Usage of native/viper emitters for libraries #11157
Replies: 2 comments 3 replies
-
|
Your assumption is correct. |
Beta Was this translation helpful? Give feedback.
-
|
Optimize for speed, or size? , you get to pick one only. Assuming you are after speed, there are more options than just adding a decorator. https://docs.micropython.org/en/latest/reference/speed_python.html#maximising-micropython-speed |
Beta Was this translation helpful? Give feedback.
-
|
Mainly speed. Thanks yes I have been working through the additional items there. I was intrigued by doubling in speed. Considering the native emitter complies to standard python, why not apply it to everything? |
Beta Was this translation helpful? Give feedback.
-
The native emitter produces larger code than bytecode. In my experience the native emitter often does not give much of a speed boost. Viper can be astonishingly fast if applied to the right type of code, notably bit manipulations on integers where it can be almost as fast as assembler. Like all optimisations the emitters are best applied selectively. An application will usually have a single bottleneck. First identify where it is, measure performance, and apply optimisations measuring at each stage. See also opt_level. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the tips @peterhinch |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
If I wanted to try and optimise a library, would I need modify and decorate every function or is there a shortcut?
I did a simple test and all three returned the same time so I imagine the emitter only applies to the code immediately inside the decorated function?
Beta Was this translation helpful? Give feedback.
All reactions