Conference turris::pascal

2668.0. "Percision Question" by NYOSS1::BUONOMO (Show me the MONKEY) Tue Apr 08 1997 00:08

    Hi All,
    
    Can anyone comment on these questions from my customer.
    
    Thanks,
    Lou
    
    
    
Lou,

We have a major problem with the Pascal compiler on Alpha for double
precision support.  We have raised with DEC (Asia) here for nearly half a year
without getting anywhere.  As you seem to have better support from DEC in
the States, wonder if you could raise this problem with them to help us
resolve.  Most feeds here now require double precision for some of the
fields, and the inability to support it on TQS will severely hamper the
migration.

Many thanks for your help in advance.

You can get more details from Tim if needed.

Regards
GH

--- Forwarded Mail Item: 1.2 -----------------------------------

Subject:   TQS Double Precision Support in Feed Migration


Mon 7 Apr 1997


TQS DOUBLE PRECISION SUPPORT IN FEED MIGRATION

All,
    Note that when TQS had to withdraw double precision support from
V1.0 it was due to a fundamental AXP PASCAL problem, not to do with
TQS s/w or defintions, capacity or issue types.

PASCAL is unable to correctly process DOUBLE, T_DOUBLE or T_RQS_DOUBLE
types, i.e. any 64-bit double precision REAL numbers.

Manoj has discovered a feed that uses DOUBLE processing. Pls review
your plans and rules.

As advised to Manoj, all instances of double precision mathematics
should be flagged to systems (myself, TQS team, Joe?) with details of
purpose, inputs and required output, so we may be able to assist in
removing the dependancy.

The current plan in the TQS roadmap is a move to C++ marketrules, so
removing the use of a very untrustworthy PASCAL compiler. Your
thoughts, support/non=support and feedback on this strategy is
required before I can proceed.

rgds
Tim
    

    re .0
    
    What platform? operating system? version? compiler version?
    
    Since there does not appear to be much specification of the problem
    perhaps a small but complete demonstration program would be in order.
    It seems likely that one would need also to mention what compile
    'options' are in use.
    
    Meanwhile what about "DOUBLE" do you need? i.e. what range of numbers
    needs to be represented? what precision do these numbers need to be
    represented to?

    
    I just asked the customer for all that info.
    About all I know its a VMS system.....
    
    Thanks,
    Lou

    I wonder if this is a D_float vs. G_float problem.
    
    				Steve

    Is this Reuters in Singapore?  I have exchanged mail with Mong-Kia Tan
    from note #2666 with a conversation very similar to this.  Namely,
    the customer thinks they can't trust DOUBLE precision on OpenVMS Alpha
    as well as finding other "problems".  However, they have never sent the
    problems/questions to us (or at least they never reached us).  
    
    As for the comment in .0 from the unknown customer:  We know of no
    problems with DOUBLE precision floating types on any DEC Pascal
    platform.  Please have the customer provide more detailed information
    (including sample programs) that shows the question.  If there are
    problems, we will fix them.
    
    				-John

    John,
    
    Good Guess. Yes this is from Reuters Singapore, it has now been
    transfered to Reuters America... so now it's my problem. I will look
    at note 2666 and get a complete "problem" statement from them and post
    it. This group is a bit trying to deal with. I'll keep you posted.
    
    Cheers,
    Lou