Changes between Initial Version and Version 1 of Ticket #2109, comment 37
- Timestamp:
- Mar 4, 2016, 3:42:14 PM (8 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #2109, comment 37
initial v1 1 I attached a patch that should i ncrease the precision of sincos_approx (from ~4*10^-4^ to ~1.5*10^-4^) while reducing the amount of multiplications used. It uses a similar approach to the "original" implementation and as such, there is no improvement for atan (arctan(z) has singularities for z=+-i and thus polynomial approximations are much worse when compared to those of sin(z) resp. cos(z) (no singularities)).1 I attached a patch that should improve the precision of sincos_approx (from ~4*10^-4^ to ~1.5*10^-4^) while reducing the amount of multiplications used. It uses a similar approach to the "original" implementation and as such, there is no improvement for atan (arctan(z) has singularities for z=+-i and thus polynomial approximations are much worse when compared to those of sin(z) resp. cos(z) (no singularities)). 2 2 3 3 Perhaps the CORDIC algorithms are better, in the meantime we can use this.
