| 8 | | * [890] Do the Taylor series on `x / π` instead of `x`: results in a measurable 2% speed increase in `sin()` (ppu-lv2-g++) |
| 9 | | * [891] Do the Taylor series in two steps, for better pipelining at the cost of one extra multiplication: a notable 11% speed increase in `sin()` (ppu-lv2-g++) |
| | 8 | * [890] Do the Taylor series on `x / π` instead of `x`: results in a measurable speed increase in `sin()` (2% with ppu-lv2-g++, almost 1% with SNC) |
| | 9 | * [891] Do the Taylor series in two steps, for better pipelining at the cost of one extra multiplication: a notable speed increase in `sin()` (11% with ppu-lv2-g++, 2% with SNC) |
