I know it can build an AI model target cpu, finally it have two part: host(cpu) and device(such as X86). if one device is MCU(single core), can iree build the an AI model for deploy at MCU? thanks.
can anybody reply this question? maybe this problem is non-professional ， thanks very much
It may be good to ask on the IREE mailing list too, just for higher visibility.
It sounds like you are asking if IREE can be used on an embedded/bare system, e.g., like the tinyIREE approach (paper on arxiv). But perhaps you’ve already seen that work and are asking a different question.
Note if it’s a custom architecture, then you’d need to utilize some lowerings somewhere (well custom VM could be used but that really depends on overheads/ granularity of offload)
thanks for your detailed guidance, just be curious after read part source code of Iree. thanks again