Ideas for improving performance of function calls

I saw that simple function calls in Brython take around 5x the time of CPython, so I looked into it a bit to see if there were some quick ways to speed it up. I found a few things that together led to a 10x speedup. I’m not sure if two of them are safe, so I’m just listing them here instead of making a PR.

1. Add if(frame[1].$has_yield_in_cm){ test around the call to exit_ctx_managers_in_generators in $B.leave_frame_exec. This saves a ton of time (when you are running inside of an exec), and is analogous to the code in $B.leave_frame. Hopefully it is possible to do this, but I don’t know for sure.

2. Every function body contains this line $locals.__annotations__ = _b_.dict.$factory(), but I don’t think it is possible to annotate local variables inside the function, so it is never used. Can it be removed?

3. In $B.leave_frame, remove the call to $B.del_exc and add frame[1].$current_exception = undefined after var frame = $B.frames_stack.pop().

4. In $B.enter_frame, return null everywhere you currently use _b_.None, and change the autogenerated trace code from if($locals.$f_trace !== _b_.None){$B.trace_line()} to if($locals.$f_trace){$B.trace_line()}. This didn’t make nearly as large a difference as the first three, but it’s a simple change that also reduces the size of the generated code, so it still seems worth doing.

Thanks for taking a look at these!

[schmave]

You can specify annotations on local variables, but I believe that they are not available at runtime (from PEP 526):

Store variable annotations also in function scope: The value of having the annotations available locally is just not enough to significantly offset the cost of creating and populating the dictionary on each function call.

[schmave]

Looks good to me! When I run the benchmark named “function call” using http://localhost:8000/speed on the master branch, it takes 1900-2000ms for 1,000,000 calls to f(). When I run it on the function_call_speed branch, it takes 150ms. I disabled caching in my browser and cleared local storage and indexed DB between runs.

In CPython, I see that __annotations__ inside a function seems to refer to the module level __annotations__, and I think this will continue to work in Brython:

x: int = 5
print(__annotations__)
def f():
  print('in f', __annotations__)

f()

def g():
   y: int = 7
   print('in g', __annotations__)
 
g()

CPython 3.8.0 output:

{'x': <class 'int'>}
in f {'x': <class 'int'>}
in g {'x': <class 'int'>}