In the past, we stored booleans as integer 0 or 1. At some point, we changed to storing them as 0 or some non-zero value. The Sandybridge specific code for emitting IF instructions with embedded conditional modifiers was not updated to account for this change.
For code such as if (bool_a == bool_b) we cannot simply emit an equality comparison: even if both operands are true, they may be represented by two different non-zero values. We must first normalize them to 0 or 1 so that they can be compared properly. Fixes piglit test spec/glsl-1.10/execution/fs-bool-less-compare-true. Fixes at least 9 subcases of oglconform's shad-compiler test: - TestLessThanEquali - TestGreaterThan - TestGreaterThani - TestGreaterThanEqual - TestGreaterThanEquali - TestEqualb - TestNotEqual - TestNotEquali - TestNotEqualb NOTE: This is a candidate for the 9.0 branch (but not 8.0). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54709 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=48629 Signed-off-by: Kenneth Graunke <kenn...@whitecape.org> --- src/mesa/drivers/dri/i965/brw_fs_visitor.cpp | 3 +++ 1 file changed, 3 insertions(+) diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp index d3cbde3..320e2f0 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp @@ -1650,6 +1650,9 @@ fs_visitor::emit_if_gen6(ir_if *ir) case ir_binop_all_equal: case ir_binop_nequal: case ir_binop_any_nequal: + resolve_bool_comparison(expr->operands[0], &op[0]); + resolve_bool_comparison(expr->operands[1], &op[1]); + inst = emit(BRW_OPCODE_IF, reg_null_d, op[0], op[1]); inst->conditional_mod = brw_conditional_for_comparison(expr->operation); -- 1.7.11.4 _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
