again: gen gas vs. gen gcc !

[D.J.Peters]

I was thinking or dreaming :-) the problem gas vs gcc was fixed in the past but isn't.
(we saw the problem with fbraytracer.bas before)

I disabled randomize(XXX) so the results of gas and gcc should not be the same but near the same.
You can see in the second image the bottom left scene isn't rendered with -gen gcc
(fbc 32-bit -gen gcc and fbc 64-bit produce the same result)

Give it a try if you like.
right click save as: gasvsgcc.bas

Joshy
fbc -gen gas gasvsgcc.bas

fbc -gen gcc gasvsgcc.bas

[frisian]

Replace the function Crater.sphereFunc with one of the three new versions.
Change the test condition in if then.

Code: Select all

function Crater.sphereFunc(x as REAL, z as REAL, r as REAL) as REAL
  x/=r : z/=r
  dim as REAL tmp =1 - x*x - z*z
  if tmp > 0 then tmp = rSqr(tmp)
  tmp=rMax(tmp,0)
  return tmp*r
end function

Swapped two lines.

Code: Select all

function Crater.sphereFunc(x as REAL, z as REAL, r as REAL) as REAL
  x/=r : z/=r
  dim as REAL tmp =1 - x*x - z*z
  tmp=rMax(tmp,0)
  if tmp then tmp = rSqr(tmp)
  return tmp*r
end function

Rewrote the if then statement.

Code: Select all

function Crater.sphereFunc(x as REAL, z as REAL, r as REAL) as REAL
  x/=r : z/=r
  dim as REAL tmp =1 - x*x - z*z
  If tmp > 0 Then
      Return rSqr(tmp) * r
  Else
      Return 0
  End If
end function

The handling of a negative number in the square root algorithm seems to be different in FB and C, both work with negative numbers.
Negative numbers need to be handled by the programmer, they are not handled by the FB or C.