Accumath™
Weekly benchmarks
Seed gold benches from the engine on the build machine — updated about once a week so you can see trends as version 1 grows.
Latest run — 2026-09-19
Machine: Intel Core i9-14900KF (32 threads), Pop!_OS Linux x86_64. SoftFloat precision: 256 bits. Profile: cargo run -p engine_core --release --bin record_bench_baseline.
Method: warmup + median of 7 runs, same seed set as benches/engine_bench.rs. Not a CI fail gate. Absolute numbers move with turbo and load; week-to-week trends matter more.
| Bench | Id | Time | ns |
|---|---|---|---|
| Simplify: fold exp(ln(x)) | cas_simplify.fold_exp_log_exp_ln_x | 78 ns | 78 |
| Simplify: polynomial | cas_simplify.simplify_poly | 223.7 µs | 223,687 |
| Integrate: ln(x)/x (Risch) | cas_integrate.ln_over_x | 1.240 ms | 1,240,121 |
| Integrate: 1/(x ln(x)) | cas_integrate.one_over_x_ln_x | 1.110 ms | 1,109,719 |
| Integrate: exp(x)/x | cas_integrate.exp_over_x | 2.300 ms | 2,299,813 |
| Integrate: exp poly | cas_integrate.exp_poly | 2.287 ms | 2,287,173 |
| Factor: minpoly √2+√3 | cas_factor.minpoly_sqrt2_plus_sqrt3 | 144.7 µs | 144,699 |
| Factor: minpoly √2+∛2 | cas_factor.minpoly_sqrt2_plus_cbrt2 | 916.5 µs | 916,458 |
| Solve: x³−1 | cas_solve.solve_x3_minus_1 | 378.3 µs | 378,313 |
| Eval SoftFloat constants | num_eval.eval_soft_constants | 23.5 µs | 23,510 |
| Specials: erf(1) | num_specials.eval_erf_1 | 56.1 µs | 56,140 |
| Specials: Γ(1/2) | num_specials.eval_gamma_half | 25.629 ms | 25,628,517 |
| nsolve: x²−2 | num_linalg.nsolve_x2_minus_2 | 1.303 ms | 1,303,421 |
| Quad: ∫ 1/(1+x²) on [0,1] @ 1e-8 | num_quad.quad_atan_unit | 480.5 µs | 480,475 |
How we record
- Same seed set each week (simplify, integrate, factor, solve, SoftFloat eval, specials, nsolve, quad).
- Release build on the Accumath desktop; median of seven timed runs after one warmup.
- History lives in
benchmarks.jsonon this site so prior weeks stay visible as we add them.