Re: Small PIC Fixes

["Vladimir 'φ-coder/phcoder' Serbinenko" <phcoder@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1458 bytes --]

On 01/05/2011 06:58 AM, ehem+grub@m5p.com wrote:
> Perhaps not the biggest deal, but I do like to get low-hanging fixes out
> of the way if they appear.
>
> One very significant item I found. It appears GCC is fine with %rbx being
> clobbered when building PIC in 64-bit mode, even though it has problems
> with %ebx being clobbered when building PIC in 32-bit mode.
>
>   
This patch only increases the number of possible ways preprocessor
defines can be resolved, which in turn increases the maintenance cost.
Restoring %ebx/%rbx unconditionally is cheap. Maintaining exponentially
growing number of the way #if's can be resolved isn't.
I don't see any problem to enduser with these 2 small instructions
always being there.
> One other item I did notice. Are there really any processors in the amd64
> class that *don't* support CPUID? I'd like to hardcode
> grub_cpu_is_cpuid_supported() to return 1 if __x86_64__ is defined, but
> I'm a tad worried I'll be unpleasantly surprised.
>
>   
Similar problems. Maintaining something that is always the same is
easier than something with loads of #if's.
Rule of thumb is: "if it works and your improvement isn't visible to any
enduser, don't touch it".
>   
>
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/grub-devel
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]