Skip to main content
S
StuartBruff23-Emerald V
23-Emerald V
November 15, 2025
Question

MEP - Change Mathcad numerical infinity to IEEE 754 Infinity

  • November 15, 2025
  • 3 replies
  • 487 views

Mathcad Enhancement Proposal:  Change Mathcad's numerical representation of infinity to the IEEE 754 floating-point standard.

 

When dealing with large numbers, representing infinity as 10^307 can lead to numerical errors and yield ambiguous or unreadable results.

 

2025 11 15 A.png

 

Note: Python math, NumPy, Mathematica and Matlab all give correct results.

 

I don't know why 2^1020 is allowable, but 2^n, where n:=1020, raises an error.

 

Stuart

 

(Numerical ∞ = 10307 has annoyed me since Mathsoft Mathcad days.  At least change ∞ to ⧜ (U+29DC Incomplete Infinity)

    3 replies

    S
    StuartBruff23-Emerald VAuthor
    23-Emerald V
    November 16, 2025

    Interesting.

     

    2025 11 16 A.png

     

    Math Result display set to "General".

     

    Stuart

    L
    LucMeekes23-Emerald IV
    23-Emerald IV
    November 16, 2025

    The 'Infinity' unit was introduced in Prime 6. Up to and including Prime 5 you'd get:

    LucMeekes_0-1763294390507.png

    But the funny behaviour with 2^1020 was already present in Prime 2.

     

    Success!
    Luc

     

      S
      StuartBruff23-Emerald VAuthor
      23-Emerald V
      November 16, 2025

      Thanks, Luc.

       

      Perhaps PTC can fix that problem at the same time as they implement an IEEE 754 Infinity.

       

      Stuart

      S
      StuartBruff23-Emerald VAuthor
      23-Emerald V
      November 27, 2025

      By the Power of Grayvector and by the Magic of Mathcad’s Unit System, Werner’s work in another thread has inspired a technique for saving us from Thanos by transmogrifying an infinity stone into a very large number of kilograms,

       

      2025 11 26 C.png

       

      I knew I'd find a use for a finite infinity.  Previously, I could only destroy them two at a time before the structure of the 64-bit floating-point universe threw an overflow error (which wouldn't happen with a proper IEEE 754 infinity).   But now I can handle all of them at once.

       

      I’m not sure it’s a good enough reason to retain 10^307 as infinity, though. 

      W
      Werner_E25-Diamond I
      25-Diamond I
      November 27, 2025

      Unfortunately only one "Right' out of six:

      Werner_E_0-1764224151268.png

       

      BTW, you get correct results with the full version and using the symbolics:

      Werner_E_0-1764234129248.png

      in Prime as well as in MC15

      Werner_E_1-1764234294189.png

       

        S
        StuartBruff23-Emerald VAuthor
        23-Emerald V
        November 27, 2025

        Thanks, Werner.  Looks like a copy & paste error on my part (I forgot to change the equality operator to division).

         

        Compatibility with symbolic results (where meaningful) is another good reason to implement a proper IEEE 754 infinity rather than an arbitrarily large number, say, 𒌋𒐗 𒎙𒐖 𒑪 𒑪𒐘 𒑪𒐝 𒐝 𒎙𒐝 𒑪𒐜 𒎙𒐚 𒑩𒐗 𒌋𒐛 𒌍𒐕 𒑪𒐕 𒐚 𒑩.

         

        (And if you're going to pick a big FP number, why 10307 rather than 253-1?)

         

        Stuart

        Terms & ConditionsCookie settingsAccessibility statement