Re: [Plplot-devel] Recently spotted issues with libnn and libcsa

[Arjen M. <arj...@wl...>]

Jo=E3o Cardoso wrote:
>=20
> On Thursday 25 September 2003 01:45, Alan W. Irwin wrote:
> |  On 2003-09-25 00:27+0100 Jo=E3o Cardoso wrote:
> |  > On Wednesday 24 September 2003 20:26, Alan W. Irwin wrote:
> |  > |  You misunderstood what I said. I am not advocating incorporatin=
g the
> |  > |  relevant part of lapack source into plplot. As you say, that is=
 much
> |  > | too big, and also I would be concerned about supporting it.  Ins=
tead,
> |  > | I am advocating simply *linking* to lapack which should be
> |  > | straightforward with the help of cfortran.h.
> |  >
> |  > Yes, that is the fast and simplest solution. But overkill; after a=
ll we
> |  > are replacing a 230 lines function by a whole library! It would be=
 much
> |  > better if we could find a single file svd.c.
> |
> |  If you could find ~200-line SVD code written in C
>=20
> I was talking about the SVD NR code included in csa.c, it's 232 lines.
>=20
> I had already searched netlib and gams, but stoped at (c)lapack; howeve=
r,
> eispack has an almost self-contained svd.f:
> http://www.netlib.org/cgi-bin/netlibfiles.pl?filename=3D/eispack/svd.f
>=20
> After converting it to C with the help of f2c, it resulted in a 577 lin=
es file
> (the original fortrans code has 358 lines of code). However I get an
> undefined symbol when trying to link with a dummy main():
>=20
> jcard@linux:~/Desktop/eispack> gcc test.c svd.o pythag.o -o test -lm
> svd.o(.text+0x363): In function `svd_':
> : undefined reference to `d_sign'
> svd.o(.text+0x784): In function `svd_':
> : undefined reference to `d_sign'
> svd.o(.text+0x16ae): In function `svd_':
> : undefined reference to `d_sign'
> collect2: ld returned 1 exit status
>=20
> Of course using f2c means linking with -lf2c, but perhaps that could be
> avoided in our case, as we don't need entry/exit points. Your knowledge=
 of
> fortran can be helpfull -- I don't even know how to call the subrotine =
:(
> Can you help?
>=20

>From the Fortran code I see that dsign() is a function. I do not know
what f2c makes out of this, but this is my educated guess:

double d_sign( double v, double w )
{
   if { w >=3D 0.0 } {=20
      return abs(v) ;=20
   } else {
      return -abs(v) ;
   }
}

(You transfer the sign of w to the first argument)

Regards,

Arjen