Skip to content

README.md

Latest commit

 

History

History
73 lines (50 loc) · 2.03 KB

README.md

File metadata and controls

73 lines (50 loc) · 2.03 KB
Raw

🧪 Android Custom Image Loader Benchmark

A no-library, from-scratch low-tech image loader built to understand how libraries like Glide and Coil might work under the hood.

🎯 Why I Built This

Instead of using Glide blindly, I built my own image loader to explore:

  • What happens without downsampling?
  • Can I reuse bitmaps safely?
  • Does cancelling offscreen image loads help?
  • What do GC and memory stats reveal during scroll?

🛠️ Benchmark Setup

  • Loads shuffled high-res images from Picsum

  • Displays in a RecyclerView

  • Benchmark flags:

    data class BenchmarkConfig(
    val usePool: Boolean,
    val useDownsampling: Boolean,
    val useCancellation: Boolean
)

  • Measures scroll time, memory usage, GC events, and network stats

  • Automated scroll flow for consistent results

🧵 Image Loading Flow

  1. Cancel task if ViewHolder is recycled
  2. Decode via BitmapFactory
  3. Downsample based on ImageView size
  4. Try to reuse bitmap from custom pool
  5. Update UI only if still bound to correct image

📊 Key Findings

Feature Impact
❌ No Optimization High memory, GC storms, janky scroll
✅ Downsampling Lower memory, faster decode, fewer GC events
✅ Bitmap Pooling Reduces allocations, tricky to get right
✅ Task Cancellation Huge win during flings, saves CPU & memory
🏆 All Combined Smoothest scroll, minimal GC, lowest network use

🧠 Lessons Learned

  • Reusing bitmaps requires exact match + inMutable=true
  • Downsampling is a no-brainer
  • Cancellation improves UX and performance drastically

❤️ Final Thoughts

Don’t use this in production — use Glide or Coil.

But if you're curious, bored, or want to learn: build one. You'll walk away with a deep understanding of bitmaps, memory, and image loading internals.