I don’t think that we need to adopt the same stuff...
# datascience
k
I don’t think that we need to adopt the same stuff for all platforms.
t
khud: thats my thinking
Spark is not running on Android locally anytime soon
k
im more interested in supporting native than android, for the performance and interop possibilities
k
Kotlin native may provide great experience for scientific computations on desktop. But in many cases we can do the same with JVM (almost without performance loses)
k
I'd really love to be able to write numerical code and deploy at least a subset of capabilities on both. ive found for linalg computations theres a significant performance hit for going through EJML, MTJ, JBlas et al vs bare metal, especially for certain load types. but I also want the JVM in many cases for its huge library availability.
I don't really use ML personally so availability of that elsewhere wouldnt affect me, but the more that we can make things pluggable the better i think