Research 137: the batch executor's stage 4 (columns read in place, comparisons counting their lanes): knapsack 10.07 to 8.25 ms (bar 8.6; b9 11.46, the compiled code 7.9), its instructions 12.2% below b9's; what is left of knapsack's gap is the per-batch interpretation (about 3 ms), not the kernels (ADD 1.7, GT 1.7, the masked set 0.55 ms); the others the same by instructions; make check's server_view_workers timed out once (the known flake), passed 3/3 alone and in a full rerun; bank queries' instructions within 0.1% of b9's, TPC-C -0.3%; the two large gaps left (matmul's nested rounds, BFS's candidate tail) the user's to choose

1 files changed, 4 insertions, 0 deletions

diff --git a/research/experiments/040-against-c/2026-10-08-queries-s16-12.jsonl b/research/experiments/040-against-c/2026-10-08-queries-s16-12.jsonl
new file mode 100644
index 0000000..4a3ce77
--- /dev/null
+++ b/research/experiments/040-against-c/2026-10-08-queries-s16-12.jsonl
@@ -0,0 +1,4 @@
+{"query": "per_branch", "rows": 10, "accounts": 100000, "pure_median_ms": 5.28, "pure_fastest_ms": 5.18, "sqlite_median_ms": 45.5, "sqlite_fastest_ms": 45.27, "sqlite_indexed_median_ms": 14.15, "sqlite_indexed_fastest_ms": 14.06}
+{"query": "in_credit", "rows": 1, "accounts": 100000, "pure_median_ms": 1.19, "pure_fastest_ms": 1.04, "sqlite_median_ms": 6.29, "sqlite_fastest_ms": 6.26, "sqlite_indexed_median_ms": 0.16, "sqlite_indexed_fastest_ms": 0.16}
+{"query": "large", "rows": 212, "accounts": 100000, "pure_median_ms": 1.06, "pure_fastest_ms": 0.7, "sqlite_median_ms": 1.87, "sqlite_fastest_ms": 1.51, "sqlite_indexed_median_ms": 1.48, "sqlite_indexed_fastest_ms": 1.47}
+{"query": "ranks", "rows": 404, "accounts": 100000, "pure_median_ms": 8.2, "pure_fastest_ms": 6.77, "sqlite_median_ms": 2883.56, "sqlite_fastest_ms": 2837.2, "sqlite_indexed_median_ms": 7.61, "sqlite_indexed_fastest_ms": 7.52}