Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

cellmatch: New pass for picking out standard cells automatically #4300

Merged
merged 6 commits into from
May 6, 2024

Conversation

povik
Copy link
Member

@povik povik commented Mar 27, 2024


    cellmatch -lib <design> [module selection]

This pass identifies functionally equivalent counterparts between each of the
selected modules and a module from the secondary design <design>. For every such
correspondence found, a techmap rule is generated for mapping instances of the
former to instances of the latter. This techmap rule is saved in yet another
design called '$cellmatch', which is created if non-existent.

This pass restricts itself to combinational modules. Modules are functionally
equivalent as long as their truth tables are identical upto a permutation of
inputs and outputs. The supported number of inputs is limited to 6.

@QuantamHD
Copy link
Contributor

You're a machine @povik!

@povik
Copy link
Member Author

povik commented Mar 27, 2024

This is useful for writing low-level structural Verilog with some degree of portability between cell libraries.

Say you write a Kogge-Stone carry lookahead unit out of AND, AOI2 gates with the expectation that the target cell library will contain cells resembling those.

module LCU(P, G, CI, CO);
	parameter WIDTH = 4;

	(* force_downto *)
	input [WIDTH-1:0] P, G;
	input CI;

	(* force_downto *)
	output [WIDTH-1:0] CO;

	localparam STAGES = $clog2(WIDTH);

	integer i, j;
	(* force_downto *)
	reg [WIDTH-1:0] p [STAGES:0], g [STAGES:0];

	assign p[0] = P;
	assign g[0] = G | (P[0] & CI);

	generate
		genvar i, j;

		for (i = 0; i < $clog2(WIDTH); i = i + 1) begin
			for (j = 0; j < 2**i; j = j + 1) begin
				assign g[i + 1][j] = g[i][j];
				assign p[i + 1][j] = p[i][j];
			end
			for (j = 2**i; j < WIDTH; j = j + 1) begin
				AOI2 aoi2(.A(g[i][j]), .B(p[i][j]), .C(g[i][j - 2**i]), .Y(g[i + 1][j]));
				AND and_(.A(p[i][j]), .B(p[i][j - 2**i]), .Y(p[i + 1][j]));
			end
		end
	endgenerate

	assign CO = g[STAGES];
endmodule

module AND(A, B, Y);
	input wire A;
	input wire B;
	output wire Y = A & B;
endmodule

module AOI2(A, B, C, Y);
	input wire A;
	input wire B;
	input wire C;
	output wire Y = A | B & C;
endmodule

You can map to those cells with the following script (here taking a SKY130 target as an example):

read_liberty -ignore_miss_func sky130_fd_sc_hd__tt_025C_1v80.lib
aigmap
design -save gatelib
design -reset

read_verilog design.v
prep
aigmap AND AOI2
cellmatch -lib gatelib AND AOI2
techmap -map %$cellmatch
dump

@povik
Copy link
Member Author

povik commented Mar 27, 2024

@QuantamHD Hah, you are quick to notice the PR. :)

Copy link
Member

@whitequark whitequark left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This could use comb cell eval instead of limiting itself to AIGs, no?

@QuantamHD
Copy link
Contributor

I think this is super cool, and really helpful to users! Thanks for working on it

@povik
Copy link
Member Author

povik commented Mar 27, 2024

This could use comb cell eval instead of limiting itself to AIGs, no?

Yes, but I leave dealing with the eval API for another day...

@povik povik marked this pull request as ready for review April 13, 2024 15:22
@povik
Copy link
Member Author

povik commented Apr 13, 2024

This could use comb cell eval instead of limiting itself to AIGs, no?

I made this change, also I updated the quoted help message above to what it is now.

@widlarizer
Copy link
Collaborator

widlarizer commented May 2, 2024

Please review and consider pulling in my comments
widlarizer@24bf8a5
EDIT: instead, I created PR #4366

@widlarizer widlarizer merged commit 68c7fc4 into main May 6, 2024
35 checks passed
@povik povik deleted the cellmatch branch May 6, 2024 13:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants