It's a combination of the hypervisor software that uses emulation drivers (so that it can present well known hardware abstractions to the guest OS installed) as well as hardware features like Intel's VT-x and VT-d to accelerate this mapping from the hypervisor to the actual platform.  One important distinction is whether you are talking about a type 1 or type 2 hypervisor - i.e. one that runs on bear metal or one that runs on an OS and in turn allows you to install guest OS's, most often used for testing and convenience.  A simple way to learn about this is to just download Oracle Virtual Box (https://www.virtualbox.org/) and start installing guest OS's on you native Windows/Linux/Mac-OS-X system, but this is type 2.  To learn about a bare metal type 1, you'll need to try out VmWare ESX or Xen, or others that support bare metal install.  The type 2 are great for learning, teaching, testing, and general convenience.

There are many techniques to achieve this. The hardware can be fully (full virtualization) or partly (partial virtualization) simulated. The guest OSs are generally managed by the hypervisor or virtual machine manager which may fully access the hardware (full virtualization) or control only part of it (paravirtualization). You can read some basics about virtualization here http://en.wikipedia.org/wiki/Virtualization

Perhaps a slightly more detailed description can be found at http://en.wikipedia.org/wiki/Virtual_machine - here an OS that supports multiple guest operating system images is referred to as a 'System virtual machine'.

You might find this research paper very enlightening: http://www.cs.cornell.edu/people/egs/papers/kimera-sosp99.pdf. Also see http://www.amazon.com/Virtual-Machine-Design-Implementation-Blunden/dp/1556229038.

BTW, IBM's mainframe VM OS is described here: http://en.wikipedia.org/wiki/VM_(operating_system)#Overview.

I couldn't find any clear description of IBM mainframe LPAR's that did not confuse them with VM OS's. IBM initially released its fully virtualized VM OS in 1972 - it allowed booting multiple OS systems a guest machines, each to some extent sharing nearly all of the host computer system's hardware facilities.

In 1988 IBM offered the Processor Resource/System Management facility as a hardware/firmware feature on its 3090 series of mainframe processors. This feature allowed static partitioning of a computer system's memory, peripheral I/O channels and CPU's to up to 4 logical partitions, or LPAR's (this number was later increased, but remains limited). While CPU's can be shared between partitions or dedicated, there is significant sharing overhead due to the cost of firmware context switching between partitions that do not share memory.  A rudimentary firmware dispatcher handles time-sliced and interrupt-driven partition switching.

The principal benefit of PR/SM processor partitioning compared to running multiple guest systems under the VM OS is that the system overhead incurred by firmware partitioning is much less that than that for a hypervisor OS, and OS system support is reduced (especially when VM OS support is not otherwise needed).

It's a combination of the hypervisor software that uses emulation drivers (so that it can present well known hardware abstractions to the guest OS installed) as well as hardware features like Intel's VT-x and VT-d to accelerate this mapping from the hypervisor to the actual platform.  One important distinction is whether you are talking about a type 1 or type 2 hypervisor - i.e. one that runs on bear metal or one that runs on an OS and in turn allows you to install guest OS's, most often used for testing and convenience.  A simple way to learn about this is to just download Oracle Virtual Box (https://www.virtualbox.org/) and start installing guest OS's on you native Windows/Linux/Mac-OS-X system, but this is type 2.  To learn about a bare metal type 1, you'll need to try out VmWare ESX or Xen, or others that support bare metal install.  The type 2 are great for learning, teaching, testing, and general convenience.

I recommend you to read the following 2 papers, they will help you to better understand the principles behind vitalization and explain must of the concept that you want to know.   Xen is a para virtual environment, and VMware is a fully virtual enviroment.  

Paul Barham, Boris Dragovic, Keir Fraser, Steven Hand, Tim Harris, Alex Ho, Rolf

Neugebauer, Ian Pratt, Andrew Warfield, "Xen and the Art of Virtualization ", SOSP

2003.

Carl Waldspurger, “Memory Resource Management in VMware ESX Server”, OSDI,

2002.

FWIW, Wine in your question is not a virtualization product, but rather maps the Windows API to the Linux API.