[PATCH] PPC64 iSeries: cleanup viopath

Linas Vepstas linas at austin.ibm.com
Wed Mar 16 04:43:10 EST 2005


On Wed, Mar 16, 2005 at 02:53:39AM +1100, Stephen Rothwell was heard to remark:
> On Tue, 15 Mar 2005 08:32:27 -0600 Hollis Blanchard <hollis at penguinppc.org> wrote:
> > 
> > Why not use a byte instead of a full int (reordering the members for 
> > alignment)?
> 
> Because "classical" boleans are ints.
> 
> Because I don't know the relative speed of accessing single byte variables.
> 
> Because it was easy.
> 
> Because we only allocate 32 of these structures.  Changing them really
> only adds four bytes per structure.  I guess using bytes and rearranging
> the structure could actually save 4 bytes per structure.

FWIW, keep in mind that a cache miss due to large structures not fitting
is a zillion times more expensive than byte-aligning in the cpu 
(even if byte operands had a cpu perf overhead, which I don't think 
they do on ppc).

> It really makes little difference, 

Yep. So my apologies for making you read this email.

--linas




More information about the Linuxppc64-dev mailing list