Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!coelho!drc
From: drc@coelho (David R. Coelho)
Newsgroups: comp.lang.vhdl
Subject: Re: Behavioral Model Effort Estimate
Message-ID: <220@coelho>
Date: 16 Apr 91 05:00:22 GMT
References: <6018@trantor.harris-atd.com>
Organization: Coelho Publications, Fremont, CA
Lines: 24



> estimate the effort required in writing VHDL behavioral models. The 
> models I have in mind would be detailed enough to provide the
> correct function at the module I/O pins so that tests written for the 
> behavioral model could also be run on the structural VHDL model for
> the same module to be written later.

A rough rule of thumb goes as follows:
    ssi, msi (<100 gates equiv):    1 man-day
    lsi (<1000 gate equiv):         1 man-week
    vlsi (<10,000 gate equiv):      1 man-month
    big vlsi (10,000+ gate equiv):  up to 1 man-year

Past experience modelling the 68000 using the Helix language resulted in a
1 man-year effort amounting to roughly 15,000 lines of code. Helix is very
similar to VHDL, and I would expect a similar result in effort and code in
VHDL.

> Also, how should timing and set-up and hold checks be included in 
> the behavioral model?

Your best bet is to use the VHDL "assertion" statement to establish timing
conditions which must be met.