Subject: Re: Reflector topics
From: Alex Zamfirescu (hxml@pacbell.net)
Date: Thu Nov 20 2003 - 12:59:16 PST
Jim:
You might have used the wrong reflector to reply to my E-mail.
I have to post a reply here to make sure all who read
your e-mail have the opportunity to read my reply too.
There is a lot of information regarding what works
and what does not that can be exchanged without
any reference to specific vendor tools. You can
also use "vendor A" and "vendor B", if you need
to talk about correlated findings. In any case mentioning
company names is not the way to go.
Keep in mind that IEEE reflector discussion is or
might become public information, and that members
in the IEEE groups are from different companies,
and should not make in any way, voluntarily or
involuntarily any public statements about commercial
products, features, bugs, prices etc.
I do not know why other reflectors accept discussion
about tool bugs or features (a la Cooley), but I know that
the IEEE ones will not allow that. As a chair of the group
I have the responsibility to advice you, to stay under
the safe guidelines of the IEEE.
As you well know the 1076.3 group does not have a
sponsor (like Accellera). We are all pure volunteers and the
standard is developed under the IEEE. Groups who
have coalition feeders can rely on the bylaws of the
feeder organization to conduct some industry validation.
We have to do it all by ourselves, and we do this for
basic and fundamental standard information (the numerics
semantic). This is not a cry for help, it is just to point that
your question is anchored in reality.
There is another answer to the issue. That is to go higher
in the level of abstraction of the specification and have
compliant package implementations be worked by
industry groups. Those groups can make whatever rules
they can find, that still comply with anti-trust regulations.
Unfortunately, such approach did not seem to get
consideration from the DASC and/or DASC SC.
Last time I tried to place this topic on the agenda
it was reordered to the end of the topics list and
then not considered from lack of time.
Note that at the same meeting more than half the time
was used to discuss who is a member and who is not
a member of DASC and to chose next meeting venues.
So, until better times, we will have to continue on the current
path, observing all IEEE rules.
Jim, please use the fphdl reflector for the discussions of the
issues related to the group. If you wish to go higher to address the
DASC or even the DASC SC please formulate your own
questions and proposals, and better, do not forward entire
threads from fphdl.
Please contact me if there is any problem.
Best regards,
Alex Zamfirescu
Chair 1076.3 WG
CTO ASC
IEC USNC Technical Advisor
Phone: 650 473 1067
Cell: 415 412 6903
E-mail: hxml<at>pacbell<dot>net
or azro<at>onebox<dot>com
or alexz<at>ascinc<dot>com
(replace <at> by @ and <dot> by .)
----- Original Message -----
From: "Jim Lewis" <Jim@synthworks.com>
To: "DASC" <stds-dasc@dasc.org>
Sent: Thursday, November 20, 2003 10:54 AM
Subject: Reflector topics
> groups, packages are
> being developed and tested. In some of the VHDL working As people are beta
> testing these packages we need to be able to talk
> about what in the packages works with which tools
> (and which version of a particular tool).
>
> Are we allowed to have this type of discussion on an
> EDA.org reflector that is the offical reflector for
> the working group? Are we allowed to post this type
> of information on the website?
>
> If not, how do we communicate this information?
> How can we do any beta testing of the packages
> before standardizing them if we don't discuss
> this type of information?
>
> Best Regards,
> Jim Lewis
>
>
>
> -------- Original Message --------
> Subject: Re: Synthesizer problemo!!
> Date: Wed, 19 Nov 2003 20:22:11 -0800
> From: Alex Zamfirescu <hxml@pacbell.net>
> To: <dbishop@vhdl.org>, "Javed" <javed_tata@yahoo.com>, <fphdl@eda.org>
> References: <20031119043325.99029.qmail@web11708.mail.yahoo.com>
<3FBB7310.70903@vhdl.org>
>
> Dave:
>
> I know we are trying to come up with packages that are useful to
> the industry. In doing that we should not forget about some rules to
> follow regarding company names and products discussed over
> the IEEE reflector. Please refrain answering questions about
> specific products. You can use reports like for example,
> "it works on two out of three implementations I tried",
> or "it does not work on any implementation."
> The line has to be drown at the place where our technical
> discussion might be interpreted as providing information
> that will distinguish specific products. If you have any
> questions please contact me at 415 412 6903.
>
> By the way, great job in keeping this volunteer effort
> alive!
>
> Best regards,
>
> Alex Zamfirescu
> Chair 1076.3 WG
> CTO ASC
> IEC USNC Technical Advisor
> Phone: 650 473 1067
> Cell: 415 412 6903
> E-mail: hxml<at>pacbell<dot>net
> or azro<at>onebox<dot>com
> or alexz<at>ascinc<dot>com
> (replace <at> by @ and <dot> by .)
>
> ----- Original Message -----
> From: "David Bishop" <dbishop@vhdl.org>
> To: "Javed" <javed_tata@yahoo.com>; <fphdl@eda.org>
> Sent: Wednesday, November 19, 2003 5:41 AM
> Subject: Re: Synthesizer problemo!!
>
>
> > You have hit upon a basic "compromise" I had to make in
> > the floating point packages.
> >
> > Most of the major synthesis tools do not handle unsigned
> > division, unless it is a power of 2. I continually complain
> > to Synopsys, Exemplar, and Synplicity about this.
> >
> > To get around this issue: You will find a commented out
> > unsigned divide routine in the "division" and the "oneoverx"
> > algorithm. Uncomment them and division will be synthesizable.
> > All I did here was implement a subtraction tree.
> >
> > In my testcases, I have "_reduced" version of the main package
> > body which works with Synopsys (Presto), Exemplar, and Synplicity
> > with a few reported bugs. Should I post it to the page? Useful?
> >
> > I also have an "_synopsys2002" version for people using the older
> > Synopsys compiler.
> >
> > Javed wrote:
> > > Hi David,
> > > I tried to implement the basic functions using
> > > Leonardospectrum level 3 synthesizer...but it does not
> > > recognise the division at all and thus functions like
> > > division, exp, sqrt, sin cos were not
> > > synthesize...whereas those that dont involve division
> > > like addition sub and multi went thru...I would like
> > > to know which synthesizer would synthesize it and is
> > > there any way that I could synthesize using leonardo??
> > > Thnaks for your continous support
> > > Javed
> >
> > --
> > David W. Bishop dbishop@vhdl.org All standard disclaimers apply.
> >
> >
> >
>
This archive was generated by hypermail 2b28 : Thu Nov 20 2003 - 13:04:43 PST