projects / mesa.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 9954f01) | patch
gallivm: optimize repeat linear npot code in the aos int path
authorJeff Muizelaar <jmuizelaar@mozilla.com>
Sat, 15 Feb 2014 00:47:12 +0000 (01:47 +0100)
committerRoland Scheidegger <sroland@vmware.com>
Fri, 14 Mar 2014 18:41:18 +0000 (19:41 +0100)
commitff1e850eec7a18d2cc5984848977112f3c4ad4cf
treead1da405fc3840b9fe9ef657c1c223a949f9dcc0tree
parent9954f0149799b726b8d57d723328ee5bec2b5070commit | diff
gallivm: optimize repeat linear npot code in the aos int path

Similar to the other cases, shift some weight/coord calculations to int
space. This should be slightly faster (on x86 sse it should actually safe one
instruction, and generally int instructions are cheaper).
src/gallium/auxiliary/gallivm/lp_bld_sample_aos.c diff | blob | history