|
|
 |
 |
 |
 |
Fortran Programming Language
|
 |
 |
 |
 |
|
|
|
|
please help
does this program work, i dont know did i call in the right way function
deqn
the task is to solve 20 linear equations with 20 unknowns, and to write the
solution of x2
this is my program:
program matrice
real(8), dimension(20,20)::a
real(8), dimension(20)::b
integer::i,j,ifail
dimension IR(20)
do i=1,20
do j=1,20
if (i==j) then
a(i,j)=real(i)
else if (j==(i-1)) then
a(i,j)=0.5*sqrt(real(i+1))
else if (j==(i+1)) then
a(i,j)=0.5*sqrt(real(i+1))
else
a(i,j)=0.0
end if
end do
end do
b(1)=1.0
do i=2,20
b(i)=0.0
end do
call deqn(20,a,20,IR,IFAIL,20,b)
print*,b(2)
end program
fake wrote: > does this program work, i dont know did i call in the right way function > deqn > the task is to solve 20 linear equations with 20 unknowns, and to write the
> solution of x2
> this is my program:
> program matrice
> real(8), dimension(20,20)::a
> real(8), dimension(20)::b
> integer::i,j,ifail
> dimension IR(20)
> do i=1,20
> do j=1,20
> if (i==j) then
> a(i,j)=real(i)
> else if (j==(i-1)) then
> a(i,j)=0.5*sqrt(real(i+1))
> else if (j==(i+1)) then
> a(i,j)=0.5*sqrt(real(i+1))
> else
> a(i,j)=0.0
> end if
> end do
> end do
> b(1)=1.0
> do i=2,20
> b(i)=0.0
> end do
> call deqn(20,a,20,IR,IFAIL,20,b)
> print*,b(2)
> end program
deqn? you should at least print out the value of IFAIL.
It also would help if you would use a different name or e-mail address.
There is so much spam on the internet that I almost deleted your message
without reading it. Even a title like "please help with subroutine
deqn" would be better.
Dick Hendrickson
> What happened when you tried it? Do you have the documentation for
> deqn? you should at least print out the value of IFAIL.
i cant try because i dont have that documentation... ifail is declared in
that function i know that..
> It also would help if you would use a different name or e-mail address.
> There is so much spam on the internet that I almost deleted your message
> without reading it. Even a title like "please help with subroutine deqn"
> would be better.
sorry for that, this is my first time that i am here, i am from croatia, and
we dont have so much spam here, as maybe on english pages
On May 17, 5:21 pm, "fake" <a@a.a> wrote:
> > What happened when you tried it? Do you have the documentation for > > deqn? you should at least print out the value of IFAIL. > i cant try because i dont have that documentation... ifail is declared in
> that function i know that..
> > It also would help if you would use a different name or e-mail address.
> > There is so much spam on the internet that I almost deleted your message
> > without reading it. Even a title like "please help with subroutine deqn"
> > would be better.
> sorry for that, this is my first time that i am here, i am from croatia, and
> we dont have so much spam here, as maybe on english pages
norm of (A * x - b) is effectively zero?
On May 17, 6:21 pm, "fake" <a@a.a> wrote:
> > What happened when you tried it? Do you have the documentation for > > deqn? you should at least print out the value of IFAIL. > i cant try because i dont have that documentation... ifail is declared in
> that function i know that..
> > It also would help if you would use a different name or e-mail address.
> > There is so much spam on the internet that I almost deleted your message
> > without reading it. Even a title like "please help with subroutine deqn"
> > would be better.
> sorry for that, this is my first time that i am here, i am from croatia, and
> we dont have so much spam here, as maybe on english pages
DEQN eventually leads to
http://dollywood.itp.tuwien.ac.at/cernlib/shortwrupsdir/f010/top.html
You *may* have an argument mismatch here. In particular, look at the
value of K. If K is too large for your problem, your subroutine may
write over other variables.
A google search may also show you source code for the DEQN routine,
but a quick look shows that it appears to differ from the
documentation listed above.
-- e-mail: epc8 at juno dot com
-- elliot
On 18 May 2007 10:32:06 -0700, e p chandler <e@juno.com>
wrote in <1179509526.436240.265@k79g2000hse.googlegroups.com>:
> This looks like a routine from CERN. A google search on CERNLIB and > DEQN eventually leads to > http://dollywood.itp.tuwien.ac.at/cernlib/shortwrupsdir/f010/top.html > You *may* have an argument mismatch here. In particular, look at the > value of K. If K is too large for your problem, your subroutine may > write over other variables. > A google search may also show you source code for the DEQN routine, > but a quick look shows that it appears to differ from the > documentation listed above.
as REAL R(N). It's not used in the routine, but passed on to DFACT() and
DFEQN(). In both of them the corresponding array is again INTEGER so it
should be a harmless pass-through, providing compiler flags aren't playing
games with variable sizes. Note it is a work-space so it could be dimensioned
larger than IDIM, but smaller sizes would cause problems.
--
Ivan Reid, School of Engineering & Design, _____________ CMS Collaboration,
Brunel University. Ivan.Reid@[brunel.ac.uk|cern.ch] Room 40-1-B12, CERN
KotPT -- "for stupidity above and beyond the call of duty".
On Thu, 17 May 2007 22:36:02 +0200, fake <a@a.a>
wrote in <f2iebm$nr@news1.carnet.hr>:
> does this program work, i dont know did i call in the right way function > deqn > the task is to solve 20 linear equations with 20 unknowns, and to write the
> solution of x2
> this is my program:
> program matrice
> real(8), dimension(20,20)::a
> real(8), dimension(20)::b
> integer::i,j,ifail
> dimension IR(20)
...
> call deqn(20,a,20,IR,IFAIL,20,b)
> print*,b(2)
SUBROUTINE DEQN(N,A,IDIM,R,IFAIL,K,B)
REAL R(N),T1,T2,T3
DOUBLE PRECISION A(IDIM,N),B(IDIM,K),DET,S,TEMP,
The value of K (parameter 6) should be 1, n'est-ce pas? I've
mentioned in another post that the mismatch in declaration of IR/R should
not be a problem as it's a scratch array, and passed through to other
routines which do declare it integer anyway.
Note that in general I recommend against passing literal constants
to library routines. This helps avoid problems with type
mismatches/conversions.
--
Ivan Reid, School of Engineering & Design, _____________ CMS Collaboration,
Brunel University. Ivan.Reid@[brunel.ac.uk|cern.ch] Room 40-1-B12, CERN
KotPT -- "for stupidity above and beyond the call of duty".
On May 18, 2:52 pm, "Dr Ivan D. Reid" <Ivan.R@brunel.ac.uk> wrote:
> On Thu, 17 May 2007 22:36:02 +0200, fake <a @a.a> > wrote in <f2iebm$nr @news1.carnet.hr>: > > does this program work, i dont know did i call in the right way function
> > deqn
> > the task is to solve 20 linear equations with 20 unknowns, and to write the
> > solution of x2
> > this is my program:
> > program matrice
> > real(8), dimension(20,20)::a
> > real(8), dimension(20)::b
> > integer::i,j,ifail
> > dimension IR(20)
> ...
> > call deqn(20,a,20,IR,IFAIL,20,b)
> > print*,b(2)
> This is the library declaration (2004 version of KERNLIB):
> SUBROUTINE DEQN(N,A,IDIM,R,IFAIL,K,B)
> REAL R(N),T1,T2,T3
> DOUBLE PRECISION A(IDIM,N),B(IDIM,K),DET,S,TEMP,
> The value of K (parameter 6) should be 1, n'est-ce pas? I've
> mentioned in another post that the mismatch in declaration of IR/R should
> not be a problem as it's a scratch array, and passed through to other
> routines which do declare it integer anyway.
> Note that in general I recommend against passing literal constants
> to library routines. This helps avoid problems with type
> mismatches/conversions.
> --
> Ivan Reid, School of Engineering & Design, _____________ CMS Collaboration,
> Brunel University. Ivan.Reid@[brunel.ac.uk|cern.ch] Room 40-1-B12, CERN
> KotPT -- "for stupidity above and beyond the call of duty".
call deqn(20,a,20,ir,ifail,1,b)
A * X is reasonably close to B. (checked in Excel).
Note:
I compiled the test program plus the minimal set of source files
needed (after making necessary changes). The libraries provided on
CERN's web site for Windows use DVF, not a 'GNU' toolset. Google shows
libraries built on Cygwin but not MinGW. Has someone built these for
(MinGW) g77, g95 or gfortran?
-- elliot
On 19 May 2007 09:43:49 -0700, e p chandler <e@juno.com>
wrote in <1179593029.327277.26@p77g2000hsh.googlegroups.com>:
> On May 18, 2:52 pm, "Dr Ivan D. Reid" <Ivan.R @brunel.ac.uk> wrote: >> The value of K (parameter 6) should be 1, n'est-ce pas? I've >> mentioned in another post that the mismatch in declaration of IR/R should >> not be a problem as it's a scratch array, and passed through to other >> routines which do declare it integer anyway. > Setting K to 1 seems to solve the OP's problem: > call deqn(20,a,20,ir,ifail,1,b)
> A * X is reasonably close to B. (checked in Excel).
> Note:
> I compiled the test program plus the minimal set of source files
> needed (after making necessary changes). The libraries provided on
> CERN's web site for Windows use DVF, not a 'GNU' toolset. Google shows
> libraries built on Cygwin but not MinGW. Has someone built these for
> (MinGW) g77, g95 or gfortran?
When I first started using CERNLIB in anger (back about 1990 at PSI)
their licence was fairly broad, but restricted to organisations allied with
CERN (ISTR I had to get some sort of authorisation to get the MINUIT sources).
This was well before the GPL (or at least before it became popular) -- maybe
they've changed to that nowadays. In any case, I use full-blown Cygwin
when I'm not using Scientific Linux CERN so I'm not able to offer MinGW
libs unless someone gives me a few pointers.
I'm pretty sure it'd all be g77, too. The modern Fortrans didn't
appear until SLC 4.0; most people are still at 3.0.6, though this will
change in the next six months or so.
--
Ivan Reid, School of Engineering & Design, _____________ CMS Collaboration,
Brunel University. Ivan.Reid@[brunel.ac.uk|cern.ch] Room 40-1-B12, CERN
KotPT -- "for stupidity above and beyond the call of duty".
yesterday i saw that mistake..
i wrote:
program matrice
implicit none
real(8), dimension(20,20)::a
real(8), dimension(20)::b
integer::i,j,ifail
dimension IR(20)
do i=1,20
do j=1,20
if (i==j) then
a(i,j)=real(i)
else if (j==(i-1)) then
a(i,j)=0.5*sqrt(real(i+1))
else if (j==(i+1)) then
a(i,j)=0.5*sqrt(real(i+1))
else
a(i,j)=0.0
end if
end do
end do
b(1)=1.0
do i=2,20
b(i)=0.0
end do
call deqn(20,a,20,IR,IFAIL,1,b)
print*,b(2)
end program
and i get -0.811863376101259
i hope that it is the right answer for the solution of x2
"Dr Ivan D. Reid" <Ivan.R@brunel.ac.uk> wrote in message
news:slrnf4uchg.3co.Ivan.Reid@loki.brunel.ac.uk...
> When I first started using CERNLIB in anger (back about 1990 at PSI)
> their licence was fairly broad, but restricted to organisations allied
> with
> CERN (ISTR I had to get some sort of authorisation to get the MINUIT
> sources).
> This was well before the GPL (or at least before it became popular) --
> maybe
> they've changed to that nowadays. In any case, I use full-blown Cygwin
> when I'm not using Scientific Linux CERN so I'm not able to offer MinGW
> libs unless someone gives me a few pointers.
NAG's first f90 compiler. After NAG's bugs had been ironed out, I had
250,000 lines of code working with it. Of course, CERNLIB never was
converted to f90, but that's a (very) long story.
Regards,
Mike Metcalf
On May 19, 3:47 pm, "fake" <a@a.a> wrote:
> yesterday i saw that mistake.. > i wrote:
> program matrice
> implicit none
> real(8), dimension(20,20)::a
> real(8), dimension(20)::b
> integer::i,j,ifail
integer:: IR
> dimension IR(20) > do i=1,20 > do j=1,20 > if (i==j) then > a(i,j)=real(i) > else if (j==(i-1)) then > a(i,j)=0.5*sqrt(real(i+1)) > else if (j==(i+1)) then > a(i,j)=0.5*sqrt(real(i+1)) > else > a(i,j)=0.0 > end if > end do > end do > b(1)=1.0 > do i=2,20 > b(i)=0.0 > end do > call deqn(20,a,20,IR,IFAIL,1,b) > print*,b(2) > end program > and i get -0.811863376101259
> i hope that it is the right answer for the solution of x2
Compiling with g95 (4.1) on XP, my output is:
-0.8118633413147692
Although the calculation is carried out in double precision, the
result may only be good to single precision.
As a check, multiplying A * X (in Excel) gives for B:
1
-2.9976E-15
-2.52576E-15
-9.88792E-16
6.00214E-16
-7.80626E-17
-1.0842E-17
-4.11997E-18
1.08759E-18
-2.3484E-19
2.76874E-20
-1.31224E-20
6.00533E-22
-2.92305E-22
1.12962E-23
-3.6383E-24
5.51319E-25
-4.50345E-26
-8.70902E-28
1.21169E-27
-- e-mail: epc8 at juno dot com
|
|
|
|
|
|
Add to del.icio.us |
Digg this |
Stumble it |
Powered by Megasolutions Inc