Re: CP-010 in address decoding

Subject: Re: CP-010 in address decoding
From: Jim Lewis (jim@synthworks.com)
Date: Sat Dec 07 2002 - 08:35:40 PST

Farrell,
    The issue is not the match function itself.
Historically std_match was implemented in numeric_std
for sul, slv, and sulv. Going forward, it would be
nice to have the std_match function which overloads
sul, slv, and sulv in std_logic_1164.

    The issue is if we move it from one to the other,
it may break some code (because it is correct to call
the match function by numeric_std.std_match(...))

    So now we really don't want to break code. How
do we do it? CP-010 suggests using a new function
named match.

     This opens a new issue. For going forward it would
be nice if everyone called match the same name. Hence,
perhaps we should put match in numeric_std (which is
also open).

     Based on your example, perhaps we should be
considering two types of match functions. One
that returns boolean and one that returns std_logic.

  SEL <= '1' when match(A(0 to AB-MAB-1), "1101...01")
         else '0';

  SEL <= EQUAL(A(0 to AB-MAB-1), "1101...01") ;

I tenatively gave the one which returns std_logic
a different name. I suspect this may be necessary
to avoid overloading issues. However we may also
want to consider the possibility of equal being
match.

Cheers,
Jim Lewis

Farrell Ostler wrote:
> Let AB be the number of address bits in a system.
> A memory is to be mapped into this space with the address size
> of the memory variable up to AB, in which case the memory would
> occupy the entire address space. (The actual size of the
> memory would be determined at elaboration when the system
> generics are evaluated.)
>
> The match function is an attractive primitive for doing the
> address decoding for the memory's select signal. Here are
> two reasons:
>
> 1. Match won't generate simulation warnings during the period of time
> from simulation start until the system address gets initialized
> out of the 'U' state, as happens, for example, with the overloaded
> "=" operator in numeric_std.
>
> 2. If there is a desire to economize the address decoding at the
> expense of aliasing the memory in a larger address range, the
> "match on don't care" behavior caters nicely.
>
> Let MAB be the number of address bits that the memory actually
> needs, let A: std_logic_vector(0 to AB-1) be the system
> address signal and let SEL: std_logic be the memory select signal.
> Then, the decoding could be done as, e.g.,
>
> SEL <= '1' when match(A(0 to AB-MAB-1), "1101...01")
> else '0';
>
> The address decoder would work fine for all MAB < AB
> but then fail when MAB = AB, although this later case is a perfectly
> valid system.
>
> Farrell Ostler
> Xilinx
>
> P.S. For reference, I have copied the proposed match implementation,
> below.
>
>
> FUNCTION match ( l, r : std_logic_vector ) RETURN BOOLEAN IS
> ALIAS lv : std_logic_vector(1 TO l'LENGTH) IS l;
> ALIAS rv : std_logic_vector(1 TO r'LENGTH) IS r;
> BEGIN
> IF ((l'LENGTH < 1) OR (r'LENGTH < 1)) THEN
> ASSERT FALSE
> REPORT "STD_LOGIC_1164.MATCH: null detected, returning FALSE"
> SEVERITY WARNING;
> RETURN FALSE;
> END IF;
> IF lv'LENGTH /= rv'LENGTH THEN
> ASSERT FALSE
> REPORT "STD_LOGIC_1164.MATCH: L'LENGTH /= R'LENGTH, returning FALSE"
> SEVERITY WARNING;
> RETURN FALSE;
> ELSE
> FOR i IN lv'LOW TO lv'HIGH LOOP
> IF NOT (match_table(lv(i), rv(i))) THEN
> RETURN FALSE;
> END IF;
> END LOOP;
> RETURN TRUE;
> END IF;
> END match;
>
> 

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jim Lewis
Director of Training             mailto:Jim@SynthWorks.com
SynthWorks Design Inc.           http://www.SynthWorks.com
1-503-590-4787

Expert VHDL Training for Hardware Design and Verification
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This archive was generated by hypermail 2b28 : Sat Dec 07 2002 - 08:36:12 PST