--- title: "How close can pure FP get to C?" slug: how-close-can-pure-fp-get-to-c url: https://listedarticles.com/articles/how-close-can-pure-fp-get-to-c canonical_url: https://rtfeldman.com/fp-vs-c content_type: blog_post language: en published_at: 2026-10-22T00:00:00.000Z updated_at: 2026-10-23T14:18:57.143Z author: "Richard Feldman" authored_by: human publisher: "Richard Feldman" publisher_url: https://rtfeldman.com/ topics: ["programming", "functional-programming", "performance", "compilers"] license: all-rights-reserved word_count: 9413 reading_minutes: 41 citation: "Richard Feldman, Richard Feldman. \"How close can pure FP get to C?.\" 22 Oct 2026. https://rtfeldman.com/fp-vs-c (all-rights-reserved)" # The full text follows. The web page shows an extract and sends readers # to the source above; quote the citation and link the canonical URL. --- # How close can pure FP get to C? > Richard Feldman takes up the lean-zip versus libdeflate result and explores how far a DEFLATE implementation in Roc, a pure functional language, can be pushed toward C performance, with benchmarks, compiler work, and reproduction scripts. # [How close can pure FP get to C?](#how-close-can-pure-fp-get-to-c) *October 22, 2026* In [Why Lean is Faster Than Rust](https://kim-em.github.io/blog/2026-7-24-why-lean-is-faster-than-rust/), the author shares an impressive result: a [DEFLATE compression](https://en.wikipedia.org/wiki/Deflate) library written in [Lean 4](https://github.com/leanprover/lean4), a functional programming language and proof assistant, outperforming an equivalent Rust package that [crates.io says](https://crates.io/search?q=miniz_oxide) has been downloaded *over a billion times.* From the article: > `lean-zip` beats every single compression level `miniz_oxide` provides, i.e. we're capable of providing the same compression at higher throughput. At its fastest setting we're 15% quicker and the output is 8.8% smaller. However, the article also notes that while they managed to surpass Rust's `miniz_oxide`, they were no match for the end boss of DEFLATE performance: [libdeflate](https://github.com/ebiggers/libdeflate), written in C. > `lean-zip` is certainly not the best here: `libdeflate` unsurprisingly blows it out of the water (unsurprisingly because this is a very carefully tuned implementation using architecture-specific SIMD, that we can't touch in Lean). Enter [Roc](https://roc-lang.org), a functional programming language (but not a theorem prover) with an explicit focus on runtime performance. Roc's tagline is "Fast, Friendly, Functional," and [Fast comes first](https://roc-lang.org/fast) because both running fast *and* compiling fast have always been central goals. Roc is not a mature language (it's still weeks away from its first numbered release, 0.1.0), and its nascent package ecosystem has had almost no optimization work done on it. Still, between its language design, [longstanding optimizations](https://youtu.be/vzfy4EKwG_Y?si=oYa6k2pFdsRJcQW7), and [brand-new compiler implementation](https://rtfeldman.com/rust-to-zig), I figured Roc already had enough of a foundation in place to take on the champ: libdeflate. So I spent a bunch of time grinding through performance optimizations on [roc-deflate](https://github.com/niclas-ahden/roc-deflate), a package [Niclas Ahden](https://www.youtube.com/@NiclasAhden) wrote (and [uses in paid production work](https://youtu.be/12yVcgQHAK0?si=JETe6lec5y4jCLVL)) for DEFLATE compression. I didn't expect to beat libdeflate's performance across the board, but I wanted to see how close it could get on the non-streaming APIs (as roc-deflate hasn't implemented streaming yet), using the classic Silesia corpus for benchmarking and verifying that it produced identical outputs. ## [Under 10% Slower](#under-10-slower) Here's where the benchmarks stand today, running on the 2021 M1 MacBook Pro where I did all the optimization work, as well as on an AMD Ryzen 9 3900X Ubuntu machine for an x64 comparison. roc-deflate ran these 144 benchmarks slower on average than libdeflate, in the median case **8.9%** slower on the Ryzen 3900X, and **2.6%** on the M1 MacBook Pro. ### [Compression on Silesia corpus](#chart) M1 MacBook Pro Ryzen 3900X Ubuntu MB/s 700 600 500 400 300 200 100 0 file that ran fastest file that ran slowest L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 libdeflate roc-deflate libdeflate at L1 compression 230.57 MB/s compressing dickens (median across 15 runs) libdeflate at L1 compression 248.31 MB/s compressing mozilla (median across 15 runs) libdeflate at L1 compression 312.26 MB/s compressing mr (median across 15 runs) libdeflate at L1 compression 601.03 MB/s compressing nci (median across 15 runs) libdeflate at L1 compression 169.98 MB/s compressing ooffice (median across 15 runs) libdeflate at L1 compression 284.01 MB/s compressing osdb (median across 15 runs) libdeflate at L1 compression 281.14 MB/s compressing reymont (median across 15 runs) libdeflate at L1 compression 344.00 MB/s compressing samba (median across 15 runs) libdeflate at L1 compression 160.21 MB/s compressing sao (median across 15 runs) libdeflate at L1 compression 268.58 MB/s compressing webster (median across 15 runs) libdeflate at L1 compression 145.93 MB/s compressing x-ray (median across 15 runs) libdeflate at L1 compression 476.83 MB/s compressing xml (median across 15 runs) roc-deflate at L1 compression 202.92 MB/s compressing dickens (median across 15 runs) roc-deflate at L1 compression 225.79 MB/s compressing mozilla (median across 15 runs) roc-deflate at L1 compression 277.57 MB/s compressing mr (median across 15 runs) roc-deflate at L1 compression 529.49 MB/s compressing nci (median across 15 runs) roc-deflate at L1 compression 153.99 MB/s compressing ooffice (median across 15 runs) roc-deflate at L1 compression 257.14 MB/s compressing osdb (median across 15 runs) roc-deflate at L1 compression 245.80 MB/s compressing reymont (median across 15 runs) roc-deflate at L1 compression 309.79 MB/s compressing samba (median across 15 runs) roc-deflate at L1 compression 145.28 MB/s compressing sao (median across 15 runs) roc-deflate at L1 compression 239.18 MB/s compressing webster (median across 15 runs) roc-deflate at L1 compression 134.66 MB/s compressing x-ray (median across 15 runs) roc-deflate at L1 compression 419.07 MB/s compressing xml (median across 15 runs) libdeflate at L2 compression 161.68 MB/s compressing dickens (median across 15 runs) libdeflate at L2 compression 184.55 MB/s compressing mozilla (median across 15 runs) libdeflate at L2 compression 204.86 MB/s compressing mr (median across 15 runs) libdeflate at L2 compression 453.14 MB/s compressing nci (median across 15 runs) libdeflate at L2 compression 128.05 MB/s compressing ooffice (median across 15 runs) libdeflate at L2 compression 203.10 MB/s compressing osdb (median across 15 runs) libdeflate at L2 compression 207.66 MB/s compressing reymont (median across 15 runs) libdeflate at L2 compression 249.57 MB/s compressing samba (median across 15 runs) libdeflate at L2 compression 112.78 MB/s compressing sao (median across 15 runs) libdeflate at L2 compression 198.86 MB/s compressing webster (median across 15 runs) libdeflate at L2 compression 117.18 MB/s compressing x-ray (median across 15 runs) libdeflate at L2 compression 352.59 MB/s compressing xml (median across 15 runs) roc-deflate at L2 compression 143.60 MB/s compressing dickens (median across 15 runs) roc-deflate at L2 compression 168.37 MB/s compressing mozilla (median across 15 runs) roc-deflate at L2 compression 182.07 MB/s compressing mr (median across 15 runs) roc-deflate at L2 compression 409.82 MB/s compressing nci (median across 15 runs) roc-deflate at L2 compression 118.08 MB/s compressing ooffice (median across 15 runs) roc-deflate at L2 compression 185.41 MB/s compressing osdb (median across 15 runs) roc-deflate at L2 compression 183.69 MB/s compressing reymont (median across 15 runs) roc-deflate at L2 compression 228.02 MB/s compressing samba (median across 15 runs) roc-deflate at L2 compression 105.68 MB/s compressing sao (median across 15 runs) roc-deflate at L2 compression 181.07 MB/s compressing webster (median across 15 runs) roc-deflate at L2 compression 107.00 MB/s compressing x-ray (median across 15 runs) roc-deflate at L2 compression 323.17 MB/s compressing xml (median across 15 runs) libdeflate at L3 compression 130.85 MB/s compressing dickens (median across 15 runs) libdeflate at L3 compression 169.26 MB/s compressing mozilla (median across 15 runs) libdeflate at L3 compression 171.11 MB/s compressing mr (median across 15 runs) libdeflate at L3 compression 415.29 MB/s compressing nci (median across 15 runs) libdeflate at L3 compression 118.92 MB/s compressing ooffice (median across 15 runs) libdeflate at L3 compression 183.10 MB/s compressing osdb (median across 15 runs) libdeflate at L3 compression 169.42 MB/s compressing reymont (median across 15 runs) libdeflate at L3 compression 223.27 MB/s compressing samba (median across 15 runs) libdeflate at L3 compression 100.60 MB/s compressing sao (median across 15 runs) libdeflate at L3 compression 173.36 MB/s compressing webster (median across 15 runs) libdeflate at L3 compression 116.62 MB/s compressing x-ray (median across 15 runs) libdeflate at L3 compression 324.65 MB/s compressing xml (median across 15 runs) roc-deflate at L3 compression 122.04 MB/s compressing dickens (median across 15 runs) roc-deflate at L3 compression 158.29 MB/s compressing mozilla (median across 15 runs) roc-deflate at L3 compression 159.07 MB/s compressing mr (median across 15 runs) roc-deflate at L3 compression 382.21 MB/s compressing nci (median across 15 runs) roc-deflate at L3 compression 111.14 MB/s compressing ooffice (median across 15 runs) roc-deflate at L3 compression 171.27 MB/s compressing osdb (median across 15 runs) roc-deflate at L3 compression 157.19 MB/s compressing reymont (median across 15 runs) roc-deflate at L3 compression 209.86 MB/s compressing samba (median across 15 runs) roc-deflate at L3 compression 95.51 MB/s compressing sao (median across 15 runs) roc-deflate at L3 compression 162.75 MB/s compressing webster (median across 15 runs) roc-deflate at L3 compression 105.93 MB/s compressing x-ray (median across 15 runs) roc-deflate at L3 compression 300.31 MB/s compressing xml (median across 15 runs) libdeflate at L4 compression 119.94 MB/s compressing dickens (median across 15 runs) libdeflate at L4 compression 158.07 MB/s compressing mozilla (median across 15 runs) libdeflate at L4 compression 156.88 MB/s compressing mr (median across 15 runs) libdeflate at L4 compression 376.70 MB/s compressing nci (median across 15 runs) libdeflate at L4 compression 113.68 MB/s compressing ooffice (median across 15 runs) libdeflate at L4 compression 164.59 MB/s compressing osdb (median across 15 runs) libdeflate at L4 compression 152.16 MB/s compressing reymont (median across 15 runs) libdeflate at L4 compression 207.97 MB/s compressing samba (median across 15 runs) libdeflate at L4 compression 94.48 MB/s compressing sao (median across 15 runs) libdeflate at L4 compression 158.64 MB/s compressing webster (median across 15 runs) libdeflate at L4 compression 116.63 MB/s compressing x-ray (median across 15 runs) libdeflate at L4 compression 294.31 MB/s compressing xml (median across 15 runs) roc-deflate at L4 compression 113.71 MB/s compressing dickens (median across 15 runs) roc-deflate at L4 compression 148.77 MB/s compressing mozilla (median across 15 runs) roc-deflate at L4 compression 149.00 MB/s compressing mr (median across 15 runs) roc-deflate at L4 compression 356.47 MB/s compressing nci (median across 15 runs) roc-deflate at L4 compression 107.59 MB/s compressing ooffice (median across 15 runs) roc-deflate at L4 compression 156.41 MB/s compressing osdb (median across 15 runs) roc-deflate at L4 compression 142.45 MB/s compressing reymont (median across 15 runs) roc-deflate at L4 compression 197.17 MB/s compressing samba (median across 15 runs) roc-deflate at L4 compression 89.37 MB/s compressing sao (median across 15 runs) roc-deflate at L4 compression 151.15 MB/s compressing webster (median across 15 runs) roc-deflate at L4 compression 106.32 MB/s compressing x-ray (median across 15 runs) roc-deflate at L4 compression 278.53 MB/s compressing xml (median across 15 runs) libdeflate at L5 compression 95.14 MB/s compressing dickens (median across 15 runs) libdeflate at L5 compression 133.06 MB/s compressing mozilla (median across 15 runs) libdeflate at L5 compression 129.45 MB/s compressing mr (median across 15 runs) libdeflate at L5 compression 287.09 MB/s compressing nci (median across 15 runs) libdeflate at L5 compression 95.87 MB/s compressing ooffice (median across 15 runs) libdeflate at L5 compression 140.08 MB/s compressing osdb (median across 15 runs) libdeflate at L5 compression 120.63 MB/s compressing reymont (median across 15 runs) libdeflate at L5 compression 170.75 MB/s compressing samba (median across 15 runs) libdeflate at L5 compression 83.22 MB/s compressing sao (median across 15 runs) libdeflate at L5 compression 125.60 MB/s compressing webster (median across 15 runs) libdeflate at L5 compression 102.48 MB/s compressing x-ray (median across 15 runs) libdeflate at L5 compression 229.70 MB/s compressing xml (median across 15 runs) roc-deflate at L5 compression 90.27 MB/s compressing dickens (median across 15 runs) roc-deflate at L5 compression 125.87 MB/s compressing mozilla (median across 15 runs) roc-deflate at L5 compression 122.34 MB/s compressing mr (median across 15 runs) roc-deflate at L5 compression 311.35 MB/s compressing nci (median across 15 runs) roc-deflate at L5 compression 89.73 MB/s compressing ooffice (median across 15 runs) roc-deflate at L5 compression 133.37 MB/s compressing osdb (median across 15 runs) roc-deflate at L5 compression 115.52 MB/s compressing reymont (median across 15 runs) roc-deflate at L5 compression 167.01 MB/s compressing samba (median across 15 runs) roc-deflate at L5 compression 76.99 MB/s compressing sao (median across 15 runs) roc-deflate at L5 compression 123.12 MB/s compressing webster (median across 15 runs) roc-deflate at L5 compression 90.40 MB/s compressing x-ray (median across 15 runs) roc-deflate at L5 compression 237.29 MB/s compressing xml (median across 15 runs) libdeflate at L6 compression 70.60 MB/s compressing dickens (median across 15 runs) libdeflate at L6 compression 107.04 MB/s compressing mozilla (median across 15 runs) libdeflate at L6 compression 96.85 MB/s compressing mr (median across 15 runs) libdeflate at L6 compression 221.32 MB/s compressing nci (median across 15 runs) libdeflate at L6 compression 82.91 MB/s compressing ooffice (median across 15 runs) libdeflate at L6 compression 122.89 MB/s compressing osdb (median across 15 runs) libdeflate at L6 compression 79.91 MB/s compressing reymont (median across 15 runs) libdeflate at L6 compression 129.19 MB/s compressing samba (median across 15 runs) libdeflate at L6 compression 67.48 MB/s compressing sao (median across 15 runs) libdeflate at L6 compression 96.76 MB/s compressing webster (median across 15 runs) libdeflate at L6 compression 102.50 MB/s compressing x-ray (median across 15 runs) libdeflate at L6 compression 178.60 MB/s compressing xml (median across 15 runs) roc-deflate at L6 compression 72.83 MB/s compressing dickens (median across 15 runs) roc-deflate at L6 compression 107.54 MB/s compressing mozilla (median across 15 runs) roc-deflate at L6 compression 98.92 MB/s compressing mr (median across 15 runs) roc-deflate at L6 compression 246.59 MB/s compressing nci (median across 15 runs) roc-deflate at L6 compression 81.43 MB/s compressing ooffice (median across 15 runs) roc-deflate at L6 compression 122.02 MB/s compressing osdb (median across 15 runs) roc-deflate at L6 compression 84.85 MB/s compressing reymont (median across 15 runs) roc-deflate at L6 compression 134.11 MB/s compressing samba (median across 15 runs) roc-deflate at L6 compression 66.36 MB/s compressing sao (median across 15 runs) roc-deflate at L6 compression 101.16 MB/s compressing webster (median across 15 runs) roc-deflate at L6 compression 90.67 MB/s compressing x-ray (median across 15 runs) roc-deflate at L6 compression 194.35 MB/s compressing xml (median across 15 runs) libdeflate at L7 compression 54.23 MB/s compressing dickens (median across 15 runs) libdeflate at L7 compression 75.96 MB/s compressing mozilla (median across 15 runs) libdeflate at L7 compression 66.46 MB/s compressing mr (median across 15 runs) libdeflate at L7 compression 157.22 MB/s compressing nci (median across 15 runs) libdeflate at L7 compression 70.52 MB/s compressing ooffice (median across 15 runs) libdeflate at L7 compression 117.53 MB/s compressing osdb (median across 15 runs) libdeflate at L7 compression 46.77 MB/s compressing reymont (median across 15 runs) libdeflate at L7 compression 91.00 MB/s compressing samba (median across 15 runs) libdeflate at L7 compression 59.49 MB/s compressing sao (median across 15 runs) libdeflate at L7 compression 71.34 MB/s compressing webster (median across 15 runs) libdeflate at L7 compression 102.43 MB/s compressing x-ray (median across 15 runs) libdeflate at L7 compression 128.20 MB/s compressing xml (median across 15 runs) roc-deflate at L7 compression 56.93 MB/s compressing dickens (median across 15 runs) roc-deflate at L7 compression 79.18 MB/s compressing mozilla (median across 15 runs) roc-deflate at L7 compression 70.19 MB/s compressing mr (median across 15 runs) roc-deflate at L7 compression 174.91 MB/s compressing nci (median across 15 runs) roc-deflate at L7 compression 70.27 MB/s compressing ooffice (median across 15 runs) roc-deflate at L7 compression 115.95 MB/s compressing osdb (median across 15 runs) roc-deflate at L7 compression 51.20 MB/s compressing reymont (median across 15 runs) roc-deflate at L7 compression 96.46 MB/s compressing samba (median across 15 runs) roc-deflate at L7 compression 59.40 MB/s compressing sao (median across 15 runs) roc-deflate at L7 compression 75.99 MB/s compressing webster (median across 15 runs) roc-deflate at L7 compression 90.26 MB/s compressing x-ray (median across 15 runs) roc-deflate at L7 compression 141.60 MB/s compressing xml (median across 15 runs) libdeflate at L8 compression 37.50 MB/s compressing dickens (median across 15 runs) libdeflate at L8 compression 39.17 MB/s compressing mozilla (median across 15 runs) libdeflate at L8 compression 43.21 MB/s compressing mr (median across 15 runs) libdeflate at L8 compression 83.68 MB/s compressing nci (median across 15 runs) libdeflate at L8 compression 52.45 MB/s compressing ooffice (median across 15 runs) libdeflate at L8 compression 106.76 MB/s compressing osdb (median across 15 runs) libdeflate at L8 compression 24.87 MB/s compressing reymont (median across 15 runs) libdeflate at L8 compression 55.69 MB/s compressing samba (median across 15 runs) libdeflate at L8 compression 51.86 MB/s compressing sao (median across 15 runs) libdeflate at L8 compression 45.55 MB/s compressing webster (median across 15 runs) libdeflate at L8 compression 91.30 MB/s compressing x-ray (median across 15 runs) libdeflate at L8 compression 80.95 MB/s compressing xml (median across 15 runs) roc-deflate at L8 compression 39.56 MB/s compressing dickens (median across 15 runs) roc-deflate at L8 compression 43.00 MB/s compressing mozilla (median across 15 runs) roc-deflate at L8 compression 45.70 MB/s compressing mr (median across 15 runs) roc-deflate at L8 compression 91.28 MB/s compressing nci (median across 15 runs) roc-deflate at L8 compression 52.69 MB/s compressing ooffice (median across 15 runs) roc-deflate at L8 compression 101.66 MB/s compressing osdb (median across 15 runs) roc-deflate at L8 compression 27.81 MB/s compressing reymont (median across 15 runs) roc-deflate at L8 compression 60.04 MB/s compressing samba (median across 15 runs) roc-deflate at L8 compression 51.65 MB/s compressing sao (median across 15 runs) roc-deflate at L8 compression 48.78 MB/s compressing webster (median across 15 runs) roc-deflate at L8 compression 81.01 MB/s compressing x-ray (median across 15 runs) roc-deflate at L8 compression 86.27 MB/s compressing xml (median across 15 runs) libdeflate at L9 compression 34.82 MB/s compressing dickens (median across 15 runs) libdeflate at L9 compression 27.51 MB/s compressing mozilla (median across 15 runs) libdeflate at L9 compression 38.60 MB/s compressing mr (median across 15 runs) libdeflate at L9 compression 58.64 MB/s compressing nci (median across 15 runs) libdeflate at L9 compression 47.26 MB/s compressing ooffice (median across 15 runs) libdeflate at L9 compression 106.10 MB/s compressing osdb (median across 15 runs) libdeflate at L9 compression 20.65 MB/s compressing reymont (median across 15 runs) libdeflate at L9 compression 43.95 MB/s compressing samba (median across 15 runs) libdeflate at L9 compression 51.18 MB/s compressing sao (median across 15 runs) libdeflate at L9 compression 40.57 MB/s compressing webster (median across 15 runs) libdeflate at L9 compression 91.69 MB/s compressing x-ray (median across 15 runs) libdeflate at L9 compression 68.13 MB/s compressing xml (median across 15 runs) roc-deflate at L9 compression 36.73 MB/s compressing dickens (median across 15 runs) roc-deflate at L9 compression 30.48 MB/s compressing mozilla (median across 15 runs) roc-deflate at L9 compression 40.95 MB/s compressing mr (median across 15 runs) roc-deflate at L9 compression 64.32 MB/s compressing nci (median across 15 runs) roc-deflate at L9 compression 48.05 MB/s compressing ooffice (median across 15 runs) roc-deflate at L9 compression 101.46 MB/s compressing osdb (median across 15 runs) roc-deflate at L9 compression 22.85 MB/s compressing reymont (median across 15 runs) roc-deflate at L9 compression 47.64 MB/s compressing samba (median across 15 runs) roc-deflate at L9 compression 50.68 MB/s compressing sao (median across 15 runs) roc-deflate at L9 compression 43.36 MB/s compressing webster (median across 15 runs) roc-deflate at L9 compression 80.68 MB/s compressing x-ray (median across 15 runs) roc-deflate at L9 compression 73.29 MB/s compressing xml (median across 15 runs) libdeflate at L10 compression 12.07 MB/s compressing dickens (median across 15 runs) libdeflate at L10 compression 14.87 MB/s compressing mozilla (median across 15 runs) libdeflate at L10 compression 16.07 MB/s compressing mr (median across 15 runs) libdeflate at L10 compression 7.51 MB/s compressing nci (median across 15 runs) libdeflate at L10 compression 18.09 MB/s compressing ooffice (median across 15 runs) libdeflate at L10 compression 16.97 MB/s compressing osdb (median across 15 runs) libdeflate at L10 compression 8.64 MB/s compressing reymont (median across 15 runs) libdeflate at L10 compression 14.74 MB/s compressing samba (median across 15 runs) libdeflate at L10 compression 24.08 MB/s compressing sao (median across 15 runs) libdeflate at L10 compression 10.10 MB/s compressing webster (median across 15 runs) libdeflate at L10 compression 34.15 MB/s compressing x-ray (median across 15 runs) libdeflate at L10 compression 11.46 MB/s compressing xml (median across 15 runs) roc-deflate at L10 compression 12.97 MB/s compressing dickens (median across 15 runs) roc-deflate at L10 compression 14.67 MB/s compressing mozilla (median across 15 runs) roc-deflate at L10 compression 15.47 MB/s compressing mr (median across 15 runs) roc-deflate at L10 compression 7.19 MB/s compressing nci (median across 15 runs) roc-deflate at L10 compression 18.02 MB/s compressing ooffice (median across 15 runs) roc-deflate at L10 compression 17.23 MB/s compressing osdb (median across 15 runs) roc-deflate at L10 compression 9.63 MB/s compressing reymont (median across 15 runs) roc-deflate at L10 compression 14.59 MB/s compressing samba (median across 15 runs) roc-deflate at L10 compression 22.93 MB/s compressing sao (median across 15 runs) roc-deflate at L10 compression 11.18 MB/s compressing webster (median across 15 runs) roc-deflate at L10 compression 32.15 MB/s compressing x-ray (median across 15 runs) roc-deflate at L10 compression 10.73 MB/s compressing xml (median across 15 runs) libdeflate at L11 compression 8.85 MB/s compressing dickens (median across 15 runs) libdeflate at L11 compression 10.27 MB/s compressing mozilla (median across 15 runs) libdeflate at L11 compression 8.76 MB/s compressing mr (median across 15 runs) libdeflate at L11 compression 4.82 MB/s compressing nci (median across 15 runs) libdeflate at L11 compression 13.56 MB/s compressing ooffice (median across 15 runs) libdeflate at L11 compression 12.91 MB/s compressing osdb (median across 15 runs) libdeflate at L11 compression 6.49 MB/s compressing reymont (median across 15 runs) libdeflate at L11 compression 8.46 MB/s compressing samba (median across 15 runs) libdeflate at L11 compression 19.43 MB/s compressing sao (median across 15 runs) libdeflate at L11 compression 6.77 MB/s compressing webster (median across 15 runs) libdeflate at L11 compression 26.05 MB/s compressing x-ray (median across 15 runs) libdeflate at L11 compression 6.60 MB/s compressing xml (median across 15 runs) roc-deflate at L11 compression 10.10 MB/s compressing dickens (median across 15 runs) roc-deflate at L11 compression 10.53 MB/s compressing mozilla (median across 15 runs) roc-deflate at L11 compression 8.19 MB/s compressing mr (median across 15 runs) roc-deflate at L11 compression 4.74 MB/s compressing nci (median across 15 runs) roc-deflate at L11 compression 14.03 MB/s compressing ooffice (median across 15 runs) roc-deflate at L11 compression 13.62 MB/s compressing osdb (median across 15 runs) roc-deflate at L11 compression 7.71 MB/s compressing reymont (median across 15 runs) roc-deflate at L11 compression 8.68 MB/s compressing samba (median across 15 runs) roc-deflate at L11 compression 18.87 MB/s compressing sao (median across 15 runs) roc-deflate at L11 compression 8.07 MB/s compressing webster (median across 15 runs) roc-deflate at L11 compression 24.93 MB/s compressing x-ray (median across 15 runs) roc-deflate at L11 compression 6.51 MB/s compressing xml (median across 15 runs) libdeflate at L12 compression 7.35 MB/s compressing dickens (median across 15 runs) libdeflate at L12 compression 7.97 MB/s compressing mozilla (median across 15 runs) libdeflate at L12 compression 4.26 MB/s compressing mr (median across 15 runs) libdeflate at L12 compression 3.51 MB/s compressing nci (median across 15 runs) libdeflate at L12 compression 11.45 MB/s compressing ooffice (median across 15 runs) libdeflate at L12 compression 11.40 MB/s compressing osdb (median across 15 runs) libdeflate at L12 compression 5.59 MB/s compressing reymont (median across 15 runs) libdeflate at L12 compression 5.92 MB/s compressing samba (median across 15 runs) libdeflate at L12 compression 18.10 MB/s compressing sao (median across 15 runs) libdeflate at L12 compression 5.82 MB/s compressing webster (median across 15 runs) libdeflate at L12 compression 23.50 MB/s compressing x-ray (median across 15 runs) libdeflate at L12 compression 4.19 MB/s compressing xml (median across 15 runs) roc-deflate at L12 compression 8.59 MB/s compressing dickens (median across 15 runs) roc-deflate at L12 compression 8.22 MB/s compressing mozilla (median across 15 runs) roc-deflate at L12 compression 3.83 MB/s compressing mr (median across 15 runs) roc-deflate at L12 compression 3.48 MB/s compressing nci (median across 15 runs) roc-deflate at L12 compression 11.99 MB/s compressing ooffice (median across 15 runs) roc-deflate at L12 compression 12.21 MB/s compressing osdb (median across 15 runs) roc-deflate at L12 compression 6.87 MB/s compressing reymont (median across 15 runs) roc-deflate at L12 compression 6.03 MB/s compressing samba (median across 15 runs) roc-deflate at L12 compression 17.74 MB/s compressing sao (median across 15 runs) roc-deflate at L12 compression 7.11 MB/s compressing webster (median across 15 runs) roc-deflate at L12 compression 22.60 MB/s compressing x-ray (median across 15 runs) roc-deflate at L12 compression 4.18 MB/s compressing xml (median across 15 runs) Compression Level MB/s 700 600 500 400 300 200 100 0 file that ran fastest file that ran slowest L1 L2 L3 L4 L5 L6 L7 L8 L9 L10 L11 L12 libdeflate roc-deflate libdeflate at L1 compression 191.10 MB/s compressing dickens (median across 7 runs) libdeflate at L1 compression 235.50 MB/s compressing mozilla (median across 7 runs) libdeflate at L1 compression 261.70 MB/s compressing mr (median across 7 runs) libdeflate at L1 compression 387.30 MB/s compressing nci (median across 7 runs) libdeflate at L1 compression 167.90 MB/s compressing ooffice (median across 7 runs) libdeflate at L1 compression 254.70 MB/s compressing osdb (median across 7 runs) libdeflate at L1 compression 229.60 MB/s compressing reymont (median across 7 runs) libdeflate at L1 compression 291.00 MB/s compressing samba (median across 7 runs) libdeflate at L1 compression 157.40 MB/s compressing sao (median across 7 runs) libdeflate at L1 compression 207.00 MB/s compressing webster (median across 7 runs) libdeflate at L1 compression 142.20 MB/s compressing x-ray (median across 7 runs) libdeflate at L1 compression 411.30 MB/s compressing xml (median across 7 runs) roc-deflate at L1 compression 182.30 MB/s compressing dickens (median across 7 runs) roc-deflate at L1 compression 204.50 MB/s compressing mozilla (median across 7 runs) roc-deflate at L1 compression 228.90 MB/s compressing mr (median across 7 runs) roc-deflate at L1 compression 428.60 MB/s compressing nci (median across 7 runs) roc-deflate at L1 compression 144.60 MB/s compressing ooffice (median across 7 runs) roc-deflate at L1 compression 222.30 MB/s compressing osdb (median across 7 runs) roc-deflate at L1 compression 217.30 MB/s compressing reymont (median across 7 runs) roc-deflate at L1 compression 270.30 MB/s compressing samba (median across 7 runs) roc-deflate at L1 compression 138.50 MB/s compressing sao (median across 7 runs) roc-deflate at L1 compression 213.90 MB/s compressing webster (median across 7 runs) roc-deflate at L1 compression 130.80 MB/s compressing x-ray (median across 7 runs) roc-deflate at L1 compression 353.80 MB/s compressing xml (median across 7 runs) libdeflate at L2 compression 138.50 MB/s compressing dickens (median across 7 runs) libdeflate at L2 compression 162.80 MB/s compressing mozilla (median across 7 runs) libdeflate at L2 compression 153.40 MB/s compressing mr (median across 7 runs) libdeflate at L2 compression 286.50 MB/s compressing nci (median across 7 runs) libdeflate at L2 compression 120.40 MB/s compressing ooffice (median across 7 runs) libdeflate at L2 compression 175.30 MB/s compressing osdb (median across 7 runs) libdeflate at L2 compression 156.80 MB/s compressing reymont (median across 7 runs) libdeflate at L2 compression 210.50 MB/s compressing samba (median across 7 runs) libdeflate at L2 compression 104.20 MB/s compressing sao (median across 7 runs) libdeflate at L2 compression 163.50 MB/s compressing webster (median across 7 runs) libdeflate at L2 compression 111.60 MB/s compressing x-ray (median across 7 runs) libdeflate at L2 compression 258.40 MB/s compressing xml (median across 7 runs) roc-deflate at L2 compression 121.20 MB/s compressing dickens (median across 7 runs) roc-deflate at L2 compression 144.50 MB/s compressing mozilla (median across 7 runs) roc-deflate at L2 compression 145.30 MB/s compressing mr (median across 7 runs) roc-deflate at L2 compression 326.40 MB/s compressing nci (median across 7 runs) roc-deflate at L2 compression 105.10 MB/s compressing ooffice (median across 7 runs) roc-deflate at L2 compression 155.00 MB/s compressing osdb (median across 7 runs) roc-deflate at L2 compression 146.00 MB/s compressing reymont (median across 7 runs) roc-deflate at L2 compression 188.40 MB/s compressing samba (median across 7 runs) roc-deflate at L2 compression 94.10 MB/s compressing sao (median across 7 runs) roc-deflate at L2 compression 151.60 MB/s compressing webster (median across 7 runs) roc-deflate at L2 compression 95.50 MB/s compressing x-ray (median across 7 runs) roc-deflate at L2 compression 248.60 MB/s compressing xml (median across 7 runs) libdeflate at L3 compression 112.90 MB/s compressing dickens (median across 7 runs) libdeflate at L3 compression 152.80 MB/s compressing mozilla (median across 7 runs) libdeflate at L3 compression 132.10 MB/s compressing mr (median across 7 runs) libdeflate at L3 compression 304.30 MB/s compressing nci (median across 7 runs) libdeflate at L3 compression 113.00 MB/s compressing ooffice (median across 7 runs) libdeflate at L3 compression 164.00 MB/s compressing osdb (median across 7 runs) libdeflate at L3 compression 128.00 MB/s compressing reymont (median across 7 runs) libdeflate at L3 compression 191.40 MB/s compressing samba (median across 7 runs) libdeflate at L3 compression 94.50 MB/s compressing sao (median across 7 runs) libdeflate at L3 compression 143.70 MB/s compressing webster (median across 7 runs) libdeflate at L3 compression 112.20 MB/s compressing x-ray (median across 7 runs) libdeflate at L3 compression 209.90 MB/s compressing xml (median across 7 runs) roc-deflate at L3 compression 100.80 MB/s compressing dickens (median across 7 runs) roc-deflate at L3 compression 132.50 MB/s compressing mozilla (median across 7 runs) roc-deflate at L3 compression 125.30 MB/s compressing mr (median across 7 runs) roc-deflate at L3 compression 301.90 MB/s compressing nci (median across 7 runs) roc-deflate at L3 compression 97.90 MB/s compressing ooffice (median across 7 runs) roc-deflate at L3 compression 148.80 MB/s compressing osdb (median across 7 runs) roc-deflate at L3 compression 120.80 MB/s compressing reymont (median across 7 runs) roc-deflate at L3 compression 169.40 MB/s compressing samba (median across 7 runs) roc-deflate at L3 compression 84.80 MB/s compressing sao (median across 7 runs) roc-deflate at L3 compression 133.90 MB/s compressing webster (median across 7 runs) roc-deflate at L3 compression 95.10 MB/s compressing x-ray (median across 7 runs) roc-deflate at L3 compression 231.70 MB/s compressing xml (median across 7 runs) libdeflate at L4 compression 100.80 MB/s compressing dickens (median across 7 runs) libdeflate at L4 compression 144.50 MB/s compressing mozilla (median across 7 runs) libdeflate at L4 compression 123.40 MB/s compressing mr (median across 7 runs) libdeflate at L4 compression 254.00 MB/s compressing nci (median across 7 runs) libdeflate at L4 compression 109.70 MB/s compressing ooffice (median across 7 runs) libdeflate at L4 compression 151.90 MB/s compressing osdb (median across 7 runs) libdeflate at L4 compression 118.90 MB/s compressing reymont (median across 7 runs) libdeflate at L4 compression 178.70 MB/s compressing samba (median across 7 runs) libdeflate at L4 compression 89.10 MB/s compressing sao (median across 7 runs) libdeflate at L4 compression 129.70 MB/s compressing webster (median across 7 runs) libdeflate at L4 compression 111.30 MB/s compressing x-ray (median across 7 runs) libdeflate at L4 compression 201.50 MB/s compressing xml (median across 7 runs) roc-deflate at L4 compression 91.90 MB/s compressing dickens (median across 7 runs) roc-deflate at L4 compression 126.60 MB/s compressing mozilla (median across 7 runs) roc-deflate at L4 compression 117.60 MB/s compressing mr (median across 7 runs) roc-deflate at L4 compression 282.80 MB/s compressing nci (median across 7 runs) roc-deflate at L4 compression 95.50 MB/s compressing ooffice (median across 7 runs) roc-deflate at L4 compression 138.00 MB/s compressing osdb (median across 7 runs) roc-deflate at L4 compression 106.80 MB/s compressing reymont (median across 7 runs) roc-deflate at L4 compression 159.20 MB/s compressing samba (median across 7 runs) roc-deflate at L4 compression 81.50 MB/s compressing sao (median across 7 runs) roc-deflate at L4 compression 125.00 MB/s compressing webster (median across 7 runs) roc-deflate at L4 compression 95.60 MB/s compressing x-ray (median across 7 runs) roc-deflate at L4 compression 215.50 MB/s compressing xml (median across 7 runs) libdeflate at L5 compression 82.40 MB/s compressing dickens (median across 7 runs) libdeflate at L5 compression 120.00 MB/s compressing mozilla (median across 7 runs) libdeflate at L5 compression 103.10 MB/s compressing mr (median across 7 runs) libdeflate at L5 compression 250.40 MB/s compressing nci (median across 7 runs) libdeflate at L5 compression 89.30 MB/s compressing ooffice (median across 7 runs) libdeflate at L5 compression 128.80 MB/s compressing osdb (median across 7 runs) libdeflate at L5 compression 94.60 MB/s compressing reymont (median across 7 runs) libdeflate at L5 compression 152.10 MB/s compressing samba (median across 7 runs) libdeflate at L5 compression 74.80 MB/s compressing sao (median across 7 runs) libdeflate at L5 compression 112.20 MB/s compressing webster (median across 7 runs) libdeflate at L5 compression 93.70 MB/s compressing x-ray (median across 7 runs) libdeflate at L5 compression 198.40 MB/s compressing xml (median across 7 runs) roc-deflate at L5 compression 73.50 MB/s compressing dickens (median across 7 runs) roc-deflate at L5 compression 105.70 MB/s compressing mozilla (median across 7 runs) roc-deflate at L5 compression 96.60 MB/s compressing mr (median across 7 runs) roc-deflate at L5 compression 247.90 MB/s compressing nci (median across 7 runs) roc-deflate at L5 compression 77.50 MB/s compressing ooffice (median across 7 runs) roc-deflate at L5 compression 116.60 MB/s compressing osdb (median across 7 runs) roc-deflate at L5 compression 86.20 MB/s compressing reymont (median across 7 runs) roc-deflate at L5 compression 133.00 MB/s compressing samba (median across 7 runs) roc-deflate at L5 compression 68.10 MB/s compressing sao (median across 7 runs) roc-deflate at L5 compression 100.10 MB/s compressing webster (median across 7 runs) roc-deflate at L5 compression 82.10 MB/s compressing x-ray (median across 7 runs) roc-deflate at L5 compression 182.10 MB/s compressing xml (median across 7 runs) libdeflate at L6 compression 65.40 MB/s compressing dickens (median across 7 runs) libdeflate at L6 compression 102.90 MB/s compressing mozilla (median across 7 runs) libdeflate at L6 compression 81.60 MB/s compressing mr (median across 7 runs) libdeflate at L6 compression 203.10 MB/s compressing nci (median across 7 runs) libdeflate at L6 compression 78.50 MB/s compressing ooffice (median across 7 runs) libdeflate at L6 compression 118.80 MB/s compressing osdb (median across 7 runs) libdeflate at L6 compression 71.00 MB/s compressing reymont (median across 7 runs) libdeflate at L6 compression 120.10 MB/s compressing samba (median across 7 runs) libdeflate at L6 compression 64.40 MB/s compressing sao (median across 7 runs) libdeflate at L6 compression 91.80 MB/s compressing webster (median across 7 runs) libdeflate at L6 compression 94.80 MB/s compressing x-ray (median across 7 runs) libdeflate at L6 compression 168.30 MB/s compressing xml (median across 7 runs) roc-deflate at L6 compression 57.80 MB/s compressing dickens (median across 7 runs) roc-deflate at L6 compression 89.40 MB/s compressing mozilla (median across 7 runs) roc-deflate at L6 compression 74.10 MB/s compressing mr (median across 7 runs) roc-deflate at L6 compression 188.20 MB/s compressing nci (median across 7 runs) roc-deflate at L6 compression 70.60 MB/s compressing ooffice (median across 7 runs) roc-deflate at L6 compression 107.80 MB/s compressing osdb (median across 7 runs) roc-deflate at L6 compression 60.90 MB/s compressing reymont (median across 7 runs) roc-deflate at L6 compression 105.40 MB/s compressing samba (median across 7 runs) roc-deflate at L6 compression 59.20 MB/s compressing sao (median across 7 runs) roc-deflate at L6 compression 81.30 MB/s compressing webster (median across 7 runs) roc-deflate at L6 compression 83.20 MB/s compressing x-ray (median across 7 runs) roc-deflate at L6 compression 149.50 MB/s compressing xml (median across 7 runs) libdeflate at L7 compression 49.60 MB/s compressing dickens (median across 7 runs) libdeflate at L7 compression 76.10 MB/s compressing mozilla (median across 7 runs) libdeflate at L7 compression 56.30 MB/s compressing mr (median across 7 runs) libdeflate at L7 compression 150.20 MB/s compressing nci (median across 7 runs) libdeflate at L7 compression 69.10 MB/s compressing ooffice (median across 7 runs) libdeflate at L7 compression 113.80 MB/s compressing osdb (median across 7 runs) libdeflate at L7 compression 39.90 MB/s compressing reymont (median across 7 runs) libdeflate at L7 compression 85.90 MB/s compressing samba (median across 7 runs) libdeflate at L7 compression 53.10 MB/s compressing sao (median across 7 runs) libdeflate at L7 compression 66.10 MB/s compressing webster (median across 7 runs) libdeflate at L7 compression 94.20 MB/s compressing x-ray (median across 7 runs) libdeflate at L7 compression 119.90 MB/s compressing xml (median across 7 runs) roc-deflate at L7 compression 45.20 MB/s compressing dickens (median across 7 runs) roc-deflate at L7 compression 67.20 MB/s compressing mozilla (median across 7 runs) roc-deflate at L7 compression 52.40 MB/s compressing mr (median across 7 runs) roc-deflate at L7 compression 141.70 MB/s compressing nci (median across 7 runs) roc-deflate at L7 compression 61.10 MB/s compressing ooffice (median across 7 runs) roc-deflate at L7 compression 106.10 MB/s compressing osdb (median across 7 runs) roc-deflate at L7 compression 36.20 MB/s compressing reymont (median across 7 runs) roc-deflate at L7 compression 76.50 MB/s compressing samba (median across 7 runs) roc-deflate at L7 compression 52.80 MB/s compressing sao (median across 7 runs) roc-deflate at L7 compression 59.40 MB/s compressing webster (median across 7 runs) roc-deflate at L7 compression 81.60 MB/s compressing x-ray (median across 7 runs) roc-deflate at L7 compression 108.40 MB/s compressing xml (median across 7 runs) libdeflate at L8 compression 32.10 MB/s compressing dickens (median across 7 runs) libdeflate at L8 compression 41.40 MB/s compressing mozilla (median across 7 runs) libdeflate at L8 compression 35.80 MB/s compressing mr (median across 7 runs) libdeflate at L8 compression 80.70 MB/s compressing nci (median across 7 runs) libdeflate at L8 compression 50.40 MB/s compressing ooffice (median across 7 runs) libdeflate at L8 compression 97.80 MB/s compressing osdb (median across 7 runs) libdeflate at L8 compression 20.60 MB/s compressing reymont (median across 7 runs) libdeflate at L8 compression 53.40 MB/s compressing samba (median across 7 runs) libdeflate at L8 compression 46.60 MB/s compressing sao (median across 7 runs) libdeflate at L8 compression 40.00 MB/s compressing webster (median across 7 runs) libdeflate at L8 compression 78.30 MB/s compressing x-ray (median across 7 runs) libdeflate at L8 compression 72.60 MB/s compressing xml (median across 7 runs) roc-deflate at L8 compression 30.00 MB/s compressing dickens (median across 7 runs) roc-deflate at L8 compression 36.80 MB/s compressing mozilla (median across 7 runs) roc-deflate at L8 compression 33.90 MB/s compressing mr (median across 7 runs) roc-deflate at L8 compression 75.50 MB/s compressing nci (median across 7 runs) roc-deflate at L8 compression 45.60 MB/s compressing ooffice (median across 7 runs) roc-deflate at L8 compression 90.20 MB/s compressing osdb (median across 7 runs) roc-deflate at L8 compression 19.50 MB/s compressing reymont (median across 7 runs) roc-deflate at L8 compression 49.00 MB/s compressing samba (median across 7 runs) roc-deflate at L8 compression 44.50 MB/s compressing sao (median across 7 runs) roc-deflate at L8 compression 37.10 MB/s compressing webster (median across 7 runs) roc-deflate at L8 compression 72.10 MB/s compressing x-ray (median across 7 runs) roc-deflate at L8 compression 66.40 MB/s compressing xml (median across 7 runs) libdeflate at L9 compression 29.00 MB/s compressing dickens (median across 7 runs) libdeflate at L9 compression 29.10 MB/s compressing mozilla (median across 7 runs) libdeflate at L9 compression 32.50 MB/s compressing mr (median across 7 runs) libdeflate at L9 compression 58.50 MB/s compressing nci (median across 7 runs) libdeflate at L9 compression 45.00 MB/s compressing ooffice (median across 7 runs) libdeflate at L9 compression 95.30 MB/s compressing osdb (median across 7 runs) libdeflate at L9 compression 16.90 MB/s compressing reymont (median across 7 runs) libdeflate at L9 compression 41.60 MB/s compressing samba (median across 7 runs) libdeflate at L9 compression 45.60 MB/s compressing sao (median across 7 runs) libdeflate at L9 compression 33.90 MB/s compressing webster (median across 7 runs) libdeflate at L9 compression 78.60 MB/s compressing x-ray (median across 7 runs) libdeflate at L9 compression 57.90 MB/s compressing xml (median across 7 runs) roc-deflate at L9 compression 27.10 MB/s compressing dickens (median across 7 runs) roc-deflate at L9 compression 26.30 MB/s compressing mozilla (median across 7 runs) roc-deflate at L9 compression 30.60 MB/s compressing mr (median across 7 runs) roc-deflate at L9 compression 54.40 MB/s compressing nci (median across 7 runs) roc-deflate at L9 compression 41.20 MB/s compressing ooffice (median across 7 runs) roc-deflate at L9 compression 90.80 MB/s compressing osdb (median across 7 runs) roc-deflate at L9 compression 15.90 MB/s compressing reymont (median across 7 runs) roc-deflate at L9 compression 38.70 MB/s compressing samba (median across 7 runs) roc-deflate at L9 compression 43.30 MB/s compressing sao (median across 7 runs) roc-deflate at L9 compression 31.70 MB/s compressing webster (median across 7 runs) roc-deflate at L9 compression 72.70 MB/s compressing x-ray (median across 7 runs) roc-deflate at L9 compression 54.80 MB/s compressing xml (median across 7 runs) libdeflate at L10 compression 12.10 MB/s compressing dickens (median across 7 runs) libdeflate at L10 compression 14.80 MB/s compressing mozilla (median across 7 runs) libdeflate at L10 compression 16.20 MB/s compressing mr (median across 7 runs) libdeflate at L10 compression 7.30 MB/s compressing nci (median across 7 runs) libdeflate at L10 compression 18.20 MB/s compressing ooffice (median across 7 runs) libdeflate at L10 compression 16.10 MB/s compressing osdb (median across 7 runs) libdeflate at L10 compression 8.70 MB/s compressing reymont (median across 7 runs) libdeflate at L10 compression 14.40 MB/s compressing samba (median across 7 runs) libdeflate at L10 compression 23.30 MB/s compressing sao (median across 7 runs) libdeflate at L10 compression 10.00 MB/s compressing webster (median across 7 runs) libdeflate at L10 compression 34.70 MB/s compressing x-ray (median across 7 runs) libdeflate at L10 compression 10.80 MB/s compressing xml (median across 7 runs) roc-deflate at L10 compression 12.30 MB/s compressing dickens (median across 7 runs) roc-deflate at L10 compression 13.70 MB/s compressing mozilla (median across 7 runs) roc-deflate at L10 compression 14.70 MB/s compressing mr (median across 7 runs) roc-deflate at L10 compression 6.80 MB/s compressing nci (median across 7 runs) roc-deflate at L10 compression 16.80 MB/s compressing ooffice (median across 7 runs) roc-deflate at L10 compression 15.50 MB/s compressing osdb (median across 7 runs) roc-deflate at L10 compression 9.00 MB/s compressing reymont (median across 7 runs) roc-deflate at L10 compression 13.60 MB/s compressing samba (median across 7 runs) roc-deflate at L10 compression 20.60 MB/s compressing sao (median across 7 runs) roc-deflate at L10 compression 10.20 MB/s compressing webster (median across 7 runs) roc-deflate at L10 compression 29.40 MB/s compressing x-ray (median across 7 runs) roc-deflate at L10 compression 9.80 MB/s compressing xml (median across 7 runs) libdeflate at L11 compression 8.80 MB/s compressing dickens (median across 7 runs) libdeflate at L11 compression 10.30 MB/s compressing mozilla (median across 7 runs) libdeflate at L11 compression 9.30 MB/s compressing mr (median across 7 runs) libdeflate at L11 compression 4.70 MB/s compressing nci (median across 7 runs) libdeflate at L11 compression 13.70 MB/s compressing ooffice (median across 7 runs) libdeflate at L11 compression 12.30 MB/s compressing osdb (median across 7 runs) libdeflate at L11 compression 6.50 MB/s compressing reymont (median across 7 runs) libdeflate at L11 compression 8.40 MB/s compressing samba (median across 7 runs) libdeflate at L11 compression 18.90 MB/s compressing sao (median across 7 runs) libdeflate at L11 compression 6.70 MB/s compressing webster (median across 7 runs) libdeflate at L11 compression 26.80 MB/s compressing x-ray (median across 7 runs) libdeflate at L11 compression 6.20 MB/s compressing xml (median across 7 runs) roc-deflate at L11 compression 9.50 MB/s compressing dickens (median across 7 runs) roc-deflate at L11 compression 9.90 MB/s compressing mozilla (median across 7 runs) roc-deflate at L11 compression 8.30 MB/s compressing mr (median across 7 runs) roc-deflate at L11 compression 4.50 MB/s compressing nci (median across 7 runs) roc-deflate at L11 compression 13.00 MB/s compressing ooffice (median across 7 runs) roc-deflate at L11 compression 12.30 MB/s compressing osdb (median across 7 runs) roc-deflate at L11 compression 7.10 MB/s compressing reymont (median across 7 runs) roc-deflate at L11 compression 8.30 MB/s compressing samba (median across 7 runs) roc-deflate at L11 compression 17.00 MB/s compressing sao (median across 7 runs) roc-deflate at L11 compression 7.30 MB/s compressing webster (median across 7 runs) roc-deflate at L11 compression 23.00 MB/s compressing x-ray (median across 7 runs) roc-deflate at L11 compression 6.00 MB/s compressing xml (median across 7 runs) libdeflate at L12 compression 7.40 MB/s compressing dickens (median across 7 runs) libdeflate at L12 compression 8.00 MB/s compressing mozilla (median across 7 runs) libdeflate at L12 compression 4.70 MB/s compressing mr (median across 7 runs) libdeflate at L12 compression 3.40 MB/s compressing nci (median across 7 runs) libdeflate at L12 compression 11.50 MB/s compressing ooffice (median across 7 runs) libdeflate at L12 compression 10.90 MB/s compressing osdb (median across 7 runs) libdeflate at L12 compression 5.60 MB/s compressing reymont (median across 7 runs) libdeflate at L12 compression 5.90 MB/s compressing samba (median across 7 runs) libdeflate at L12 compression 17.70 MB/s compressing sao (median across 7 runs) libdeflate at L12 compression 5.80 MB/s compressing webster (median across 7 runs) libdeflate at L12 compression 24.60 MB/s compressing x-ray (median across 7 runs) libdeflate at L12 compression 4.00 MB/s compressing xml (median across 7 runs) roc-deflate at L12 compression 8.10 MB/s compressing dickens (median across 7 runs) roc-deflate at L12 compression 7.80 MB/s compressing mozilla (median across 7 runs) roc-deflate at L12 compression 4.10 MB/s compressing mr (median across 7 runs) roc-deflate at L12 compression 3.30 MB/s compressing nci (median across 7 runs) roc-deflate at L12 compression 11.20 MB/s compressing ooffice (median across 7 runs) roc-deflate at L12 compression 11.10 MB/s compressing osdb (median across 7 runs) roc-deflate at L12 compression 6.30 MB/s compressing reymont (median across 7 runs) roc-deflate at L12 compression 5.80 MB/s compressing samba (median across 7 runs) roc-deflate at L12 compression 15.80 MB/s compressing sao (median across 7 runs) roc-deflate at L12 compression 6.40 MB/s compressing webster (median across 7 runs) roc-deflate at L12 compression 21.00 MB/s compressing x-ray (median across 7 runs) roc-deflate at L12 compression 3.90 MB/s compressing xml (median across 7 runs) Compression Level So while we didn't outperform the champ on average, a median slowdown percentage in the single digits compared to such a carefully-tuned SIMD-powered C library is *wild.* It's frankly not a result I knew was achievable going in, since Roc is a functional language with automatic memory management racing against code that doesn't have to do bounds checks. (Bounds checks have a cost, but then again, the last time a skipped bounds check in C compression code led to a major Remote Code Execution vulnerability [was yesterday](https://nodejs.org/en/blog/vulnerability/october-2026-security-releases#nodejs-brotli-decoder-out-of-bounds-write-permits-remote-code-execution-cve-2026-111497---high).) The graph shows more detail than just the median gap. On the worst-case inputs, roc-deflate was 18% slower than libdeflate, but on the best-case inputs, roc-deflate was slightly over 18% *faster!* That's not because Roc has some magical optimization that C can't access; it's just that the two implementations are different, and although on average roc-deflate is slower, for certain inputs it happens to be faster. So while it's *possible* to get a speedup by switching from libdeflate to roc-deflate, you shouldn't expect it on average. To be clear, C's performance ceiling is higher than Roc's. For any input where roc-deflate currently outperforms libdeflate, C code can close that gap. The reverse is not true; Roc code is memory-safe, does automatic reference counting, checks bounds, and supports no Rust-style `unsafe` to opt out of any of that. These properties give Roc code stronger safety guarantees than code written in C, Zig, Rust, and other languages that allow full access to the hardware—and the price it pays for those guarantees is a lower performance ceiling. Obviously we can't extrapolate from this one result about how close Roc can get to C in the general case, but now something has been demonstrated that nobody knew was achievable before. As far as I know, this is the most optimization effort that has been put into any Roc package to date, so how close others can get to this result is a complete unknown right now. So while I'm thrilled by this result, I'm fully aware that we won't know how well it generalizes until we've optimized a wider sample of real-world Roc code bases! ## [What the Performance Gap Buys](#what-the-performance-gap-buys) An obvious question is "Why settle for *any* amount slower?" If you *could* get C levels of performance, why not use libdeflate itself and close that remaining gap? A pure-Roc implementation has several advantages over a pure-C one: - **Memory safety.** Roc code is fully memory-safe and has no language primitives that let you opt out of it. This code is only able to approach C levels of performance because Roc's optimization pipeline determined that many bounds checks and reference counting operations were provably safe to optimize away. Compression packages written in pure Roc don't have access to the unchecked memory operation that caused yesterday's [Brotli Remote Code Execution vulnerability](https://nodejs.org/en/blog/vulnerability/october-2026-security-releases#nodejs-brotli-decoder-out-of-bounds-write-permits-remote-code-execution-cve-2026-111497---high). - **No side effects.** Roc's type system [infers which functions are pure](https://roc-lang.org/docs/main/langref/functions/) and which can perform side effects. Pure functions' types use a thin arrow, like `Bool -> Bool`, whereas *effectful functions* use a thick arrow, like `Bool => Bool`. (As with memory safety, Roc enforces this and supports no way to opt out.) So when you see that roc-deflate's functions have only thin arrows in their annotations, you know they really are pure functions. - **Supply-Chain Security.** Roc does not have an arbitrary [FFI](https://en.wikipedia.org/wiki/Foreign_function_interface), meaning you know for sure that roc-deflate's pure functions can't even attempt to do I/O or access arbitrary memory. Given the explosion in [supply chain attacks](https://shattered.io/npm-supply-chain-attacks-2026/), it's valuable to know that the damage a malicious actor can do with a pure function in a compromised Roc package is limited to returning the wrong value, crashing, or wasting system resources. Roc packages also don't support build scripts, which was the vector used to [successfully compromise the popular Rust `arrayref` package](https://blog.rust-lang.org/2026/08/20/supply-chain-attack-on-arrayref/) two months ago. - **Portability.** Roc pure functions are designed to be extremely deterministic: they always return the same values when given the same arguments—not just when run on the same machine, but also on any target you compile that code to. (The functions may also crash or fail to terminate, but that behavior should also be deterministic across targets.) The only way a Roc pure function should be expected to behave differently when run on two different systems is if one system runs out of memory mid-execution. If any of these guarantees don't hold in practice, it can only be due to a defect in Roc's compiler, the [platform](https://www.roc-lang.org/docs/main/langref/platforms/) the application is running on, or the underlying hardware. So when you see in [roc-deflate's docs](https://niclas-ahden.github.io/roc-deflate/0.4.0/) that the entire package is 100% pure functions, you can be confident you are getting all of the above properties from everything the package has to offer. The type system enforces it! The reason we've been working so hard on all this might become clearer if you look at the proposition in reverse: would you give up guaranteed memory safety, guaranteed lack of side effects, guaranteed deterministic behavior across all targets, and the security benefits of having no exploitable build scripts in packages...for an average speedup percentage in the single digits? That answer can legitimately vary by use case—Roc's own compiler, for example, wants [the maximum performance the hardware can deliver](https://rtfeldman.com/rust-to-zig)—but I bet a lot of people reading this can see why we're so excited to have achieved this result. ## [Roc-Specific Optimizations](#roc-specific-optimizations) Most of the optimizations that improved roc-deflate's numbers came from general optimization techniques that are widely beneficial across languages. (I always recommend [Casey Muratori's Performance-Aware Programming Series](https://www.computerenhance.com/p/table-of-contents), which I paid for and learned a lot from.) Some of the optimization techniques were Roc-specific, though, and I wanted to mention them here as a reference for others. Even if you're not writing Roc code, this can also give you an idea of [how Roc achieves its performance](https://youtu.be/vzfy4EKwG_Y?si=oYa6k2pFdsRJcQW7). ### [Avoid Unnecessary Sharing](#avoid-unnecessary-sharing) One of Roc's most effective optimizations is *opportunistic mutation* (which [Koka](https://koka-lang.github.io/koka/doc/book.html#why-perceus) calls ["Functional but In-Place"](https://koka-lang.github.io/koka/doc/book.html#sec-fbip)). You can learn how it works in Roc's [docs on opportunistic mutation](https://www.roc-lang.org/docs/main/langref/expressions/#opportunistic-mutation), but from a performance optimization perspective, it's critically important to avoid unnecessary sharing so that values remain unique and therefore eligible for in-place mutation. ### [Hoist Bounds Checks](#hoist-bounds-checks) You can't opt out of bounds checks (e.g. getting a value out of a dynamically-sized collection) in Roc like you can in C, Zig, and Rust, but Roc's compiler can optimize them away if it knows they'd be guaranteed to pass. Bounds checks are most costly inside loops, so you can get a lot of mileage out of adding something like `if collection.len() <= highest_index_used { return 0 }` before the loop starts. This adds an additional check that looks redundant, but it can allow the optimizer to safely delete costlier checks inside the loop, once it can tell that failures would not be reachable after that early return. ### [Build Lookup Tables at Compile Time](#build-lookup-tables-at-compile-time) [Roc's compile-time evaluation](https://www.roc-lang.org/docs/main/langref/compile-time/) can be used to assemble lookup tables at compile time as long as nothing depends on runtime inputs. Even better, since their lengths are known at compile time, you can do things like `Ok(entry) = lookup_table.get(5)` to do an [exhaustive pattern match](https://www.roc-lang.org/docs/main/langref/pattern-matching/#exhaustiveness) that Roc's compiler will know is valid thanks to compile-time evaluation. This avoids runtime bounds checks that non-constant tables might have to pay for. ### [Consider Wrapping Arithmetic](#consider-wrapping-arithmetic) In Roc, `a + b` desugars to `a.plus(b)`. For an integer like [`U64`](https://www.roc-lang.org/docs/main/Num/#U64), [`U64.plus`](https://www.roc-lang.org/docs/main/Num/#U64.plus) includes an overflow check, whereas [`plus_wrap`](https://www.roc-lang.org/docs/main/Num/#U64.plus_wrap) does not. Although `plus` is the safer default, [overflow checks have a runtime cost](https://danluu.com/integer-overflow), as CPUs can do wrapping integer arithmetic in one instruction. When Roc's compiler can prove that overflow won't occur, both `plus` and `plus_wrap` will optimize to the same instructions, but when it can't, you may get a speedup by choosing `plus_wrap`. ## [Room for Improvement](#room-for-improvement) I'm sure there are still opportunities to improve roc-deflate's performance—especially on x64 CPUs—as well as the optimization pipeline in Roc's compiler. At this point I feel reasonably confident that all the low-hanging fruit has been picked, but if anyone wants to try it out for themselves, [here's the script I ran to set up and run the benchmark suite on Linux](https://gist.github.com/rtfeldman/7b6830809149867b79761ffb8e0456c3). (Note that Claude Opus 5.5 generated this script, and I haven't vetted it for security; run it outside a sandbox at your own risk!) It pins both Roc compiler revisions and roc-deflate versions, so you'll want to adjust those commits if you're trying out potential improvements. For my part, I'll be focusing on the upcoming Roc 0.1.0 release from now until it's out. It's been eight years in the making, so I'll be very happy for us to finally have a numbered release! If you'd like to learn more about Roc, check out [roc-lang.org](https://roc-lang.org) or [come say hi on Roc Zulip](https://roc.zulipchat.com/#narrow/channel/231634-beginners)!