Migrating

Every number on this page is measured. The share of results off by a wei or more, and the worst error, come from the WAD results; the gas comes from the per-op benchmarks. Hover a row for the exact call.

From Solady

BlockulatorWad has the names, signatures, rounding directions and error selectors of Solady's FixedPointMathLib, so the move is one import:

import {BlockulatorWad as FixedPointMathLib} from "@blockulator/contracts/wad/BlockulatorWad.sol";

Solady's multiply and divide functions return exact values. Their share wrong is all reverts where x · y (or x · 10¹⁸ for a quotient) passes 2²⁵⁶ though the result fits, so their worst error is 0. The worst error is absolute: near 10⁷⁴, one part in 10¹⁸ is many wei.

BlockulatorWad returns the exact value rounded once, and its domains only widen: mulWad and divWad revert only when the result does not fit, powWad(0, y) is defined, and expWad reaches one wei further (full list).

It needs the CLZ opcode: compile with evm_version = "osaka" and solc 0.8.31 or later. Keep Solady's integer helpers (abs, min, avg and the rest); they have no WAD meaning.

From PRBMath

PRBMath's values are user-defined types over the same 18-decimal integers, so a call moves by unwrapping. exp, ln and pow keep PRBMath's names. mul, div and sqrt become mulWad or sMulWad, divWad or sDivWad, and sqrtWad. log2 and log10 become log2Wad and log10Wad.

// SD59x18 x, y;
int256 r = BlockulatorWad.pow(x.unwrap(), y.unwrap());   // was x.pow(y)

From ABDK

Each of ABDK's two formats has a Blockulator counterpart of the same kind, and both use one int256 per value:

• ABDKMath64x64 moves to FP127: 127.128 fixed point, with 64 more fractional bits. x << 64 converts.
• ABDKMathQuad's binary128 moves to FP19: a 256-bit binary float, with 237 bits of significand against 113.

Gas is the inline library on both sides. FP127 and FP19 are also deployed as contracts you can staticcall; the library pages have the addresses.