micheleandClaude Sonnet 5
63cac6a7e5
Implements optimization #2 from the final benchmark campaign's own recommendation, on top of DEC-0015's word-level burst rewrite: a new shared activation_cache.v module fetches a given activation (X) vector from PSRAM once instead of once per neuron sharing it - the exact redundant traffic pattern the dense-layer workloads in this project's benchmark suite exhibit. Each memory_manager's own prefetch_engine now fetches WEIGHTS only; the activation half is requested from the shared cache instead (single-tag, tile-granular, N_SLOTS request ports, its own real word-level PSRAM backend via a new dedicated arbiter port). dataflow_core.v/slot_mem_arbiter.v/neural_multiprocessor.v widened to N_SLOTS+1 ports to arbitrate the cache's traffic alongside each slot's weight traffic. Two real bugs found and fixed during implementation (ERR-0010): a target-bank/pending-bank race in memory_manager.v's activation-cache wiring (the same bug class ERR-0006 already fixed once for pf_target_bank - a later handoff's queued request can overwrite which bank an earlier, still-in-flight request's ack applies to), and a repeat of ERR-0009's N_SLOTS=1 zero-width replication bug in activation_cache.v itself. Real, measured results: the full final-benchmark campaign (24/24 workload/config combinations) re-verified bit-exact. D-Stress cycles fall a further 1.66-2.00x on top of DEC-0015 (~4x combined vs the original byte-level baseline). But the cache's real Fmax cost is much steeper than DEC-0015's own: N_SLOTS=2 (the recommended default, DEC-0014) drops from 133.58 to 87.72 MHz (-34%, margin over 80MHz shrinks from +67% to +9.7%), and N_SLOTS=4 drops to 65.01 MHz - now FAILING the 80MHz target it previously passed. Combined real wall-clock speedup vs the original baseline: N=1 3.86x, N=2 2.45x (both real net wins); N=4 is a real regression once its own now-failing Fmax is honestly used, though N=4 was never the recommended configuration. N_SLOTS=2 remains the recommended default (DEC-0014 unaffected) with a thinner but still real Fmax margin. Cache hit-detection pipelining is flagged as concrete follow-up work if N_SLOTS>2 is ever needed with the cache active - not attempted this round. Logged: simulation/synthesis/timing/benchmark/decisions (DEC-0016)/ experiments (EXP-0016)/errors (ERR-0010)/development.log, ROADMAP.md updated. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013xXuuRUWZScuo1DeYJxs3v
Formato dei log V2
Regola non negoziabile (docs/v2-description.md §25-29): ogni attività
significativa (modifica, simulazione, sintesi, benchmark, decisione, errore)
deve essere registrata. Nessun log viene mai sovrascritto o troncato — solo
append. Nessun ID esperimento (EXP-XXXX) o decisione (DEC-XXXX) viene mai
riutilizzato, anche se il risultato è un FAIL.
File
development.log— log principale di sviluppo, un'entry per ogni sessione di lavoro/milestone (creazione file, refactor, avanzamento roadmap).architecture.log— decisioni e note di architettura a grana fine (non scelte finali — quelle vanno indecisions.log— ma esplorazioni, alternative considerate, vincoli scoperti).simulation.log— ogni run di simulazione (Icarus/Verilator): test, vettori, cicli, PASS/FAIL, confronto bit-exact con V1, stall/memory-wait.synthesis.log— ogni run Yosys: LUT/FF/DSP/BRAM, warning, problemi CHECK.timing.log— ogni run nextpnr-ecp5: Fmax, percorso critico, WNS/TNS se disponibili. Fmax "ufficiale" di una configurazione = solo da qui, mai da simulazione o stima.benchmark.log— tabelle di confronto per configurazione (Fmax, MAC/cycle, cycles/neuron, utilization, ecc.), sempre con etichetta THEORETICAL/SIMULATED/SYNTHESIZED/POST-P&R.decisions.log— decisioni architetturali importanti, formatoDEC-XXXX(vedidocs/v2-description.md§27).experiments.log— registro principale, unEXP-XXXXper ogni esperimento end-to-end (config → sim/synth/timing → risultato), rimanda areports/experiments/EXP-XXXX/.errors.log— errori/bug/regressioni incontrati durante lo sviluppo V2 stesso (non i bug V1, già chiusi inhardware/v1/docs/validation/bugs.md).
Campi minimi per entry (§26)
timestamp, experiment_id (se applicabile), git_commit, session/agent,
module, configuration, action, reason, command, result, errors, decision,
next_action
Per synthesis/timing aggiungere: LUT, FF, DSP, BRAM, Fmax, critical path, WNS/TNS. Per simulazione: test, vectors, cycles, PASS/FAIL, bit-exact result, stall cycles, memory wait, utilization.