[Question] Performance f32 vs f64
Claus1 opened this issue · comments
How different is the library performance vs go standard math in average?
I have not tested it but I imagine on 64bit systems this library would perform as similar or worse than math
package.
My understanding is that this library does not seek to beat math
speed, which is already optimized with assembly all over the place. math32
is a option for when you need to work with float32
's i.e:
- Working with an external interface that accepts float32, such as graphics libraries (OpenGL, WebGL)
- Working on <=32bit systems with no FPU such as microcontrollers (see TinyGo)
I wasn't aware of it before but working with float32 slices can speed up programs in certain cases since more data fits on a cache line.