Re: Bus error/ segmentation fault--help?

a.e.pace@xxxxxxxxx wrote:
All:
I have a code that successfully compiles on my machine, but does
not run properly. When I compile with the intel fortran compiler, I
get the following error:

forrtl: severe (180): SIGBUS, bus error occurred

Stack trace terminated abnormally.

When I compile with gfortran, I get the following error:
"Bus error"

I believe that the error lies somewhere in how I allocate the array "Y"
and pass it to various subroutines. I will attach my code after this
message. Any thoughts or assistance on the matter would be greatly
appreciated.

My warmest thanks,

Alexander Pace
Georgia Institute of Technology
School of Aerospace Engineering.

Code:
program visc
c use global
allocatable Y(:,:)
c.... enter number of subdivisions from n=0->4.8
numsubs= 100
c.... step size
dn=4.8/numsubs

c... define upper and lower bounds for y3(0) (xu and xl)
xl=0
xu=2

c.... Start bisection method
do while (xu-xl .gt. 0.000001)
xm=(xu-xl)/2

c.... Perform Runge-Kutta Integrations to get values of y2(4.8) at top,
c.... lower, and midpoint

c allocate (Y(3,1000))
call runge(Y,xu,numsubs,dn)
gu = Y(2,numsubs) - 1

call runge(Y,xl,numsubs,dn)
gl = Y(2,numsubs) - 1

call runge(Y,xm,numsubs,dn)
gm = Y(2,numsubs) - 1

c.... Check for location of root, change bounds
if (gl*gm .lt. 0) then
xu = xm
else if (gl*gm .gt. 0) then
xl = xm
else if (gl*gm .eq. 0) then
xl = xu
endif
end do

print *, "y3(0)=", Y(3,1)

end



c.... This subroutine integrates the function y3 up one step, dn

subroutine integratey3(y1i,y2i,y3i,y3n,dn)
implicit double precision (a-h, o-z)

g1= y2i**2-y1i*y3i-1
g2= (y2i+.5*g1)**2-(y1i+.5*g1)*(y3i+.5*g1)-1
g3= (y2i+.5*g2)**2-(y1i+.5*g2)*(y3i+.5*g2)-1
g4= (y2i+g3)**2-(y1i+g3)*(y3i+g3)-1

y3n= y3i+dn*(g1+2*g2+2*g3+g4)/6

return
end

c.... This subroutine integrates y2 up one step, dn

subroutine integratey2(y2i,y3i,y2n,dn)
implicit double precision (a-h, o-z)

g1=y3i
g2=y3i+.5*g1
g3=y3i+.5*g2
g4=y3i+g3

y2n= y2i+dn*(g1+2*g2+2*g3+g4)/6

return
end

c.... This subroutine integrates y1 up one step, dn

subroutine integratey1(y1i,y2i,y1n,dn)
implicit double precision (a-h, o-z)

g1=y2i
g2=y2i+.5*g1
g3=y2i+.5*g2
g4=y2i+g3

y1n= y1i+dn*(g1+2*g2+2*g3+g4)/6

return
end

c.... This subroutine inputs the matrix Y and an initial value for
c.... y3(0) and performs the runge kutta integration from n=0->4.8


subroutine runge(Y,y3init,numsubs,dn)
c use global

real Y(:,:)

When you call a subroutine that has an array with an assumed shape,
like Y here, you need to give the compiler more information. In
Fortran jargon, you need an explicit interface. Basically, you need to
put subroutine runge in a module and USE the module. As a general rule,
for programs of this smallish size, you should put all of the
subroutines in a module. This ensures that the compiler will have
the necessary information to process calls and almost always ensures
that argument mismatches like this can't happen.

*** hendrickson

c.... define initial conditions
Y(1,1)=0
Y(2,1)=0
Y(3,1)=y3init
y1n=0
y2n=0
y3n=0

c.... let 'er rip
do i=2, numsubs-1
y1i=Y(1,i)
y2i=Y(2,i)
y3i=Y(3,i)
call integratey3(y1i,y2i,y3i,y3n,dn)
Y(3,i+1)=y3n
call integratey2(y2i,y3i,y2n,dn)
Y(2,i+1)=y2n
call integratey1(y1i,y2i,y1n,dn)
Y(1,i+1)=y1n
end do

return
end

.