If it's possible that the entire E1 can be compiled for a single device then it makes sense to treat it as device code. In such case, moving `alloc` to the host could be treated as an "optimization" that is specific to this target. However, if E1 has a non-composite target, how would that optimization know what "target host" to use? Where would it get it from?
--- [Visit Topic](https://discuss.tvm.ai/t/rfc-tvm-target-specification/6844/17) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.ai/email/unsubscribe/a361be618a8aab2108184ec9c04af59fe1a18d11a0262cbf5753290bbb57bfb6).
