Lean Kernel Arena / perf/proj-app-lazy-delta

Test "perf/proj-app-lazy-delta"

Expected: 👍 accept · Size: 122.0 KB · Lines: 2.2 k · lean4export: 3.1.0 · Lean: 4.36.0 · Timeout: 60.0 s · 📄 Declaration · 🔗 Source

The declaration to check is fun a h => h at type

∀ a, a ∈ small → a ∈ big N

where PSet is a structure with one field Mem : α → Prop (like Mathlib's Set), a ∈ s unfolds to the applied projection s.1 a, small unfolds to big N in one delta step, and reducing big N to a constructor takes Θ(N) steps.

The two sides are applied projections small.1 a and (big N).1 a. Unfolding small makes them syntactically equal, for Θ(1). A checker that instead reduces the projected structures to constructors before comparing evaluates big N, for Θ(N).

Official compares bare projections s.1 =?= t.1 lazily, but not applied ones: for those it reduces the structures with whnf first (leanprover/lean4#16613). The issue uses N = 2^(2^32), on which official gives up; N=100000 in the Lean source, where it is merely slow.

Checker Result ⏱️ 🧠
nano-lean 👍 1.39 ms (÷1498) 47.9 MB (÷13)
gruszka 👍 2.86 ms (÷727) 51.9 MB (÷12)
sokonanoda 👍 2.20 ms (÷946) 53.6 MB (÷11)
rpylean 👍 2.14 ms (÷971) 12.6 MB (÷49)
pestka 👍 4.35 ms (÷478) 19.8 MB (÷31)
kefir 👍 5.65 ms (÷368) 6.4 MB (÷96)
palladium 👍 804 ms (÷2.6) 670.3 MB (+10%)
con-ron 👍 1.77 s (-15%) 496.6 MB (-19%)
kefir-wasm 👍 224 ms (÷9.3) 108.1 MB (÷5.7)
con-agua 👍 2.25 s (+8%) 499.9 MB (-18%)
nanobruijn 👍 1.88 s (-10%) 565.9 MB (-8%)
nanoda 👍 2.12 s (+2%) 546.4 MB (-11%)
con-leche ⌛ — —
pestka-wasm 👍 133 ms (÷16) 120.8 MB (÷5.1)
tenet-jvm 👍 3.04 s (+46%) 751.3 MB (+23%)
official-nightly 👍 1.55 s (-26%) 629.7 MB (+3%)
eink0rn 👍 6.13 ms (÷340) 20.1 MB (÷30)
official 👍 2.08 s 612.1 MB
ind-models 👍 2.14 s (+3%) 628.0 MB (+3%)
tenet 👍 2.46 s (+18%) 1.5 GB (×2.4)
tenet-hs 👍 3.37 s (+62%) 384.5 MB (-37%)
kiota 👍 6.60 ms (÷316) 11.1 MB (÷55)
ocaml4lean ⌛ — —
logos-kernel 👍 543 ms (÷3.8) 864.6 MB (+41%)
evmlean ⌛ — —
mini 🚫 41.3 ms 65.0 MB
overfull ⌛ — —
lean4cobol 🚫 36.8 s 276.9 MB
lean4lean ✋ 924 ms 221.8 MB
canonical-min 🚫 394 ms 1.0 GB
nanoclo 👍 3.05 ms (÷683) 61.3 MB (÷10.0)
vow-lean-kernel 👍 18.2 s (×8.7) 80.2 MB (÷7.6)
lazylean 👍 375 ms (÷5.5) 412.6 MB (-33%)
metalean-slow 🚫 2.51 s 69.6 MB
metalean-fast 👍 54.3 s (×26) 746.3 MB (+22%)
nanoclo-fortran 👍 7.61 ms (÷274) 5.3 MB (÷114)
mathgraph 👍 2.12 ms (÷980) 73.1 MB (÷8.4)