"XCM Part III: Execution and Error Management", by Gavin Wood

In the first two articles (Part I, Part II) I wrote on XCM, I introduced the basics of its design and versioning structure. In this article, we will take a deeper look into its underlying design and execution model. Since XCM is based around the instruction set of the XCVM, a very high-level virtual machine, this amounts to becoming familiar with this machine architecture.

The XCVM is a very high level, non-Turing complete virtual machine. It is register-based (rather than stack-based) and has several special-purpose registers, most of which hold highly structured data. Unlike general-purpose processors, the XCVM’s registers are not free to be set to arbitrary values, but have strict mechanics governing how they may change. Beyond certain means of interacting with the local chain state (such as the WithdrawAsset and DepositAsset instructions which we have already seen) there is no additional “memory”. There is no possibility of looping and no explicit branch instructions.

Read full article