Subject: Comments on Proposals
From: Stephen Bailey (stephen@srbailey.com)
Date: Thu Jul 03 2003 - 18:34:52 PDT
FIFOs:
Can you add a call to get the largest size of an unconstrained FIFO? I know this is probably more important for high-level
modeling than for verification, but it should be easy to add. Perhaps call it Fifo_object'Max?
Since you can declare FIFO signals, then we need to define what constitutes an event or a transaction on the FIFO. Either that or
limit sensitivity on FIFO-typed signals to the 'Pop with wait or conditional checks such as wait until FIFO'empty or FIFO'full.
EVENTs:
Shouldn't events be type-less (or perhaps more accurately, shouldn't we treat *event* as a native type in VHDL):
signal sevent : event;
While the answer is implied, we need to ensure that the LRM integration disallows initial values specified in the declaration and
the referencing of attributes that return a value. Also, what attributes would be supported? I could see supporting 'Event,
'Transaction (same as 'Event), 'Delayed, 'Quiet and 'Stable. Others?
LISTs:
'UNIQUE should be defined analogous to 'SORT. Specifically, 'UNIQUE should be valid only if the element type has a visible "="
function defined for it. (Also, 'SORT should be more specific in identifying which relational operator "<", ">"?)
See FIFO discussion above for the need to identify events/transactions on signals of a LIST type (meaning of things like
'delayed). Need to provide ways to wait on a list as FIFO has defined.
With LISTs (may apply to FIFOs as well) consider enhancing the portion of the declaration that specifies the size of the list to
also allow specification of helper functions such as "=" for key lookup. Sure, if we had full OO capabilities in VHDL, this would
fall out naturally. However, the reality is that it is easier to add a few special cases than it is to add full OO capability. In
any case, it is not much of a stretch to expect templated classes (if they are ever to be added to VHDL) would be used in a similar
manner as this.
Keep up the good work.
--------------------------------------
Stephen Bailey
stephen@srbailey.com
---------------------------------------
This archive was generated by hypermail 2b28 : Thu Jul 03 2003 - 18:25:28 PDT