Ambigous operator '&'

[Mohammed A khader]

HI all,

While compiling my design in Quartus II 4.2 , I got following errors.
The piece of code is given below....

Error: VHDL error at Ctrl_Ram.Vhd(66): can't determine definition of
operator ""&"" -- found 2 possible definitions
Error: Verilog HDL or VHDL error at Ctrl_Ram.Vhd(66): Unconverted
VHDL-1402: ambiguous type: 'Regfile' or 'SIGNED'
Error: VHDL Type Conversion error at Ctrl_Ram.Vhd(66): can't determine
type of object or expression near text or symbol "UNSIGNED"
Error: Ignored construct Ctrl_Ram_Arch at Ctrl_Ram.Vhd(42) because of
previous errors


-- These subtypes are in a package....

subtype WORD is signed(15 downto 0);
subtype DWORD is unsigned(31 downto 0);


-- In the Architecture declaration part I am defining follwing
signals..


type Regfile is array (natural range<>) of WORD;
signal Regfile_Ctrl_High : Regfile(0 to 15); -- High Bank Registe File
signal Regfile_Ctrl_Low : Regfile(0 to 15); -- Low Bank Register File

signal data_out_high : WORD;
signal data_out_low : WORD;


-- And these are the concurrent statements in Architecture body...

-- Regfile Read Assignments
data_out_low <= Regfile_Ctrl_Low(TO_INTEGER(Addrs_In));
data_out_high <= Regfile_Ctrl_High(TO_INTEGER(Addrs_In));

-- Concatenation of Low and High to form 32 Control Word
Data32_Out <= unsigned(data_out_high & data_out_low); -- ERROR IS IN
THIS LINE...

Whats wrong with the last statment. I expect '&' operator to
concatenate the two signals. What could be the other meaning of '&'
operator. I think I am doing a silly mistake some where . Please help
me in resolving this..

Thanks.
-- Mohammed A Khader.

 

[Laurent Gauch]

-- Concatenation of Low and High to form 32 Control Word
Data32_Out <= unsigned(data_out_high & data_out_low); -- ERROR IS IN
THIS LINE...

-- just try
Data32_Out <= unsigned(data_out_high) & unsigned(data_out_low);

regards,
laurent gauch
www.amontec.com

 

[Mohammed A khader]

Thanks ! It Worked .. But both are logically equal. What was my
mistake ?

 

[Sebastian Weiser]

-- Concatenation of Low and High to form 32 Control Word
Data32_Out <= unsigned(data_out_high & data_out_low); -- ERROR IS IN
THIS LINE...

Whats wrong with the last statment. I expect '&' operator to
concatenate the two signals. What could be the other meaning of '&'
operator. I think I am doing a silly mistake some where . Please help
me in resolving this..

With '&' you can (1) concatenate two arrays. For example, when using two
arrays of std_login (aka std_logic_vector):
"000" & "111"
This will create a 6 element long array.
With '&' you can (2) append or (3) prepend a single element to an array:
"000" & '1'
'0' & "111"
This will create a 4 element long array. Note that it is syntactically
different from "0" & "111", which I described in (1).
With '&' you can (4) create a new array by concatenating two elements:
'0' & '1'
This will create a 2 element array.


Now, this works not only with std_logic_vector, but with any array. For
example, with this one:

type Regfile is array (natural range<>) of WORD;

Now we have two different interpretations of
data_out_high & data_out_low
It can either produce a signed(0 to 31) (VHDL-93) according to (1), or a
Regfile(0 to 1) according to (4).


Laurent showed one way to avoid this problem. My first guess would have
been something like
Data32_Out <= unsigned(signed(data_out_high) & signed(data_out_low));
which is closer to your original code. Maybe the following would work:
Data32_Out <= unsigned(signed'(data_out_high & data_out_low));
But I'm not sure. (Please note the tick.)


[xp and f'up2 comp.lang.vhdl]

Sebastian

 

[Mike Treseler]

But both are logically equal.
What was my mistake ?

Using & on your own array types
without writing an overload for "&".
Probably easier to use the predefined &
for unsigned as Laurent suggests.

If these are process statements, you could also say:

Data32_Out(31 downto 16) <= unsigned(data_out_high);
Data32_Out(15 downto 0) <= unsigned(data_out_low);

-- Mike Treseler

 

[Mohammed A khader]

HI Mike,

Using & on your own array types without writing an overload for "&".

my signals 'data_out_high' and 'data_out_low' are subtype of signed.
Hence I expect that sysnthesis tool does'nt consider it as a new type.

 

[Mike Treseler]

my signals 'data_out_high' and 'data_out_low' are subtype of signed.
Hence I expect that sysnthesis tool does'nt consider it as a new type.

When you declare an array type like Regfile,
you automatically get new "&" functions
to combine arrays and to add
an element to the array.
In this case, an element of Regfile
is indistinguishable from a WORD.

Data32_Out <= unsigned(data_out_high) & unsigned(data_out_low);

Works in all cases
because & for unsigned
is the only possibility

Data32_Out <= unsigned'(data_out_high & data_out_low);

Works only if you change the
WORD declaration to
unsigned(31 downto 0)
otherwise:
Ambiguous type: 'regfile' or 'signed'.


Data32_Out <= unsigned(data_out_high & data_out_low);

Ambiguous type: 'regfile' or 'signed'.
No easy way to make this do what
you expect without removing the
Regfile type declaration.

-- Mike Treseler

 

[Sebastian Weiser]

Using & on your own array types
without writing an overload for "&".

Hmm, that differs from my explanation.

I've never seen an explicit overloading of the "&" operator, since it is
AFAIK implicitly given for any array type.


Sebastian

 

[Mike Treseler]

I've never seen an explicit overloading of the "&" operator, since it is
AFAIK implicitly given for any array type.

I agree.
The OP's problem was
too many overloads,
not too few.

-- Mike Treseler