On the swift-dev list they announced the 2.2 branch is now under
restrictive change control. If you're been putting off checking it out, I'd
suggest the next snapshot is worth a look.
I did some benchmarks on the last snapshot using a Parks-McClellan filter
design algorithm. The algorithm is heavy on floating point arithmetic with
a little trig while iterating over arrays.
C: 2.07ms
Swift 2.2: 2.10ms
Swift 2.1: 2.70ms
Go 1.5.2: 2.93ms
Ruby 2.0.0: 85.50ms
Math performance is important to my plans for Swift, so this makes me
happy. Ruby and Go haven't improved on this benchmark in two years.
The Parks-McClellan filter design algorithm is hosted here:
Unfortunately, compile times are longer in 2.2. But one day we'll have it
all.
On the swift-dev list they announced the 2.2 branch is now under restrictive change control. If you're been putting off checking it out, I'd suggest the next snapshot is worth a look.
I did some benchmarks on the last snapshot using a Parks-McClellan filter design algorithm. The algorithm is heavy on floating point arithmetic with a little trig while iterating over arrays.
C: 2.07ms
Swift 2.2: 2.10ms
Swift 2.1: 2.70ms
Go 1.5.2: 2.93ms
Ruby 2.0.0: 85.50ms
Math performance is important to my plans for Swift, so this makes me happy. Ruby and Go haven't improved on this benchmark in two years.
The Parks-McClellan filter design algorithm is hosted here: GitHub - AE9RB/firpm
Nice to hear!
Unfortunately, compile times are longer in 2.2. But one day we'll have it all.
You might want to file a bug so we can investigate the build time regression. Erik's been working on some infrastructure improvements to make whole-module builds more parallel, so if you want to live on the bleeding edge, you might give the 3.0 snapshots a try and see if build time is more survivable.
-Joe
···
On Feb 2, 2016, at 11:10 AM, David Turnbull via swift-users <swift-users@swift.org> wrote: