[Bug fortran/100561] New: Error while computing square of a real number (8 and 16 bytes)

[Bug fortran/100561] New: Error while computing square of a real number (8 and 16 bytes)

[h0bb_88 at protonmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100561

            Bug ID: 100561
           Summary: Error while computing square of a real number (8 and
                    16 bytes)
           Product: gcc
           Version: og10 (devel/omp/gcc-10)
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: h0bb_88 at protonmail dot com
  Target Milestone: ---

This simple example didn't give the same result than wolfram or any calculator
:
- result from program : 9.869604950...
- result from other calculators : 9.869604401...

The rounding is not the cause, because of error from 10° decimal, in 16-byte
floating point. This error exist too in 8-byte floating-points number. 

Simple example :

program main
real(16), parameter :: pi =
3.1415926535897932384626433832795028841971693993751058209749445923
write(*,*) pi**2, pi*pi
end program

[Bug fortran/100561] Error while computing square of a real number (8 and 16 bytes)

[h0bb_88 at protonmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100561

--- Comment #1 from h0bb_88 at protonmail dot com ---
Created attachment 50797
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50797&action=edit
Source code

[Bug fortran/100561] Error while computing square of a real number (8 and 16 bytes)

[h0bb_88 at protonmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100561

--- Comment #2 from h0bb_88 at protonmail dot com ---
Created attachment 50798
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=50798&action=edit
Comparison between f90 result and bc computation

[Bug fortran/100561] Error while computing square of a real number (8 and 16 bytes)

[h0bb_88 at protonmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100561

--- Comment #3 from h0bb_88 at protonmail dot com ---
System:    Kernel: 5.8.0-53-generic x86_64 bits: 64 compiler: N/A Desktop:
Cinnamon 4.8.6 
           wm: muffin dm: LightDM Distro: Linux Mint 20.1 Ulyssa base: Ubuntu
20.04 focal 
Machine:   Type: Desktop Mobo: ASUSTeK model: Z170-K v: Rev X.0x serial:
<filter> 
           UEFI [Legacy]: American Megatrends v: 3805 date: 05/16/2018 
CPU:       Topology: Quad Core model: Intel Core i7-6700K bits: 64 type: MT MCP
arch: Skylake-S 
           rev: 3 L2 cache: 8192 KiB 
           flags: avx avx2 lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx
bogomips: 63999 
           Speed: 3900 MHz min/max: 800/4200 MHz Core speeds (MHz): 1: 3900 2:
3900 3: 3900 
           4: 3900 5: 3900 6: 3900 7: 3900 8: 3900

[Bug fortran/100561] Error while computing square of a real number (8 and 16 bytes)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100561

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
It looks like gfortran parses the constant in single-precision, which might be
an issue with your fortran source.

[Bug fortran/100561] Error while computing square of a real number (8 and 16 bytes)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100561

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
it is.

real(8), parameter :: pi = 3.1415926535d0

works for example.

[Bug fortran/100561] Error while computing square of a real number (8 and 16 bytes)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100561

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This is invalid because you wrote:
3.1415926535897932384626433832795028841971693993751058209749445923

And not:
3.1415926535897932384626433832795028841971693993751058209749445923_16

The first is a real type which is already rounded to 64bit.

When I used the _16, I get the correct answer.