-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Timothy Miller wrote: > On 8/25/06, Hamish Marson <[EMAIL PROTECTED]> wrote: > >> A linked list is definately better. Although it requires more >> programming around it... And would also be prone to application >> errors causing the graphics engine to go wandering off into >> never/never land executing garbage... That could probably be >> alleiviated somewhat if the graphics engine is made to reset to a >> known good state if it tries to execute an illegal instruction... >> (Rather than the old 6502 method of executing it anyway, More >> like the 68k method of raising an interrupt for illegal >> instruction). > > Usually, the GPU would be accessing a ring buffer linearly. Since > it's linearly, bursts on the bus are possible. If you have a > linked list, then DMA has to jump around, wasting bus cycles to > issue new addresses. Linked lists have their advantages, but > they're a major performance hit. I was thinking linked lists of whole blocks. Sized in multiples of DMZ burst lengths (Sorry, wasn't sure how big that was). I was assuming DMA had to be setup for each burst anyway...
H -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFE7zgD/3QXwQQkZYwRAmdZAJ9nAgmOwkbbpSIisZryhWC7BQm/UACgiyxD +z1/ZymyOgMfEBdcxxqxaqo= =p5s1 -----END PGP SIGNATURE----- _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
