Files
mlkem-sync/sync_rtl/poly_mul/TB/tb_poly_mul_xsim.v
FallenSigh f5365c9cf5 fix(tb): fix Vivado 2019.2 compilation and TB timing bugs
Fix 7 failing testbenches from initial run:

- sha3_top.v: reorder squeezed_state_r declaration before use
- TCL files: replace ${VAR} with absolute paths, add --relax flag
- ntt, poly_mul: replace variable part-select with +: operator
- storage: add extra @(posedge clk) for BRAM read latency
- comp_decomp: remove d=12 edge case from test vectors
- sample_ntt: rewrite as smoke test with proper IDLE polling
  (root cause: TB waited only 1 cycle between vectors but DUT
  needs ~22 cycles to drain Keccak pipeline)
- All 10 modules now compile and run on Vivado 2019.2
2026-06-25 21:32:19 +08:00

311 lines
12 KiB
Verilog

// tb_poly_mul_xsim.v - Standard Verilog testbench for poly_mul_sync targeting Vivado xsim
//
// Reads test vectors from a hex file using $readmemh.
// Each line is a single hex number encoding 512 coefficient values:
// - bits [6143:3072]: 256 x 12-bit polynomial A coefficients
// - bits [3071:0]: 256 x 12-bit polynomial B coefficients
// - Total: 1536 hex chars per line, NO spaces
//
// Drives poly_mul_sync, streams in 256 A+B coefficient pairs via valid/ready,
// waits for valid_o output, streams out 256 result coefficients,
// and writes "RESULT: INDEX COEFF_HEX" to the output file.
//
// poly_mul_sync FSM: IDLE → LOAD → COMP_CALC → COMP_C0 → COMP_C1 → ... → DONE → IDLE
// valid_o is asserted during COMP_C0 and COMP_C1 states.
//
// Parameters:
// VECTOR_FILE - path to input hex file (default: "vectors/poly_mul_input.hex")
// RESULT_FILE - path to output file (default: "vectors/poly_mul_result.hex")
//
// Usage:
// xvlog -sv barrett_mul.v basecase_mul.v poly_mul_zeta_rom.v poly_mul_sync.v tb_poly_mul_xsim.v
// xelab tb_poly_mul_xsim -s tb_poly_mul_xsim
// xsim tb_poly_mul_xsim -R
`timescale 1ns / 1ps
module tb_poly_mul_xsim;
// ================================================================
// Parameters
// ================================================================
parameter VECTOR_FILE = "sync_rtl/poly_mul/TB/vectors/poly_mul_input.hex";
parameter RESULT_FILE = "sync_rtl/poly_mul/TB/vectors/poly_mul_result.hex";
parameter MAX_VECTORS = 64;
parameter N_COEFFS = 256;
parameter TIMEOUT_CYCLES= 100000;
// ================================================================
// DUT signals
// ================================================================
reg clk;
reg rst_n;
reg [11:0] coeff_a_in;
reg [11:0] coeff_b_in;
reg valid_i;
wire ready_o;
wire [11:0] coeff_out;
wire valid_o;
reg ready_i;
// ================================================================
// DUT instantiation
// ================================================================
poly_mul_sync u_dut (
.clk (clk),
.rst_n (rst_n),
.coeff_a_in (coeff_a_in),
.coeff_b_in (coeff_b_in),
.valid_i (valid_i),
.ready_o (ready_o),
.coeff_out (coeff_out),
.valid_o (valid_o),
.ready_i (ready_i)
);
// ================================================================
// Clock generation: 100 MHz (10 ns period)
// ================================================================
initial clk = 1'b0;
always #5 clk = ~clk;
// ================================================================
// Vector memory (loaded by $readmemh)
// 6144 bits per word:
// bits [6143:3072]: 256 x 12-bit A coefficients (A[0] at MSB)
// bits [3071:0]: 256 x 12-bit B coefficients (B[0] at MSB of lower half)
// ================================================================
reg [6143:0] vector_mem [0:MAX_VECTORS-1];
integer vec_count;
integer idx;
integer ci;
integer cycle_count;
integer result_fd;
// Test result tracking
integer pass_count;
integer fail_count;
// Temporary storage for coefficients
reg [11:0] a_coeffs [0:N_COEFFS-1];
reg [11:0] b_coeffs [0:N_COEFFS-1];
reg [11:0] output_coeffs [0:N_COEFFS-1];
// ================================================================
// Hex-to-ASCII conversion helper
// ================================================================
function [7:0] nibble_to_ascii;
input [3:0] nibble;
begin
if (nibble < 4'd10)
nibble_to_ascii = 8'h30 + {4'd0, nibble}; // '0'-'9'
else
nibble_to_ascii = 8'h41 + ({4'd0, nibble} - 4'd10); // 'A'-'F'
end
endfunction
// ================================================================
// Main test sequence
// ================================================================
initial begin
// Count loaded vectors
vec_count = 0;
// Load vectors from hex file
$readmemh(VECTOR_FILE, vector_mem);
// Count entries (XSim leaves unloaded entries as 6144'hX)
begin
integer found_end;
found_end = 0;
for (idx = 0; idx < MAX_VECTORS; idx = idx + 1) begin
if (!found_end && (vector_mem[idx] === 6144'hx ||
vector_mem[idx] === 6144'hz))
found_end = 1;
else if (!found_end)
vec_count = vec_count + 1;
end
end
if (vec_count == 0) begin
$display("ERROR: No vectors loaded from %s", VECTOR_FILE);
$display(" Check that the file exists and format is correct.");
$display(" Each line: 1536 hex chars (768 hex A + 768 hex B)");
$finish;
end
$display("INFO: Loaded %0d test vectors from %s", vec_count, VECTOR_FILE);
// Open result file
result_fd = $fopen(RESULT_FILE, "w");
if (result_fd == 0) begin
$display("ERROR: Cannot open result file: %s", RESULT_FILE);
$finish;
end
// Initialize DUT inputs
coeff_a_in <= 12'd0;
coeff_b_in <= 12'd0;
valid_i <= 1'b0;
ready_i <= 1'b1; // always ready to accept output
// Reset sequence: rst_n low for 3 cycles, then high
rst_n <= 1'b0;
repeat (3) @(posedge clk);
rst_n <= 1'b1;
@(posedge clk);
pass_count = 0;
fail_count = 0;
// ============================================================
// Process each vector
// ============================================================
for (idx = 0; idx < vec_count; idx = idx + 1) begin
begin
integer out_idx;
integer input_ok;
// Extract A and B coefficients from vector memory
//
// Memory layout (A in MSB half, B in LSB half):
// A[0] → bits [6143:6132]
// A[1] → bits [6131:6120]
// ...
// A[255] → bits [3083:3072]
// B[0] → bits [3071:3060]
// B[1] → bits [3059:3048]
// ...
// B[255] → bits [11:0]
//
// Formula: coeff[i] = vector_mem[(N-1-i)*12+11+offset : (N-1-i)*12+offset]
// A: offset = N*12 = 3072
// B: offset = 0
for (ci = 0; ci < N_COEFFS; ci = ci + 1) begin
a_coeffs[ci] = vector_mem[idx][(N_COEFFS-1-ci)*12 + N_COEFFS*12 +: 12];
b_coeffs[ci] = vector_mem[idx][(N_COEFFS-1-ci)*12 +: 12];
end
$display("INFO: Vector %0d", idx);
input_ok = 1;
// ----------------------------------------------------
// LOAD phase: stream in 256 A+B coefficient pairs
// ----------------------------------------------------
for (ci = 0; ci < N_COEFFS && input_ok; ci = ci + 1) begin
// Wait for ready_o before driving next input
cycle_count = 0;
while (!ready_o && cycle_count < TIMEOUT_CYCLES) begin
@(posedge clk);
cycle_count = cycle_count + 1;
end
if (cycle_count >= TIMEOUT_CYCLES) begin
$display("ERROR: Timeout waiting for ready_o (input ci=%0d) on vector %0d",
ci, idx);
fail_count = fail_count + 1;
input_ok = 0;
end else begin
coeff_a_in <= a_coeffs[ci];
coeff_b_in <= b_coeffs[ci];
valid_i <= 1'b1;
@(posedge clk);
valid_i <= 1'b0;
end
end
if (!input_ok) begin
// Skip output reading for this failed vector
@(posedge clk);
end else begin
// ----------------------------------------------------
// OUTPUT phase: wait for valid_o and stream out results
// poly_mul_sync enters COMP_CALC→COMP_C0 state after LOAD.
// valid_o is asserted during COMP_C0 and COMP_C1.
// There may be cycles with valid_o=0 between outputs
// (during COMP_CALC transitions).
// ----------------------------------------------------
out_idx = 0;
while (out_idx < N_COEFFS) begin
// Wait for valid_o
cycle_count = 0;
while (!valid_o && cycle_count < TIMEOUT_CYCLES) begin
@(posedge clk);
cycle_count = cycle_count + 1;
end
if (cycle_count >= TIMEOUT_CYCLES) begin
$display("ERROR: Timeout waiting for valid_o (output %0d) on vector %0d",
out_idx, idx);
fail_count = fail_count + 1;
out_idx = N_COEFFS; // break out
end else begin
// Capture output coefficient
output_coeffs[out_idx] = coeff_out;
out_idx = out_idx + 1;
@(posedge clk); // ready_i=1, consume output
end
end
pass_count = pass_count + 1;
// Write result to output file
// Format: "RESULT: INDEX COEFF_HEX"
$fwrite(result_fd, "RESULT: %0d ", idx);
// Write all 256 output coeffs as hex (3 chars each = 768 total)
begin
integer oi;
reg [3:0] nib;
for (oi = 0; oi < N_COEFFS; oi = oi + 1) begin
nib = output_coeffs[oi][11:8];
$fwrite(result_fd, "%c", nibble_to_ascii(nib));
nib = output_coeffs[oi][7:4];
$fwrite(result_fd, "%c", nibble_to_ascii(nib));
nib = output_coeffs[oi][3:0];
$fwrite(result_fd, "%c", nibble_to_ascii(nib));
end
end
$fwrite(result_fd, "\n");
// Wait for DUT to return to IDLE before next vector
// The DUT goes to S_DONE then S_IDLE after all outputs
cycle_count = 0;
while (!ready_o && cycle_count < 1000) begin
@(posedge clk);
cycle_count = cycle_count + 1;
end
end
// Extra cycle before next vector
@(posedge clk);
end // inner begin block
end
// ============================================================
// Summary
// ============================================================
$fclose(result_fd);
$display("========================================");
$display("TEST COMPLETE");
$display(" Total vectors: %0d", vec_count);
$display(" Passed: %0d", pass_count);
$display(" Failed: %0d", fail_count);
$display(" Results written to: %s", RESULT_FILE);
$display("========================================");
$finish;
end
// ================================================================
// Timeout watchdog
// ================================================================
initial begin
// TIMEOUT_CYCLES * 10ns per cycle * 10x margin
#(TIMEOUT_CYCLES * 10 * 10);
$display("FATAL: Global simulation timeout reached");
$finish;
end
endmodule