aboutsummaryrefslogtreecommitdiffstats
path: root/research
diff options
context:
space:
mode:
authoreunhhu <[email protected]>2026-10-08 15:52:44 +0100
committereunhhu <[email protected]>2026-10-08 15:52:44 +0100
commit6828a7d07f9de815c858a295124f6bcde9bd861a (patch)
tree7a7ae86228c20ff8dd041588614d2d58b81fc3fc /research
parentbdcc6f66e3addf1980b3b065609cc0f5075a4194 (diff)
downloadpure-6828a7d07f9de815c858a295124f6bcde9bd861a.tar.gz
pure-6828a7d07f9de815c858a295124f6bcde9bd861a.zip
Research 135: batch plans' stage 2 measured (b11 = 7959885 against b10, interleaved): matmul 38.18 to 36.10 ms with 4.5% fewer instructions (segments in its nested rounds), knapsack's per-batch cost 0.68 to 0.63 us but knapsack 10.35 to 10.51 ms with 1% fewer instructions (its per-element loops placed elsewhere), sum0 1,743 instructions a batch: short of 025's stage-2 bars (9.2 ms, 0.45 us, 33 ms): 025 priced a segment's instruction at 50 cycles, the counts show about 25 instructions, and a batch's cost sits in finds, bounds, read records, lane setup, masks and the deferred store's check; a list's cells once a batch measured at 3% and not made; next, per element: stage 3's four-byte MUL (matmul 21 ms)
Diffstat (limited to 'research')
-rw-r--r--research/experiments/041-batches/2026-10-08-counts.jsonl10
-rw-r--r--research/experiments/041-batches/2026-10-08-pairs.jsonl90
2 files changed, 100 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 546030b..0e89a46 100644
--- a/research/experiments/041-batches/2026-10-08-counts.jsonl
+++ b/research/experiments/041-batches/2026-10-08-counts.jsonl
@@ -36,3 +36,13 @@
{"version":"b10","way":"rounds","work":"roots","instructions":2755945365,"cycles":989112526}
{"version":"b10","way":"batches","work":"join","instructions":1868533943,"cycles":1049050190}
{"version":"b10","way":"rounds","work":"join","instructions":7669661641,"cycles":3314577656}
+{"version":"b11","way":"batches","work":"bfs","instructions":3381283015,"cycles":1683369887}
+{"version":"b11","way":"rounds","work":"bfs","instructions":4715633866,"cycles":2086008861}
+{"version":"b11","way":"batches","work":"knapsack","instructions":373838059,"cycles":121140642}
+{"version":"b11","way":"rounds","work":"knapsack","instructions":21688102965,"cycles":7548998383}
+{"version":"b11","way":"batches","work":"matmul","instructions":924133786,"cycles":313909094}
+{"version":"b11","way":"rounds","work":"matmul","instructions":21436716024,"cycles":6624677320}
+{"version":"b11","way":"batches","work":"roots","instructions":166257657,"cycles":113412601}
+{"version":"b11","way":"rounds","work":"roots","instructions":2755945430,"cycles":987962207}
+{"version":"b11","way":"batches","work":"join","instructions":1862305226,"cycles":1068457401}
+{"version":"b11","way":"rounds","work":"join","instructions":7669661824,"cycles":3430154017}
diff --git a/research/experiments/041-batches/2026-10-08-pairs.jsonl b/research/experiments/041-batches/2026-10-08-pairs.jsonl
index 1d16abf..03d777c 100644
--- a/research/experiments/041-batches/2026-10-08-pairs.jsonl
+++ b/research/experiments/041-batches/2026-10-08-pairs.jsonl
@@ -118,3 +118,93 @@
{"version":"b10-256","way":"batches","turn":5,"work":"join","pure_median_ms":118.650,"c_median_ms":31.727,"ratio":3.740,"agree":true}
{"version":"b10-256","way":"batches","turn":5,"work":"matmul","pure_median_ms":38.295,"c_median_ms":7.578,"ratio":5.053,"agree":true}
{"version":"b10-256","way":"batches","turn":5,"work":"bfs","pure_median_ms":25.631,"c_median_ms":10.087,"ratio":2.541,"agree":true}
+{"version":"b10","way":"batches","turn":1,"work":"roots","pure_median_ms":8.952,"c_median_ms":6.875,"ratio":1.302,"agree":true}
+{"version":"b10","way":"batches","turn":1,"work":"knapsack","pure_median_ms":10.346,"c_median_ms":6.240,"ratio":1.658,"agree":true}
+{"version":"b10","way":"batches","turn":1,"work":"join","pure_median_ms":116.423,"c_median_ms":30.240,"ratio":3.850,"agree":true}
+{"version":"b10","way":"batches","turn":1,"work":"matmul","pure_median_ms":38.182,"c_median_ms":7.544,"ratio":5.061,"agree":true}
+{"version":"b10","way":"batches","turn":1,"work":"bfs","pure_median_ms":24.217,"c_median_ms":10.015,"ratio":2.418,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"roots","pure_median_ms":8.815,"c_median_ms":6.870,"ratio":1.283,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"knapsack","pure_median_ms":10.500,"c_median_ms":6.228,"ratio":1.686,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"join","pure_median_ms":114.853,"c_median_ms":31.476,"ratio":3.649,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"matmul","pure_median_ms":36.167,"c_median_ms":7.582,"ratio":4.770,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"bfs","pure_median_ms":24.524,"c_median_ms":10.185,"ratio":2.408,"agree":true}
+{"version":"b10","way":"batches","turn":2,"work":"roots","pure_median_ms":8.818,"c_median_ms":6.863,"ratio":1.285,"agree":true}
+{"version":"b10","way":"batches","turn":2,"work":"knapsack","pure_median_ms":10.357,"c_median_ms":6.226,"ratio":1.663,"agree":true}
+{"version":"b10","way":"batches","turn":2,"work":"join","pure_median_ms":117.499,"c_median_ms":32.088,"ratio":3.662,"agree":true}
+{"version":"b10","way":"batches","turn":2,"work":"matmul","pure_median_ms":38.138,"c_median_ms":7.527,"ratio":5.067,"agree":true}
+{"version":"b10","way":"batches","turn":2,"work":"bfs","pure_median_ms":24.247,"c_median_ms":10.089,"ratio":2.403,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"roots","pure_median_ms":8.827,"c_median_ms":6.865,"ratio":1.286,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"knapsack","pure_median_ms":10.547,"c_median_ms":6.232,"ratio":1.693,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"join","pure_median_ms":115.034,"c_median_ms":31.988,"ratio":3.596,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"matmul","pure_median_ms":36.076,"c_median_ms":7.580,"ratio":4.760,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"bfs","pure_median_ms":23.961,"c_median_ms":10.116,"ratio":2.369,"agree":true}
+{"version":"b10","way":"batches","turn":3,"work":"roots","pure_median_ms":8.815,"c_median_ms":6.861,"ratio":1.285,"agree":true}
+{"version":"b10","way":"batches","turn":3,"work":"knapsack","pure_median_ms":10.354,"c_median_ms":6.229,"ratio":1.662,"agree":true}
+{"version":"b10","way":"batches","turn":3,"work":"join","pure_median_ms":119.922,"c_median_ms":31.283,"ratio":3.834,"agree":true}
+{"version":"b10","way":"batches","turn":3,"work":"matmul","pure_median_ms":38.389,"c_median_ms":7.542,"ratio":5.090,"agree":true}
+{"version":"b10","way":"batches","turn":3,"work":"bfs","pure_median_ms":24.170,"c_median_ms":9.991,"ratio":2.419,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"roots","pure_median_ms":8.846,"c_median_ms":6.865,"ratio":1.289,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"knapsack","pure_median_ms":10.483,"c_median_ms":6.222,"ratio":1.685,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"join","pure_median_ms":116.393,"c_median_ms":33.627,"ratio":3.461,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"matmul","pure_median_ms":36.096,"c_median_ms":7.536,"ratio":4.790,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"bfs","pure_median_ms":24.333,"c_median_ms":10.073,"ratio":2.416,"agree":true}
+{"version":"b10","way":"batches","turn":4,"work":"roots","pure_median_ms":8.811,"c_median_ms":6.864,"ratio":1.284,"agree":true}
+{"version":"b10","way":"batches","turn":4,"work":"knapsack","pure_median_ms":10.321,"c_median_ms":6.228,"ratio":1.657,"agree":true}
+{"version":"b10","way":"batches","turn":4,"work":"join","pure_median_ms":117.535,"c_median_ms":32.601,"ratio":3.605,"agree":true}
+{"version":"b10","way":"batches","turn":4,"work":"matmul","pure_median_ms":38.172,"c_median_ms":7.522,"ratio":5.075,"agree":true}
+{"version":"b10","way":"batches","turn":4,"work":"bfs","pure_median_ms":24.055,"c_median_ms":10.087,"ratio":2.385,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"roots","pure_median_ms":8.818,"c_median_ms":6.860,"ratio":1.285,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"knapsack","pure_median_ms":10.517,"c_median_ms":6.229,"ratio":1.688,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"join","pure_median_ms":116.684,"c_median_ms":31.468,"ratio":3.708,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"matmul","pure_median_ms":36.241,"c_median_ms":7.531,"ratio":4.812,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"bfs","pure_median_ms":24.066,"c_median_ms":9.976,"ratio":2.412,"agree":true}
+{"version":"b10","way":"batches","turn":5,"work":"roots","pure_median_ms":8.808,"c_median_ms":6.865,"ratio":1.283,"agree":true}
+{"version":"b10","way":"batches","turn":5,"work":"knapsack","pure_median_ms":10.365,"c_median_ms":6.229,"ratio":1.664,"agree":true}
+{"version":"b10","way":"batches","turn":5,"work":"join","pure_median_ms":117.192,"c_median_ms":29.216,"ratio":4.011,"agree":true}
+{"version":"b10","way":"batches","turn":5,"work":"matmul","pure_median_ms":38.254,"c_median_ms":7.555,"ratio":5.063,"agree":true}
+{"version":"b10","way":"batches","turn":5,"work":"bfs","pure_median_ms":24.286,"c_median_ms":9.995,"ratio":2.430,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"roots","pure_median_ms":8.824,"c_median_ms":6.865,"ratio":1.285,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"knapsack","pure_median_ms":10.499,"c_median_ms":6.226,"ratio":1.686,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"join","pure_median_ms":114.614,"c_median_ms":30.930,"ratio":3.706,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"matmul","pure_median_ms":36.021,"c_median_ms":7.558,"ratio":4.766,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"bfs","pure_median_ms":24.170,"c_median_ms":10.017,"ratio":2.413,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"roots","pure_median_ms":8.824,"c_median_ms":6.861,"ratio":1.286,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"knapsack","pure_median_ms":10.487,"c_median_ms":6.227,"ratio":1.684,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"join","pure_median_ms":116.628,"c_median_ms":32.802,"ratio":3.556,"agree":true}
+{"version":"b11","way":"batches","turn":1,"work":"bfs","pure_median_ms":24.080,"c_median_ms":9.958,"ratio":2.418,"agree":true}
+{"version":"b11-256","way":"batches","turn":1,"work":"roots","pure_median_ms":10.443,"c_median_ms":6.863,"ratio":1.522,"agree":true}
+{"version":"b11-256","way":"batches","turn":1,"work":"knapsack","pure_median_ms":19.649,"c_median_ms":6.232,"ratio":3.153,"agree":true}
+{"version":"b11-256","way":"batches","turn":1,"work":"join","pure_median_ms":120.220,"c_median_ms":32.290,"ratio":3.723,"agree":true}
+{"version":"b11-256","way":"batches","turn":1,"work":"bfs","pure_median_ms":25.666,"c_median_ms":9.996,"ratio":2.568,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"roots","pure_median_ms":8.827,"c_median_ms":6.860,"ratio":1.287,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"knapsack","pure_median_ms":10.632,"c_median_ms":6.323,"ratio":1.682,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"join","pure_median_ms":118.936,"c_median_ms":33.003,"ratio":3.604,"agree":true}
+{"version":"b11","way":"batches","turn":2,"work":"bfs","pure_median_ms":27.830,"c_median_ms":12.270,"ratio":2.268,"agree":true}
+{"version":"b11-256","way":"batches","turn":2,"work":"roots","pure_median_ms":10.466,"c_median_ms":6.872,"ratio":1.523,"agree":true}
+{"version":"b11-256","way":"batches","turn":2,"work":"knapsack","pure_median_ms":19.664,"c_median_ms":6.249,"ratio":3.147,"agree":true}
+{"version":"b11-256","way":"batches","turn":2,"work":"join","pure_median_ms":118.582,"c_median_ms":33.762,"ratio":3.512,"agree":true}
+{"version":"b11-256","way":"batches","turn":2,"work":"bfs","pure_median_ms":25.601,"c_median_ms":10.025,"ratio":2.554,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"roots","pure_median_ms":8.827,"c_median_ms":6.865,"ratio":1.286,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"knapsack","pure_median_ms":10.507,"c_median_ms":6.308,"ratio":1.666,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"join","pure_median_ms":116.772,"c_median_ms":33.051,"ratio":3.533,"agree":true}
+{"version":"b11","way":"batches","turn":3,"work":"bfs","pure_median_ms":24.173,"c_median_ms":10.328,"ratio":2.340,"agree":true}
+{"version":"b11-256","way":"batches","turn":3,"work":"roots","pure_median_ms":10.516,"c_median_ms":6.863,"ratio":1.532,"agree":true}
+{"version":"b11-256","way":"batches","turn":3,"work":"knapsack","pure_median_ms":19.639,"c_median_ms":6.235,"ratio":3.150,"agree":true}
+{"version":"b11-256","way":"batches","turn":3,"work":"join","pure_median_ms":119.911,"c_median_ms":30.332,"ratio":3.953,"agree":true}
+{"version":"b11-256","way":"batches","turn":3,"work":"bfs","pure_median_ms":25.588,"c_median_ms":10.012,"ratio":2.556,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"roots","pure_median_ms":8.816,"c_median_ms":6.866,"ratio":1.284,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"knapsack","pure_median_ms":10.517,"c_median_ms":6.232,"ratio":1.688,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"join","pure_median_ms":114.556,"c_median_ms":31.417,"ratio":3.646,"agree":true}
+{"version":"b11","way":"batches","turn":4,"work":"bfs","pure_median_ms":24.061,"c_median_ms":10.088,"ratio":2.385,"agree":true}
+{"version":"b11-256","way":"batches","turn":4,"work":"roots","pure_median_ms":10.547,"c_median_ms":6.875,"ratio":1.534,"agree":true}
+{"version":"b11-256","way":"batches","turn":4,"work":"knapsack","pure_median_ms":19.628,"c_median_ms":6.237,"ratio":3.147,"agree":true}
+{"version":"b11-256","way":"batches","turn":4,"work":"join","pure_median_ms":122.063,"c_median_ms":30.966,"ratio":3.942,"agree":true}
+{"version":"b11-256","way":"batches","turn":4,"work":"bfs","pure_median_ms":25.620,"c_median_ms":9.975,"ratio":2.568,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"roots","pure_median_ms":8.814,"c_median_ms":6.864,"ratio":1.284,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"knapsack","pure_median_ms":10.527,"c_median_ms":6.235,"ratio":1.688,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"join","pure_median_ms":115.441,"c_median_ms":29.829,"ratio":3.870,"agree":true}
+{"version":"b11","way":"batches","turn":5,"work":"bfs","pure_median_ms":24.336,"c_median_ms":10.132,"ratio":2.402,"agree":true}
+{"version":"b11-256","way":"batches","turn":5,"work":"roots","pure_median_ms":10.442,"c_median_ms":6.865,"ratio":1.521,"agree":true}
+{"version":"b11-256","way":"batches","turn":5,"work":"knapsack","pure_median_ms":19.625,"c_median_ms":6.227,"ratio":3.151,"agree":true}
+{"version":"b11-256","way":"batches","turn":5,"work":"join","pure_median_ms":117.410,"c_median_ms":32.998,"ratio":3.558,"agree":true}
+{"version":"b11-256","way":"batches","turn":5,"work":"bfs","pure_median_ms":25.714,"c_median_ms":9.954,"ratio":2.583,"agree":true}