commit dd0e35d724614df8ab9635de91e1c46f9f72aff9 parent cf4585fe2597879d18541426fd6e3d5bca746942 Author: Jared Tobin <jared@jtobin.io> Date: Sat, 16 May 2026 11:57:30 -0230 readme: perf update Diffstat:
| M | README.md | | | 29 | +++++++++++++---------------- |
1 file changed, 13 insertions(+), 16 deletions(-)
diff --git a/README.md b/README.md @@ -31,26 +31,23 @@ Haddocks (API documentation, etc.) are hosted at ## Performance -The aim is best-in-class performance for pure, highly-auditable Haskell -code. We could go slightly faster by using direct allocation and writes, -but we get pretty close to the best impure versions with only builders. +The aim is best-in-class performance. -Current benchmark figures on 1kb inputs on a relatively-beefy NixOS VPS look -like (use `cabal bench` to run the benchmark suite): +Current benchmark figures on 1kb inputs on my M4 MacBook Air look like +(use `cabal bench -fllvm` to run the benchmark suite): ``` - benchmarking ppad-base16 - time 2.997 μs (2.988 μs .. 3.009 μs) + benchmarking ppad-base16/encode + time 295.9 ns (295.4 ns .. 296.4 ns) + 1.000 R² (0.999 R² .. 1.000 R²) + mean 296.8 ns (296.4 ns .. 297.2 ns) + std dev 1.367 ns (1.181 ns .. 1.619 ns) + + benchmarking ppad-base16/decode + time 270.8 ns (270.6 ns .. 271.1 ns) 1.000 R² (1.000 R² .. 1.000 R²) - mean 3.024 μs (3.012 μs .. 3.035 μs) - std dev 39.14 ns (36.12 ns .. 42.87 ns) - variance introduced by outliers: 11% (moderately inflated) - - benchmarking ppad-base16 - time 599.0 ns (597.3 ns .. 601.3 ns) - 1.000 R² (1.000 R² .. 1.000 R²) - mean 597.3 ns (596.3 ns .. 598.5 ns) - std dev 3.493 ns (2.623 ns .. 5.111 ns) + mean 270.9 ns (270.7 ns .. 271.1 ns) + std dev 627.8 ps (515.3 ps .. 766.1 ps) ``` ## Security