aboutsummaryrefslogtreecommitdiffstats
path: root/research/experiments/041-batches
diff options
context:
space:
mode:
authoreunhhu <[email protected]>2026-10-08 16:53:29 +0100
committereunhhu <[email protected]>2026-10-08 16:53:29 +0100
commitbe51c8a15d4f220ea8322f3c562f1f0c51b1e8e7 (patch)
treeec49b9aeb46f9ff1b00542d8ea774ef152919865 /research/experiments/041-batches
parente7760db551322810de7824f25a38e9f8f737b0ce (diff)
downloadpure-be51c8a15d4f220ea8322f3c562f1f0c51b1e8e7.tar.gz
pure-be51c8a15d4f220ea8322f3c562f1f0c51b1e8e7.zip
Research 136: the batch executor's stage 3 (four-byte MUL, reductions with no lanes while every lane's e is the same, a list's cells once a batch): against b9, interleaved, matmul 40.01 to 25.97 ms, knapsack 11.47 to 10.07, roots 8.97 to 8.29, join 117.2 to 115.8, BFS and sort the same; sum0 8.69 to 1.01 (bar 1.5); matmul short of its bar (21): the nested rounds' interpretation (about 14 ms) and the matrices' making (4-5) were left out of the arithmetic; the executor alone and the bank queries' instructions as b9 (0.08%), TPC-C +0.7%; stage 4 next (knapsack's six passes to three)
Diffstat (limited to 'research/experiments/041-batches')
-rw-r--r--research/experiments/041-batches/2026-10-08-counts.jsonl10
-rw-r--r--research/experiments/041-batches/2026-10-08-pairs.jsonl92
2 files changed, 102 insertions, 0 deletions
diff --git a/research/experiments/041-batches/2026-10-08-counts.jsonl b/research/experiments/041-batches/2026-10-08-counts.jsonl
index d259259..6770a54 100644
--- a/research/experiments/041-batches/2026-10-08-counts.jsonl
+++ b/research/experiments/041-batches/2026-10-08-counts.jsonl
@@ -66,3 +66,13 @@
{"version":"b13","way":"rounds","work":"roots","instructions":2755945389,"cycles":987937588}
{"version":"b13","way":"batches","work":"join","instructions":1862428182,"cycles":1070892757}
{"version":"b13","way":"rounds","work":"join","instructions":7669661642,"cycles":3261839112}
+{"version":"b14","way":"batches","work":"bfs","instructions":3367636863,"cycles":1668001492}
+{"version":"b14","way":"rounds","work":"bfs","instructions":4711634132,"cycles":2090008896}
+{"version":"b14","way":"batches","work":"knapsack","instructions":372331557,"cycles":117955187}
+{"version":"b14","way":"rounds","work":"knapsack","instructions":21688103016,"cycles":7544708847}
+{"version":"b14","way":"batches","work":"matmul","instructions":718548121,"cycles":242024080}
+{"version":"b14","way":"rounds","work":"matmul","instructions":21436715931,"cycles":6621517691}
+{"version":"b14","way":"batches","work":"roots","instructions":156139120,"cycles":109411693}
+{"version":"b14","way":"rounds","work":"roots","instructions":2755945510,"cycles":988490613}
+{"version":"b14","way":"batches","work":"join","instructions":1868676355,"cycles":1061229557}
+{"version":"b14","way":"rounds","work":"join","instructions":7669661592,"cycles":3636019220}
diff --git a/research/experiments/041-batches/2026-10-08-pairs.jsonl b/research/experiments/041-batches/2026-10-08-pairs.jsonl
index 7d1343f..36dd15a 100644
--- a/research/experiments/041-batches/2026-10-08-pairs.jsonl
+++ b/research/experiments/041-batches/2026-10-08-pairs.jsonl
@@ -308,3 +308,95 @@
{"version":"b13","way":"batches","turn":5,"work":"join","pure_median_ms":118.155,"c_median_ms":30.847,"ratio":3.830,"agree":true}
{"version":"b13","way":"batches","turn":5,"work":"matmul","pure_median_ms":26.718,"c_median_ms":7.599,"ratio":3.516,"agree":true}
{"version":"b13","way":"batches","turn":5,"work":"bfs","pure_median_ms":24.025,"c_median_ms":10.005,"ratio":2.401,"agree":true}
+{"version":"b13","way":"batches","turn":1,"work":"roots","pure_median_ms":8.316,"c_median_ms":6.869,"ratio":1.211,"agree":true}
+{"version":"b13","way":"batches","turn":1,"work":"knapsack","pure_median_ms":10.680,"c_median_ms":6.296,"ratio":1.696,"agree":true}
+{"version":"b13","way":"batches","turn":1,"work":"join","pure_median_ms":114.792,"c_median_ms":29.910,"ratio":3.838,"agree":true}
+{"version":"b13","way":"batches","turn":1,"work":"matmul","pure_median_ms":26.754,"c_median_ms":7.542,"ratio":3.547,"agree":true}
+{"version":"b13","way":"batches","turn":1,"work":"bfs","pure_median_ms":24.127,"c_median_ms":10.223,"ratio":2.360,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"roots","pure_median_ms":8.282,"c_median_ms":6.865,"ratio":1.206,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"knapsack","pure_median_ms":10.048,"c_median_ms":6.224,"ratio":1.614,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"join","pure_median_ms":115.119,"c_median_ms":30.225,"ratio":3.809,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"matmul","pure_median_ms":25.998,"c_median_ms":7.535,"ratio":3.450,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"bfs","pure_median_ms":24.029,"c_median_ms":9.958,"ratio":2.413,"agree":true}
+{"version":"b13","way":"batches","turn":2,"work":"roots","pure_median_ms":8.355,"c_median_ms":6.879,"ratio":1.215,"agree":true}
+{"version":"b13","way":"batches","turn":2,"work":"knapsack","pure_median_ms":10.627,"c_median_ms":6.281,"ratio":1.692,"agree":true}
+{"version":"b13","way":"batches","turn":2,"work":"join","pure_median_ms":114.690,"c_median_ms":32.100,"ratio":3.573,"agree":true}
+{"version":"b13","way":"batches","turn":2,"work":"matmul","pure_median_ms":26.708,"c_median_ms":7.540,"ratio":3.542,"agree":true}
+{"version":"b13","way":"batches","turn":2,"work":"bfs","pure_median_ms":23.920,"c_median_ms":10.065,"ratio":2.377,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"roots","pure_median_ms":8.285,"c_median_ms":6.865,"ratio":1.207,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"knapsack","pure_median_ms":10.050,"c_median_ms":6.224,"ratio":1.615,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"join","pure_median_ms":115.790,"c_median_ms":31.902,"ratio":3.630,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"matmul","pure_median_ms":25.976,"c_median_ms":7.538,"ratio":3.446,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"bfs","pure_median_ms":23.984,"c_median_ms":9.983,"ratio":2.403,"agree":true}
+{"version":"b13","way":"batches","turn":3,"work":"roots","pure_median_ms":8.309,"c_median_ms":6.865,"ratio":1.210,"agree":true}
+{"version":"b13","way":"batches","turn":3,"work":"knapsack","pure_median_ms":10.540,"c_median_ms":6.227,"ratio":1.693,"agree":true}
+{"version":"b13","way":"batches","turn":3,"work":"join","pure_median_ms":115.973,"c_median_ms":31.762,"ratio":3.651,"agree":true}
+{"version":"b13","way":"batches","turn":3,"work":"matmul","pure_median_ms":26.773,"c_median_ms":7.518,"ratio":3.561,"agree":true}
+{"version":"b13","way":"batches","turn":3,"work":"bfs","pure_median_ms":24.051,"c_median_ms":10.067,"ratio":2.389,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"roots","pure_median_ms":8.280,"c_median_ms":6.866,"ratio":1.206,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"knapsack","pure_median_ms":10.034,"c_median_ms":6.229,"ratio":1.611,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"join","pure_median_ms":117.243,"c_median_ms":31.426,"ratio":3.731,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"matmul","pure_median_ms":25.945,"c_median_ms":7.523,"ratio":3.449,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"bfs","pure_median_ms":24.154,"c_median_ms":10.190,"ratio":2.370,"agree":true}
+{"version":"b13","way":"batches","turn":4,"work":"roots","pure_median_ms":8.312,"c_median_ms":6.861,"ratio":1.211,"agree":true}
+{"version":"b13","way":"batches","turn":4,"work":"knapsack","pure_median_ms":10.576,"c_median_ms":6.226,"ratio":1.699,"agree":true}
+{"version":"b13","way":"batches","turn":4,"work":"join","pure_median_ms":116.402,"c_median_ms":29.137,"ratio":3.995,"agree":true}
+{"version":"b13","way":"batches","turn":4,"work":"matmul","pure_median_ms":26.717,"c_median_ms":7.663,"ratio":3.487,"agree":true}
+{"version":"b13","way":"batches","turn":4,"work":"bfs","pure_median_ms":24.260,"c_median_ms":10.011,"ratio":2.423,"agree":true}
+{"version":"b14","way":"batches","turn":4,"work":"roots","pure_median_ms":8.287,"c_median_ms":6.863,"ratio":1.207,"agree":true}
+{"version":"b14","way":"batches","turn":4,"work":"knapsack","pure_median_ms":10.163,"c_median_ms":6.448,"ratio":1.576,"agree":true}
+{"version":"b14","way":"batches","turn":4,"work":"join","pure_median_ms":115.463,"c_median_ms":30.471,"ratio":3.789,"agree":true}
+{"version":"b14","way":"batches","turn":4,"work":"matmul","pure_median_ms":25.997,"c_median_ms":7.540,"ratio":3.448,"agree":true}
+{"version":"b14","way":"batches","turn":4,"work":"bfs","pure_median_ms":24.003,"c_median_ms":10.093,"ratio":2.378,"agree":true}
+{"version":"b13","way":"batches","turn":5,"work":"roots","pure_median_ms":8.308,"c_median_ms":6.861,"ratio":1.211,"agree":true}
+{"version":"b13","way":"batches","turn":5,"work":"knapsack","pure_median_ms":10.533,"c_median_ms":6.267,"ratio":1.681,"agree":true}
+{"version":"b13","way":"batches","turn":5,"work":"join","pure_median_ms":115.249,"c_median_ms":31.876,"ratio":3.616,"agree":true}
+{"version":"b13","way":"batches","turn":5,"work":"matmul","pure_median_ms":26.792,"c_median_ms":7.545,"ratio":3.551,"agree":true}
+{"version":"b13","way":"batches","turn":5,"work":"bfs","pure_median_ms":24.135,"c_median_ms":10.019,"ratio":2.409,"agree":true}
+{"version":"b14","way":"batches","turn":5,"work":"roots","pure_median_ms":8.300,"c_median_ms":6.873,"ratio":1.207,"agree":true}
+{"version":"b14","way":"batches","turn":5,"work":"knapsack","pure_median_ms":10.102,"c_median_ms":6.227,"ratio":1.622,"agree":true}
+{"version":"b14","way":"batches","turn":5,"work":"join","pure_median_ms":117.360,"c_median_ms":31.328,"ratio":3.746,"agree":true}
+{"version":"b14","way":"batches","turn":5,"work":"matmul","pure_median_ms":25.966,"c_median_ms":7.539,"ratio":3.444,"agree":true}
+{"version":"b14","way":"batches","turn":5,"work":"bfs","pure_median_ms":24.080,"c_median_ms":10.060,"ratio":2.394,"agree":true}
+{"version":"b9","way":"batches","turn":1,"work":"roots","pure_median_ms":8.962,"c_median_ms":6.872,"ratio":1.304,"agree":true}
+{"version":"b9","way":"batches","turn":1,"work":"knapsack","pure_median_ms":11.500,"c_median_ms":6.240,"ratio":1.843,"agree":true}
+{"version":"b9","way":"batches","turn":1,"work":"sort","pure_median_ms":56.898,"c_median_ms":26.876,"ratio":2.117,"agree":true}
+{"version":"b9","way":"batches","turn":1,"work":"join","pure_median_ms":119.329,"c_median_ms":31.836,"ratio":3.748,"agree":true}
+{"version":"b9","way":"batches","turn":1,"work":"words","pure_median_ms":216.527,"c_median_ms":38.123,"ratio":5.680,"agree":true}
+{"version":"b9","way":"batches","turn":1,"work":"matmul","pure_median_ms":40.022,"c_median_ms":7.521,"ratio":5.322,"agree":true}
+{"version":"b9","way":"batches","turn":1,"work":"bfs","pure_median_ms":24.087,"c_median_ms":9.953,"ratio":2.420,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"roots","pure_median_ms":8.286,"c_median_ms":6.863,"ratio":1.207,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"knapsack","pure_median_ms":10.067,"c_median_ms":6.231,"ratio":1.616,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"sort","pure_median_ms":56.739,"c_median_ms":26.883,"ratio":2.111,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"join","pure_median_ms":115.954,"c_median_ms":31.075,"ratio":3.731,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"words","pure_median_ms":217.494,"c_median_ms":38.742,"ratio":5.614,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"matmul","pure_median_ms":25.983,"c_median_ms":7.546,"ratio":3.443,"agree":true}
+{"version":"b14","way":"batches","turn":1,"work":"bfs","pure_median_ms":24.120,"c_median_ms":10.133,"ratio":2.380,"agree":true}
+{"version":"b9","way":"batches","turn":2,"work":"roots","pure_median_ms":8.965,"c_median_ms":6.869,"ratio":1.305,"agree":true}
+{"version":"b9","way":"batches","turn":2,"work":"knapsack","pure_median_ms":11.418,"c_median_ms":6.231,"ratio":1.832,"agree":true}
+{"version":"b9","way":"batches","turn":2,"work":"sort","pure_median_ms":56.910,"c_median_ms":26.892,"ratio":2.116,"agree":true}
+{"version":"b9","way":"batches","turn":2,"work":"join","pure_median_ms":116.789,"c_median_ms":30.559,"ratio":3.822,"agree":true}
+{"version":"b9","way":"batches","turn":2,"work":"words","pure_median_ms":214.333,"c_median_ms":39.039,"ratio":5.490,"agree":true}
+{"version":"b9","way":"batches","turn":2,"work":"matmul","pure_median_ms":40.032,"c_median_ms":7.601,"ratio":5.266,"agree":true}
+{"version":"b9","way":"batches","turn":2,"work":"bfs","pure_median_ms":24.051,"c_median_ms":10.104,"ratio":2.380,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"roots","pure_median_ms":8.286,"c_median_ms":6.867,"ratio":1.207,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"knapsack","pure_median_ms":10.107,"c_median_ms":6.230,"ratio":1.622,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"sort","pure_median_ms":56.756,"c_median_ms":26.876,"ratio":2.112,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"join","pure_median_ms":115.280,"c_median_ms":30.595,"ratio":3.768,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"words","pure_median_ms":220.892,"c_median_ms":38.701,"ratio":5.708,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"matmul","pure_median_ms":25.951,"c_median_ms":7.535,"ratio":3.444,"agree":true}
+{"version":"b14","way":"batches","turn":2,"work":"bfs","pure_median_ms":24.336,"c_median_ms":9.945,"ratio":2.447,"agree":true}
+{"version":"b9","way":"batches","turn":3,"work":"roots","pure_median_ms":8.963,"c_median_ms":6.863,"ratio":1.306,"agree":true}
+{"version":"b9","way":"batches","turn":3,"work":"knapsack","pure_median_ms":11.456,"c_median_ms":6.247,"ratio":1.834,"agree":true}
+{"version":"b9","way":"batches","turn":3,"work":"sort","pure_median_ms":57.041,"c_median_ms":26.902,"ratio":2.120,"agree":true}
+{"version":"b9","way":"batches","turn":3,"work":"join","pure_median_ms":117.863,"c_median_ms":31.866,"ratio":3.699,"agree":true}
+{"version":"b9","way":"batches","turn":3,"work":"words","pure_median_ms":213.690,"c_median_ms":39.481,"ratio":5.412,"agree":true}
+{"version":"b9","way":"batches","turn":3,"work":"matmul","pure_median_ms":39.951,"c_median_ms":7.982,"ratio":5.005,"agree":true}
+{"version":"b9","way":"batches","turn":3,"work":"bfs","pure_median_ms":24.316,"c_median_ms":10.086,"ratio":2.411,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"roots","pure_median_ms":8.287,"c_median_ms":6.863,"ratio":1.208,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"knapsack","pure_median_ms":10.071,"c_median_ms":6.228,"ratio":1.617,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"sort","pure_median_ms":56.973,"c_median_ms":26.895,"ratio":2.118,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"join","pure_median_ms":115.795,"c_median_ms":32.216,"ratio":3.594,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"words","pure_median_ms":219.174,"c_median_ms":38.390,"ratio":5.709,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"matmul","pure_median_ms":25.934,"c_median_ms":7.564,"ratio":3.428,"agree":true}
+{"version":"b14","way":"batches","turn":3,"work":"bfs","pure_median_ms":27.887,"c_median_ms":10.578,"ratio":2.636,"agree":true}