[Nottingham] Compiler optimisation flags (gcc, g++)

[Martin]

Jon Masters wrote:
> Martin wrote:
> 
>> -falign-functions=4 -preferred-stack-boundary=4 
> 
> 
> These might take some emperical measurement to verify since you might 
> find aligning on higher boundaries better fits CPU cachelines.

Supposedly, aligning to 2^4 bytes gives the AthlonXP cache an easy time.


> This is all theorectical stuff because in reality you are unlikely to 
> notice a huge difference between the flags given and a simple march.

All in the hope of gaining a few % performance boost before the next 
upgrade. Opterons next when they mature to the 2nd revision.


>> Are programmers silly enough to use computed goto or other sillies 
>> that will cause problems with the above optimisations?
> 
> 
> I cannot recall a situation in which I have seen it used other than in 
> the google search I did for it before replying :-).

Then, all looks hopeful!


The first big test is when I get round to kernel compiling...


All good fun,

Cheers,
Martin

-- 
----------------
Martin Lomas
martin at ml1.co.uk
----------------