Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion MODIFICATIONS.md
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ New code, not derived from MaxCDCL.
| Area | Description |
|------|-------------|
| QLDPC encoding | MaxSAT formulation for stabilizer distance (Hx/Hz/Gx/Gz matrices) |
| Fork + pipe | Child runs solver; parent enforces wall-clock timeout (`-cpu-lim`) via `SIGKILL` |
| Fork + pipe | Child runs solver; parent enforces wall-clock timeout (`-cpu-lim`) via `SIGKILL`; the parent waits in `poll()` on the pipe and a SIGCHLD self-pipe, so it sees the child's exit at once (GH-224; was 10 ms `usleep` polling) |
| Bounds sync | Parent reads `TRY` / `LB` / `UB` / `RESULT` lines from pipe |
| Progress output | Default mode prints `c trying d:`, `c d_lb:`, `c d_ub:`, `c d:`, `o` |
| Quiet / debug | Default `verb=0`; child stdout to `/dev/null`; `-v` / `-debug` for solver log |
Expand Down
66 changes: 66 additions & 0 deletions optimization/experiments/GH-224/analysis.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,66 @@
# GH-224: parent wait loop without the 10 ms polling floor

Change: `solve_in_child_fork` (`src/core/distqldpc.cc`) used to wait for the solver child with `usleep(10000)` between checks. It now waits in `poll()` on the bounds pipe and on a SIGCHLD self-pipe. The timeout is the remaining `-cpu-lim` time, capped at 100 ms. The solver child is unchanged.

Binaries (yfclab2, frozen): old = main fecee8c9 (`frozen/server-gh224-main-fecee8c`), new = 2587b9de (`frozen/server-gh224-2587b9d`).

## Tier 0
- `make test-cc`: PASS on yfclab2 (`raw/t0tests/test_cc_new.txt`) and on the Mac (`raw/t0tests/mac_test_cc.txt`). The output equals main's except one timing-dependent counter: partE "levels abandoned" is 37 on main and 38 on new. This test never calls the changed function.
- `make test-cc-flags`: 566 checks, 0 failed (server and Mac).
- `scripts/smoke_test.sh` and `scripts/test_partition_soft_literals.py`: pass.
- Behaviour (`raw/behaviour/behaviour.txt`), old vs new, with identical output lines:
- `-cpu-lim=3` timeout, MaxSAT and CC: TIMEOUT at 3.03 s with the same bounds.
- Tighter `prlimit --as` (200 MB MaxSAT, 60 MB CC): the same UNKNOWN.
- Child killed from outside with SIGKILL: UNKNOWN within 6-8 ms.
- Child SIGSTOP/SIGCONT: finishes normally.
- Answers: 9450 + 216 panel runs, all rc 0 and all equal to the stem distance.

## Tier 0 + Tier 1 panel (`raw/t01/`)
Setup:
- Layout of main fecee8c9: 26 tier0 and 9 tier1 codes.
- Modes: MaxSAT default (mto), MaxSAT `-no-card`, CC default.
- Seeds 0-4, 9 repetitions.
- old and new run interleaved on one pinned core per (code, mode, seed), with rotating order; cores 72-101 only, each with an idle SMT sibling.
- Cell = median over repetitions, then median over seeds. Ratio = new / old.

| Tier | Mode | cells | geomean | worst cell | cells > 1.25 |
|---|---|---:|---:|---|---:|
| 0 | cc | 26 | 0.666 | 0.964 | 0 |
| 0 | mto | 26 | 0.849 | 1.032 | 0 |
| 0 | nocard | 26 | 0.853 | 1.086 | 0 |
| 1 | cc | 9 | 0.879 | 0.995 | 0 |
| 1 | mto | 9 | 0.978 | 0.987 | 0 |
| 1 | nocard | 9 | 0.981 | 0.993 | 0 |

**Logs.** stdout is byte-identical in every run of both arms for 522 of 525 units. The other 3 (all GB_144_12_12 CC) differ only in the order of a `c d_ub` line relative to the level lines. That order is timing-dependent: main's own runs vary the same way.

**Floor.** Wait slack = wall - user - sys, over runs shorter than 0.2 s (about 3300 per arm):

| Arm | median | p90 |
|---|---:|---:|
| old | 11.9 ms | 16.2 ms |
| new | 4.8 ms | 7.8 ms |

Wall times of the CC cells, in 1 ms bins:
- old LP_442 splits into two groups, 27-31 ms and 34-44 ms;
- new LP_442 is one contiguous group, 25-33 ms;
- old BB_186 is spread over 28-60 ms;
- new BB_186 is 21-30 ms.

The MaxSAT cells of 1-10 s also run about 2 % faster in CPU time: the parent no longer wakes up 100 times a second on the solver's core.

## Tier 2/3 spot check (`raw/spot_analysis.md`)
Tiers 2-5 are unaffected by construction: the solver child and the deadline path are unchanged, and the parent wakes up less often. This was spot-checked, not run in full; there was no Tier 5 run.

Setup: 3 seeds x 3 repetitions, interleaved.
- Tier 2: LP_544, TN_108 and BB_144_14_14, in all three modes.
- Tier 3: TN_250, LP_1020 and HGP_481, CC only.

| Tier | geomean | worst cell |
|---|---:|---|
| 2 | 0.952 | 1.001 (TN_108 CC) |
| 3 | 0.959 | 1.002 (TN_250 CC, 50.6 s) |

Every answer is correct. The log differences are the same timing-dependent line-order cases.

Runners: `runner/interleave.py`, `runner/analyze.py` and `runner/behaviour.sh`.
12 changes: 12 additions & 0 deletions optimization/experiments/GH-224/raw/behaviour/behaviour.txt
Original file line number Diff line number Diff line change
@@ -0,0 +1,12 @@
timeout-maxsat old rc=1 wall=3.031 c d_ub: 15|c d_lb: 2|c d_lb: 3|c d_lb: 5|c d_lb: 5|c d_ub: 15|c status: TIMEOUT (child killed after -cpu-lim)|c d : UNKNOWN|s UNKNOWN|
timeout-maxsat new rc=1 wall=3.025 c d_ub: 15|c d_lb: 2|c d_lb: 3|c d_lb: 5|c d_lb: 5|c d_ub: 15|c status: TIMEOUT (child killed after -cpu-lim)|c d : UNKNOWN|s UNKNOWN|
timeout-cc old rc=1 wall=3.031 c d_ub: 52|c d_lb: 2|c d_ub: 40|c d_lb: 4|c d_lb: 6|c d_lb: 8|c d_lb: 10|c d_lb: 12|c d_lb: 14|c d_lb: 16|c d_lb: 18|c d_lb: 20|c d_lb: 20|c d_ub: 40|c status: TIMEOUT (child killed after -cpu-lim)|c d : UNKNOWN|s UNKNOWN|
timeout-cc new rc=1 wall=3.031 c d_ub: 52|c d_lb: 2|c d_ub: 40|c d_lb: 4|c d_lb: 6|c d_lb: 8|c d_lb: 10|c d_lb: 12|c d_lb: 14|c d_lb: 16|c d_lb: 18|c d_lb: 20|c d_lb: 20|c d_ub: 40|c status: TIMEOUT (child killed after -cpu-lim)|c d : UNKNOWN|s UNKNOWN|
memcap-200M old rc=1 wall=2.913 c d_lb: -|c d_ub: -|c status: UNKNOWN|c d : UNKNOWN|s UNKNOWN|
memcap-200M new rc=1 wall=3.060 c d_lb: -|c d_ub: -|c status: UNKNOWN|c d : UNKNOWN|s UNKNOWN|
memcap-60M-cc old rc=1 wall=1.383 c d_ub: 50|c d_lb: 2|c d_lb: 2|c d_ub: 50|c status: UNKNOWN|c d : UNKNOWN|s UNKNOWN|
memcap-60M-cc new rc=1 wall=1.281 c d_ub: 50|c d_lb: 2|c d_lb: 2|c d_ub: 50|c status: UNKNOWN|c d : UNKNOWN|s UNKNOWN|
child-sigkill old rc=1 after-kill=0.008 c status: UNKNOWN|s UNKNOWN|
child-sigkill new rc=1 after-kill=0.006 c status: UNKNOWN|s UNKNOWN|
child-stop-cont old rc=0 wall=7.373 o 10|
child-stop-cont new rc=0 wall=7.220 o 10|
44 changes: 44 additions & 0 deletions optimization/experiments/GH-224/raw/spot_analysis.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,44 @@
## Cells (new vs old; median over reps, then over seeds; wall in s)

| Tier | Mode | Code | seeds | old wall | new wall | wall ratio | cpu ratio |
|---|---|---|---:|---:|---:|---:|---:|
| tier2 | cc | BB_144_14_14 | 3 | 0.6231 | 0.6204 | 0.996 | 1.006 |
| tier2 | cc | LP_544_80_12 | 3 | 0.0438 | 0.0333 | 0.759 | 0.934 |
| tier2 | cc | TN_108_2_12 | 3 | 1.8107 | 1.8118 | 1.001 | 1.001 |
| tier2 | mto | BB_144_14_14 | 3 | 13.3797 | 13.1300 | 0.981 | 0.977 |
| tier2 | mto | LP_544_80_12 | 3 | 14.5664 | 14.1177 | 0.969 | 0.968 |
| tier2 | mto | TN_108_2_12 | 3 | 19.8998 | 19.4537 | 0.978 | 0.979 |
| tier2 | nocard | BB_144_14_14 | 3 | 13.1625 | 12.6378 | 0.960 | 0.961 |
| tier2 | nocard | LP_544_80_12 | 3 | 13.9501 | 13.4979 | 0.968 | 0.968 |
| tier2 | nocard | TN_108_2_12 | 3 | 19.9104 | 19.4988 | 0.979 | 0.979 |
| tier3 | cc | HGP_481_1_16 | 3 | 0.1725 | 0.1561 | 0.905 | 0.946 |
| tier3 | cc | LP_1020_136_16 | 3 | 0.2804 | 0.2725 | 0.972 | 0.975 |
| tier3 | cc | TN_250_10_15 | 3 | 50.5596 | 50.6515 | 1.002 | 1.002 |

## Verdict per tier and mode

| Tier | Mode | cells | geomean | worst cell | cells > 1.25 | verdict |
|---|---|---:|---:|---|---:|---|
| tier2 | cc | 3 | 0.911 | 1.001 (TN_108_2_12) | 0 | PASS |
| tier2 | mto | 3 | 0.976 | 0.981 (BB_144_14_14) | 0 | PASS |
| tier2 | nocard | 3 | 0.969 | 0.979 (TN_108_2_12) | 0 | PASS |
| tier3 | cc | 3 | 0.959 | 1.002 (TN_250_10_15) | 0 | PASS |
| tier2 | all | 9 | 0.952 | 1.001 | 0 | PASS |
| tier3 | all | 3 | 0.959 | 1.002 | 0 | PASS |

## Logs

units with byte-identical stdout in every run of both arms: 34; other units: 2 (of which old itself varies across reps: 1; identical up to line order in every run: 1)
- ('tier3', 'cc', 'LP_1020_136_16', 2): distinct stdouts old 2, new 2, same set: False
- ('tier3', 'cc', 'TN_250_10_15', 0): distinct stdouts old 1, new 2, same set: False

## Answers

runs: 216; rc != 0 or answer != stem: 0 []

## Wait slack (wall - user - sys) of runs with wall < 0.2 s, in ms

| Arm | runs | median | p90 | max |
|---|---:|---:|---:|---:|
| old | 20 | 12.27 | 14.79 | 15.59 |
| new | 21 | 3.37 | 4.25 | 4.99 |
140 changes: 140 additions & 0 deletions optimization/experiments/GH-224/raw/t01/analysis.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,140 @@
## Cells (new vs old; median over reps, then over seeds; wall in s)

| Tier | Mode | Code | seeds | old wall | new wall | wall ratio | cpu ratio |
|---|---|---|---:|---:|---:|---:|---:|
| tier0 | cc | AMC_108_6_9 | 5 | 0.0253 | 0.0209 | 0.825 | 1.174 |
| tier0 | cc | AMC_216_6_12 | 5 | 0.0293 | 0.0242 | 0.827 | 0.905 |
| tier0 | cc | AMC_330_6_14 | 5 | 0.0333 | 0.0321 | 0.964 | 0.993 |
| tier0 | cc | BB_108_8_10 | 5 | 0.0274 | 0.0167 | 0.611 | 0.840 |
| tier0 | cc | BB_120_8_12 | 5 | 0.0253 | 0.0205 | 0.809 | 1.076 |
| tier0 | cc | BB_144_12_12 | 5 | 0.0248 | 0.0210 | 0.848 | 1.141 |
| tier0 | cc | BB_30_4_6 | 5 | 0.0223 | 0.0157 | 0.706 | 1.196 |
| tier0 | cc | BB_72_12_6 | 5 | 0.0231 | 0.0155 | 0.674 | 1.091 |
| tier0 | cc | BB_90_8_10 | 5 | 0.0250 | 0.0163 | 0.651 | 0.992 |
| tier0 | cc | COLOR_19_1_5 | 5 | 0.0230 | 0.0133 | 0.578 | 0.929 |
| tier0 | cc | COLOR_7_1_3 | 5 | 0.0231 | 0.0129 | 0.561 | 0.939 |
| tier0 | cc | GB_144_12_8 | 5 | 0.0267 | 0.0182 | 0.682 | 0.957 |
| tier0 | cc | HGP_106_16_4 | 5 | 0.0268 | 0.0156 | 0.582 | 0.906 |
| tier0 | cc | HGP_113_1_8 | 5 | 0.0279 | 0.0161 | 0.577 | 0.846 |
| tier0 | cc | HGP_13_1_3 | 5 | 0.0231 | 0.0121 | 0.525 | 0.900 |
| tier0 | cc | HGP_306_36_5 | 5 | 0.0352 | 0.0299 | 0.848 | 1.128 |
| tier0 | cc | HGP_41_1_5 | 5 | 0.0256 | 0.0160 | 0.625 | 0.939 |
| tier0 | cc | HGP_45_9_3 | 5 | 0.0222 | 0.0152 | 0.682 | 1.160 |
| tier0 | cc | LP_136_32_4 | 5 | 0.0260 | 0.0181 | 0.697 | 1.039 |
| tier0 | cc | LP_238_44_6 | 5 | 0.0250 | 0.0229 | 0.915 | 1.095 |
| tier0 | cc | LP_34_20_2 | 5 | 0.0242 | 0.0137 | 0.567 | 0.880 |
| tier0 | cc | TN_36_8_3 | 5 | 0.0268 | 0.0136 | 0.507 | 0.840 |
| tier0 | cc | TN_36_8_4 | 5 | 0.0239 | 0.0123 | 0.512 | 0.836 |
| tier0 | cc | TN_54_11_4 | 5 | 0.0288 | 0.0143 | 0.495 | 0.860 |
| tier0 | cc | TN_72_14_4 | 5 | 0.0284 | 0.0144 | 0.508 | 0.814 |
| tier0 | cc | TN_72_8_8 | 5 | 0.0236 | 0.0211 | 0.894 | 0.977 |
| tier0 | mto | AMC_108_6_9 | 5 | 0.1000 | 0.0968 | 0.968 | 1.011 |
| tier0 | mto | AMC_216_6_12 | 5 | 0.2981 | 0.2937 | 0.985 | 0.999 |
| tier0 | mto | AMC_330_6_14 | 5 | 0.6915 | 0.6771 | 0.979 | 0.987 |
| tier0 | mto | BB_108_8_10 | 5 | 0.0492 | 0.0464 | 0.942 | 0.995 |
| tier0 | mto | BB_120_8_12 | 5 | 0.5875 | 0.5649 | 0.961 | 0.971 |
| tier0 | mto | BB_144_12_12 | 5 | 0.1203 | 0.1235 | 1.027 | 1.018 |
| tier0 | mto | BB_30_4_6 | 5 | 0.0340 | 0.0280 | 0.823 | 1.087 |
| tier0 | mto | BB_72_12_6 | 5 | 0.0323 | 0.0276 | 0.855 | 1.039 |
| tier0 | mto | BB_90_8_10 | 5 | 0.0549 | 0.0487 | 0.887 | 1.014 |
| tier0 | mto | COLOR_19_1_5 | 5 | 0.0244 | 0.0180 | 0.738 | 1.000 |
| tier0 | mto | COLOR_7_1_3 | 5 | 0.0229 | 0.0138 | 0.603 | 0.990 |
| tier0 | mto | GB_144_12_8 | 5 | 0.0682 | 0.0620 | 0.909 | 0.996 |
| tier0 | mto | HGP_106_16_4 | 5 | 0.0590 | 0.0513 | 0.869 | 0.980 |
| tier0 | mto | HGP_113_1_8 | 5 | 0.0700 | 0.0598 | 0.855 | 0.944 |
| tier0 | mto | HGP_13_1_3 | 5 | 0.0251 | 0.0135 | 0.537 | 0.903 |
| tier0 | mto | HGP_306_36_5 | 5 | 0.3514 | 0.3452 | 0.983 | 1.000 |
| tier0 | mto | HGP_41_1_5 | 5 | 0.0261 | 0.0215 | 0.823 | 1.000 |
| tier0 | mto | HGP_45_9_3 | 5 | 0.0242 | 0.0231 | 0.957 | 1.036 |
| tier0 | mto | LP_136_32_4 | 5 | 0.0550 | 0.0471 | 0.857 | 1.012 |
| tier0 | mto | LP_238_44_6 | 5 | 0.2506 | 0.2585 | 1.032 | 1.028 |
| tier0 | mto | LP_34_20_2 | 5 | 0.0247 | 0.0147 | 0.593 | 0.845 |
| tier0 | mto | TN_36_8_3 | 5 | 0.0273 | 0.0213 | 0.779 | 0.946 |
| tier0 | mto | TN_36_8_4 | 5 | 0.0269 | 0.0219 | 0.815 | 0.893 |
| tier0 | mto | TN_54_11_4 | 5 | 0.0478 | 0.0387 | 0.810 | 0.927 |
| tier0 | mto | TN_72_14_4 | 5 | 0.0597 | 0.0484 | 0.810 | 0.911 |
| tier0 | mto | TN_72_8_8 | 5 | 0.6436 | 0.6200 | 0.963 | 0.974 |
| tier0 | nocard | AMC_108_6_9 | 5 | 0.0897 | 0.0910 | 1.014 | 1.024 |
| tier0 | nocard | AMC_216_6_12 | 5 | 0.2769 | 0.2720 | 0.982 | 0.985 |
| tier0 | nocard | AMC_330_6_14 | 5 | 0.6725 | 0.6664 | 0.991 | 0.990 |
| tier0 | nocard | BB_108_8_10 | 5 | 0.0477 | 0.0439 | 0.919 | 1.064 |
| tier0 | nocard | BB_120_8_12 | 5 | 0.5319 | 0.5160 | 0.970 | 0.986 |
| tier0 | nocard | BB_144_12_12 | 5 | 0.1193 | 0.1179 | 0.989 | 1.005 |
| tier0 | nocard | BB_30_4_6 | 5 | 0.0225 | 0.0244 | 1.086 | 1.112 |
| tier0 | nocard | BB_72_12_6 | 5 | 0.0326 | 0.0258 | 0.793 | 1.025 |
| tier0 | nocard | BB_90_8_10 | 5 | 0.0483 | 0.0415 | 0.860 | 1.028 |
| tier0 | nocard | COLOR_19_1_5 | 5 | 0.0242 | 0.0178 | 0.739 | 1.008 |
| tier0 | nocard | COLOR_7_1_3 | 5 | 0.0234 | 0.0161 | 0.688 | 1.049 |
| tier0 | nocard | GB_144_12_8 | 5 | 0.0668 | 0.0612 | 0.917 | 0.978 |
| tier0 | nocard | HGP_106_16_4 | 5 | 0.0578 | 0.0464 | 0.803 | 0.952 |
| tier0 | nocard | HGP_113_1_8 | 5 | 0.0621 | 0.0510 | 0.822 | 0.959 |
| tier0 | nocard | HGP_13_1_3 | 5 | 0.0240 | 0.0135 | 0.561 | 0.904 |
| tier0 | nocard | HGP_306_36_5 | 5 | 0.3096 | 0.3028 | 0.978 | 0.998 |
| tier0 | nocard | HGP_41_1_5 | 5 | 0.0259 | 0.0198 | 0.761 | 1.020 |
| tier0 | nocard | HGP_45_9_3 | 5 | 0.0238 | 0.0203 | 0.854 | 1.039 |
| tier0 | nocard | LP_136_32_4 | 5 | 0.0548 | 0.0493 | 0.901 | 1.019 |
| tier0 | nocard | LP_238_44_6 | 5 | 0.2436 | 0.2490 | 1.022 | 1.014 |
| tier0 | nocard | LP_34_20_2 | 5 | 0.0254 | 0.0190 | 0.749 | 0.945 |
| tier0 | nocard | TN_36_8_3 | 5 | 0.0269 | 0.0200 | 0.744 | 0.892 |
| tier0 | nocard | TN_36_8_4 | 5 | 0.0354 | 0.0258 | 0.728 | 0.955 |
| tier0 | nocard | TN_54_11_4 | 5 | 0.0485 | 0.0374 | 0.770 | 0.979 |
| tier0 | nocard | TN_72_14_4 | 5 | 0.0553 | 0.0447 | 0.809 | 0.945 |
| tier0 | nocard | TN_72_8_8 | 5 | 0.5531 | 0.5362 | 0.969 | 0.976 |
| tier1 | cc | AMC_438_6_17 | 5 | 0.0778 | 0.0699 | 0.898 | 0.965 |
| tier1 | cc | BB_186_10_14 | 5 | 0.0414 | 0.0234 | 0.564 | 0.829 |
| tier1 | cc | BB_288_12_18 | 5 | 0.3146 | 0.3130 | 0.995 | 1.007 |
| tier1 | cc | COLOR_127_1_13 | 5 | 0.0627 | 0.0587 | 0.936 | 1.008 |
| tier1 | cc | GB_144_12_12 | 5 | 0.0709 | 0.0654 | 0.923 | 1.007 |
| tier1 | cc | HGP_265_1_12 | 5 | 0.0308 | 0.0255 | 0.828 | 0.921 |
| tier1 | cc | HGP_610_64_6 | 5 | 0.1471 | 0.1357 | 0.922 | 0.969 |
| tier1 | cc | LP_340_56_8 | 5 | 0.0272 | 0.0268 | 0.985 | 1.114 |
| tier1 | cc | LP_442_68_10 | 5 | 0.0298 | 0.0285 | 0.957 | 0.972 |
| tier1 | mto | AMC_438_6_17 | 5 | 7.9938 | 7.8562 | 0.983 | 0.977 |
| tier1 | mto | BB_186_10_14 | 5 | 1.1637 | 1.1350 | 0.975 | 0.978 |
| tier1 | mto | BB_288_12_18 | 5 | 9.9731 | 9.7827 | 0.981 | 0.980 |
| tier1 | mto | COLOR_127_1_13 | 5 | 1.3208 | 1.2798 | 0.969 | 0.975 |
| tier1 | mto | GB_144_12_12 | 5 | 1.0759 | 1.0499 | 0.976 | 0.978 |
| tier1 | mto | HGP_265_1_12 | 5 | 1.7066 | 1.6676 | 0.977 | 0.980 |
| tier1 | mto | HGP_610_64_6 | 5 | 8.1101 | 8.0034 | 0.987 | 0.999 |
| tier1 | mto | LP_340_56_8 | 5 | 1.8571 | 1.8146 | 0.977 | 0.981 |
| tier1 | mto | LP_442_68_10 | 5 | 4.9794 | 4.8794 | 0.980 | 0.976 |
| tier1 | nocard | AMC_438_6_17 | 5 | 7.7806 | 7.5688 | 0.973 | 0.966 |
| tier1 | nocard | BB_186_10_14 | 5 | 1.0473 | 1.0235 | 0.977 | 0.981 |
| tier1 | nocard | BB_288_12_18 | 5 | 10.0105 | 9.8113 | 0.980 | 0.980 |
| tier1 | nocard | COLOR_127_1_13 | 5 | 1.4968 | 1.4582 | 0.974 | 0.977 |
| tier1 | nocard | GB_144_12_12 | 5 | 1.0738 | 1.0582 | 0.985 | 0.984 |
| tier1 | nocard | HGP_265_1_12 | 5 | 1.7180 | 1.6769 | 0.976 | 0.983 |
| tier1 | nocard | HGP_610_64_6 | 5 | 8.6281 | 8.5126 | 0.987 | 0.988 |
| tier1 | nocard | LP_340_56_8 | 5 | 2.1456 | 2.1094 | 0.983 | 0.983 |
| tier1 | nocard | LP_442_68_10 | 5 | 4.0491 | 4.0201 | 0.993 | 0.976 |

## Verdict per tier and mode

| Tier | Mode | cells | geomean | worst cell | cells > 1.25 | verdict |
|---|---|---:|---:|---|---:|---|
| tier0 | cc | 26 | 0.666 | 0.964 (AMC_330_6_14) | 0 | PASS |
| tier0 | mto | 26 | 0.849 | 1.032 (LP_238_44_6) | 0 | PASS |
| tier0 | nocard | 26 | 0.853 | 1.086 (BB_30_4_6) | 0 | PASS |
| tier1 | cc | 9 | 0.879 | 0.995 (BB_288_12_18) | 0 | PASS |
| tier1 | mto | 9 | 0.978 | 0.987 (HGP_610_64_6) | 0 | PASS |
| tier1 | nocard | 9 | 0.981 | 0.993 (LP_442_68_10) | 0 | PASS |
| tier0 | all | 78 | 0.784 | 1.086 | 0 | PASS |
| tier1 | all | 27 | 0.945 | 0.995 | 0 | PASS |

## Logs

units with byte-identical stdout in every run of both arms: 522; other units: 3 (of which old itself varies across reps: 2; identical up to line order in every run: 3)
- ('tier1', 'cc', 'GB_144_12_12', 0): distinct stdouts old 2, new 2, same set: False
- ('tier1', 'cc', 'GB_144_12_12', 2): distinct stdouts old 2, new 2, same set: True
- ('tier1', 'cc', 'GB_144_12_12', 3): distinct stdouts old 1, new 2, same set: False

## Answers

runs: 9450; rc != 0 or answer != stem: 0 []

## Wait slack (wall - user - sys) of runs with wall < 0.2 s, in ms

| Arm | runs | median | p90 | max |
|---|---:|---:|---:|---:|
| old | 3331 | 11.88 | 16.15 | 49.91 |
| new | 3337 | 4.77 | 7.82 | 19.38 |
50 changes: 50 additions & 0 deletions optimization/experiments/GH-224/raw/t01/config.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,50 @@
{
"arms": {
"old": [
"/home/yfc/mac-agent-20261009/frozen/server-gh224-main-fecee8c/distqldpc",
[]
],
"new": [
"/home/yfc/mac-agent-20261009/frozen/server-gh224-2587b9d/distqldpc",
[]
]
},
"cpus": [
72,
73,
74,
75,
76,
77,
78,
79,
80,
81,
82,
83,
84,
85,
86,
87,
88,
89,
90,
91,
92,
93,
94,
95,
96,
97,
98,
99,
100,
101
],
"reps": 9,
"seeds": 5,
"modes": "mto,nocard,cc",
"limit": 600,
"units": 525,
"started": 1793815873.1662354
}
Binary file not shown.
4 changes: 4 additions & 0 deletions optimization/experiments/GH-224/raw/t01/skips.log
Original file line number Diff line number Diff line change
@@ -0,0 +1,4 @@
1793815874 cpu 84 busy, unit ('AMC_438_6_17', 'cc', 2)
1793816085 cpu 101 busy, unit ('BB_108_8_10', 'nocard', 0)
1793816086 cpu 97 busy, unit ('BB_108_8_10', 'cc', 2)
1793816109 cpu 85 busy, unit ('COLOR_19_1_5', 'cc', 0)
Loading
Loading