intel: Use the NIR lowering for isign.
authorEric Anholt <eric@anholt.net>
Wed, 6 Feb 2019 21:26:17 +0000 (13:26 -0800)
committerEric Anholt <eric@anholt.net>
Thu, 14 Feb 2019 00:32:30 +0000 (00:32 +0000)
commit8f3694e1abfd702730014253af327060ac03aca1
tree9124f9b3c031cacd60bc5167102e5b38639da9aa
parent3f22b35a43cc842bf97ef51b864f80489194bb16
intel: Use the NIR lowering for isign.

Drops one instruction from fs-sign-int.shader_test.  No change in
shader-db due to it having 0 instances of sign(genIType).  This may hurt
isign64 if algebraic runs before int64 lowering, but I wasn't sure how to
mark the algebraic opt as "every bit size but 64".

v2: Update commit message about shader-db.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> (v1)
src/intel/compiler/brw_compiler.c
src/intel/compiler/brw_fs_nir.cpp
src/intel/compiler/brw_vec4_nir.cpp