Fork me on GitHub
#core-matrix
<
2016-01-18
>
kephale08:01:08

@mikera: any thoughts on allowing core.matrix to be called from java?

mikera09:01:18

Hmmmm when using Java I just use Vectorz directly

mikera09:01:48

It has all the core.matrix features and you get the added benefits of type safety and faster dispatch

kephale10:01:15

the thought was to take advantage of the multiple implementations

kephale10:01:01

it might not be worth the effort of getting everything to work in java though...

mikera10:01:50

Well... you can always use the usual Java -> Clojure API

kephale10:01:34

yeah, i’m currently in germany with a bunch of image processing folks and we’re debating on migrating to a new matrix library

kephale10:01:16

since there isn’t a clear winner across the board when it comes to decompositions and solving, i’m suggesting that there might be some value in being able to swap implementations

kephale10:01:23

(java image processing folks)

kephale10:01:08

it doesn’t seem like you’ve gotten Vectorz into the running on http://lessthanoptimal.github.io/Java-Matrix-Benchmark/runtime/2015_07_XeonQuad/

kephale10:01:25

though, since you said you based it on EJML, i’m guessing the performance is relatively similar