The extract pass¶
Like the
techmap
pass, theextract
pass is called with a map file. It compares the circuits inside the modules of the map file with the design and looks for sub-circuits in the design that match any of the modules in the map file.If a match is found, the
extract
pass will replace the matching subcircuit with an instance of the module from the map file.In a way the
extract
pass is the inverse of the techmap pass.
Example code can be found in docs/source/code_examples/macc
.
read_verilog macc_simple_test.v
hierarchy -check -top test;;
extract -constports -map macc_simple_xmap.v;;
module test(a, b, c, d, y);
input [15:0] a, b;
input [31:0] c, d;
output [31:0] y;
assign y = a * b + c + d;
endmodule
module macc_16_16_32(a, b, c, y);
input [15:0] a, b;
input [31:0] c;
output [31:0] y;
assign y = a*b + c;
endmodule
module test(a, b, c, d, x, y);
input [15:0] a, b, c, d;
input [31:0] x;
output [31:0] y;
assign y = a*b + c*d + x;
endmodule
module test(a, b, c, d, x, y);
input [15:0] a, b, c, d;
input [31:0] x;
output [31:0] y;
assign y = a*b + (c*d + x);
endmodule
The wrap-extract-unwrap method¶
Often a coarse-grain element has a constant bit-width, but can be used to implement operations with a smaller bit-width. For example, a 18x25-bit multiplier can also be used to implement 16x20-bit multiplication.
A way of mapping such elements in coarse grain synthesis is the wrap-extract-unwrap method:
- wrap
Identify candidate-cells in the circuit and wrap them in a cell with a constant wider bit-width using
techmap
. The wrappers use the same parameters as the original cell, so the information about the original width of the ports is preserved. Then use theconnwrappers
command to connect up the bit-extended in- and outputs of the wrapper cells.- extract
Now all operations are encoded using the same bit-width as the coarse grain element. The
extract
command can be used to replace circuits with cells of the target architecture.- unwrap
The remaining wrapper cell can be unwrapped using
techmap
.
Example: DSP48_MACC¶
This section details an example that shows how to map MACC operations of arbitrary size to MACC cells with a 18x25-bit multiplier and a 48-bit adder (such as the Xilinx DSP48 cells).
Preconditioning: macc_xilinx_swap_map.v
Make sure A
is the smaller port on all multipliers
(* techmap_celltype = "$mul" *)
module mul_swap_ports (A, B, Y);
parameter A_SIGNED = 0;
parameter B_SIGNED = 0;
parameter A_WIDTH = 1;
parameter B_WIDTH = 1;
parameter Y_WIDTH = 1;
input [A_WIDTH-1:0] A;
input [B_WIDTH-1:0] B;
output [Y_WIDTH-1:0] Y;
wire _TECHMAP_FAIL_ = A_WIDTH <= B_WIDTH;
\$mul #(
.A_SIGNED(B_SIGNED),
.B_SIGNED(A_SIGNED),
.A_WIDTH(B_WIDTH),
.B_WIDTH(A_WIDTH),
.Y_WIDTH(Y_WIDTH)
) _TECHMAP_REPLACE_ (
.A(B),
.B(A),
.Y(Y)
);
endmodule
Wrapping multipliers: macc_xilinx_wrap_map.v
(* techmap_celltype = "$mul" *)
module mul_wrap (A, B, Y);
parameter A_SIGNED = 0;
parameter B_SIGNED = 0;
parameter A_WIDTH = 1;
parameter B_WIDTH = 1;
parameter Y_WIDTH = 1;
input [A_WIDTH-1:0] A;
input [B_WIDTH-1:0] B;
output [Y_WIDTH-1:0] Y;
wire [17:0] A_18 = A;
wire [24:0] B_25 = B;
wire [47:0] Y_48;
assign Y = Y_48;
wire [1023:0] _TECHMAP_DO_ = "proc; clean";
reg _TECHMAP_FAIL_;
initial begin
_TECHMAP_FAIL_ <= 0;
if (A_SIGNED || B_SIGNED)
_TECHMAP_FAIL_ <= 1;
if (A_WIDTH < 4 || B_WIDTH < 4)
_TECHMAP_FAIL_ <= 1;
if (A_WIDTH > 18 || B_WIDTH > 25)
_TECHMAP_FAIL_ <= 1;
if (A_WIDTH*B_WIDTH < 100)
_TECHMAP_FAIL_ <= 1;
end
\$__mul_wrapper #(
.A_SIGNED(A_SIGNED),
.B_SIGNED(B_SIGNED),
.A_WIDTH(A_WIDTH),
.B_WIDTH(B_WIDTH),
.Y_WIDTH(Y_WIDTH)
) _TECHMAP_REPLACE_ (
.A(A_18),
.B(B_25),
.Y(Y_48)
);
endmodule
Wrapping adders: macc_xilinx_wrap_map.v
(* techmap_celltype = "$add" *)
module add_wrap (A, B, Y);
parameter A_SIGNED = 0;
parameter B_SIGNED = 0;
parameter A_WIDTH = 1;
parameter B_WIDTH = 1;
parameter Y_WIDTH = 1;
input [A_WIDTH-1:0] A;
input [B_WIDTH-1:0] B;
output [Y_WIDTH-1:0] Y;
wire [47:0] A_48 = A;
wire [47:0] B_48 = B;
wire [47:0] Y_48;
assign Y = Y_48;
wire [1023:0] _TECHMAP_DO_ = "proc; clean";
reg _TECHMAP_FAIL_;
initial begin
_TECHMAP_FAIL_ <= 0;
if (A_SIGNED || B_SIGNED)
_TECHMAP_FAIL_ <= 1;
if (A_WIDTH < 10 && B_WIDTH < 10)
_TECHMAP_FAIL_ <= 1;
end
\$__add_wrapper #(
.A_SIGNED(A_SIGNED),
.B_SIGNED(B_SIGNED),
.A_WIDTH(A_WIDTH),
.B_WIDTH(B_WIDTH),
.Y_WIDTH(Y_WIDTH)
) _TECHMAP_REPLACE_ (
.A(A_48),
.B(B_48),
.Y(Y_48)
);
endmodule
Extract: macc_xilinx_xmap.v
module DSP48_MACC (a, b, c, y);
input [17:0] a;
input [24:0] b;
input [47:0] c;
output [47:0] y;
assign y = a*b + c;
endmodule
… simply use the same wrapping commands on this module as on the design to
create a template for the extract
command.
Unwrapping multipliers: macc_xilinx_unwrap_map.v
module \$__mul_wrapper (A, B, Y);
parameter A_SIGNED = 0;
parameter B_SIGNED = 0;
parameter A_WIDTH = 1;
parameter B_WIDTH = 1;
parameter Y_WIDTH = 1;
input [17:0] A;
input [24:0] B;
output [47:0] Y;
wire [A_WIDTH-1:0] A_ORIG = A;
wire [B_WIDTH-1:0] B_ORIG = B;
wire [Y_WIDTH-1:0] Y_ORIG;
assign Y = Y_ORIG;
\$mul #(
.A_SIGNED(A_SIGNED),
.B_SIGNED(B_SIGNED),
.A_WIDTH(A_WIDTH),
.B_WIDTH(B_WIDTH),
.Y_WIDTH(Y_WIDTH)
) _TECHMAP_REPLACE_ (
.A(A_ORIG),
.B(B_ORIG),
.Y(Y_ORIG)
);
endmodule
Unwrapping adders: macc_xilinx_unwrap_map.v
module \$__add_wrapper (A, B, Y);
parameter A_SIGNED = 0;
parameter B_SIGNED = 0;
parameter A_WIDTH = 1;
parameter B_WIDTH = 1;
parameter Y_WIDTH = 1;
input [47:0] A;
input [47:0] B;
output [47:0] Y;
wire [A_WIDTH-1:0] A_ORIG = A;
wire [B_WIDTH-1:0] B_ORIG = B;
wire [Y_WIDTH-1:0] Y_ORIG;
assign Y = Y_ORIG;
\$add #(
.A_SIGNED(A_SIGNED),
.B_SIGNED(B_SIGNED),
.A_WIDTH(A_WIDTH),
.B_WIDTH(B_WIDTH),
.Y_WIDTH(Y_WIDTH)
) _TECHMAP_REPLACE_ (
.A(A_ORIG),
.B(B_ORIG),
.Y(Y_ORIG)
);
endmodule
module test1(a, b, c, d, e, f, y);
input [19:0] a, b, c;
input [15:0] d, e, f;
output [41:0] y;
assign y = a*b + c*d + e*f;
endmodule
module test2(a, b, c, d, e, f, y);
input [19:0] a, b, c;
input [15:0] d, e, f;
output [41:0] y;
assign y = a*b + (c*d + e*f);
endmodule
Wrapping in test1
:
techmap -map macc_xilinx_wrap_map.v
connwrappers -unsigned $__mul_wrapper Y Y_WIDTH \
-unsigned $__add_wrapper Y Y_WIDTH;;
Wrapping in test2
:
techmap -map macc_xilinx_wrap_map.v
connwrappers -unsigned $__mul_wrapper Y Y_WIDTH \
-unsigned $__add_wrapper Y Y_WIDTH;;
Extract in test1
:
design -push
read_verilog macc_xilinx_xmap.v
techmap -map macc_xilinx_swap_map.v
techmap -map macc_xilinx_wrap_map.v;;
design -save __macc_xilinx_xmap
design -pop
extract -constports -ignore_parameters \
-map %__macc_xilinx_xmap \
-swap $__add_wrapper A,B ;;
Extract in test2
:
design -push
read_verilog macc_xilinx_xmap.v
techmap -map macc_xilinx_swap_map.v
techmap -map macc_xilinx_wrap_map.v;;
design -save __macc_xilinx_xmap
design -pop
extract -constports -ignore_parameters \
-map %__macc_xilinx_xmap \
-swap $__add_wrapper A,B ;;
Unwrap in test2
:
techmap -map macc_xilinx_unwrap_map.v;;