Author Topic: double-double trouble  (Read 373 times)

TimoVJL

  • Member
  • ****
  • Posts: 663
double-double trouble
« on: November 27, 2020, 11:45:25 PM »
This site offers a double-double library
https://www.davidhbailey.com/dhbsoftware/

but with asmc it should be able to create a library for 128 bit float ?
May the source be with you

jack

  • Member
  • **
  • Posts: 74
Re: double-double trouble
« Reply #1 on: November 28, 2020, 12:49:52 AM »
Hi TimoVJL
some years ago I did some fiddling with double-double/quad-double and compared it with soft-float http://www.jhauser.us/arithmetic/SoftFloat.html
at the time of my tests I found that soft-float was quite a bit faster but with soft-float you have to write your own elementary transcendental functions, I adapted some of 128-bit transcendental functions from http://www.moshier.net/ to FreeBasic, but after a while I got bored
for input/output I used ioldoubl.c from http://www.moshier.net/smldbl12.zip and gdtoa from https://www.netlib.org/fp/

nidud

  • Member
  • *****
  • Posts: 2032
    • https://github.com/nidud/asmc
Re: double-double trouble
« Reply #2 on: November 28, 2020, 01:56:47 AM »
Asmc use __float128 internally so that is included in LIBC. There is also a standalone
library (using the same source) for Win 32/64 and Linux.

quadmath.inc

See also:
Large integers and floats

TimoVJL

  • Member
  • ****
  • Posts: 663
Re: double-double trouble
« Reply #3 on: November 28, 2020, 04:11:19 AM »
Thanks nidud :thumbsup:
May the source be with you