Association list in component instantiations

[Thomas Heller]

I do not understand what I can use in the association list of a
component instantiation. Say, I have a component declaration like this:

entity spislave is
Port ( sysclock : in STD_LOGIC;
data_out : out STD_LOGIC_VECTOR(7 downto 0);
data_strobe : out STD_LOGIC;
sclk : in STD_LOGIC;
mosi : in STD_LOGIC;
cs : in STD_LOGIC);
end spislave;

Then a typical instantiation is:

U1: spislave PORT MAP(
sysclock => clk64,
data_out => spi_data,
data_strobe => spi_strobe,
sclk => spi_clk,
mosi => spi_mosi,
cs => spi_cs
);

If I want to connect the data_out signal to parts of a databus then
I cannot write
data_out => data_bus(7 downto 0),
although I can write
sclk => not pin_7;

Can someone point me to a reference that explains which kind of 'things'
I can use in the instantiation?

Thanks,
Thomas

 

[Jonathan Bromley]

I do not understand what I can use in the association list of a
component instantiation. Say, I have a component declaration like this:

entity spislave is
Port ( sysclock : in STD_LOGIC;
data_out : out STD_LOGIC_VECTOR(7 downto 0);
data_strobe : out STD_LOGIC;
sclk : in STD_LOGIC;
mosi : in STD_LOGIC;
cs : in STD_LOGIC);
end spislave;

Then a typical instantiation is:

U1: spislave PORT MAP(
sysclock => clk64,
data_out => spi_data,
data_strobe => spi_strobe,
sclk => spi_clk,
mosi => spi_mosi,
cs => spi_cs
);

If I want to connect the data_out signal to parts of a databus then
I cannot write
data_out => data_bus(7 downto 0),

Why not? Works for me. Who's complaining? Did you *really*
say (7 downto 0), or was there something non-static going on
in your slice range?

although I can write
sclk => not pin_7;

That's slightly different; the actual association can be
any monadic function of a signal. This was done to allow
for type conversions in the port map, but it works for
any monadic function and the "not" operator is, of course,
nothing more than the monadic function
function "not"(v: in std_logic) return std_logic;

Can someone point me to a reference that explains which kind of 'things'
I can use in the instantiation?

For an input port:
signals or slices thereof; expressions yielding a constant value.
For an output or inout port:
signals or slices thereof.
Conversion functions can be applied: they must be monadic,
and the syntax is different for input and output ports:

port map (some_input => F1(in_sig),
F2(some_output) => out_sig,
F3out(some_inout) => F3in(io_sig))

The actual must be a "static name" but that should be OK...

In VHDL-2008 you can put arbitrary expressions to an input port,
but your port will suffer an added delta cycle delay from the
implied process that computes the expression.

 

[Thomas Heller]

Am 30.01.2011 22:09, schrieb Jonathan Bromley:

Why not? Works for me. Who's complaining? Did you *really*
say (7 downto 0), or was there something non-static going on
in your slice range?

I was confused. Of course it works.

What I really want and what didn't work is connecting some bits of an
input signal on the component (say, the data bus of an ADC) to a signal,
and other bits to a constant. Example:

X: myinstance PORT MAP(
...
data_input => data_bus(7 downto 0) & "00000000",
...)

Is it possible to implement this without using another signal, like this:

adc_datainput <= data_bus(7 downto 0) & "00000000";
X: myinstance PORT MAP(
...
data_input => adc_input,
...)

That's slightly different; the actual association can be
any monadic function of a signal. This was done to allow
for type conversions in the port map, but it works for
any monadic function and the "not" operator is, of course,
nothing more than the monadic function
function "not"(v: in std_logic) return std_logic;


For an input port:
signals or slices thereof; expressions yielding a constant value.
For an output or inout port:
signals or slices thereof.
Conversion functions can be applied: they must be monadic,
and the syntax is different for input and output ports:

port map (some_input => F1(in_sig),
F2(some_output) => out_sig,
F3out(some_inout) => F3in(io_sig))

The actual must be a "static name" but that should be OK...

In VHDL-2008 you can put arbitrary expressions to an input port,
but your port will suffer an added delta cycle delay from the
implied process that computes the expression.

Thanks for this useful info,
Thomas

 

[Jonathan Bromley]

What I really want and what didn't work is connecting some bits of an
input signal on the component (say, the data bus of an ADC) to a signal,
and other bits to a constant. Example:

X: myinstance PORT MAP(
...
data_input => data_bus(7 downto 0) & "00000000",
...)

Is it possible to implement this without using another signal

Only in VHDL-2008, as far as I'm aware.

Have you tried

data_input(15 downto 8) => data_bus(7 downto 0),
data_input(7 downto 0) => "00000000"

??? I know you can do that to split a vector port out to
several different signals, but I'm not sure whether it's
legal to mix constants and signals in that way.

 

[Thomas Heller]

Am 03.02.2011 22:38, schrieb Jonathan Bromley:

Only in VHDL-2008, as far as I'm aware.

Have you tried

data_input(15 downto 8) => data_bus(7 downto 0),
data_input(7 downto 0) => "00000000"

??? I know you can do that to split a vector port out to
several different signals, but I'm not sure whether it's
legal to mix constants and signals in that way.

Seems to work. Cool!

Thanks,
Thomas

 

[rickman]

That's slightly different; the actual association can be
any monadic function of a signal.

Why did you use the term "monadic" rather than "unary"? Do I
misunderstand the meaning of either or both? It took me five minutes
to find a simple definition of "monadic function". That may sound
like a pretty short time, but when was the last time you spent five
minutes looking for a word definition? "Unary" is a much more common
term and I think more clear in this context unless there is some
useful connotation to "monadic" I don't appreciate.

Rick

 

[Jonathan Bromley]

Why did you use the term "monadic" rather than "unary"?

Thanks - that's a bit of jargon that I must have
incorrectly absorbed from somewhere. Now you've
provoked me into looking, I see that it's not even
strictly correct, at least not if you're a
mathematician or a (Haskell-style) functional
programmer. As you may guess, I am neither.

(Side note: some programming languages do tend to
use "monadic" and "dyadic" (and even "variadic")
to describe the number of arguments of a function.
APL is certainly one such. Maybe they're wrong too.)

"Unary" is a much more common term

And rigorously accurate too. The style-checker
in my head (which usually serves me tolerably well)
has no problem with "unary operator", but finds
the sound of "unary function" rather strange.
Time for a re-calibrate, maybe.

Ho hum. Still getting things wrong after all
these years :-(

 

[Andy]

The style-checker
in my head (which usually serves me tolerably well)
has no problem with "unary operator", but finds
the sound of "unary function" rather strange.
Time for a re-calibrate, maybe.

I agree, unary more commonly applies to operators with only one
operand. While all operators are functions (in vhdl), not all
functions are operators, and perhaps that would be a/the problem.
Would "unary function" imply a function that was invoked as a unary
operator?

What what would be the term for a function with zero arguments?
Nonadic? ;^)

OK, I did not look up monadic, but I took a lucky guess... and I knew
what dyadic meant, which helped a lot.

Andy

 

[rickman]

I agree, unary more commonly applies to operators with only one
operand. While all operators are functions (in vhdl), not all
functions are operators, and perhaps that would be a/the problem.
Would "unary function" imply a function that was invoked as a unary
operator?

What what would be the term for a function with zero arguments?
Nonadic? ;^)

OK, I did not look up monadic, but I took a lucky guess... and I knew
what dyadic meant, which helped a lot.

Andy

Maybe you should have looked up Nonadic...

nonadic (comparative more nonadic, superlative most nonadic)

1. of or pertaining to an nonad; ninefold


Rick

 

[rickman]

I agree, unary more commonly applies to operators with only one
operand. While all operators are functions (in vhdl), not all
functions are operators, and perhaps that would be a/the problem.
Would "unary function" imply a function that was invoked as a unary
operator?

What what would be the term for a function with zero arguments?
Nonadic? ;^)

OK, I did not look up monadic, but I took a lucky guess... and I knew
what dyadic meant, which helped a lot.

Andy

Or maybe you should have looked up "unary function"?

Rick