Peter, yes, that is exactly what I was suggesting. It sounds like VM has this as an inherent capability; it would only need to be "decoupled" from LGR and made into a customer option.
To be clear, this is an academic thought on my part -- not something I am contemplating an RFE for. Speaking as someone with years of ISV experience, if we wanted to claim z10 compatibility, it would be nice to be able to test on a (virtual) z10. I think you can do that with a zPDT and certainly with Hercules (neglecting legalities and so forth). Why a z10? z10 is supported by z/OS V2R2, and at least at my former company, our policy was to claim support for every box that was supported by a supported level of z/OS. (Not including extended z/OS service. Our thought was that if a customer said "we're paying IBM for extended service" then we would say "then you can pay us for a back-hardware-level version." It never came up.) Charles -----Original Message----- From: IBM Mainframe Assembler List [mailto:[email protected]] On Behalf Of Peter Relson Sent: Friday, March 29, 2019 5:29 AM To: [email protected] Subject: Re: Determing the Presence of an Instruction The differentiator is typically not privileged vs non-privileged. It's interceptable vs non-interceptable. That's the "normal" state of affairs for z/VM. But many instructions (such as those in the miscellaneous instruction execution facilities that have come into play over the course of many machine generations) are not interceptable. For such instructions, z/VM sees nothing about them. But the virtual architecture level must control whether those instructions are made available or get an operation exception. It does seem like there must be some overhead with accommodating that. Anyway, back to "dial down", according to Tony T you define a relocation domain. If that domain needs to consist of actual machines, then I guess you can't do what is desired. If you could define a domain to include a machine that does not actually exist (but for which you could identify what machine it was supposed to be) that would work, because of course you would never truly relocate there. I suppose it might not let you do that... What you really want is to be able to define the virtual architecture level that a guest is to use.
