[rfc] improving 32bit user performance/experience...

[Daniel J Blueman]

On Wed, May 20, 2009 at 12:00 PM, John McCabe-Dansted <gmatht at gmail.com> wrote:
> On Wed, May 20, 2009 at 1:55 AM, Daniel J Blueman
> <daniel.blueman at gmail.com> wrote:
>> I was trying to raise a more general point about the minimum spec
>> across the board, including the embedded and old-server hardware.
>>
>> I challenge anyone to find someone using Ubuntu 8.10/9.04 on a
>> processor which doesn't support the full i586 instruction set (eg
>> i386/i486 or something with incomplete i586 support).
>
> This would be difficult since Ubuntu has dropped support for i486 ages ago:
> https://lists.ubuntu.com/archives/ubuntu-devel-discuss/2008-September/005385.html

It's true that the kernel is targeted at the i586-compatible
instruction set and tuning, thus won't boot on anything less. At that
point, there is nothing preventing using -march=i586 and -mtune=i686
for the other packages.

[snip]
> It would appear that Ubuntu has preemptively implemented your suggestion :)

Once the kernel has faulted pages in, much of the speed is down to
libraries and application efficiency, which is where the opportunity
exists (glibc has internal i5/686 optimisations, selected at runtime)
- so only part of the solution is there today.

Eg, image processing/coding, data movement, video coding benefit from
additional MMX instructions and registers, and everything benefits
from scheduling for the deeper i686 pipeline.

I'll open a LP page with the proposal and see what how this can be
integrated neatly.

Daniel
-- 
Daniel J Blueman