Lang

Cool new features in 3.11. Performance 1.2x faster generally, thanks to an adaptive interpreter (PEP659) that optimises byte-code based on observed behaviour and usage. Take for example the LOAD_ATTR instruction, which under 3.11 can be replaced by LOAD_ATTR_ADAPTIVE. This will replace the call to the most optimised instruction based on what is being done, such as: LOAD_ATTR_INSTANCE_VALUE LOAD_ATTR_MODULE LOAD_ATTR_SLOT Disassembling some code: def feet_to_meters(feet): return 0.3048 * feet for feet in (1.0, 10.0, 100.0, 1000.0, 2000.0, 3000.0, 4000.0): print(f"{feet:7.1f} feet = {feet_to_meters(feet):7.1f} meters") import dis dis.dis(feet_to_meters, adaptive=True) # 1 0 RESUME 0 # # 2 2 LOAD_CONST 1 (0.3048) # 4 LOAD_FAST 0 (feet) # 6 BINARY_OP 5 (*) # 10 RETURN_VALUE However, when the interpreter is given more concrete to work with its able to optimise. For example, outside the loop context when given a float, floating point instructions are put to work: ...