warefeats
Source

Benchmarks for developer tools, with the runs attached.

I pin the versions, hold the corpus still, warm the cache, and fork a fresh process for every pass. You get the number, the machine it came from, and every sample behind it.

Published

Newest first inside each category. The ratio is the mean of the fastest tool against the slower one, with its propagated σ.

Desktop

  1. Sep 10, 2026Tauri vs ElectronTauri moved JSON across the boundary 1.44 times faster than Electron on the M2 Max and 1.81 times slower on the Windows box: 'Tauri' is two different shells, and the webview decides which one you getApple M2 Max · 4 runs

Tile servers

  1. Sep 8, 2026Martin vs go-pmtiles vs tileserver-gl vs mbtileserver vs BBOXMartin served 12,600 tiles a second at 100 clients from the PMTiles archive, more than any other server from either format; mbtileserver was fastest on a single tile by a tenth of a millisecond, go-pmtiles reached a third of Martin's throughput, and tileserver-gl paid 12 ms per tile for re-compressing every oneApple M2 Max · 20 runs
  2. Sep 2, 2026Martin vs Tegola vs BBOX vs pg_tileserv vs TiPg vs ldproxyMartin led single-tile latency, throughput and cold start; the five servers that encode in PostGIS finished within 2 ms of each other on one tile, and ldproxy's opt-in PGIS_TILES mode cut its gap from 9x to 1.4xApple M2 Max · 20 runs

Web maps

  1. Sep 2, 2026MapLibre GL JS vs Mapbox GL JSMapLibre GL JS 5 painted a warm frame in 3.1 ms to Mapbox GL JS 1.13's 6.5 ms and held 919 MB to its 1,448 MB after the same camera path; 1.13 drew its first complete frame in 193 ms to MapLibre's 335, starting six workers to MapLibre's oneAMD Ryzen 7 6800H with Radeon Graphics · 15 runs

Caches

  1. Aug 31, 2026Redis vs Valkey vs DragonflyDragonfly led every section; Redis and Valkey tracked within a few percent of each otherAWS Graviton3 · 10 runs

Bundlers

  1. Aug 30, 2026Vite vs esbuild vs tsupesbuild ran 10.68 ± 0.57 times faster than ViteApple M2 Max · 20 runs

Caching proxies

  1. Aug 29, 2026HTTP caching proxies for HLS deliveryNo single engine dominated all four workloadsApple M2 Max · 20 runs

Code quality

  1. Aug 26, 2026ESLint vs BiomeBiome ran 1.41 ± 0.05 times faster than ESLintApple M2 Max · 20 runs

Queue

What runs next, in no promised order. Each entry becomes a page when its runner is reproducible.

  1. Published:Caching proxiesVarnish vs Vinyl vs NGINX (HLS delivery)TTFB, coalescing, grace, throughput across topologies
  2. Published:CachesRedis vs Valkey vs DragonflyThroughput, tail latency, memory pressure
  3. Published:BundlersVite vs esbuild vs tsupCold builds, output weight
  4. Published:DesktopTauri vs ElectronIPC across the boundary, webview parity, startup and soak memory, size
  5. Published:Web mapsMapLibre vs MapboxFrame time, memory, visual parity
  6. Published:Tile serversMartin vs Tegola vs BBOX vs pg_tileserv vs TiPg vs ldproxySingle-tile latency, throughput at 10 and 100 clients, cold start, memory
  7. Published:Tile serversMartin vs go-pmtiles vs tileserver-gl vs mbtileserver vs BBOXServing one pre-built archive as MBTiles and PMTiles: single-tile latency, throughput, random walk, cold start, memory