iverilog-pls vs iverilog-v11

README.md

iverilog-pls vs iverilog-v11

About PLS:

The project is developed based on icarus iverilog. Special thanks to Stephen Williams (steve@icarus.com).

PLS is a Verilog simulator, compiles verilog into a much faster optimized model, which executes about 20-100x faster than interpreted Verilog simulators such as icarus iverilog vvp.

Some performance metrics from authors.

Prepare container

Put Dockerfile and tb.sv in the same folder.

# clone repos to the same folder:
git clone -b v11-branch git@github.com:steveicarus/iverilog.git
git clone git@github.com:pytec8800/pint_iverilog.git

# build Docker container named 'iverilog_bench'
docker build -t iverilog_bench  .

Run containter

Run this to get an interactive console for further experiments:

docker run -i -v $(pwd):/project -u $(id -u):$(id -g) -t iverilog_bench

Another way is to run needed commands directly:

# simulate in iverilog-v11
docker run -i -v $(pwd):/project -u $(id -u):$(id -g) -t iverilog_bench iverilog-v11 tb.sv -g2012 && vvp-v11 -v a.out

# simulate in iverilog-pls
docker run -i -v $(pwd):/project -u $(id -u):$(id -g) -t iverilog_bench iverilog-pls tb.sv -g2012 && ./a.out

Or just

Results

My results for quite a synthetic testbench:

• iverilog-v11: ~2h
• iverilog-pls: ~3s

Dockerfile

# Prepare system
FROM ubuntu:20.04
RUN apt-get update
RUN apt-get install -y build-essential flex gperf bison libreadline6-dev libncurses5-dev autoconf

# Copy sources
COPY ./iverilog /src/iverilog
COPY ./pint_iverilog /src/pint_iverilog
WORKDIR /src

# Build iverilog-pls
RUN cd pint_iverilog && \
    sh autoconf.sh && \
    ./configure --enable-suffix="-pls" && \
    make -j$(nproc) && \
    make install

# Build iverilog-v11
RUN cd iverilog && \
    sh autoconf.sh && \
    ./configure --enable-suffix="-v11" && \
    make -j$(nproc) && \
    make install

# Prepare for work
WORKDIR /project
CMD ["/bin/bash"]

tb.sv

module xor_unit
#(
    parameter  DATA_W = 32
)(
  input  logic              clk,
  input  logic              rst_n,
  input  logic [DATA_W-1:0] in_a,
  input  logic [DATA_W-1:0] in_b,
  output logic [DATA_W-1:0] res
);

always_ff @(posedge clk or negedge rst_n) begin
  if (!rst_n) begin
    res <= '0;
  end else begin
    res <= in_a ^ in_b;
  end
end

endmodule


module tb;

localparam XOR_NUM = 512;
localparam DATA_W  = 32;

bit clk = 0;
always #5 clk = ~clk;

bit rst_n = 0;
initial #42 rst_n = 1;

logic [XOR_NUM-1:0][DATA_W-1:0] in_a;
logic [XOR_NUM-1:0][DATA_W-1:0] in_b;
logic [XOR_NUM-1:0][DATA_W-1:0] res;

logic [DATA_W-1:0] drv_a;
logic [DATA_W-1:0] drv_b;

always_ff @(posedge clk or negedge rst_n) begin
  if (!rst_n) begin
    drv_a <= '0;
    drv_b <= '1;
  end else begin
    drv_a <= drv_a + 7;
    drv_b <= drv_b - 13;
  end
end

for (genvar i = 0; i < XOR_NUM; i++) begin
  xor_unit #(
    .DATA_W(DATA_W)
  ) xor_i (
    .clk    (clk),
    .rst_n  (rst_n),
    .in_a   (in_a[i]),
    .in_b   (in_b[i]),
    .res    (res[i])
  );

  assign in_a[i] = (i == 0) ? drv_a : drv_a + res[i-1];
  assign in_b[i] = (i == 0) ? drv_b : drv_b + res[i-1];
end

logic [DATA_W-1:0] final_res;
always_comb begin
  final_res = '0;
  for (int i = 0; i < XOR_NUM; i++) begin
    final_res += res[i];
  end
end

initial begin
  #1000;
  $display("%0t Done! Result is %0x", $time, final_res);
  $finish;
end

endmodule