#!/usr/bin/env python3 """Reduce a heatmovie session (samples.jsonl + events.jsonl) to cycle-averaged per-sensor movies. For every run: align its cycles on heat_on, subtract each sensor's mean over the 2 s before, and average the cycles in 0.25 s frames from -2 s to +19 s. Writes data.json for the page. """ import json import sys STEP, T0, T1, BASE = 0.25, -2.0, 19.0, 2.0 LABELS = { "core-S14": "One core (1 of 32) in shire S14, the centre", "shire-S0": "One shire, S0: the NW corner", "shire-S31": "One shire, S31: the SE corner", "shire-S14": "One shire, S14: the centre", "block-2x2-centre": "A 2×2 block in the centre (S13, S14, S21, S22)", "quadrant-NW": "The NW quadrant (9 shires)", "quadrant-SE": "The SE quadrant (8 shires)", "half-west": "The west half (18 shires)", "half-east": "The east half (14 shires)", "quadrant-NE": "The NE quadrant (6 shires)", "quadrant-SW": "The SW quadrant (9 shires)", } # Die frame (rows r0-r5, columns c1-c6) from the lab's placements.json; "M" = master/spare, "IO", "P" = PCIe. GRID = [["S0", "S8", "S3", "M32", "IO", "P"], ["S24", "S16", "S4", "S12", "S20", "S28"], ["S9", "S1", "S13", "S21", "S29", "S5"], ["S25", "S17", "S14", "S22", "S30", "S6"], ["S2", "S10", "S18", "S26", "S15", "S7"], ["S11", "S19", "S27", "M33", "S23", "S31"]] def main(d): S = [] for line in open(d + "/samples.jsonl"): r = json.loads(line) if "c" in r and r.get("st", 0) == 0 and len(r["c"]) == 34: S.append(r) E = [json.loads(line) for line in open(d + "/events.jsonl")] times = [r["t"] for r in S] io_last = None for r in S: # forward-fill the I/O sensor (read once a second) as sensor 34 if r.get("io") is not None: io_last = r["io"] r["v"] = r["c"] + [io_last if io_last is not None else r["c"][0]] import bisect def window_mean(t_a, t_b): i, j = bisect.bisect_left(times, t_a), bisect.bisect_left(times, t_b) if j <= i: return None n = j - i return [sum(S[k]["v"][s] for k in range(i, j)) / n for s in range(35)] runs, order = {}, [] for e in E: if e["ev"] == "heat_on": if e["run"] not in runs: runs[e["run"]] = {"name": e["run"], "shires": e["shires"], "per_shire": e["per_shire"], "mask": e["mask"], "cycles": []} order.append(e["run"]) runs[e["run"]]["cycles"].append({"on": e["t"]}) elif e["ev"] == "heat_off": runs[e["run"]]["cycles"][-1]["off"] = e["t"] nframes = int(round((T1 - T0) / STEP)) + 1 rel = [round(T0 + i * STEP, 2) for i in range(nframes)] out_runs = [] for name in order: R = runs[name] acc = [[0.0] * 35 for _ in range(nframes)] cnt = [0] * nframes base_acc, nb, changed = [0.0] * 35, 0, [0] * 35 heat_s = [] for c in R["cycles"]: base = window_mean(c["on"] - BASE, c["on"]) if base is None: continue nb += 1 for s in range(35): base_acc[s] += base[s] heat_s.append(c.get("off", c["on"]) - c["on"]) lo_ix = bisect.bisect_left(times, c["on"]) hi_ix = bisect.bisect_left(times, c["on"] + 10.0) for s in range(34): vals = {S[k]["c"][s] for k in range(max(lo_ix - 40, 0), hi_ix)} changed[s] += 1 if len(vals) > 1 else 0 for f, tr in enumerate(rel): m = window_mean(c["on"] + tr - STEP / 2, c["on"] + tr + STEP / 2) if m is None: continue cnt[f] += 1 for s in range(35): acc[f][s] += m[s] - base[s] frames = [[round(acc[f][s] / cnt[f], 3) if cnt[f] else 0.0 for s in range(35)] for f in range(nframes)] heated = [s for s in R["shires"]] others = [s for s in range(32) if s not in heated] hmean = [round(sum(fr[s] for s in heated) / len(heated), 3) for fr in frames] omean = [round(sum(fr[s] for s in others) / len(others), 3) for fr in frames] gmean = [round(sum(fr[:34]) / 34, 3) for fr in frames] contrast = [round(h - o, 3) for h, o in zip(hmean, omean)] ion = [i for i, t in enumerate(rel) if 0 <= t <= 10] fpk = max(ion, key=lambda i: contrast[i]) pk_sensor = max(range(34), key=lambda s: frames[fpk][s]) out_runs.append({ "name": name, "label": LABELS.get(name, name), "shires": heated, "per_shire": R["per_shire"], "mask": R["mask"], "cycles": nb, "heat_s": round(sum(heat_s) / max(len(heat_s), 1), 2), "frames": frames, "heated_mean": hmean, "others_mean": omean, "global_mean": gmean, "contrast": contrast, "peak_contrast": contrast[fpk], "peak_t": rel[fpk], "peak_sensor": pk_sensor, "peak_sensor_heated": pk_sensor in heated, "peak_heated_dT": max(hmean[i] for i in ion), "peak_global_dT": max(gmean[i] for i in ion), "baseline": [round(base_acc[s] / max(nb, 1), 2) for s in range(35)], "sensors_changed": sum(1 for s in range(34) if changed[s] > 0), }) dts = [b - a for a, b in zip(times, times[1:])] meta = {"samples": len(S), "median_dt_s": round(sorted(dts)[len(dts) // 2], 4), "t_start": times[0], "t_end": times[-1], "step_s": STEP, "t0": T0, "t1": T1, "base_s": BASE, "device_us_median": sorted(r["us"] for r in S)[len(S) // 2], "max_c": max(max(r["c"]) for r in S), "min_c": min(min(r["c"]) for r in S)} json.dump({"meta": meta, "grid": GRID, "times": rel, "runs": out_runs}, open(d + "/data.json", "w")) for r in out_runs: print("%-17s cyc %d heat %.1fs heated peak %+.2f global %+.2f contrast %+.2f at %.2fs peak sensor S%d %s changed %d/34" % (r["name"], r["cycles"], r["heat_s"], r["peak_heated_dT"], r["peak_global_dT"], r["peak_contrast"], r["peak_t"], r["peak_sensor"], "(heated)" if r["peak_sensor_heated"] else "", r["sensors_changed"])) print(json.dumps(meta)) if __name__ == "__main__": main(sys.argv[1])