[COMMITTED,frange] Revert relation handling in LTGT_EXPR.
Checks
Commit Message
In trying to come up with a missing testcase for commit 979e0fbf53cd,
I've realized the patch doesn't catch anything.
A relation of VREL_EQ in foperator_ltgt::fold_range() is either both
arguments the same (x LTGT_EXPR x), which we should never emit, or two
arguments that are actually the same, in which case !NAN applies, and the
whole thing can be handled as NE_EXPR further down.
The whole FP relation thing is getting a face lift shortly, thanks to
comments from both Jakub and Andrew, but there's no need in keeping
useless code around.
gcc/ChangeLog:
* range-op-float.cc (foperator_ltgt::fold_range): Do not special
case VREL_EQ nor call frelop_early_resolve.
---
gcc/range-op-float.cc | 15 ---------------
1 file changed, 15 deletions(-)
@@ -2206,21 +2206,6 @@ public:
const frange &op1, const frange &op2,
relation_trio trio = TRIO_VARYING) const final override
{
- relation_kind rel = trio.op1_op2 ();
-
- // VREL_EQ is really VREL_(UN)EQ because we could have a NAN in
- // the operands, but since LTGT_EXPR is really a NE_EXPR without
- // the NAN, VREL_EQ & LTGT_EXPR is an impossibility.
- if (rel == VREL_EQ)
- {
- r = range_false (type);
- return true;
- }
- // ...otherwise pretend we're trying to resolve a NE_EXPR and
- // everything will "just work".
- if (frelop_early_resolve (r, type, op1, op2, trio, VREL_NE))
- return true;
-
if (op1.known_isnan () || op2.known_isnan ())
{
r = range_false (type);