Add JMH benchmark comparing various sort algorithms specifically for sorting ScoreDoc[] of varying lengths#15950

Open

mikemccand wants to merge 28 commits intoapache:mainfrom

mikemccand:lucene_sort_benchy

Commits on Apr 10, 2026

Claude Code (Opus 4.6) helped me write a simple JMH benchy to test the many options for sorting a Lucene ScoreDoc[] by docid, helpful often for post-processing hits from a search. For example, if …
mikemccand
committed
Add javadoc cross-reference to jmh-table.py + tidy formatting

claude
and
mikemccand
committed
Fix config banner: use Unicode multiply char, one setting per row

claude
and
mikemccand
committed
Make table cells deep-linkable via URL hash fragment

claude
and
mikemccand
committed
Expand ScoreDocSortBenchmark javadoc with build/run/visualize instructions

claude
and
mikemccand
committed
Show JVM details and split harness vs fork JVM args in config banner

claude
and
mikemccand
committed
Preserve sort column and direction in URL hash

claude
and
mikemccand
committed
Require source file argument in jmh-table.py

claude
and
mikemccand
committed
Add sizes 50/500, adaptive int[]/long[] primitive extract sort

claude
and
mikemccand
committed
human made a direct change to increase jvm fork count from 8 to 10
mikemccand
committed
Add run-benchmark.sh to compile-then-run JMH in one step

claude
and
mikemccand
committed
Update javadoc to document run-benchmark.sh as primary way to run

claude
and
mikemccand
committed
Make ScoreDocSortBenchmark thread-safe for concurrent JMH execution

claude
and
mikemccand
committed
Use Arrays.copyOf for shallow work array copy

claude
and
mikemccand
committed
use 8 threads, turn on detailed json reporting for jmh sort benchy
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed
radixSort2Pass: switch from 16-bit to 8-bit radix; log-scale heatmap; --skip flag for jmh-table.py

claude
and
mikemccand
committed
default distribution @Param to random only

claude
and
mikemccand
committed
back tgo single thread
mikemccand
committed
fix sign-bit overflow in jdkSortPrimitiveExtractAdaptive int packing

claude
and
mikemccand
committed
make verify() non-fatal so one broken algo doesn't block all benchmarks

claude
and
mikemccand
committed
Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode
and
mikemccand
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add JMH benchmark comparing various sort algorithms specifically for sorting ScoreDoc[] of varying lengths#15950

Add JMH benchmark comparing various sort algorithms specifically for sorting ScoreDoc[] of varying lengths#15950
mikemccand wants to merge 28 commits intoapache:mainfrom
mikemccand:lucene_sort_benchy

Commits on Apr 10, 2026

Claude Code (Opus 4.6) helped me write a simple JMH benchy to test the many options for sorting a Lucene ScoreDoc[] by docid, helpful often for post-processing hits from a search. For example, if …

Add javadoc cross-reference to jmh-table.py + tidy formatting

Fix config banner: use Unicode multiply char, one setting per row

Make table cells deep-linkable via URL hash fragment

Expand ScoreDocSortBenchmark javadoc with build/run/visualize instructions

Show JVM details and split harness vs fork JVM args in config banner

Preserve sort column and direction in URL hash

Require source file argument in jmh-table.py

Add sizes 50/500, adaptive int[]/long[] primitive extract sort

human made a direct change to increase jvm fork count from 8 to 10

Add run-benchmark.sh to compile-then-run JMH in one step

Update javadoc to document run-benchmark.sh as primary way to run

Make ScoreDocSortBenchmark thread-safe for concurrent JMH execution

Use Arrays.copyOf for shallow work array copy

use 8 threads, turn on detailed json reporting for jmh sort benchy

Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode <noreply@google.com>

Gemini AI Thinking Mode <noreply@google.com>

radixSort2Pass: switch from 16-bit to 8-bit radix; log-scale heatmap; --skip flag for jmh-table.py

default distribution @Param to random only

back tgo single thread

fix sign-bit overflow in jdkSortPrimitiveExtractAdaptive int packing

make verify() non-fatal so one broken algo doesn't block all benchmarks

Gemini AI Thinking Mode <noreply@google.com>