Just for fun, a back-of-napkin calculation indicates that 400 0.1% improvements building upon each other will get you almost a 50% speedup, and around 700 0.1% improvements will get you to 2x. So, to grossly overgeneralize, small improvements over the lifetime of a project can eventually end up making a big difference.
WebKit for example has a zero regression policy, but I’m not sure how strictly it’s enforced. 0.1% is within the margin of error unless you are testing very rigorously.
I work on SQLCipher so I’ve become accustom to using the command line shell interface, however there slowly becoming alternative GUI options available for folks.
I asked over twitter what he attributed the performance to and he responded:
“… hundreds of micro-optimizations, each yielding between 0.1% to 1.0% improvement (usually closer to 0.1%), add up over time.” - @DRichardHipp
Just for fun, a back-of-napkin calculation indicates that 400 0.1% improvements building upon each other will get you almost a 50% speedup, and around 700 0.1% improvements will get you to 2x. So, to grossly overgeneralize, small improvements over the lifetime of a project can eventually end up making a big difference.
Note that in order to benefit from this, you need to prevent 0.1% regressions.
WebKit for example has a zero regression policy, but I’m not sure how strictly it’s enforced. 0.1% is within the margin of error unless you are testing very rigorously.
This is always great news!
sqlite is amazingly light and fast as it is… But since it’s use is so universal, this will make a difference in life.
ps: If you ever need a gui front end to sqlite, the firefox addon sqlitemanager is fantastic.
Quite the best datamining tool available.
I work on SQLCipher so I’ve become accustom to using the command line shell interface, however there slowly becoming alternative GUI options available for folks.
Can you recommend any Open Source app for SQLite on OS X?
Also: 3.8.11(.1) has been released.
I thought the subject sounded familiar. 3.8.7 had a similar note about lots of cumulative microoptimizations.