This is the mail archive of the gsl-discuss@sourceware.cygnus.com mailing list for the GSL project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Undefined identifiers for alphaev6-dec-osf4.0e


In regard to: Re: Undefined identifiers for alphaev6-dec-osf4.0e, PETER...:

>at least on Digital Unix 4.0c, there is no such macro in
>/usr/include/machines/fpu.h. There are 5 similar macros (IEEE_TRAP_ENABLE_INV,
>IEEE_TRAP_ENABLE_DZE, IEEE_TRAP_ENABLE_OVF, IEEE_TRAP_ENABLE_UNF,
>IEEE_TRAP_ENABLE_INE, IEEE_TRAP_ENABLE_MASK), but IEEE_TRAP_ENABLE_DNO. Could
>this be a difference between DU 4.0c and the later versions?

Peter-

You're right.  4.0D doesn't have it either.  There's also no mention of
denormalized operands in the man page for ieee_set_fp_control.  This means
that the capability apparently doesn't even exist in 4.0D and earlier.  4.0f
and later does have it -- 4.0e might -- I don't know as I don't have access
to a 4.0e system anywhere.  I don't see the change mentioned in the 4.0f
release notes, so I have to assume the change was made at 4.0e.

This means that fp-tru64.c won't compile on 4.0d and earlier.  It could be
made to compile (without support for GSL_IEEE_MASK_DENORMALIZED) on 4.0d
and earlier if we had a 

#if !defined(HAVE_IEEE_TRAP_ENABLE_DNO)
/* At 4.0d and earlier it was still called Digital Unix, so use that name in
 * the error.
 */
GSL_ERROR ("Sorry, this version of Digital Unix does not support denormalized operands.", GSL_EUNSUP) ;
#else
	/* use the code already in place, if IEEE_TRAP_ENABLE_DNO is available */

#endif



The simple/lame way to define the HAVE_IEEE_TRAP_ENABLE_DNO symbol (or
something shorter, suggestions welcome) would be to do so for all versions
of Tru64 of 4.0e and higher.  That could be done with a

	case "$host_os" in

		...

	esac

construct.

The better fix would be to AC_TRY_COMPILE a test program that tries to use
the IEEE_TRAP_ENABLE_DNO define.

Unless someone else is anxious to tackle this, I'll add a test to detect
this problem and submit a patch.  I'll be gone for more than a week starting
this coming weekend, so if I don't get it done before Saturday (not likely,
at this point) it may be 2 or 3 weeks before I get to it.

Tim
-- 
Tim Mooney                              mooney@dogbert.cc.ndsu.NoDak.edu
Information Technology Services         (701) 231-1076 (Voice)
Room 242-J1, IACC Building              (701) 231-8541 (Fax)
North Dakota State University, Fargo, ND 58105-5164




Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]