[sldev] RFC: Vectorisation control patch

[Paul TBBle Hampson]

On Wed, Aug 15, 2007 at 08:48:06AM -0700, Dzonatas wrote:
> Paul TBBle Hampson wrote:
> >>We have discovered the results are too random.

> >It wasn't random at all. It was very consistent, over several runs.

> That depends on which machine you run it on. On mine, the results are
> the same every time. On a Xeon, the results varied every time. Two
> similar HT systems were also tested, and the results were almost the
> same every time on one but the other showed totally different results.
> There were more tested. Looking at it from all the different systems
> and how these results overall are not consistent across the board, it
> is too random.

Not a single one of those is an Altivec machine, which is what I was
talking about.

Apologies, I may have misplaced my context earlier in the thread.

Either way, I think it's a good vote in support of trying all the
relevant hand-tuned vectorisations, the gcc autovectorisation, and the
non-vectorised code at each run. I'm not totally clear on how expensive
that is, but I presume not very, given that to be an optimisation target
this routine would have to be called an awful lot anyway.

-- 
-----------------------------------------------------------
Paul "TBBle" Hampson, B.Sc, LPI, MCSE
Very-later-year Asian Studies student, ANU
The Boss, Bubblesworth Pty Ltd (ABN: 51 095 284 361)
Paul.Hampson at Pobox.com

Of course Pacman didn't influence us as kids. If it did,
we'd be running around in darkened rooms, popping pills and
listening to repetitive music.
 -- Kristian Wilson, Nintendo, Inc, 1989

License: http://creativecommons.org/licenses/by/2.1/au/
-----------------------------------------------------------
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.secondlife.com/pipermail/sldev/attachments/20070817/402ec881/attachment.pgp