From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <3669C883.E34637EE@jlc.net> Date: Sat, 05 Dec 1998 18:57:55 -0500 From: Dan Malek MIME-Version: 1.0 To: Gabriel Paubert CC: Benjamin Herrenschmidt , linuxppc-dev@lists.linuxppc.org, Cort Dougan Subject: Re: Mapping frame buffer using BAT References: Content-Type: text/plain; charset=us-ascii Sender: owner-linuxppc-dev@lists.linuxppc.org List-Id: Gabriel Paubert wrote: > Not on PPC (at least 6xx/7xx), remember that the TLB are not.... You missed my points, so let me try again a little more specifically. First, the MMU requires system resources to support, whether it is processor registers, memory for page tables, bus cycles to load translations, and processor instruction cycles to maintain all of that. This thread started by suggesting a modification that reduced this system support. My request was to simply consider all PPC cores and a longer term better solution. Second, a later discussion indiciated it was easy to dismiss a particular enhancement. My point was that you need to evaluate that enchancement using more than one experiment, perhaps trying to find one that specifically causes performance challenges, and that it may be difficult to quantify. > ...by adding code (which represent additional cache lines to fetch). Just make sure you don't load them later, at a higher cost. If the cache doesn't provide a benefit, such as additional hits or improved bus timing, then perhaps it should be disabled for those parts of the system. -- Dan [[ This message was sent via the linuxppc-dev mailing list. Replies are ]] [[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]] [[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]] [[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]