On Tue, Feb 2, 2010 at 11:48 AM, Arthur HUILLET <arthur.huil...@free.fr> wrote: > Hi, > > On Tue, Feb 02, 2010 at 09:45:32AM +0200, Pauli Nieminen wrote: >> I bisected the rendering problem to commit adding xmm0 zeroing code to >> transform functions. > > Unfortunately, xmm0 happens to be undefined at least sometimes, for > example on one SDL + OpenGL game I tried. It made the game crash in a > reproductible manner (all the time). > > Please refer to my original comment for details about why I believe my > change is correct. If it so happens that the upper qword of %xmm0 does > sometimes contain fog factor, then it probably is because of luck rather > than a voluntary thing, because you cannot rely on %xmm0 keeping its > value between two calls to the GL driver. > > As a result, reverting my patch is *not* the proper way to fix the bug > you are encountering. Properly loading %xmm0 with the value you want to > see in there is the correct solution. >
Is some application leaving NaN to xmm0 which is causing the crash? What application was causing your crash? Because undefined value in xmm0 should cause any problems. We are anyway loading known value to it. Also all math later are only operating in the part where we have the known value. Pauli ------------------------------------------------------------------------------ The Planet: dedicated and managed hosting, cloud storage, colocation Stay online with enterprise data centers and the best network in the business Choose flexible plans and management services without long-term contracts Personal 24x7 support from experience hosting pros just a phone call away. http://p.sf.net/sfu/theplanet-com _______________________________________________ Mesa3d-dev mailing list Mesa3d-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
