please help me!!after parallel the Execution efficiency is 10 times slower than serial

[wert7588]

my parallel source is:
do ii = 1, nion-1
i = iion(ii)
in = jion(ii)
ic = kion(ii)
xic = x(ic)
yic = y(ic)
zic = z(ic)
xi = x(i) - xic
yi = y(i) - yic
zi = z(i) - zic
fi = f * pchg(ii)
iuse = (use(i) .or. use(ic))
c$omp parallel
c$omp sections
c$omp single
do j = 1, nion
cscale(iion(j)) = 1.0d0
end do
c$omp end single
c$omp section
do j = 1, n12(in)
cscale(i12(j,in)) = c2scale
end do
c$omp section
do j = 1, n13(in)
cscale(i13(j,in)) = c3scale
end do
c$omp section
do j = 1, n14(in)
cscale(i14(j,in)) = c4scale
end do
c$omp section
do j = 1, n15(in)
cscale(i15(j,in)) = c5scale
end do
c$omp end sections
c$omp do
c$omp& private(k,kn,kc,proceed,xr,yr,zr)
c$omp& reduction( + : ec, dec, vir )
do kk = ii+1, nion
k = iion(kk)
kn = jion(kk)
kc = kion(kk)
proceed = .true.
if (use_group) call groups (proceed,fgrp,i,k,0,0,0,0)
if (proceed) proceed = (iuse .or. use(k) .or. use(kc))
if (proceed) then
xc = xic - x(kc)
yc = yic - y(kc)
zc = zic - z(kc)
if (use_image) call image (xc,yc,zc,0)
rc2 = xc*xc + yc*yc + zc*zc
if (rc2 .le. off2) then
xr = xc + xi - x(k) + x(kc)
yr = yc + yi - y(k) + y(kc)
zr = zc + zi - z(k) + z(kc)
r2 = xr*xr + yr*yr + zr*zr
r = sqrt(r2)
fik = fi * pchg(kk) * cscale(kn)
e = fik / r
de = -fik / r2
dc = 0.0d0
shift = fik / (0.5d0*(off+cut))
e = e - shift
if (rc2 .gt. cut2) then
rc = sqrt(rc2)
rc3 = rc2 * rc
rc4 = rc2 * rc2
rc5 = rc2 * rc3
rc6 = rc3 * rc3
rc7 = rc3 * rc4
taper = c5*rc5 + c4*rc4 + c3*rc3
& + c2*rc2 + c1*rc + c0
dtaper = 5.0d0*c5*rc4 + 4.0d0*c4*rc3
& + 3.0d0*c3*rc2 + 2.0d0*c2*rc + c1
trans = fik * (f7*rc7 + f6*rc6 + f5*rc5 + f4*rc4
& + f3*rc3 + f2*rc2 + f1*rc + f0)
dtrans = fik * (7.0d0*f7*rc6 + 6.0d0*f6*rc5
& + 5.0d0*f5*rc4 + 4.0d0*f4*rc3
& + 3.0d0*f3*rc2 + 2.0d0*f2*rc + f1)
dc = (e * dtaper + dtrans) / rc
de = de * taper
e = e * taper + trans
end if
if (use_group) then
e = e * fgrp
de = de * fgrp
dc = dc * fgrp
end if
de = de / r
dedx = de * xr
dedy = de * yr
dedz = de * zr
dedxc = dc * xc
dedyc = dc * yc
dedzc = dc * zc
ec = ec + e
dec(1,i) = dec(1,i) + dedx
dec(2,i) = dec(2,i) + dedy
dec(3,i) = dec(3,i) + dedz
dec(1,ic) = dec(1,ic) + dedxc
dec(2,ic) = dec(2,ic) + dedyc
dec(3,ic) = dec(3,ic) + dedzc
dec(1,k) = dec(1,k) - dedx
dec(2,k) = dec(2,k) - dedy
dec(3,k) = dec(3,k) - dedz
dec(1,kc) = dec(1,kc) - dedxc
dec(2,kc) = dec(2,kc) - dedyc
dec(3,kc) = dec(3,kc) - dedzc
vxx = xr*dedx + xc*dedxc
vyx = yr*dedx + yc*dedxc
vzx = zr*dedx + zc*dedxc
vyy = yr*dedy + yc*dedyc
vzy = zr*dedy + zc*dedyc
vzz = zr*dedz + zc*dedzc
vir(1,1) = vir(1,1) + vxx
vir(2,1) = vir(2,1) + vyx
vir(3,1) = vir(3,1) + vzx
vir(1,2) = vir(1,2) + vyx
vir(2,2) = vir(2,2) + vyy
vir(3,2) = vir(3,2) + vzy
vir(1,3) = vir(1,3) + vzx
vir(2,3) = vir(2,3) + vzy
vir(3,3) = vir(3,3) + vzz
if (molcule(i) .ne. molcule(k)) then
einter = einter + e
end if
end if
end if
end do
c$omp end do
c$omp end parallel
end do
but after parallel the Execution efficiency is 10 times slower than serialso i want to know ,why?
especially in reduction( + : vir,ec,dec)
please help me,please tell me
for example:
if nion=1000 Serial time=56 seconds,but parallel time =560 seconds
if nion=180,serial time=0.24 seconds,but parallel time = 2.4 seconds
if nion larger, the parallel time is more larger than serial time

[KitturGanesh]

Hi,
You had posted this issue to the Parallel Studio Forum instead of the Fortran forum. I've moved it to the appropriate forum so this issue will be addressed by those monitoring the Fortran issues, just FYI. Appreciate your patience till then.
-regards,
Kittur

[TimP]

The posted replies are still on the copy of the thread which he started on the Threading Parallel forum.

[Steven_L_Intel1]

Quoting - tim18

The posted replies are still on the copy of the thread which he started on the Threading Parallel forum.

That would be this thread. I'm closing the one here so that conversation can continue in the Threading forum.