3 Replies Latest reply on Mar 20, 2014 9:28 AM by rmm200

    Double precision

    JohnJohn

      When might we see support for double precision in the Arduino IDE?

        • 1. Re: Double precision
          rmm200

          Define double precision...  On the Uno and other Atmel based processors, double is identical to float (4 bytes), with no gain in precision.

           

          Only the Due provide 8 byte doubles.

           

          I can't answer your question, but I would at least like to understand it.

          • 2. Re: Double precision
            JohnJohn

            Double precision in the Intel Quark context means IEEE standard 8 byte floating point.

            The gcc compilers and math libraries support this precision.

            The Intel Arduino compiler does not (yet) do so.

            Hence my question.

            • 3. Re: Double precision
              rmm200

              Well the Arduino IDE does use the GCC compiler.

              Precision limitations are likely to be do to the use of uClibc. You might check there to see how double is implemented.

               

              The Arduino IDE use of UClibc will not change since it is related to the amount of rom on the Galileo.

               

              If you run an SD image based on eglibc, and run your programs directly on that image, 8 byte double precision should work for you. Just don't expect to use the IDE. SSH to the Galileo works like a dream - Mine has been up 24/7 for a month now.

               

              A handy Linux reference that gives GCC options to invoke SSE is:

               

              An Introduction to GCC Compiler Intrinsics in Vector Processing | Linux Journalto GCC operands

               

              Message was edited by: Robert Mitchell to add link to GCC operands

              1 of 1 people found this helpful