Android-on-Python would have been even more slower (Java is a _fast_ interpreted language) and without possibility for enhancements (JITs for Java are commonplace, JITs for Python are not).
Posted Sep 1, 2012 23:06 UTC (Sat) by anselm (subscriber, #2796)
[Link]
Android-on-Python would have been even more slower (Java is a _fast_ interpreted language) and without possibility for enhancements (JITs for Java are commonplace, JITs for Python are not).
PyPy is supposed to be very good indeed.
Java is modern cobol thanks to Y2K
Posted Sep 2, 2012 3:28 UTC (Sun) by Cyberax (✭ supporter ✭, #52523)
[Link]
And it requires about 20 times more RAM than Dalvik's JIT and _still_ produces inferior code.
We're using a lot of Python (and numpy) code in our R&D and so I've tested all possible ways to make Python work faster. They all suck in various ways.
Java is modern cobol thanks to Y2K
Posted Sep 7, 2012 13:33 UTC (Fri) by pboddie (subscriber, #50784)
[Link]
Care to describe "all possible ways"?
Within Google - I'm guessing that you and another prominent commenter on LWN work there given the numerous "nod and wink" references to various projects - there have apparently been a number of projects to improve Python performance and/or predictability, some being very widely known and others only barely surfacing on the radar of the most central Python core developers (many of whom also seem to be at Google), so I'd be interested to hear what you've looked at.