Floating point math library

Brought to you by Bankex Foundation

WIP, not ready for production

Fuctionality

Float point number is represented in a binary256 IEEE standard (so library operates on bytes32 and uint256[3] internally).

Included functions:

• add
• sub
• mul
• div (naive implementation, slow)
• log2
• fast inverse sqrt
• fast div (using fast inverse sqrt)

Planned functions:

• exp

Known issues:

• Non-normalized numbers are not properly represented
• Overflow is not processed

What's inside

Crude core for library and set of tests, including helper functions to encode BigNumber into bytes32. Library is in a form of contract for ease of compilation in deployment (see Tester.sol for example).

How to run

npm install
npm test

tests.js is a set of helper and test functions.

Contributors

• shamatar
• skywinder