• via Digitalmars-d-bugs (24/24) Jan 17 2017 https://issues.dlang.org/show_bug.cgi?id=17104

via Digitalmars-d-bugs <digitalmars-d-bugs puremagic.com> writes:

https://issues.dlang.org/show_bug.cgi?id=17104

          Issue ID: 17104
           Summary: "Real close to the machine" webpage error
           Product: D
           Version: D2
          Hardware: All
                OS: All
            Status: NEW
          Severity: minor
          Priority: P1
         Component: dlang.org
          Assignee: nobody puremagic.com
          Reporter: safety0ff.bugz gmail.com

The website contains the following passage:
"However, a limited form of associativity is possible if the type used for
intermediate results is larger than any of the operands (which happens on x87
and Itanium machines). If R is the intermediate type, and F is the type being
multiplied, up to min(R.max_exp/F.max_exp, R.epsilon/F.epsilon) values of type
F can be multiplied together in any order without influencing the result."

https://dlang.org/d-floating-point.html


Since R.epsilon is less than F.epsilon, R.epsilon/F.epsilon is < 1.
Therefore min(R.max_exp/F.max_exp, R.epsilon/F.epsilon) is always < 1, which
contradicts the paragraph.

--