⚡ Performance

Benchmarked against Dyalog APL v20.0 and GNU APL v2.0

Benchmarked against Dyalog APL v20.0 and GNU APL v2.0 on Apple M5. Native ARM64 + WASM (headless Chrome) side-by-side. Internal timing via ⎕AI (compute-only, no startup overhead). 40 iterations, 95% CI.

APL Run wins or ties 28 of 34 benchmarks vs Dyalog.

The AST viewer shows exactly which optimizations were applied to your code. Learn about execution badges →

Absolute Performance (msec, 95% CI)

Benchmark	Expression	APL Run	WASM	Dyalog	GNU APL
Prime Sieve
Primes N=2000	(2=+⌿0=∘.\\|⍨⍳2000)/⍳2000	1.9 ± 0.0	5.2 ± 1.4	5.6 ± 0.5	65.8 ± 2.6
Primes N=4000	(2=+⌿0=∘.\\|⍨⍳4000)/⍳4000	7.4 ± 0.0	14.1 ± 2.0	18.1 ± 0.8	—
Matrix Multiply
MatMul 100×100	M←100 100⍴⍳10000 ⋄ +/,M+.×M	0.2 ± 0.0	0.9 ± 1.0	0.8 ± 0.2	5.1 ± 0.1
MatMul 200×200	M←200 200⍴⍳40000 ⋄ +/,M+.×M	1.0 ± 0.0	2.0 ± 1.3	1.5 ± 0.2	40.6 ± 2.3
MatMul 400×400	M←400 400⍴⍳160000 ⋄ +/,M+.×M	7.4 ± 0.0	11.3 ± 2.5	8.9 ± 0.6	—
Reductions & Arithmetic
Sum ⍳2M	+/⍳2000000` (★ fused to O(1))	0.0 ± 0.0	0.4 ± 0.7	0.5 ± 0.2	9.7 ± 0.7
Sum ?2M⍴1000	+/?2000000⍴1000` (materialized)	8.7 ± 0.1	5.9 ± 1.5	10.5 ± 0.6	24.9 ± 2.2
Arith chain 10M	+/(3×⍳10000000)+(-⍳10000000)	9.5 ± 0.0	18.3 ± 4.5	11.1 ± 1.2	178.8 ± 3.9
Recursive Dfns
Collatz ⍳1K	+/c¨⍳1000` (recursive dfn)	1.3 ± 0.1	1.9 ± 0.9	33.1 ± 0.9	—
Collatz ⍳5K	+/c¨⍳5000` (recursive dfn)	8.4 ± 0.5	9.4 ± 1.2	201.8 ± 1.4	—
Fibonacci ⍳25	+/f¨⍳25` (doubly-recursive dfn)	6.6 ± 0.2	9.4 ± 1.2	316.8 ± 1.6	—
Sort & Newton
Sort Int 100K	v[⍋v]` (100K random ints)	1.3 ± 0.1	2.1 ± 1.1	1.4 ± 0.2	25.4 ± 2.0
Sort Float 1M	v[⍋v]` (1M random floats)	15.4 ± 0.1	15.6 ± 2.4	15.8 ± 0.6	—
Sort String 100K	M[⍋M;]` (100K×10 char matrix)	13.9 ± 0.7	18.4 ± 1.9	17.1 ± 0.6	—
Newton √100K	{⍵-((⍵*2)-N)÷2×⍵}⍣20⊢N	2.5 ± 0.2	3.4 ± 1.8	3.9 ± 0.5	—
Newton √1M	{⍵-((⍵*2)-N)÷2×⍵}⍣20⊢N	21.5 ± 0.2	28.0 ± 6.4	23.3 ± 0.8	—
LCS (Longest Common Substring)
LCS N=80	nested dfns, trains, ¨	3.0 ± 0.3	4.8 ± 1.5	6.2 ± 0.4	—
FFT (Cooley-Tukey)
FFT 1024	recursive dfn, complex ×/+	2.0 ± 0.1	4.1 ± 1.5	3.4 ± 0.3	—
FFT 4096	recursive dfn, complex ×/+	7.6 ± 0.0	10.9 ± 2.0	12.8 ± 0.6	—
Standard Primitives
Transpose 3K×3K	⍉M` (3000×3000 matrix)	14.9 ± 1.7	21.5 ± 3.3	7.9 ± 0.9	62.8 ± 2.8
Reverse 10M	⌽v` (10M-element vector)	4.1 ± 0.5	7.3 ± 2.8	3.3 ± 0.7	64.0 ± 3.6
Membership 2M∊1M	a∊b` (2M in 1M)	3.8 ± 0.1	6.1 ± 2.0	3.5 ± 0.4	—
Rotate 10M	100⌽v` (10M-element vector)	4.0 ± 0.4	7.0 ± 2.5	2.6 ± 0.5	105.3 ± 9.1
Erdős Primes
Erdős ⍳2500	nested dfns, prime test, ¨	3.8 ± 0.0	5.7 ± 1.3	6.3 ± 0.4	—
Erdős ⍳5000	nested dfns, prime test, ¨	8.1 ± 0.2	10.5 ± 1.7	12.9 ± 0.6	—
Lucky Numbers
Lucky N=1600	recursive sieve, replicate, modulus	1.1 ± 0.1	2.5 ± 1.3	1.2 ± 0.3	—
Lucky N=3000	recursive sieve, replicate, modulus	1.7 ± 0.0	3.8 ± 1.6	1.8 ± 0.3	—
LSWRC (Unique Substrings)
LSWRC N=117	tacit defs, trains, compose, ¨	0.3 ± 0.1	1.4 ± 1.4	0.3 ± 0.2	—
LSWRC N=567	tacit defs, trains, compose, ¨	2.4 ± 0.0	6.1 ± 1.6	7.0 ± 0.5	—
Shannon Entropy
Shannon N=1500	Key (⌸), compose bind, tacit trains	4.0 ± 0.2	8.7 ± 2.0	7.3 ± 0.3	—
Munchausen Numbers
Munchausen ⍳10K	tacit train, compose, ⍎¨⍕, *⍨	28.5 ± 0.4	36.6 ± 2.1	78.8 ± 6.4	—
Jaccard Similarity
Jaccard 300K	(≢A∩B)÷≢A∪B` (set ops, deal)	3.2 ± 0.2	6.9 ± 0.7	7.0 ± 0.5	—
Bell Triangle
Bell N=200×10	trains, scan, compose, ⍣N	1.7 ± 0.1	3.0 ± 1.4	4.4 ± 1.1	—
Multi-LCS
Multi-LCS ×2000	tacit train, 3 strings, ↑⌽∧\	9.9 ± 0.2	15.7 ± 2.6	9.6 ± 0.4	—

Relative Performance (× slower than fastest)

Benchmark	APL Run	WASM	Dyalog	GNU APL
Primes N=2000	1.0×	2.7×	2.9×	34.4×
Primes N=4000	1.0×	1.9×	2.4×	—
MatMul 100×100	1.0×	4.2×	3.4×	23.0×
MatMul 200×200	1.0×	2.1×	1.6×	42.7×
MatMul 400×400	1.0×	1.5×	1.2×	—
Sum ⍳2M	1.0×	3.8×	4.7×	97.0×
Sum ?2M⍴1000	1.5×	1.0×	1.8×	4.2×
Arith chain 10M	1.0×	1.9×	1.2×	18.9×
Collatz ⍳1K	1.0×	1.5×	25.9×	—
Collatz ⍳5K	1.0×	1.1×	24.1×	—
Fibonacci ⍳25	1.0×	1.4×	47.7×	—
Sort Int 100K	1.0×	1.6×	1.1×	19.5×
Sort Float 1M	1.0×	1.0×	1.0×	—
Sort String 100K	1.0×	1.3×	1.2×	—
Newton √100K	1.0×	1.4×	1.6×	—
Newton √1M	1.0×	1.3×	1.1×	—
LCS N=80	1.0×	1.6×	2.1×	—
FFT 1024	1.0×	2.0×	1.7×	—
FFT 4096	1.0×	1.4×	1.7×	—
Transpose 3K×3K	1.9×	2.7×	1.0×	8.0×
Reverse 10M	1.2×	2.2×	1.0×	19.4×
Membership 2M∊1M	1.1×	1.7×	1.0×	—
Rotate 10M	1.5×	2.7×	1.0×	40.1×
Erdős ⍳2500	1.0×	1.5×	1.7×	—
Erdős ⍳5000	1.0×	1.3×	1.6×	—
Lucky N=1600	1.0×	2.2×	1.1×	—
Lucky N=3000	1.0×	2.2×	1.1×	—
LSWRC N=117	1.0×	5.1×	1.2×	—
LSWRC N=567	1.0×	2.6×	3.0×	—
Shannon N=1500	1.0×	2.2×	1.8×	—
Munchausen ⍳10K	1.0×	1.3×	2.8×	—
Jaccard 300K	1.0×	2.2×	2.2×	—
Bell N=200×10	1.0×	1.7×	2.6×	—
Multi-LCS ×2000	1.0×	1.6×	1.0×	—

Last benchmarked: 2026-07-22. APL Run native = ARM64 binary (apl_cli). WASM = same Rust code compiled to WebAssembly, running in headless Chrome. Single-core (no threading). SIMD enabled (native only).