On 12/11/13 02:51, Richard Biener wrote:
+ /* If inversion is needed, first try to invert the test since + that's cheapest. */ + if (invert) + { + enum tree_code new_code + = invert_tree_comparison (cond_code, + HONOR_NANS (TYPE_MODE (TREE_TYPE (rhs))));That looks wrong - you want to look at HONOR_NANS on the mode of one of the comparison operands, not of the actual value you want to negate (it's integer and thus never has NaNs).
HONOR_NANS (TYPE_MODE (TREE_TYPE (gimple_cond_lhs (cond)))) Right? Jeff
