rounded transcendentals

[Paul.Zimmermann@xxxxxxxx (Paul Zimmermann)]

Annex D says:

   "Implementations should provide correctly-rounded versions of the functions
    listed in Table D.1, for binary32 format if supported and binary64 format
    if supported."

then a few lines later:

   "For all other basic formats, these functions should be faithfully-rounded."

I thus conclude that decimal64 and decimal128 should be faithfully-rounded,
as well as binary128.

Am I correct?

Paul Zimmermann