We are all walking around with handsets that have multiple weak cores. If we can spread the general work over those cores our handset could possibly get back to a lower power mode faster.
Any advantages gained from this are going to be lost due to the lack of a highly optimizing compiler. Unless you have a lot of cores, my general experience is that you need sequential performance first and only then can you think about achieving speedups from going multicore. Compiler optimizations matter even more on ARM than on x86 due to shorter pipelines, smaller caches, super-expensive synchronization primitives, and so forth.
Of course, to be fair to Go, the primary alternative is Java (though the performance-critical parts of Android tend to be written in C++), which as implemented in Android doesn't do a whole lot of optimization either.
I think Go could be a great choice for application development, mind you. I just don't think that reaching for it to get multicore speedups will work at the moment compared to the alternatives.
Comments
Any advantages gained from this are going to be lost due to the lack of a highly optimizing compiler. Unless you have a lot of cores, my general experience is that you need sequential performance first and only then can you think about achieving speedups from going multicore. Compiler optimizations matter even more on ARM than on x86 due to shorter pipelines, smaller caches, super-expensive synchronization primitives, and so forth.
Of course, to be fair to Go, the primary alternative is Java (though the performance-critical parts of Android tend to be written in C++), which as implemented in Android doesn't do a whole lot of optimization either.
I think Go could be a great choice for application development, mind you. I just don't think that reaching for it to get multicore speedups will work at the moment compared to the alternatives.