Compare timings of dict constructor with literal dict

dict_perf.py

import random
import string
import time

values = []
for i in range(10):
    values.append("".join(random.choices(string.ascii_letters + string.digits, k=10)))

n = 500000
print(f"Construct {n=} dicts with these random values: {values=}")
print("---")


begin_dict = time.time_ns()
for i in range(n):
    dict(
        a0=values[0],
        a1=values[1],
        a2=values[2],
        a3=values[3],
        a4=values[4],
        a5=values[5],
        a6=values[6],
        a7=values[7],
        a8=values[8],
        a9=values[9],
    )
constructor_timing = time.time_ns() - begin_dict
print(f"{'dict constructor:':>20} {constructor_timing / 10 ** 6:.1f}ms")

begin_dict = time.time_ns()
for i in range(n):
    {
        "a0": values[0],
        "a1": values[1],
        "a2": values[2],
        "a3": values[3],
        "a4": values[4],
        "a5": values[5],
        "a6": values[6],
        "a7": values[7],
        "a8": values[8],
        "a9": values[9],
    }
literal_timing = time.time_ns() - begin_dict
print(f"{'literal dict:':>20} {literal_timing / 10 ** 6:.1f}ms")
percentage_difference = (constructor_timing - literal_timing) / constructor_timing * 100
print("---")
print(f"A literal dict is {percentage_difference:.2f}% faster")

Result

Construct n=500000 dicts with these random values: values=['zlh3wqFTEA', '0QE2KPvpfc', 'uBGj31IYCG', 'tkgopdTVkF', 'H3AZwyrqKo', 'Mcg78nqIss', 'KKhhOKjfII', 'rBpQ6WsJhm', 'vMVEdBeIZz', '2CUR1yfWEQ']
---
   dict constructor: 255.3ms
       literal dict: 207.7ms
---
A literal dict is 18.65% faster

I was curious how the number of keys might effect the time difference.
I re-ran this script with 0, 1, 2, 5, 10, 20, and 50 values using Python 3.9.16 on a Mac Pro M2: https://gist.github.com/PhilMarsh/1c7c695631cd0cd1a3924371fe3764c9


¹ The "% speedup" graph has a logarithmic trend line here, but power series also fit pretty well (maybe better?).