On Wed, 12 Jun 2002, David S. Miller wrote: > Basically, someone has asked for me to default AGPMode to 4 in > our config file, which I thought was absolutely crazy. The claim > is that AGP defaulted to 1x, but changing it to 4x sped things up > dramatically. That is what lead me to believe that the "default" > being used is not the BIOS setting. > >That's almost as stupid as the current situation which defaults >to 1X AGP.
I agree, defaulting to AGP 1x right now is just not the best thing to do anymore. All current generation hardware and even the last gen or two can do AGP 2x. >Another known failure case (when user sets BIOS to 4X and X tries >to use it's default of 1X) is ALI M1647 chipsets. That also hangs. Right. >All of this points to using the BIOS supplied default as being the >logical way to go. Is this information already available inside the X server? xf86GetAGPInfo() appears it might be what I'm looking for here. info->agp_mode Is this the right bits? >Until we can be confident that every single AGP chipset is %100 >done and has %100 of the workarounds, I'd highly suggest against >trying to switch AGP modes by default. I agree. How about this scenario then: Any default is going to have certain failure cases. Lets make the default AGP mode the X server uses be the AGP mode that the BIOS has set. Whatever failure cases show up, we can blacklist and/or whitelist known good ones, etc. This makes autodetection and autoconfiguration all the simpler. Doing this with user initiated config tools IMHO is just a bad idea. I also think doing the BIOS AGPmode autodetect in a config tool, and setting AGPMode inside the X config file is bad because it assumes too much, and hard codes things at config time rather than autodetecting them at runtime. Change motherboards, you don't want problems to pop up if it can be avoided for example. >To my knowledge only the ServerWorks AGP bits are a) written by >the vendor b) have all the necessary errata workarounds. ServerWorks AGP was written by VA under contract with ServerWorks IIRC. I've got a Tyan 2567 motherboard with the ServerWorks HEsl on it personally. Unfortunately it only works in AGP 1x mode. AGP 2x is fried (or was last time I tried. Chipset/BIOS/implementation problems IMHO, not X bugs. Kimball Brown at ServerWorks was very helpful to me in getting AGP working at all on this thing though despite the faulty motherboard design and broken BIOS. It'd be nice to pull the needed info out of other vendors also. -- Mike A. Harris Shipping/mailing address: OS Systems Engineer 190 Pittsburgh Ave., Sault Ste. Marie, XFree86 maintainer Ontario, Canada, P6C 5B3 Red Hat Inc. http://www.redhat.com ftp://people.redhat.com/mharris _______________________________________________________________ Don't miss the 2002 Sprint PCS Application Developer's Conference August 25-28 in Las Vegas - http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
