gallivm: only use lp_build_conv 4x4f -> 1x16 ub fastpath with sse2
authorRoland Scheidegger <sroland@vmware.com>
Wed, 13 Oct 2010 13:26:37 +0000 (15:26 +0200)
committerRoland Scheidegger <sroland@vmware.com>
Wed, 13 Oct 2010 13:26:37 +0000 (15:26 +0200)
commitd838e4f66d585baf3577f1298dd97d1b7c444ac2
treeb73d96f94ea6d80a3fda75328827636acd73a367
parentff4b397517a374ac3d4bf437f85ae6a96171a714
gallivm: only use lp_build_conv 4x4f -> 1x16 ub fastpath with sse2

This is relying on lp_build_pack2 using the sse2 pack intrinsics which
handle clamping.
(Alternatively could have make it use lp_build_packs2 but it might
not even produce more efficient code than not using the fastpath
in the first place.)
src/gallium/auxiliary/gallivm/lp_bld_conv.c