Reduce the default page size from 256 KB to 32 KB - #868
Conversation
|
/bench-verify |
|
⏳ Benchmark started on the bench server. The recursion-guest cycle comparison adds guest builds on top of the verifier bench, longer on a cold runner. The bench server is occupied until it finishes. |
Verifier benchmark —
|
| Metric | main | PR | Δ |
|---|---|---|---|
| Verify time (ABBA, 20 pairs, per-side) | 2.849s | 4.228s | +48.40% 🔴 |
| Proof size (exact, 1 reading) | 156.30 MiB | 239.03 MiB | +52.93% 🔴 |
Per-side (PR can't deserialize the baseline's proof — proof-format change): A/B/B/A cancels machine drift but not proof-specific variance — read the Verify-time Δ as approximate.
pairs: 20 mean A (PR): 4.228s mean B (main): 2.849s
[parametric] paired-t mean +48.40% sd 1.82% se 0.41%
95% CI: [+47.55%, +49.25%] (t df=19 = 2.093)
[robust] median +48.64% Wilcoxon W+=210 W-=0 p(exact)=1.9e-06 (z=+3.90)
run-to-run jitter: A CV 1.27% B CV 0.53% (lower = steadier)
within-session drift: +0.37% over the run, 1st->2nd half +0.15%
🔴 REAL REGRESSION — PR verifies ~48.40% slower (paired-t and Wilcoxon agree).
ethrex 20-tx block · continuations, epoch 2^20 (9 epochs) · blowup=2, 219 queries
| Metric | main | PR | Δ |
|---|---|---|---|
| Verify time (ABBA, 8 pairs, per-side) | 7.926s | 8.137s | +2.67% 🔴 |
| Proof size (exact, 1 reading) | 479.35 MiB | 494.26 MiB | +3.11% 🔴 |
Per-side (PR can't deserialize the baseline's proof — proof-format change): A/B/B/A cancels machine drift but not proof-specific variance — read the Verify-time Δ as approximate.
pairs: 8 mean A (PR): 8.137s mean B (main): 7.926s
[parametric] paired-t mean +2.67% sd 0.45% se 0.16%
95% CI: [+2.29%, +3.05%] (t df=7 = 2.365)
[robust] median +2.54% Wilcoxon W+=36 W-=0 p(exact)=0.0078 (z=+2.45)
run-to-run jitter: A CV 0.20% B CV 0.43% (lower = steadier)
within-session drift: +0.10% over the run, 1st->2nd half -0.07%
🔴 REAL REGRESSION — PR verifies ~2.67% slower (paired-t and Wilcoxon agree).
Verify-time rows only: drift-free interleaved A/B/B/A, with paired-t and exact Wilcoxon — trust the verdict when the two agree. Proof sizes are single exact readings (no averaging). - = PR faster.
Recursion guest cycles — verifier running INSIDE the VM (main vs PR)
empty program · monolithic · blowup=2, 1 query (diagnostic — NOT a real verifier cost)
Single exact reading per ref — no ABBA: guest cycles are deterministic for a fixed
(guest ELF, input blob), so there is no machine drift to cancel.
| Metric | main | PR | Δ |
|---|---|---|---|
| Guest cycles | 39.8M | 41.2M | +1.3M (+3.34%) |
| Keccak calls | 3025 | 2959 | -66 |
baseline origin/main fcc4848fef guest=recursion-min.elf
PR ca58d945fc2a02a26e41fa98d6b40ef64042d2d1 ca58d945fc guest=recursion-min.elf
note: cycles reproduce to ~±100k (build codegen + proof nondeterminism);
treat sub-100k deltas as noise, not signal.
raw (exact integer counts)
ref_b_sha=fcc4848fefca60d38a9cb7d20311d89016c3f17d ref_b_elf=recursion-min.elf ref_b_cycles=39833310 ref_b_keccak=3025 ref_b_execute_wall_s=2
ref_a_sha=ca58d945fc2a02a26e41fa98d6b40ef64042d2d1 ref_a_elf=recursion-min.elf ref_a_cycles=41163399 ref_a_keccak=2959 ref_a_execute_wall_s=1
delta_cycles=1330089 delta_keccak=-66
ethrex 20-tx block · continuations, epoch 2^21 (5 epochs) · blowup=2, 219 queries (128-bit)
Single exact reading per ref — no ABBA: guest cycles are deterministic for a fixed
(guest ELF, input blob), so there is no machine drift to cancel.
| Metric | main | PR | Δ |
|---|---|---|---|
| Guest cycles | 4548.3M | 4811.6M | +263.3M (+5.79%) |
| Keccak calls | 6944771 | 7609088 | +664317 |
baseline origin/main fcc4848fef guest=recursion-cont-blowup2.elf
PR ca58d945fc2a02a26e41fa98d6b40ef64042d2d1 ca58d945fc guest=recursion-cont-blowup2.elf
note: cycles reproduce to ~±100k (build codegen + proof nondeterminism);
treat sub-100k deltas as noise, not signal.
raw (exact integer counts)
ref_b_sha=fcc4848fefca60d38a9cb7d20311d89016c3f17d ref_b_elf=recursion-cont-blowup2.elf ref_b_cycles=4548337489 ref_b_keccak=6944771 ref_b_execute_wall_s=72
ref_a_sha=ca58d945fc2a02a26e41fa98d6b40ef64042d2d1 ref_a_elf=recursion-cont-blowup2.elf ref_a_cycles=4811598498 ref_a_keccak=7609088 ref_a_execute_wall_s=78
delta_cycles=263261009 delta_keccak=664317
|
/bench-verify |
|
⏳ Benchmark started on the bench server. Two verifier arms (monolithic + continuations over an ethrex 20-tx block), then the recursion-guest cycle comparison, which adds guest builds on top — longer on a cold runner. The bench server is occupied until it finishes. |
1 participant
Description
Lowers the
DEFAULT_PAGE_SIZEto 32 KB (1<<15). Regenerates the static zero-init page commitments (blowups 2/4/8) for the new size and updates the page-count and page-boundary tests accordingly.20-tx ethrex block
Recursion cycles
Prove Time