Passing and receiving different SystemVerilog datatypes with C/C++ using DPI

dt.cc

#include <stdio.h>
#include "dt.h"

void
add_bpuv(
    const svBitVecVal* a,
    const svBitVecVal* b,
    svBitVecVal* c) {
	for(int i = 0; i < 4; i++) {

		c[i] = a[i] + b[i];
	}
}

void
and_buv(
    const svBit* a,
    const svBit* b,
    svBit* c) {
	for(int i = 0; i < 4; i++) {
		c[i] = a[i] & b[i];
	}
}

void
add_bpv(
    const svBitVecVal* a,
    const svBitVecVal* b,
    svBitVecVal* c) {
	*c = *a + *b;
}

void
add_lpv(
    const svLogicVecVal* a,
    const svLogicVecVal* b,
    svLogicVecVal* c) {
	c->aval = a->aval + b->aval;
	c->bval = 0;
}

void
add_struct(
    const struct_t* a,
    const struct_t* b,
    struct_t* c) {
	c->mem1 = a->mem1 + b->mem1;
	c->mem2[0].aval = a->mem2[0].aval + b->mem2[0].aval;
	c->mem2[0].bval = 0;
	c->mem3[0] = a->mem3[0] + b->mem3[0];
}

svBit
and_bit(
    svBit a,
    svBit b) {
	return a && b;
}

svLogic
and_logic(
    svLogic a,
    svLogic b) {
	return a && b;
}

void
or_luv(
    const svLogic* a,
    const svLogic* b,
    svLogic* c) {
	for(int i = 0; i < 4; i++) {
		c[i] = a[i] | b[i];
	}
}

dt_restart.do

run -all

Makefile

dt: dt_compile dt_csim

.PHONY: dt dt_compile dt_csim clean

dt_compile:
	if [ -d work ]; then vdel -lib work -all; fi;
	vlib work
	vmap work work
	vlog -sv -dpiheader dt.h tb_dt_dpi.sv
	g++ -shared -Bsymbolic -I/home/$(whoami)/modelsim/modeltech/include/ -fPIC -o dt.so dt.cc

dt_csim:
	LD_LIBRARY_PATH=/home/$(whoami)/modelsim/modeltech/lib64/; vsim -64 -c -sv_lib dt -do dt_restart.do -novopt tb_dt_dpi

clean:
	vdel -all

tb_dt_dpi.sv

typedef struct { 
  shortint mem1; 
  logic [7:0] mem2; 
  bit [7:0] mem3; 
} struct_t;

module tb_dt_dpi(clk);
	input reg clk;
	typedef bit bitvec3up [3:0];
	typedef logic logicvec3up [3:0];

	bit b1, b2, b3;
	logic l1, l2, l3;
	bit [3:0] b4p1, b4p2, b4p3;
	logic [3:0] l4p1, l4p2, l4p3;
	bit b4u1[3:0], b4u2[3:0], b4u3 [3:0];
	logic l4u1[3:0], l4u2[3:0], l4u3 [3:0];
	bit [3:0] b4p4u1[3:0], b4p4u2[3:0], b4p4u3 [3:0];

	struct_t struct1, struct2, struct3;

	always #10 clk = ~clk;

	initial begin
		$display("Sending and receiving bit through DPI: ");
		b1 = 1'b1;
		b2 = 1'b1;
		b3 = and_bit(b1, b2);
		$display("%d b1=%b b2=%b b3=%b", $time, b1, b2, b3);

		$display("Sending and receiving logic through DPI: ");
		l1 = 1'b1;
		l2 = 1'b1;
		l3 = and_logic(l1, l2);
		$display("%d l1=%b l2=%b l3=%b", $time, l1, l2, l3);

		$display("Sending and receiving packed bit vector(size = 4) through DPI: ");
		b4p1 = 4'b1010;
		b4p2 = 4'b0010;
		add_bpv(b4p1, b4p2, b4p3);
		$display("%d b4p1=%b b4p2=%b b4p3=%b", $time, b4p1, b4p2, b4p3);

		$display("Sending and receiving packed logic vector(size = 4) through DPI: ");
		l4p1 = 4'b1010;
		l4p2 = 4'b0010;
		add_lpv(l4p1, l4p2, l4p3);
		$display("%d l4p1=%b l4p2=%b l4p3=%b", $time, l4p1, l4p2, l4p3);

		$display("Sending and receiving unpacked bit vector(size = 4) through DPI: ");
		b4u1 = bitvec3up'(4'b1010);
		b4u2 = bitvec3up'(4'b0010);
		and_buv(b4u1, b4u2, b4u3);
		$display("%d b4u1=%p b4u2=%p b4u3=%p", $time, b4u1, b4u2, b4u3);

		$display("Sending and receiving unpacked logic vector(size = 4) through DPI: ");
		l4u1 = logicvec3up'(4'b1010);
		l4u2 = logicvec3up'(4'b0010);
		or_luv(l4u1, l4u2, l4u3);
		$display("%d l4u1=%p l4u2=%p l4u3=%p", $time, l4u1, l4u2, l4u3);

		$display("Sending and receiving packed and unpacked bit vector(size = 4) through DPI: ");
		b4p4u1 = '{5, 4, 2, 1};
		b4p4u2 = '{5, 4, 2, 1};
		add_bpuv(b4p4u1, b4p4u2, b4p4u3);
		$display("%d b4p4u1=%p b4p4u2=%p b4p4u3=%p", $time, b4p4u1, b4p4u2, b4p4u3);

		$display("Sending and receiving structures through DPI: ");
		struct1 = '{1, 2, 3};
		struct2 = '{4, 5, 6};
		add_struct(struct1, struct2, struct3);
		$display("%d struct1=%p struct2=%p struct3=%p", $time, struct1, struct2, struct3);
		$finish;
	end

	import "DPI-C" context function bit and_bit(input bit a, b);
	import "DPI-C" context function logic and_logic(input logic a, b);
	//cannot return packed bit or logic arrays
	import "DPI-C" context function void add_bpv(input bit [3:0] a, b, output bit [3:0] c);
	import "DPI-C" context function void add_lpv(input logic [3:0] a, b, output logic [3:0] c);
	//cannot return unpacked bit arrays
	import "DPI-C" context function void and_buv(input bit a[3:0], b [3:0], output bit c [3:0]);
	import "DPI-C" context function void or_luv(input logic a[3:0], b [3:0], output logic c [3:0]);
	import "DPI-C" context function void add_bpuv(input bit [0:3]a[0:3], b [0:3], output bit [0:3]c [3:0]);
	import "DPI-C" context function void add_struct(input struct_t a, b, output struct_t c);
endmodule

Can you add an example for packed arrays wider than 32 bits? e.g. "bit [255:0]c"