Prev: kernel.h panic cleanup
Next: [resend] unusual_devs.h: Fix capacity for SL11R-IDE 2.6c
From: Dave Young on 26 Feb 2010 05:00
On Fri, Feb 26, 2010 at 3:42 PM, Joe Perches <joe(a)perches.com> wrote:
> On Fri, 2010-02-26 at 15:33 +0800, Dave Young wrote:
>> It will be better to move some of them to proper place instead of kernel.h
>
> Perhaps you could describe some of the reasons why
> it's better to move them.
>
> Maybe you could post the results of a timed compile
> comparison with kernel.h broken into multiple pieces
> while you're at it.
>

Thanks joe.

Spliting is necessary from my point of view.
It will be more clean to put externs/declarations to their own header files.
At the same time compiling time benifits as well.

printk/panic/string/trace header are splited then included by kernel.h so
just hexdump/int_sqrt are actually seperated.

fresh boot and same .config, make without -j, test result as following:

with the patches applied:

time:
real 15m16.504s
user 12m51.866s
sys 1m39.700s

vmlinux .text size: 003bfcd2

without the patches applied*
time:
real 15m20.518s
user 12m54.120s
sys 1m40.190s

vmlinux .text size: 003bfcea

--
Regards
dave
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Alexey Dobriyan on 26 Feb 2010 05:00
On Fri, Feb 26, 2010 at 11:55 AM, Dave Young <hidave.darkstar(a)gmail.com> wrote:
> It will be more clean to put externs/declarations to their own header files.
> At the same time compiling time benifits as well.
>
> printk/panic/string/trace header are splited then included by kernel.h so
> just hexdump/int_sqrt are actually seperated.
>
> fresh boot

You mean cold-cache compile so results are basically random?

> and same .config, make without -j, test result as following:
>
> with the patches applied:
>
> time:
> real 15m16.504s
> user 12m51.866s
> sys 1m39.700s
>
> vmlinux .text size: 003bfcd2
>
> without the patches applied*
> time:
> real 15m20.518s
> user 12m54.120s
> sys 1m40.190s

Not scientific at all.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Dave Young on 26 Feb 2010 06:00
On Fri, Feb 26, 2010 at 5:58 PM, Alexey Dobriyan <adobriyan(a)gmail.com> wrote:
> On Fri, Feb 26, 2010 at 11:55 AM, Dave Young <hidave.darkstar(a)gmail.com> wrote:
>> It will be more clean to put externs/declarations to their own header files.
>> At the same time compiling time benifits as well.
>>
>> printk/panic/string/trace header are splited then included by kernel.h so
>> just hexdump/int_sqrt are actually seperated.
>>
>> fresh boot
>
> You mean cold-cache compile so results are basically random?

Yes, roughly test.
I have to say I will have no time to do more testing.

>
>> and same .config, make without -j, test result as following:
>>
>> with the patches applied:
>>
>> time:
>> real 15m16.504s
>> user 12m51.866s
>> sys 1m39.700s
>>
>> vmlinux .text size: 003bfcd2
>>
>> without the patches applied*
>> time:
>> real 15m20.518s
>> user 12m54.120s
>> sys 1m40.190s
>
> Not scientific at all.
>



--
Regards
dave
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Steven Rostedt on 26 Feb 2010 09:30
On Fri, 2010-02-26 at 17:55 +0800, Dave Young wrote:

> Spliting is necessary from my point of view.

> with the patches applied:
>
> time:
> real 15m16.504s
> user 12m51.866s
> sys 1m39.700s
>
> vmlinux .text size: 003bfcd2
>
> without the patches applied*
> time:
> real 15m20.518s
> user 12m54.120s
> sys 1m40.190s
>
> vmlinux .text size: 003bfcea

Why the change in text size? If moving stuff around in headers affects
the final result, I'm thinking that you did something wrong.

The merge for x86 did a lot of file manipulation, but Thomas and Ingo
would always test that the binary result was the same. If the binary
result is different, then something went wrong with the file
manipulations.

-- Steve



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Dave Young on 27 Feb 2010 05:00
On Fri, Feb 26, 2010 at 10:27 PM, Steven Rostedt <rostedt(a)goodmis.org> wrote:
> On Fri, 2010-02-26 at 17:55 +0800, Dave Young wrote:
>
>> Spliting is necessary from my point of view.
>
>> with the patches applied:
>>
>> time:
>> real 15m16.504s
>> user 12m51.866s
>> sys 1m39.700s
>>
>> vmlinux .text size: 003bfcd2
>>
>> without the patches applied*
>> time:
>> real 15m20.518s
>> user 12m54.120s
>> sys 1m40.190s
>>
>> vmlinux .text size: 003bfcea
>
> Why the change in text size? If moving stuff around in headers affects
> the final result, I'm thinking that you did something wrong.

trace/printk/panic/string stuff just being moved to another head file,
then included
in kernel.h, so there shouldn't be difference with before.

So I guess it's the hexdump/int_sqrt head file which are removed from kernel.h?
>
> The merge for x86 did a lot of file manipulation, but Thomas and Ingo
> would always test that the binary result was the same. If the binary
> result is different, then something went wrong with the file
> manipulations.
>
> -- Steve
>
>
>
>



--
Regards
dave
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: kernel.h panic cleanup
Next: [resend] unusual_devs.h: Fix capacity for SL11R-IDE 2.6c