undefined | Better HN

0 pointsgeodel9y ago0 comments

All good stuff. But I have heard JNI is already multiple times faster than Go's cgo stuff so I wonder what could be improved there.

0 comments

pjmlp9y ago

According to Mark Reinhold, Sun designed JNI to be hard on purpose, to discourage developers to write native code.

He has repeated this a few times at his JavaONE presentations, a bit hard to track down which ones.

Thanks to the work of Charles Nutter on adding FFI to JRuby, and the pressure from FinTech to improve Java's mechanical sympathy, a new project was started, project Panama, where binding to native code from Java should be something like P/Invoke on the CLR. Similarly calling into Java APIs from native side should have less ceremony setting up what to call.

Additionally there should be a standard support for integrating GPGPU into Java.

As it was deemed too complex to be ready by Java 9 timeframe, it ended up being scheduled for 10+ roadmap.

For the latest status check "Going Native" at JVM Language Summit 2016.

https://www.youtube.com/watch?v=JR1zI5gLhRM&index=13&list=PL...

geodelOP9y ago

Interesting talk. Go folks also probably want cgo similarly difficult to use. Though for them I guess it will work, as they have already written compiler/GC/runtime/ssl etc in Go plus some asm.

pjmlp9y ago

I never understood why cgo was an option instead of doing what most languages do.

The only reason I see, was having an easier way to parse header files.

j / k navigate · click thread line to collapse

0 comments

pjmlp9y ago

According to Mark Reinhold, Sun designed JNI to be hard on purpose, to discourage developers to write native code.

He has repeated this a few times at his JavaONE presentations, a bit hard to track down which ones.

Additionally there should be a standard support for integrating GPGPU into Java.

As it was deemed too complex to be ready by Java 9 timeframe, it ended up being scheduled for 10+ roadmap.

For the latest status check "Going Native" at JVM Language Summit 2016.

https://www.youtube.com/watch?v=JR1zI5gLhRM&index=13&list=PL...

geodelOP9y ago

Interesting talk. Go folks also probably want cgo similarly difficult to use. Though for them I guess it will work, as they have already written compiler/GC/runtime/ssl etc in Go plus some asm.

pjmlp9y ago

I never understood why cgo was an option instead of doing what most languages do.

The only reason I see, was having an easier way to parse header files.

j / k navigate · click thread line to collapse