gallivm: Prefer backend agnostic intrinsic for rounding.
authorJose Fonseca <jfonseca@vmware.com>
Fri, 1 Apr 2016 10:50:28 +0000 (11:50 +0100)
committerJose Fonseca <jfonseca@vmware.com>
Sun, 3 Apr 2016 21:09:07 +0000 (22:09 +0100)
commite4f01da15d8c6ce3e8c77ff3ff3d2ce2574a3f7b
tree2decca255387a937c279f7586aec9961ea9d25c1
parent324451e73fae17c2844f24c7e02000bddc260e78
gallivm: Prefer backend agnostic intrinsic for rounding.

We could unconditionally use these instrinsics, but performance with SSE2
would suck, as LLVM falls back to calling libm.

lp_test_arit.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
src/gallium/auxiliary/gallivm/lp_bld_arit.c