Kenneth Graunke <kenn...@whitecape.org> writes: > On 10/02/2012 07:52 PM, Eric Anholt wrote: >> Based on split_virtual_grfs(), we choose the same set every time, so set it >> in >> stone. This will help us avoid regenerating the somewhat expensive >> class/register set setup every compile. >> --- >> src/mesa/drivers/dri/i965/brw_fs.h | 1 + >> src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp | 101 >> +++++++++------------ >> 2 files changed, 42 insertions(+), 60 deletions(-) > > Hmm. I guess we don't have SEND from GRFs expressed at the visitor > level...only a couple of instances at the generate_code level. And > those don't use more than 4 payload registers. > > In that case, this seems reasonable.
Yeah, if we need bigger vgrfs (like the branch I have laying around), then I think we'll just statically add those other sizes. They're small numbers.
pgpQ94MET8vDU.pgp
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
