Subject: [sv-cc] Re: Calling SV from C
From: Andrzej Litwiniuk (Andrzej.Litwiniuk@synopsys.com)
Date: Wed Feb 05 2003 - 10:35:16 PST
Francoise wrote:
>
> I don't understand section 5.2 calling SV functions from C:
>
> "Note that the functions exported from SV may not have open arrays as
> arguments.
> Otherwise the same types of formal arguments may be declared in SV both for
> the exported functions and for the external functions.
Well, what I wanted to say was that generally the exported functions may have
the same types of formal arguments as the external functions, with the exception
of open arrays.
I should have added also that similar restrictions are required for the function
result type. This is definitely missing.
> A function exported from SV will have the same function header as the external
> function with the same function result type and same formal argument list."
>
>
> My understanding was that there is an export declaration of the SV function
> in the Verilog code.
Right.
I suppose that the export declaration will only provide the function name
and the hierarchical path. The path will identify both the unique instance and
the unique function declaration. There is no need to repeat the specification of formal arguments and function result.
> There must also be an extern function declaration in C indicating that this
> function is extern to the C file.
> The formal arguments types, modes and argument passing of the C function
> extern declaration must comply with the rules which were defined for a SV
> function calling a directC function.
> Therefore the user declare this function and determine the matching types
> to the Verilog function
> data types.
Right.
> Example:
> typedef struct {int i; double r;} complex;
> export logic [15:0] myfunc (input int address, output logic [31:0] r, input
> T complex)
Two modifications.
First, function result types are restricted for the exported
functions in the same way as for the external functions.
Assume for the sake of this example, that function result type is "bit [15:0]"
rather than "logic [15:0]".
Second modification: 'myfunc' will be mentioned twice: in its primary
declaration and in an export declaration.
So, SV portion may look like:
typedef struct {int i; double r;} T; // "T" rather than "complex"?
export a.b.c.myfunc; // whatever is the current syntax ...
// a.b.c identifies a unique instance of a module where myfunc is defined
...
module M;
...
function bit [15:0] myfunc (input int address, output logic [31:0] r,
input T complex);
begin ... end endfunction
...
endmodule
> The extern function declaration which will be used would be:
> extern SVlogicVec32 myfunc (int address, SVlogicVec32 *r, T *complex)
> address is passed by value, r is passed by reference, complex is passed by
> reference, the function
> return is a 32 bit logic struct.
>
> Correct?
Almost correct. With the result type changed from 'logic' to 'bit'
the extern function declaration which will be used would be:
extern svBitVec32 myfunc (int address, svLogicPackedArrRef r, const T *complex);
> the function return is a 32 bit logic struct.
actually svBitVec32, because of the modification we had to make (logic-->bit)
> address is passed by value
Right.
> r is passed by reference
Right, svLogicPackedArrRef is a reference (actually typedefed to void *);
note that the representation is implementation-dependent.
> complex is passed by reference
Right, and qualified as 'const'.
> If I am not understanding this correctly, please explain and or reword
> section 5.2
I believe you are understanding this correctly, by and large.
> Another question I have is what would be the C matching type of a output SV
> function
> argument which is output logic r[31:0]; (unpacked array of 32 elements, 1
> bit wide).
svLogic *, I suppose. The actual size of an unpacked array is obviously absent.
Andrzej
This archive was generated by hypermail 2b28 : Wed Feb 05 2003 - 10:35:58 PST