>>> Daniel Kiper <daniel.ki...@oracle.com> 07/21/15 7:24 PM >>>
>On Tue, Jul 21, 2015 at 03:35:07AM -0600, Jan Beulich wrote:
>> >>> On 20.07.15 at 16:28, <daniel.ki...@oracle.com> wrote:
>>
>> Without any explanation (description) I'm inclined to say this makes
>> things more fragile instead of improving the situation. As it looks
>> like we indeed pointlessly copy .eh_frame, but I think this would
>> better be avoided by suppressing its generation (i.e. add
>> -fno-asynchronous-unwind-tables just like Rules.mk has).
>
>Make sense, however, there is still place for two small optimizations.
>
>First of all ld generates .got.plt section and objcopy copy it to binary file.
>It is not needed because we do not link our stuff here with shared libraries.
>So, we can use -R objcopy option to remove it (if you do not like -j .text).
>This way we could save 15 bytes (at least on my machines).

.got.plt shouldn't be generated in the first place (and I don't recall having
seen one here - I'll double check once in the office), i.e. perhaps another
missing compiler option or a linker quirk?

>We could also save another 3 bytes (per one xen/arch/x86/boot C input file)
>in final Xen binary in worst case :-))). We just need generate output assembly
>files as string of .byte instead of .long.

If you feel it worth your time, go ahead.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Reply via email to