gallivm: avoid unnecessary URem in linear wrap repeat case
authorRoland Scheidegger <sroland@vmware.com>
Fri, 8 Oct 2010 19:08:49 +0000 (21:08 +0200)
committerRoland Scheidegger <sroland@vmware.com>
Fri, 8 Oct 2010 22:36:38 +0000 (00:36 +0200)
commit2cc6da85d6e60e80d0b5b86fe42f8f82073b5d51
tree4148eacacfbb1f33250c202fee70974310677d61
parent318bb080b04c961141a962d440d0e1296f8adda6
gallivm: avoid unnecessary URem in linear wrap repeat case

Haven't looked at what code this exactly generates but URem can't be fast.
Instead of using two URem only use one and replace the second one with
select/add (this is what the corresponding aos code already does).
src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c