34fa05f66b
Overall gain is about 20-22% of time on my laptop. This includes: * replacing string sort_key with a tuple (because we can and because compairing two 3-tuples is faster than comparing two 26-byte strings); * keeping track of hungriest dev in tier (allows us to use built-in dict.__getitem__ as a key instead of lambdas in couple of places); * remove unnecessary sorted() call in the innermost loop (because we don't need to sort all of them or better we don't need to sort anything there); * memoize tiers for each dev (saves just a couple of percents but why not). Performance measurments have been done using this script: http://paste.openstack.org/show/55609/ Related-Bug: #1262166 Related-Bug: #1261659 Change-Id: If38bd9fe82efc12b01e9aa146e0f4ab565fb6bea |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| builder.py | ||
| ring.py | ||
| utils.py | ||