Mobiconf 2018: Fernando Cejas, IBM
summary
Processes two 30 MB Wikipedia XML files (about 200k words) and measures real timings: single-thread run takes ~22 seconds, two threads drop to ~13–14 seconds, RxJava and Kotlin coroutines offer modest gains but reveal hidden costs. Diagnosis finds ~4s per-file parsing I/O, garbage‑collection churn from many short‑lived objects, and contention depending on concurrent hash map vs synchronized map. Final approach splits work finer, reuses threads and reduces allocations to get ~30% improvement — measure, then optimize.
AI-written summary · about
chapters (7)
- 0:00Start
- 6:17A Difference between Parallel and Concurrent
- 10:31Race Condition
- 12:56Why Multi-Threading
- 18:09Threads and Locks
- 36:35Reference Counting
- 41:21Always Use the Right Tool for the Right Job
related talks
50mWhat Mom Never Told You About Multi-threading
47mConquering a Billion Rows with Modern Java Concurrency
08h 39mLivestream | Day 2, theater 2
45mEverything Everywhere All at Once
48mConcurrency: It’s Harder (and Easier) Than You Think
47mMore Action, More Overview – From 4k to 1M Threads
35mHow we brought down our build times. Modularization impact on build speeds and metrics
30mHow we brought down our build times. General optimization tips