[llvm] [KnownFPClass] Fix fsub input denormal handling (PR #225605)

[Eric Ross via llvm-commits]

================
@@ -1176,7 +1186,7 @@ define float @fadd_self_no_pzero_maybe_undef(float nofpclass(pzero) %arg) {
 
 ; The input is not logical 0 since it's IEEE, but the output could
 ; still be flushed.
-define float @fadd_self_no_zero__output_only_is_ftz(float noundef nofpclass(zero) %arg) #7 {
----------------
ZERICO2005 wrote:

Attribute `#7` was a duplicate of Attribute `#1`.

https://github.com/llvm/llvm-project/pull/225605