Results
Correct digits left after 10,000 steps, per scenario and library.
Each scenario repeats one step N times in every library's own representation and compares the result with the same loop in 200-digit arithmetic. Daily compounding: FP127 keeps 34 digits, Solady 14. The per-second accrual a lending market runs: 29 against 8.5.
| scenario | FP256 | FP127 | Quad | Maker | ABDK | Solady | PRBMath | Balancer | Aave | Morpho |
|---|---|---|---|---|---|---|---|---|---|---|
| Daily compound interest | 67 | 34 | 29 | 23 | 15 | 14 | 14 | 14 | ||
| Compound interest in one pow | 67 | 34 | 30 | 15 | 14 | 13 | 14.5 | |||
| Square-root bonding curve | 70 | 40 | 30 | 21 | 20 | 20 | 18 | |||
| exp(ln(x)) round trip | 71.5 | 38 | 28.5 | 15 | 14 | 13 | 14 | |||
| Amortising loan | 67 | 35 | 29 | 15.5 | 15.5 | 15.5 | 15.5 | |||
| Running geometric mean | 71 | 38 | 33 | 18 | 17 | 16 | 17 | |||
| Black-Scholes d1 along a price path | 67.5 | 37 | 30 | 18 | 17 | 16 | 17 | |||
| Cumulative product of small factors | 69 | 34 | 29 | 15 | 14 | 14 | 14 | |||
| Annual compounding at 10% (README claim check) | 67 | rev | 29 | rev | rev | rev | rev | rev | ||
| Per-second accrual | 62 | 29 | 24 | 18 | 10 | 8.5 | 8 | 9 | ||
| Continuous accrual | 71 | 37 | 33 | 17.5 | 17.5 | 17 | 17 | 1 | 1 | |
| An option walking out of the money | 65 | 0.5 | 27 | 0 | 0 | 0 | 4 | |||
| Decay, the share released | 59 | 26 | 21 | 7 | 11 | 6 | 11 |
Median correct digits over each scenario's inputs at N = 10,000, against exact inputs; capped at 100. rev: every input reverted. —: the library has no model for that scenario. FP127Lib equals FP127 and is left out.
9,384 cells checked live on Sepolia at block 11,928,247, 0 mismatches. 2026-10-16, runner 0xfcdf2f18…. Re-run: uv run scripts/ladder/livecap.py