R600/SI: Use V_ADD_F32 instead of V_MOV_B32 for clamp/neg/abs modifiers.
authorMichel Danzer <michel.daenzer@amd.com>
Mon, 11 Feb 2013 15:58:21 +0000 (15:58 +0000)
committerMichel Danzer <michel.daenzer@amd.com>
Mon, 11 Feb 2013 15:58:21 +0000 (15:58 +0000)
commit10ed47f927a9079e1c21198ec18a601c5b2d2404
treea083929b5b259027907297e1a69a05e864ac65cd
parent29f9dea1ab821e9b0c7cada17802e6b3fd5a7c93
R600/SI: Use V_ADD_F32 instead of V_MOV_B32 for clamp/neg/abs modifiers.

The modifiers don't seem to have any effect with V_MOV_B32, supposedly it's
meant to just move bits untouched.

Fixes 46 piglit tests with radeonsi, though unfortunately 11 of those had
just regressed because they started using the clamp modifier.

NOTE: This is a candidate for the Mesa stable branch.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 174890
llvm/lib/Target/R600/SIISelLowering.cpp