Bugs in the built-in
Math
namespace.
- Numeric instability of transcendental functions
- Sine and cosine in V8
- Trigonometric functions in V8
- V8 not IEEE 754-2008 compliant
- Mozilla discussion on sine and cosine in V8
- V8 replaced a lookup table by computing
Math.tan
asMath.sin/Math.cos
- Browser math accuracy issues
- Mozilla attempt to address precision in new Math functions
- Mozilla's previous lack of tolerance tests for Math functions
- Mozilla
Math.expm1
accuracy - Mozilla thread on implementing ES6 math functions
- V8
Math.atanh
issues - V8
Math.acosh
issues - V8
Math.asinh
issues - V8 numeric issues in hyperbolic functions
- ES6 accuracy of special functions
- Accuracy of
Math.exp
in V8 - TC39 meeting discussing Math accuracy issues
- Spreadsheet showing trigonometric results across browsers
- Accuracy of
Math.pow
in V8 - Accumulation of errors in Mozilla
Math.pow
- Accuracy of hyperbolic trigonometric functions in V8
- ES6 shim accuracy issues
- TC39 testing of Math built-ins
- V8
Math.random
PRNG quality - V8 trig function precision issues
- WebKit
Math.random
repeat values in separate workers - Predictable
Math.random
in WebKit - Predictable
Math.random
in Chrome - Desire for fastmath implementations
- Mozilla
Math.round
issues - Mozilla
Math.pow
issue - Mozilla
Math.pow
issue - V8 deoptimizes if
Math.floor
provided-0
- V8 deoptimizes if
Math.ceil
provided0
- V8
Math.round
accuracy - Observability of distinguishable NaNs (see also get-nans and ECMA262 issue 635)
- NaN converted to -NaN
- After 10 years, Mozilla updates
Math.random
- WebKit bug to use a better PRNG
- WebKit 32-bit Math bug
- Mozilla optimizations break commutativity
- Mozilla debate on speed versus precision
- Mozilla division bug to due C library delegation
- Fixit to unify Math implementations in V8
- Fixit to unify
Math.pow
implementation in V8
- To search for V8 Math bugs, see V8 issue tracker.
- To search for WebKit Math bugs, see WebKit issue tracker.
- To search for Mozilla Math bugs, see Bugzilla.