| |||||||||||||
A brief experiment with PyPyA brief experiment with PyPyPosted May 12, 2011 1:15 UTC (Thu) by andresfreund (subscriber, #69562)Parent article: A brief experiment with PyPy
One reason for the higher amount of cache misses might simply be that a tigher execution schedule makes it harder for the prefetching units to load all the data in advance.
(Log in to post comments)
A brief experiment with PyPy Posted May 12, 2011 1:42 UTC (Thu) by jzbiciak (✭ supporter ✭, #5246) [Link]
Or freeing dead objects less aggressively. I'm willing to bet that there's a lot of temporary objects that get recycled in CPython, but don't get reaped as quickly in the PyPy version.
45M cache misses with a 64 byte line size is ~2.8GB of RAM... that's a lot of RAM to cycle through!
|
|||||||||||||