<h1 id="jeioptimizer">JEIOptimizer</h1> <p><strong>Builds JEI's ingredient filter index in parallel &mdash; cuts server join time by up to 17 seconds on large modpacks.</strong></p> <hr> <h2 id="the-story">The story</h2> <p>I was playing on a private AllTheMods server with friends. Every disconnect, every restart, every "let me grab a snack" meant 40 seconds of staring at <em>"Joining world&hellip;"</em> while everyone else was already mining.</p> <p>Got tired of it. Spent a weekend with JEI's source code open instead of playing the game. This mod is what came out.</p> <hr> <h2 id="what-it-does">What it does</h2> <p>When you connect to a modded server, JEI rebuilds its entire item search index from scratch &mdash; because the server may ship custom recipes. On a 250-mod pack with ~50,000 items, this takes <strong>12+ seconds on the main thread</strong>. Your client just sits frozen.</p> <p>JEIOptimizer parallelizes JEI's <code>IngredientFilter</code> build across worker threads. Each search prefix (item name, mod ID, tag, tooltip, &hellip;) gets its own thread, running on independent data structures &mdash; no contention, no races, no risks.</p> <hr> <h2 id="real-world-measurements">Real-world measurements</h2> <p>Tested on a 250+ mod pack (AE2, Mekanism, Create, Cobblemon, Apotheosis, AllTheCompressed, &hellip;). Ryzen 5600X &middot; 10 worker threads &middot; NVMe &middot; 8 GB RAM allocated &middot; remote server (Hetzner DC).</p> <table> <thead> <tr> <th id="phase">Phase</th> <th id="vanilla_jei">Vanilla JEI</th> <th id="+_jeioptimizer">+ JEIOptimizer</th> <th id="reduction">Reduction</th> </tr> </thead> <tbody> <tr> <td>Building ingredient filter</td> <td>12.00 s</td> <td><strong>0.79 s</strong></td> <td><strong>&minus;94% (15&times; faster)</strong></td> </tr> <tr> <td>Starting JEI total</td> <td>25.84 s</td> <td>14.17 s</td> <td>&minus;45%</td> </tr> <tr> <td><strong>Total connect time</strong></td> <td><strong>~40 s</strong></td> <td><strong>~23 s</strong></td> <td><strong>&minus;42% (&minus;17 s)</strong></td> </tr> </tbody> </table> <p>Combined with the bonus tweak below (see <em>Free 8 more seconds</em>), you can drop server connect from <strong>40 s &rarr; ~23 s</strong> consistently.</p> <hr> <h2 id="how-it-works-technical">How it works (technical)</h2> <p>JEI's <code>IngredientFilter</code> builds a search index where each "prefix" (name, modId, tag, tooltip, &hellip;) has its own <code>GeneralizedSuffixTree</code>. Independent structures &rarr; safe to build in parallel.</p> <p><strong>Vanilla JEI:</strong></p> <pre><code>for each ingredient (50,000): for each prefix (7-10): compute tokens &rarr; insert into suffix tree</code></pre> <p>One thread = ~12 seconds.</p> <p><strong>JEIOptimizer:</strong></p> <pre><code>parallelStream over prefixes: for each ingredient: compute tokens &rarr; insert into own suffix tree</code></pre> <p>6-10 cores doing 7-10 prefixes simultaneously = ~0.8 seconds.</p> <h3 id="filter-modes-in-config">Filter modes (in config)</h3> <ul> <li><code>OFF</code> &mdash; vanilla JEI behavior</li> <li><code>BATCH</code> &mdash; small cache-friendly improvement, single-threaded</li> <li><code>PARALLEL_PREFIX</code> &mdash; parallel per-prefix build (safe)</li> <li><code>PARALLEL_FULL</code> &mdash; also parallelizes tokenization within prefix (<strong>default, fastest</strong>)</li> </ul> <hr> <h2 id="free-8-more-seconds-manual-jei-tweak">Free 8 more seconds (manual JEI tweak)</h2> <p>In <code>config/jei/jei-client.ini</code>, change:</p> <pre><code class="ini language-ini">tooltipSearchMode = DISABLED</code></pre> <p>JEI indexes the tooltip text of every item on every connect. Some mods do heavy I/O in tooltip rendering &mdash; eating ~8 of those original 12 seconds for nothing. Players almost always search by name, not tooltip text.</p> <p>If players want tooltip search via <code>#prefix</code>, use <code>REQUIRE_PREFIX</code> instead.</p> <hr> <h2 id="configuration">Configuration</h2> <p><code>config/jeioptimizer-client.toml</code> (auto-generated):</p> <pre><code class="toml language-toml"> mode = "PARALLEL_FULL" # Default. Safe, fastest. async = false # Experimental: build in background log_timing = true # Log timings for verification worker_count = 0 # 0 = auto (cores - 2) parallel_phases = # Experimental &mdash; see below parallel_creative_tabs = false # Experimental &mdash; see below</code></pre> <p><strong>Safe defaults out of the box.</strong> Drop the jar in, you get the win. No config touching required.</p> <hr> <h2 id="compatibility">Compatibility</h2> <ul> <li><strong>Minecraft:</strong> 1.21.1</li> <li><strong>Mod loader:</strong> NeoForge 21.x</li> <li><strong>JEI:</strong> 19.x (1.21.1 builds)</li> <li><strong>Side:</strong> Client-only &mdash; server installation not needed</li> </ul> <hr> <h2 id="what-this-wont-speed-up">What this WON'T speed

for each ingredient (50,000):
 for each prefix (7-10):
 compute tokens → insert into suffix tree

parallelStream over prefixes:
 for each ingredient:
 compute tokens → insert into own suffix tree

tooltipSearchMode = DISABLED

 mode = "PARALLEL_FULL" # Default. Safe, fastest.
 async = false # Experimental: build in background
 log_timing = true # Log timings for verification
worker_count = 0 # 0 = auto (cores - 2)
parallel_phases = [] # Experimental — see below
 parallel_creative_tabs = false # Experimental — see below