Re: Revised PPC assembly implementation

From: Paul Mackerras <paulus@samba.org>
Date: Apr 27, 2005, 03:39 UTC
Message-ID: <17007.2390.258823.189255@cargo.ozlabs.ibm.com>
In-Reply-To: <20050427014712.13552.qmail@science.horizon.com>

linux@horizon.com writes:

Show 5 quoted lines

> Here's a massively revised version, scheduled very close to optimally for
> the G4.  (The main remaining limitation is the loading of the k value
> in %r5, which could be split up more.)
> 
> My hope is that the G5 will do decently on it as well.

Nice... your new version takes 4.413 seconds on my G5 for 1000MB, compared to 4.606 for your old version, i.e. it's about 4.4% faster. Unfortunately it gives the wrong answer, though.

On my powerbook, which has a 1.5GHz G4 (7447A), the same test takes 4.68 seconds with my version, 4.72 seconds with your old version, but only 3.90 seconds with your new version.

Care to check the code and find out why it's giving the wrong answer?

Regards, Paul.