OSDN Git Service

tnl: Optimize SSE load[23]f_1 since they don't need the identity swizzle.
authorGuillaume Melquiond <guillaume.melquiond@gmail.com>
Tue, 9 Dec 2008 21:10:56 +0000 (13:10 -0800)
committerEric Anholt <eric@anholt.net>
Tue, 9 Dec 2008 21:17:48 +0000 (13:17 -0800)
commitb66495a0d915f5d5cc5ab50c843c9c1b296a5851
tree7f0e7b1770b4590998a632168e3f0d0ed7840bef
parent6e29a3c8e2dc920b6216a0df6357abd8234f1ec4
tnl: Optimize SSE load[23]f_1 since they don't need the identity swizzle.

SSE movss from memory zeroes out everything above the destination dword, so
we get the (a, 0) or (a, 0, 0) result that these functions needed.

Bug #16520.
src/mesa/tnl/t_vertex_sse.c