why does std_logic_arith suck?

[Toby]

What are some specific problems with the std_logic_arith library? From
what I am reading in these forums, it is better to use numeric_std, why
exactly is it better? Any general answers on this issue would be
great.

Also, what I have been doing is simple arithmetic on std_logic_vectors
(using std_logic_arith library). Sometimes it works and sometimes it
doesnt. I understand that it is better to used signed or unsigned
types for arithmetic operations and I am in the process of changing my
code to do that, I am just trying to understand some of what I have
seen. For example, for std_logic_vector signals A and B (using
std_logic_arith), the following code seems to work in one part of my
code, but the exact same code doesnt work if I put it in another part
of the code:

IF (A - B > 400) Then
blah blah blah

It seemed like if B was an input port (an actual entity input), it
behaves differently than if B is a signal. Also seemed different if I
put this line of code in a different process (i.e. everything wasn't
concurrent like it is supposed to be). I dunno, anyone seen any
wierdness like this?

 

[Ralf Hildebrandt]

Toby wrote:

What are some specific problems with the std_logic_arith library?

Also, what I have been doing is simple arithmetic on std_logic_vectors
....

This is one of the problems: How do you specify if these vectors are threatened as signed
or unsigned data? (AFAIK it is done implicitely in this lirbary.)

For some arithmetic operations it does not matter, but for a comparison it does.
constant a : std_ulogic_vector:=16#ff#;
constant b : std_ulogic_vector:=16#01#;
Is a = 255 or a = -1 respectively is b greather than a or not?


Ralf

 

[Mike Treseler]

Toby wrote:

What are some specific problems with the std_logic_arith library?

http://groups.google.com/groups?q=vhdl+problems+std_logic_arith+numeric_std

It seemed like if B was an input port (an actual entity input), it
behaves differently than if B is a signal.

An input port identifier must appear
only on the right side of an assignment operator.

A variable or signal identifier may appear
on the left, right or both.
Consider getting a vhdl simulator.

-- Mike Treseler

 

[john Doef]

Ralf Hildebrandt a écrit :

For some arithmetic operations it does not matter, but for a comparison it does.
constant a : std_ulogic_vector:=16#ff#;
constant b : std_ulogic_vector:=16#01#;
Is a = 255 or a = -1 respectively is b greather than a or not?
Please, please,

this is not correct vhdl.

JD.

 

[Mike Treseler]

john Doef wrote:

Ralf Hildebrandt a écrit :
For some arithmetic operations it does not matter, but for a comparison it does.
constant a : std_ulogic_vector:=16#ff#;
constant b : std_ulogic_vector:=16#01#;
Is a = 255 or a = -1 respectively is b greather than a or not?

Please, please,
this is not correct vhdl.
JD.

Indeed a typo on the x"ff" and x"01"
but Ralf's point to the OP is a good one.
Using a plain std_logic_vector as an integer is ambiguous.

-- Mike Treseler