Robert Allan JamesandClaude Sonnet 5
66ba21adb4
doe_log.c's per-heartbeat-tick CSV gains two columns: fleet_k_q48 (vm_physics_fleet_heat_sum() over ALL live VMs -- the genuine fleet-wide conservation invariant K, not reconstructable from the 3 named-Tripod- member heat columns already logged, which omit every identity VM's own heat) and fleet_conserved (vm_physics_conserved() as 0/1). Requested explicitly after the first heartbeat-telemetry analysis pass (analysis-20260912/) omitted K entirely. Kernel rebuilt on all three architectures, full 3x9x3 campaign rerun (results-20260912-with-k/). K = 1.0000000000 (Q48.16 raw 65536) on every one of 775 heartbeat-tick observations, sd(K) = 0, 100% fleet_conserved, across amd64/aarch64/riscv64, nine identities, three replicates -- zero deviation. Also a free regression check on both recent Stadium fixes (§XVI/§XVII): neither disturbed the reservoir-transfer accounting K depends on. Found and fixed a tooling wrinkle along the way: fleet_conserved, being the CSV row's very last field with nothing after it to bound a regex match, can have a resumed trial digit merge into it with zero separator on the wire -- combine.py now derives it from fleet_k_q48 directly (same epsilon vm_physics_conserved() uses) instead of trusting the raw field. fleet_k_q48 itself is unaffected either way. Full analysis, discussion, and light/dark SVG->PDF figures written up as a proper LaTeX report (report-20260912/report/std79_doe_report.pdf), following experiments/bare_metal/analysis/report/bare_metal_doe_report.tex's established style -- supersedes analysis-20260912/'s markdown-only first pass as the primary deliverable for this dataset (kept, not discarded). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EXieurDfDSsDFdnSyusuWo
79 lines
3.5 KiB
Python
79 lines
3.5 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
combine.py -- build combined.csv from the three per-architecture
|
|
correlated CSVs (correlate_doe.py's output), decoding Q48.16 fixed-point
|
|
fields to plain floats and jitter_bits' IEEE754 bit pattern to a real
|
|
jitter_ns value, since R has no convenient native uint64/bit-punning
|
|
story worth using here.
|
|
|
|
v2 (2026-09-12): adds fleet_k (decoded fleet_k_q48 -- vm_physics_fleet_heat_sum()
|
|
over ALL live VMs, the genuine conservation invariant, requested explicitly
|
|
after the first analysis pass omitted it -- see doe_log.c's column 19/20
|
|
doc comment) and fleet_conserved (already 0/1, passed through as-is).
|
|
|
|
Usage:
|
|
python3 combine.py amd64-correlated.csv aarch64-correlated.csv riscv64-correlated.csv combined.csv
|
|
"""
|
|
import csv
|
|
import struct
|
|
import sys
|
|
|
|
Q48 = 65536.0
|
|
|
|
def main():
|
|
if len(sys.argv) < 3:
|
|
print("usage: combine.py <in1.csv> [in2.csv ...] <out.csv>", file=sys.stderr)
|
|
sys.exit(1)
|
|
*in_paths, out_path = sys.argv[1:]
|
|
|
|
fields = ["arch", "run_id", "id_idx", "id_label", "rep", "tick_number", "elapsed_ns",
|
|
"tick_interval_ns", "hot_word_count", "avg_word_heat", "window_width",
|
|
"actual_window_size", "predicted_label_hits", "jitter_ns", "apic_ticks",
|
|
"time_trust", "variance", "vm_call_depth_max", "hera_heat", "hermes_heat",
|
|
"artemis_heat", "fleet_k", "fleet_conserved"]
|
|
|
|
out_rows = []
|
|
for path in in_paths:
|
|
arch = path.split("-correlated.csv")[0].split("/")[-1]
|
|
with open(path) as f:
|
|
for row in csv.DictReader(f):
|
|
jitter_bits = int(row["jitter_bits"])
|
|
jitter_ns = struct.unpack('<d', struct.pack('<Q', jitter_bits))[0]
|
|
out = dict(row)
|
|
out["arch"] = arch
|
|
out["avg_word_heat"] = int(row["avg_word_heat_q48"]) / Q48
|
|
out["time_trust"] = int(row["time_trust_q48"]) / Q48
|
|
out["variance"] = int(row["variance_q48"]) / Q48
|
|
out["hera_heat"] = int(row["hera_heat_q48"]) / Q48
|
|
out["hermes_heat"] = int(row["hermes_heat_q48"]) / Q48
|
|
out["artemis_heat"] = int(row["artemis_heat_q48"]) / Q48
|
|
out["jitter_ns"] = jitter_ns
|
|
fleet_k_q48 = int(row["fleet_k_q48"])
|
|
out["fleet_k"] = fleet_k_q48 / Q48
|
|
# fleet_conserved is doe_log.c's very LAST printed field, with
|
|
# nothing after it in the row to bound the regex match against
|
|
# -- unlike every other field, a resumed trial digit with zero
|
|
# separator can merge into it (same failure mode
|
|
# correlate_doe.py's RUN_ID_MAP works around for id_idx,
|
|
# confirmed live 2026-09-12: raw values like
|
|
# "19223372036854775807" appeared where 0/1 was expected).
|
|
# Derive it instead from fleet_k_q48 directly, using the same
|
|
# epsilon and threshold vm_physics_conserved() uses
|
|
# (capsule_vm_physics.c: 3277/65536 = 5% of Q48_ONE) --
|
|
# correct by construction, immune to the splice.
|
|
Q48_ONE = 65536
|
|
EPSILON_Q48 = 3277
|
|
diff = abs(fleet_k_q48 - Q48_ONE)
|
|
out["fleet_conserved"] = 1 if diff < EPSILON_Q48 else 0
|
|
out_rows.append({k: out[k] for k in fields})
|
|
|
|
with open(out_path, "w", newline="") as f:
|
|
w = csv.DictWriter(f, fieldnames=fields)
|
|
w.writeheader()
|
|
w.writerows(out_rows)
|
|
|
|
print(f"rows written: {len(out_rows)}")
|
|
|
|
if __name__ == "__main__":
|
|
main()
|