Is there a good way to verify whether algorithms written in renderscript are CPU or GPU bound?
I understand that this decision is made runtime on the device, however
Nexus 5 and Nexus 10 have different GPUs (and thus different GPU drivers written by completely different software teams). They are not going to accelerate the same kinds of code well, because they are very different architectures. We are trying to get some consistent logging from those drivers, but there won't be public updates for quite a while (since we don't push out new GPU drivers for AOSP very often).