topic OK, when did 16.0.2 come out in Intel® Fortran Compiler

Another debugging issue - gives wrong values

WSinc — Tue, 23 Feb 2016 20:52:49 GMT

The following source code runs fine as far as the PRINT statements go -

But I get completely wrong values when I put a break point after the PRINT statements, and

try to look at the variable contents. It looks like it does not think they were initialized.

And about half the time VS crashes ! !

The problem goes away when I make everything real(8) instead.

program kepler
implicit none
real(16) r(3),th(3),x(3),y(3),e,p,kang
!!!!!!!!!!!!!!!!!!!!!
kang=1
e=0.
p=1.0
th(1)=.7854Q0
r(1) =p/( 1.0 - e*cos(th(1)) )
th(2)=kang/r(1)**2 ! theta dot
r(2) =-sin(th(1))/( 1.0 - e*cos(th(1))**2 )
r(2) =r(2)*p
x(1)=r(1)*cos(th(1))
y(1)=r(1)*sin(th(1))
x(2)=-r(1)*sin(th(1))*th(2)+cos(th(1))*r(2)
y(2)= r(1)*cos(th(1))*th(2)-sin(th(1))*r(2)
th(3)=-2.0*r(2)/r(1)**3
print *,"e,p=",e,p
print *,"r=",r
print *,"th=",th
print *,"x=",x
print *,"y=",y
read(*,*)
end program kepler

Real(16) is part of the standard, right ?

No, real(16) is not "part of

Steven_L_Intel1 — Tue, 23 Feb 2016 21:16:47 GMT

No, real(16) is not "part of the standard" in that the standard does not specify particular kind numbers. Furthermore, the standard requires only single and double precision. That said, this program should work ok.

I can reproduce an issue with Visual Studio crashing when one hits the breakpoint. It might be that the debug support for REAL(16) has some issues - I'll investigate further.

Something that looks a little

Harry_A_ — Wed, 24 Feb 2016 01:20:03 GMT

Something that looks a little strange to me is that you feel you need the precision of REAL(16) in your calculations, but you are using a value for pi/4 that is rounded at the 4th significant digit.

Not trying to be exact about

WSinc — Wed, 24 Feb 2016 02:18:31 GMT

Not trying to be exact about that, just trying to get approx. 45 degree polar angle.

If I wanted pi/4, I would put the whole thing in to 18 places, then divide it by four.

3.14159265358979323846, etc.

Or say atan(1.0)

I tried with VS2013+IFort 16

mecej4 — Wed, 24 Feb 2016 10:45:12 GMT

I tried with VS2013+IFort 16.0.1. I did not notice anything unusual in the debugger, after reaching a breakpoint set on the READ statement. All the variables had reasonable values.

I could reproduce this with

Steven_L_Intel1 — Wed, 24 Feb 2016 15:26:19 GMT

I could reproduce this with 16.0.0 but not 16.0.2. Please install the current version.

OK, when did 16.0.2 come out

WSinc — Wed, 24 Feb 2016 15:59:46 GMT

OK, when did 16.0.2 come out ?

I might already have it on the other PC -- -

No big deal, but thought you'd want to be apprised anyway.

Last week.

Steven_L_Intel1 — Wed, 24 Feb 2016 16:33:17 GMT

Last week.