On 2002.03.03 13:48 Keith Whitwell wrote: > ... > > The advantages of doing it this way: > - unmapping rather than copying the data is probably faster. > - no need to reformat data in kernel space. > > The disadvantages: > - you have to re-write '_vbtmp.h' and maybe '_tritmp.h", and keep > these uptodate with any bugfixes > - the driver looks less like the other drivers and it is harder > to incoperate new optimizations. > - unmapping is more complex than copying. > > It's probably a matter of personal choice... > Keith >
Since the advantages in getting DMA to work are substancially bigger than those added by what I proposed, I'll choose to give one step at at time: 1. finish the update of the mach64 driver to Mesa 4.x without any changes in the templates 2. make a mach64 emulate a d3d-like device as Keith suggested, and emulate the DMA engine with PIO as well 3. change the DMA emulation to a real DMA engine 4. decide wether is viable/worthy to eliminate d3d-like vertex translation by changing the templates. I think this way we can achive better know-how and even get to the final result faster than attempt the 4th step right now. Anyway, if this is a matter of personal choice, I would like that the choice wouldn't be mine alone, so I would like to hear more people's opinion about this. Regards, José Fonseca _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com _______________________________________________ Dri-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/dri-devel