Subject: RE: Additional Comments on TBV Proposals
From: Jayaram Bhasker (JBhasker@esilicon.com)
Date: Wed Jul 30 2003 - 07:15:04 PDT
Steve:
If we use protected data types, we would have to declare a protected data type and its associated subprograms
for each kind of base type. Looks like a lot of work a user would have to do to start using say a fifo of integers. This
looks very much like a package approach.
Or did I misinterpret something? If so, maybe you can provide an example.
- bhasker
-----Original Message-----
From: Stephen Bailey [mailto:SBailey@model.com]
Sent: Monday, July 07, 2003 4:14 PM
To: 'vhdl-200x-tbv@eda.org'
Subject: Additional Comments on TBV Proposals
Bhasker, et al,
I was thinking that it would probably be better to define LISTs and FIFOs as predefined protected types. Instead of providing the predefined attributes, there would be protected type subprograms that would define the same functionality. (The VHDL source for these protected types would be similar to existing source for NOW and TextIO subprograms as many of them cannot be implemented in VHDL. That is OK as long as we semantically define the functionality in the LRM.)
As I see it, the benefit of defining them as protected types is that it would probably make both data structures more easily forward compatible with any subsequent additions of object-oriented capabilities in VHDL. I would expect OO capabilities to build on the existing protected type encapsulation and referencing syntax.
I also think that protected types could also be applied to associative and sparse arrays.
If a protected type approach is taken, I recommend defining them in the context of package STANDARD to avoid the need for adding yet another use clause in the VHDL source.
----
Stephen A. Bailey
TME, Model Technology
1811 Pike Road, Building #2, Suite F
Longmont, CO 80501
sbailey@model.com
303-775-1655 (mobile)
720-494-1202 (office)
720-494-0457 (fax)
www.model.com
This archive was generated by hypermail 2b28 : Wed Jul 30 2003 - 07:16:27 PDT