On Fri, 2004-10-15 at 14:55, Michel D�nzer wrote: > On Thu, 2004-10-14 at 12:29 +0200, Thomas Hellstr�m wrote: > > > > is DRM_WRITEMEMORYBARRIER() the right way to make sure data has been > > flushed to AGP memory before firing it off to the DMA engine? > > I think so, assuming DRM_WRITEMEMORYBARRIER boils down to an instruction > with the LOCK prefix. This is pretty well documented in Intel's IA-32 > manuals, which you can download from their site, e.g. from > > http://www.intel.com/design/pentium4/manuals/index_new.htm
On Radeon the WRITEMEMORYBARRIER wasn't sufficient, and we're doing a read from the register before the commit of the ring write pointer, supposedly due to some PCI write posting interaction that caused hangs. Might be related to this issue, might not, or it might have been some other issue and that extra read isn't necessary any more. Just throwing this out there. -- Eric Anholt [EMAIL PROTECTED] http://people.freebsd.org/~anholt/ [EMAIL PROTECTED] ------------------------------------------------------- This SF.net email is sponsored by: IT Product Guide on ITManagersJournal Use IT products in your business? Tell us what you think of them. Give us Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more http://productguide.itmanagersjournal.com/guidepromo.tmpl -- _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel
