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
63 lines
2.9 KiB
Markdown
63 lines
2.9 KiB
Markdown
# C.7 — SPI slave + engine (`spi_slave.v`, `spi_engine.v`)
|
|
|
|
Data: 2026-09-04.
|
|
|
|
---
|
|
|
|
## 7.1 CDC, framing, opcode dispatch — CERTIFICATO (copertura pre-esistente estesa, riverificata)
|
|
|
|
Questo modulo ha già ricevuto lavoro di verifica sostanziale **in questa stessa sessione**
|
|
(non solo dichiarato in sessioni precedenti):
|
|
- **Una race reale trovata e corretta** nel meccanismo sticky di `STATUS` (Fase 4/9 di
|
|
sessioni precedenti, ma il fix e la sua verifica sono tracciabili e riverificati).
|
|
- **CDC a 2/3 stadi per `sclk`/`mosi`/`cs_n`**: `sim/spi_slave_tb.v` include un test
|
|
esplicito con rapporto SCLK/clk diverso (TEST 4, "slower SPI clock... confirms no hidden
|
|
dependency on a specific SCLK/clk ratio") — non solo un singolo rapporto a piacere.
|
|
- **`sim/spi_engine_tb.v`**: 10 test (A-J) coprono WRITE_RAM/READ_RAM round-trip, SET_BASE,
|
|
START idle/busy, STATUS live/sticky/clear-on-read, RESET, READ_OUTPUT (neuron-major),
|
|
READ_CONFIG, NOP (nessun side-effect), byte MOSI in eccesso ignorati, transazioni
|
|
back-to-back.
|
|
- **Opcode sconosciuti**: `default: begin // OP_NOP and unknown opcodes` (riga 724) —
|
|
trattati esplicitamente come NOP, nessun rischio di hang per costruzione, coerente col
|
|
pattern già verificato per `spi_flash_master.v` (opcode illegale, Fase F1).
|
|
- Rieseguito in Fase 0 di questa campagna (non solo citato): `spi_slave_tb.v` e
|
|
`spi_engine_tb.v` **PASS**, confermato dall'harness di regressione indipendente.
|
|
|
|
**Verdetto: CERTIFICATO**, con la stessa evidenza di prima (riverificata, non solo citata).
|
|
|
|
---
|
|
|
|
## 7.2 `len=0` per WRITE_RAM/READ_RAM — CERTIFICATO (guard esplicito trovato per ispezione)
|
|
|
|
Dato il pattern ricorrente in questa campagna (guard mancante su valori "reali=0" in più
|
|
moduli, BUG-002/003/004/005/006), ho controllato se lo stesso buco esistesse anche qui.
|
|
**Non esiste**: `rtl/spi_engine.v:817` ha un guard esplicito —
|
|
|
|
```verilog
|
|
if ({len_acc[7:0], rx_byte} == 16'h0) begin
|
|
state <= ST_IGNORE;
|
|
end else if (opcode == OP_WRITE_RAM) begin
|
|
state <= ST_WRITE_DATA;
|
|
...
|
|
```
|
|
|
|
`len=0` transita correttamente a `ST_IGNORE` (no-op sicuro) invece di entrare nel loop di
|
|
trasferimento — a differenza di `layer_sequencer.v`/`graph_engine.v`, qui il caso limite è
|
|
gestito esplicitamente. Non serviva un nuovo test dedicato: il guard è verificabile per
|
|
ispezione diretta, inequivocabile.
|
|
|
|
**Verdetto: CERTIFICATO.** Nota positiva per il registro: questo modulo dimostra che il
|
|
progetto **non manca sistematicamente** di guardie sui valori limite — il buco è
|
|
specifico ai moduli già segnalati (BUG-002 - BUG-006), non universale.
|
|
|
|
---
|
|
|
|
## 7.3 Verdetto complessivo C.7
|
|
|
|
| Sotto-aspetto | Verdetto |
|
|
|---|---|
|
|
| CDC, framing, dispatch opcode, opcode sconosciuti | **CERTIFICATO** (copertura estesa pre-esistente, riverificata) |
|
|
| `len=0` WRITE_RAM/READ_RAM | **CERTIFICATO** (guard esplicito confermato per ispezione) |
|
|
|
|
Nessun nuovo bug trovato in questo aspetto.
|