Mirror of forum.nim-lang.org

6885 :: 80-bit (long double) support?

• robert_m_muench (orginal) [2020-10-03T15:46:32+02:00] view original

  Since Nim doesn't seem to support 80-Bit IEEE floats, how am I supposed to deal with them?

  And yes, we have a codebase in C that uses these numbers.

---

• mratsim (orginal) [2020-10-03T17:16:22+02:00] view original

  Use

  type float80 *{.importc: "long double".} = object
  Run

---

• shirleyquirk (orginal) [2020-10-03T18:56:24+02:00] view original

  It's pretty horrible, hopefully you only need to move them and not do math with them. https://play.nim-lang.org/#ix=2zBO

---

• robert_m_muench (orginal) [2020-10-03T20:44:01+02:00] view original

  Thanks for the snippet... we currently have that code in C, and I think it's best to keep it there and access the results. It looks like a simple route with Nim because it compiles to C and should fit with the rest of the code-base.

---

• cdunn2001 (orginal) [2020-10-04T09:25:02+02:00] view original

  When I compile with C++ instead of C backend:

  /usercode/nimcache/@min.nim.cpp: In function 'void coloneq___7zlIf7POTZHxgRJdjlbnawin(long double&, NF)':
  /usercode/nimcache/@min.nim.cpp:103:3: error: invalid type argument of unary '*' (have 'long double')
    *a=b;
     ^
  Run

  Why?

---

• shirleyquirk (orginal) [2020-10-04T10:39:12+02:00] view original

  Oof, I knew that one was going to be brittle. You can't dereference a C++ reference, it's not a pointer

---