gallivm: don't try to use fast rcp for fdiv
authorRoland Scheidegger <sroland@vmware.com>
Mon, 23 Jan 2017 17:06:03 +0000 (18:06 +0100)
committerRoland Scheidegger <sroland@vmware.com>
Tue, 24 Jan 2017 18:12:46 +0000 (19:12 +0100)
commitf4df21ed9509af3d7bff3da4dc7f4b2949124b3f
tree61ae65fce5785f025eb48f7726e297b71df3ba65
parent25208949d7293aa060a3416f8cf3cdb3ca1fbfdd
gallivm: don't try to use fast rcp for fdiv

The use of fast rcp instruction is disabled, and will always fall back
to use a division instead (1 / x). Hence, if we get a division opcode,
it doesn't make much sense trying to split that into rcp/mul.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
src/gallium/auxiliary/gallivm/lp_bld_arit.c