files may be used at different architectures using #define at build time.
I can close this discussion.
thanks to all.

Exactly: there are a set of "very(!) low-level Linux modules" which are written more-or-less in architecture-specific assembly code," meant to be processed by an also architecture-specific compiler. (These modules are, so to speak, "where the rubber meets the road.") One, and only one, of these alternatives will actually be used when compiling Linux for a ("your ...") particular architecture.

BTW: I happen to find these modules quite fascinating. They are the secret of why "Linux" is now available for over twenty(!) entirely-different architectures ... "and yet, it works the same."

That is, if I may say, "a truly astonishing(!) Software Engineering Achievement!" But ... such is Linux.

As an extension to that is the GNU C compiler - which is used on all 20+ architectures... and can be used for the purpose of cross compiling from one architecture targeting another.