micheleandClaude Sonnet 5
dc0b331d3e
Begins the V2 Neural Multiprocessor / Dataflow architecture per docs/v2-description.md, per explicit user request to freeze V1 and start V2 development, copying from V1 what's needed. Scaffold: - hardware/v1/: byte-exact, read-only copy of the current V1 codebase (rtl, testbenches, tools, constraints, a representative subset of synthesis results, and reference docs) -- verified identical via diff/cmp against the live top-level tree before being made filesystem-read-only. The live top-level tree is untouched and remains the project's "production" V1 (see hardware/v1/README.md and hardware/v2/logs/decisions.log DEC-0001 for why copy-not-move). - hardware/v2/: mandatory structure (rtl/sim/constraints/synthesis/ reports/scripts/logs/docs) plus the full logging system required by the spec (development/architecture/simulation/synthesis/timing/ benchmark/decisions/experiments/errors.log). M1 -- Neural Processor (hardware/v2/rtl/neural_processor.v): - 8-stage pipelined perceptron unit (P_IN=8): input align, 8 multipliers, 3-level adder tree, accumulator, bias+activation, INT8 saturation. Genuine 1-tile/cycle throughput, not just a wider combinational datapath. - 7-state FSM (NP_IDLE..NP_ERROR per docs/v2-description.md §6, with 4 baseline states merged into NP_WAIT_OPERANDS -- see decisions.log DEC-0002); valid/ready/data/last stream interfaces per §7. - Bit-exact vs the frozen hardware/v1/rtl/neuron_parallel.v + mac8.v + mac_unit.v: 7/7 tests pass (hardware/v2/sim/tb_neural_processor.v), covering regular/mixed-sign/extreme-INT8 vectors, both activations, a zero-idle-gap back-to-back-tiles throughput check, and an 8-tile job -- verified with Verilator (see below for why). - Real synthesis + place&route (Yosys + nextpnr-ecp5): 0 CHECK problems, Fmax 183.12 MHz at ACC_WIDTH=32 (PASS at 80MHz, ~3x V1's isolated PARALLEL=8 Fmax of 61.71 MHz) and 176.21 MHz at ACC_WIDTH=24 (a user-requested comparison experiment, also bit-exact-verified; see experiments.log EXP-0001/EXP-0002 and benchmark.log). Three real bugs found and resolved during M1 development (full diagnostic record in errors.log): - Two independent, reproducible Icarus Verilog v13.0 scheduling defects (ERR-0001, ERR-0002) that silently produced wrong simulation results for standard sequential Verilog -- confirmed via Verilator 5.050 giving correct results on the same minimal repros. Verilator is now the trusted simulator for hardware/v2/ (decisions.log DEC-0004); Icarus's affected protocol-violation check was removed from the RTL and deferred architecturally to the Neural Director (DEC-0003) rather than chased further. - One real RTL bug (ERR-0003): last0 wasn't gated like valid0, letting a "last tile" tag leak into the pipeline ahead of its actual valid tile on back-to-back jobs. Fixed and verified. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013xXuuRUWZScuo1DeYJxs3v
126 lines
2.7 KiB
Verilog
126 lines
2.7 KiB
Verilog
module top #(
|
|
parameter DATA_WIDTH = 8,
|
|
parameter N_INPUTS = 256,
|
|
parameter N_NEURONS = 4,
|
|
parameter PARALLEL = 8,
|
|
parameter ACC_WIDTH = 32
|
|
)(
|
|
input wire clk,
|
|
input wire rst,
|
|
input wire start,
|
|
|
|
output wire signed [31:0] y_bus,
|
|
output wire busy,
|
|
output wire done
|
|
);
|
|
|
|
localparam X_BITS = DATA_WIDTH * N_INPUTS;
|
|
localparam W_BITS = DATA_WIDTH * N_INPUTS * N_NEURONS;
|
|
localparam B_BITS = DATA_WIDTH * N_NEURONS;
|
|
|
|
/*
|
|
* Internal neural-network data.
|
|
*
|
|
* These are deliberately registers, not parameters/constants.
|
|
* This prevents the complete datapath from disappearing during
|
|
* synthesis.
|
|
*/
|
|
|
|
reg signed [X_BITS-1:0] x_bus;
|
|
reg signed [W_BITS-1:0] weights_bus;
|
|
reg signed [B_BITS-1:0] bias_bus;
|
|
|
|
integer i;
|
|
integer n;
|
|
|
|
/*
|
|
* Deterministic initialization.
|
|
*
|
|
* The actual datapath remains present because the vectors are
|
|
* stored in registers and loaded through the clocked process.
|
|
*/
|
|
|
|
always @(posedge clk) begin
|
|
if (rst) begin
|
|
|
|
x_bus <= '0;
|
|
weights_bus <= '0;
|
|
bias_bus <= '0;
|
|
|
|
end
|
|
else if (start) begin
|
|
|
|
/*
|
|
* INT8 input vector.
|
|
*
|
|
* Pattern:
|
|
* -16 ... +15
|
|
*/
|
|
|
|
for (i = 0; i < N_INPUTS; i = i + 1) begin
|
|
x_bus[i*DATA_WIDTH +: DATA_WIDTH]
|
|
<= ((i * 17 + 3) % 31) - 15;
|
|
end
|
|
|
|
/*
|
|
* INT8 weights.
|
|
*/
|
|
|
|
for (n = 0; n < N_NEURONS; n = n + 1) begin
|
|
|
|
for (i = 0; i < N_INPUTS; i = i + 1) begin
|
|
|
|
weights_bus[
|
|
(n*N_INPUTS+i)*DATA_WIDTH
|
|
+: DATA_WIDTH
|
|
]
|
|
<= ((n * 29 + i * 13 + 5) % 31) - 15;
|
|
|
|
end
|
|
|
|
end
|
|
|
|
/*
|
|
* INT8 biases.
|
|
*/
|
|
|
|
for (n = 0; n < N_NEURONS; n = n + 1) begin
|
|
|
|
bias_bus[
|
|
n*DATA_WIDTH
|
|
+: DATA_WIDTH
|
|
]
|
|
<= ((n * 7 + 1) % 9) - 4;
|
|
|
|
end
|
|
|
|
end
|
|
end
|
|
|
|
|
|
/*
|
|
* Real neural-network layer.
|
|
*/
|
|
|
|
layer #(
|
|
.DATA_WIDTH(DATA_WIDTH),
|
|
.N_INPUTS(N_INPUTS),
|
|
.N_NEURONS(N_NEURONS),
|
|
.PARALLEL(PARALLEL),
|
|
.ACC_WIDTH(ACC_WIDTH)
|
|
) dut (
|
|
.clk(clk),
|
|
.rst(rst),
|
|
.start(start),
|
|
|
|
.x_bus(x_bus),
|
|
.weights_bus(weights_bus),
|
|
.bias_bus(bias_bus),
|
|
|
|
.y_bus(y_bus),
|
|
.busy(busy),
|
|
.done(done)
|
|
);
|
|
|
|
endmodule
|